Changeset 2e9b59b

Jenkinsfile

rba897d21	r2e9b59b
108	108
109	109	// Configure libcfa
110		sh 'make -j 8 --no-print-directory configure-libcfa'
	110	sh 'make -j $(nproc) --no-print-directory configure-libcfa'
111	111	}
112	112	}
…	…
116	116	dir (BuildDir) {
117	117	// Build driver
118		sh 'make -j 8 --no-print-directory -C driver'
	118	sh 'make -j $(nproc) --no-print-directory -C driver'
119	119
120	120	// Build translator
121		sh 'make -j 8 --no-print-directory -C src'
	121	sh 'make -j $(nproc) --no-print-directory -C src'
122	122	}
123	123	}
…	…
126	126	// Build outside of the src tree to ease cleaning
127	127	dir (BuildDir) {
128		sh "make -j 8 --no-print-directory -C libcfa/${Settings.Architecture.name}-debug"
	128	sh "make -j \$(nproc) --no-print-directory -C libcfa/${Settings.Architecture.name}-debug"
129	129	}
130	130	}
…	…
133	133	// Build outside of the src tree to ease cleaning
134	134	dir (BuildDir) {
135		sh "make -j 8 --no-print-directory -C libcfa/${Settings.Architecture.name}-nodebug"
	135	sh "make -j \$(nproc) --no-print-directory -C libcfa/${Settings.Architecture.name}-nodebug"
136	136	}
137	137	}
…	…
140	140	// Build outside of the src tree to ease cleaning
141	141	dir (BuildDir) {
142		sh ~~"make -j 8 --no-print-directory install"~~
	142	sh 'make -j $(nproc) --no-print-directory install'
143	143	}
144	144	}
…	…
161	161	Tools.BuildStage('Test: full', Settings.RunAllTests) {
162	162	dir (BuildDir) {
163		jopt = ""
	163	jopt = '-j $(nproc)'
164	164	if( Settings.Architecture.node == 'x86' ) {
165		jopt = ~~"-j2"~~
	165	jopt = '-j2'
166	166	}
167	167	//Run the tests from the tests directory

benchmark/io/http/main.cfa

-              rba897d21
+              r2e9b59b
 //============================================================================================='
+thread StatsPrinter {};
+thread StatsPrinter {
+        Worker * workers;
+        int worker_cnt;
+};
 void ?{}( StatsPrinter & this, cluster & cl ) {
         ((thread&)this){ "Stats Printer Thread", cl };
+        this.worker_cnt = 0;
+}
 void ^?{}( StatsPrinter & mutex this ) {}
+#define eng3(X) (ws(3, 3, unit(eng( X ))))
 void main(StatsPrinter & this) {
 …
                 print_stats_now( *active_cluster(), CFA_STATS_READY_Q | CFA_STATS_IO );
+                if(this.worker_cnt != 0) {
+                        uint64_t tries = 0;
+                        uint64_t calls = 0;
+                        uint64_t header = 0;
+                        uint64_t splcin = 0;
+                        uint64_t splcot = 0;
+                        struct {
+                                volatile uint64_t calls;
+                                volatile uint64_t bytes;
+                        } avgrd[zipf_cnts];
+                        memset(avgrd, 0, sizeof(avgrd));
+                        for(i; this.worker_cnt) {
+                                tries += this.workers[i].stats.sendfile.tries;
+                                calls += this.workers[i].stats.sendfile.calls;
+                                header += this.workers[i].stats.sendfile.header;
+                                splcin += this.workers[i].stats.sendfile.splcin;
+                                splcot += this.workers[i].stats.sendfile.splcot;
+                                for(j; zipf_cnts) {
+                                        avgrd[j].calls += this.workers[i].stats.sendfile.avgrd[j].calls;
+                                        avgrd[j].bytes += this.workers[i].stats.sendfile.avgrd[j].bytes;
+                                }
+                        }
+                        double ratio = ((double)tries) / calls;
+                        sout | "----- Worker Stats -----";
+                        sout | "sendfile  : " | calls | "calls," | tries | "tries (" | ratio | " try/call)";
+                        sout | "            " | header | "header," | splcin | "splice in," | splcot | "splice out";
+                        sout | " - zipf sizes:";
+                        for(i; zipf_cnts) {
+                                double written = avgrd[i].calls > 0 ? ((double)avgrd[i].bytes) / avgrd[i].calls : 0;
+                                sout | "        " | zipf_sizes[i] | "bytes," | avgrd[i].calls | "shorts," | written | "written";
+                        }
+                }
+                else {
+                        sout | "No Workers!";
+                }
+        }
+}
 …
+                        {
                                 Worker * workers = anew(options.clopts.nworkers);
+                                cl[0].prnt->workers = workers;
+                                cl[0].prnt->worker_cnt = options.clopts.nworkers;
                                 for(i; options.clopts.nworkers) {
                                         // if( options.file_cache.fixed_fds ) {
 …
+        }
+}
+const size_t zipf_sizes[] = { 102, 204, 307, 409, 512, 614, 716, 819, 921, 1024, 2048, 3072, 4096, 5120, 6144, 7168, 8192, 9216, 10240, 20480, 30720, 40960, 51200, 61440, 71680, 81920, 92160, 102400, 204800, 307200, 409600, 512000, 614400, 716800, 819200, 921600 };
+static_assert(zipf_cnts == sizeof(zipf_sizes) / sizeof(zipf_sizes[0]));

benchmark/io/http/parhttperf

-              rba897d21
+              r2e9b59b
 mkdir -p out
+rm -v out/*
+rm out/*
+echo "httperf --client [0-$(($NTHREADS - 1))]/$NTHREADS $@ > out/result.[0-$(($NTHREADS - 1))].out"
 for ((i=0; i<$NTHREADS; i++))
 do
-        # echo "httperf --client $i/$NTHREADS $@ > out/result.$i.out"
         httperf --client $i/$NTHREADS $@ > out/result.$i.out &
 done

benchmark/io/http/protocol.cfa

-              rba897d21
+              r2e9b59b
 #include "options.hfa"
+#include "worker.hfa"
 #define PLAINTEXT_1WRITE
 …
                 count -= ret;
-                offset += ret;
                 size_t in_pipe = ret;
                 SPLICE2: while(in_pipe > 0) {
 …
+}
 static inline int wait_and_process(header_g & this) {
+static inline int wait_and_process(header_g & this, sendfile_stats_t & stats) {
         wait(this.f);
 …
+        }
+        stats.header++;
         // It must be a Short read
         this.len  -= this.f.result;
 …
         io_future_t f;
         int fd; int pipe; size_t len; off_t off;
+        short zipf_idx;
         FSM_Result res;
 };
 …
         this.len = len;
         this.off = 0;
+        this.zipf_idx = -1;
+        STATS: for(i; zipf_cnts) {
+                if(len <= zipf_sizes[i]) {
+                        this.zipf_idx = i;
+                        break STATS;
+                }
+        }
+        if(this.zipf_idx < 0) mutex(serr) serr | "SPLICE IN" | len | " greated than biggest zipf file";
+}
 …
+}
 static inline int wait_and_process(splice_in_t & this) {
+static inline int wait_and_process(splice_in_t & this, sendfile_stats_t & stats ) {
         wait(this.f);
 …
                         return error(this.res, -ECONNRESET);
+                }
+                mutex(serr) serr | "SPLICE IN got" | error | ", WTF!";
+                return error(this.res, -ECONNRESET);
+        }
 …
                 return done(this.res);
+        }
+        stats.splcin++;
+        stats.avgrd[this.zipf_idx].calls++;
+        stats.avgrd[this.zipf_idx].bytes += this.f.result;
         // It must be a Short read
 …
+}
 static inline void wait_and_process(splice_out_g & this) {
+static inline void wait_and_process(splice_out_g & this, sendfile_stats_t & stats ) {
         wait(this.f);
 …
                         return error(this, -ECONNRESET);
+                }
+                mutex(serr) serr | "SPLICE OUT got" | error | ", WTF!";
+                return error(this, -ECONNRESET);
+        }
 …
 SHORT_WRITE:
+        stats.splcot++;
         // It must be a Short Write
         this.len -= this.f.result;
 …
+}
+int answer_sendfile( int pipe[2], int fd, int ans_fd, size_t fsize ) {
+int answer_sendfile( int pipe[2], int fd, int ans_fd, size_t fsize, sendfile_stats_t & stats ) {
+        stats.calls++;
         #if defined(LINKED_IO)
                 char buffer[512];
 …
                 RETRY_LOOP: for() {
+                        stats.tries++;
                         int have = need(header.res) + need(splice_in.res) + 1;
                         int idx = 0;
 …
                         // we may need to kill the connection if it fails
                         // If it already completed, this is a no-op
                         wait_and_process(splice_in);
+                        wait_and_process(splice_in, stats);
                         if(is_error(splice_in.res)) {
 …
                         // Process the other 2
                         wait_and_process(header);
                         wait_and_process(splice_out);
+                        wait_and_process(header, stats);
+                        wait_and_process(splice_out, stats);
                         if(is_done(splice_out.res)) {
 …
                 return len + fsize;
         #else
+                stats.tries++;
                 int ret = answer_header(fd, fsize);
                 if( ret < 0 ) { close(fd); return ret; }

benchmark/io/http/protocol.hfa

-              rba897d21
+              r2e9b59b
 #pragma once
+struct sendfile_stats_t;
 enum HttpCode {
 …
 int answer_plaintext( int fd );
 int answer_empty( int fd );
 int answer_sendfile( int pipe[2], int fd, int ans_fd, size_t count );
+int answer_sendfile( int pipe[2], int fd, int ans_fd, size_t count, struct sendfile_stats_t & );
 [HttpCode code, bool closed, * const char file, size_t len] http_read(int fd, []char buffer, size_t len);

benchmark/io/http/worker.cfa

-              rba897d21
+              r2e9b59b
         this.pipe[1] = -1;
         this.done = false;
+        this.stats.sendfile.calls = 0;
+        this.stats.sendfile.tries = 0;
+        this.stats.sendfile.header = 0;
+        this.stats.sendfile.splcin = 0;
+        this.stats.sendfile.splcot = 0;
+        for(i; zipf_cnts) {
+                this.stats.sendfile.avgrd[i].calls = 0;
+                this.stats.sendfile.avgrd[i].bytes = 0;
+        }
+}
 …
                         // Send the desired file
                         int ret = answer_sendfile( this.pipe, fd, ans_fd, count);
+                        int ret = answer_sendfile( this.pipe, fd, ans_fd, count, this.stats.sendfile );
                         if( ret == -ECONNRESET ) break REQUEST;

benchmark/io/http/worker.hfa

-              rba897d21
+              r2e9b59b
 //=============================================================================================
+extern const size_t zipf_sizes[];
+enum { zipf_cnts = 36, };
+struct sendfile_stats_t {
+        volatile uint64_t calls;
+        volatile uint64_t tries;
+        volatile uint64_t header;
+        volatile uint64_t splcin;
+        volatile uint64_t splcot;
+        struct {
+                volatile uint64_t calls;
+                volatile uint64_t bytes;
+        } avgrd[zipf_cnts];
+};
 thread Worker {
         int pipe[2];
 …
         int flags;
         volatile bool done;
+        struct {
+                sendfile_stats_t sendfile;
+        } stats;
 };
 void ?{}( Worker & this);

benchmark/plot.py

-              rba897d21
+              r2e9b59b
+}
 def plot(data, x, y):
+def plot(data, x, y, out):
         fig, ax = plt.subplots()
         colors = itertools.cycle(['#0095e3','#006cb4','#69df00','#0aa000','#fb0300','#e30002','#fd8f00','#ff7f00','#8f00d6','#4b009a','#ffff00','#b13f00'])
 …
         ax.yaxis.set_major_formatter( EngFormatter(unit=field_names[y].unit) )
         plt.legend(loc='upper left')
+        plt.show()
+        if out:
+                plt.savefig(out)
+        else:
+                plt.show()
 …
         parser = parser = argparse.ArgumentParser(description='Python Script to draw R.M.I.T. results')
         parser.add_argument('-f', '--file', nargs='?', type=argparse.FileType('r'), default=sys.stdin)
+        parser.add_argument('-o', '--out', nargs='?', type=str, default=None)
+        parser.add_argument('-y', nargs='?', type=str, default="")
         try:
 …
                         fields.add(label)
+        print(series)
+        print("fields")
+        for f in fields:
+                print("{}".format(f))
+        if not options.out :
+                print(series)
+                print("fields")
+                for f in fields:
+                        print("{}".format(f))
+        plot(data, "Number of processors", "ns per ops")
+        if options.y and options.y in field_names.keys():
+                plot(data, "Number of processors", options.y, options.out)
+        else:
+                if options.y:
+                        print("Could not find key '{}', defaulting to 'ns per ops'".format(options.y))
+                plot(data, "Number of processors", "ns per ops", options.out)

doc/LaTeXmacros/common.sty

-              rba897d21
+              r2e9b59b
 %% Created On       : Sat Apr  9 10:06:17 2016
 %% Last Modified By : Peter A. Buhr
 %% Last Modified On : Mon Feb  7 23:00:46 2022
 %% Update Count     : 569
+%% Last Modified On : Sat Apr  2 17:35:23 2022
+%% Update Count     : 570
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 …
   {=>}{$\Rightarrow$}2
   {->}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex\textgreater}2,
-defaultdialect={CFA},
 }% lstset
 }% CFAStyle

doc/LaTeXmacros/common.tex

-              rba897d21
+              r2e9b59b
 %% Created On       : Sat Apr  9 10:06:17 2016
 %% Last Modified By : Peter A. Buhr
 %% Last Modified On : Mon Feb  7 23:00:08 2022
 %% Update Count     : 552
+%% Last Modified On : Sat Apr  2 16:42:31 2022
+%% Update Count     : 553
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 …
   {=>}{$\Rightarrow$}2
   {->}{\makebox[1ex][c]{\raisebox{0.4ex}{\rule{0.8ex}{0.075ex}}}\kern-0.2ex\textgreater}2,
-defaultdialect={CFA},
 }% lstset
 }% CFAStyle

doc/theses/mubeen_zulfiqar_MMath/Makefile

-              rba897d21
+              r2e9b59b
+# directory for latex clutter files
+# Configuration variables
 Build = build
 Figures = figures
 Pictures = pictures
+LaTMac = ../../LaTeXmacros
+BibRep = ../../bibliography
 TeXSRC = ${wildcard *.tex}
 FigSRC = ${notdir ${wildcard ${Figures}/*.fig}}
 PicSRC = ${notdir ${wildcard ${Pictures}/*.fig}}
+BIBSRC = ${wildcard *.bib}
+TeXLIB = .:../../LaTeXmacros:${Build}: # common latex macros
+BibLIB = .:../../bibliography # common citation repository
+BibSRC = ${wildcard *.bib}
+TeXLIB = .:${LaTMac}:${Build}:
+BibLIB = .:${BibRep}:
 MAKEFLAGS = --no-print-directory # --silent
 VPATH = ${Build} ${Figures} ${Pictures} # extra search path for file names used in document
+### Special Rules:
+DOCUMENT = uw-ethesis.pdf
+BASE = ${basename ${DOCUMENT}}                  # remove suffix
+.PHONY: all clean
+.PRECIOUS: %.dvi %.ps # do not delete intermediate files
+### Commands:
+# Commands
 LaTeX = TEXINPUTS=${TeXLIB} && export TEXINPUTS && latex -halt-on-error -output-directory=${Build}
 BibTeX = BIBINPUTS=${BibLIB} bibtex
+BibTeX = BIBINPUTS=${BibLIB} && export BIBINPUTS && bibtex
 #Glossary = INDEXSTYLE=${Build} makeglossaries-lite
 ### Rules and Recipes:
+# Rules and Recipes
+DOC = uw-ethesis.pdf
+BASE = ${DOC:%.pdf=%} # remove suffix
+.PHONY : all clean                              # not file names
+.PRECIOUS: %.dvi %.ps # do not delete intermediate files
+.ONESHELL :
 all: ${DOC}
+all : ${DOCUMENT}
 clean:
         @rm -frv ${DOC} ${Build}
+clean :
+        @rm -frv ${DOCUMENT} ${Build}
 # File Dependencies #
+# File Dependencies
 ${Build}/%.dvi : ${TeXSRC} ${FigSRC:%.fig=%.tex} ${PicSRC:%.fig=%.pstex} ${BIBSRC} Makefile | ${Build}
+%.dvi : ${TeXSRC} ${FigSRC:%.fig=%.tex} ${PicSRC:%.fig=%.pstex} ${BibSRC} ${BibRep}/pl.bib ${LaTMac}/common.tex Makefile | ${Build}
         ${LaTeX} ${BASE}
         ${BibTeX} ${Build}/${BASE}
         ${LaTeX} ${BASE}
         # if nedded, run latex again to get citations
+        # if needed, run latex again to get citations
         if fgrep -s "LaTeX Warning: Citation" ${basename $@}.log ; then ${LaTeX} ${BASE} ; fi
 #       ${Glossary} ${Build}/${BASE}
 …
 ${Build}:
         mkdir $@
+        mkdir -p $@
 %.pdf : ${Build}/%.ps | ${Build}

doc/theses/mubeen_zulfiqar_MMath/allocator.tex

-              rba897d21
+              r2e9b59b
 \chapter{Allocator}
+\section{uHeap}
+uHeap is a lightweight memory allocator. The objective behind uHeap is to design a minimal concurrent memory allocator that has new features and also fulfills GNU C Library requirements (FIX ME: cite requirements).
+The objective of uHeap's new design was to fulfill following requirements:
+\begin{itemize}
+\item It should be concurrent and thread-safe for multi-threaded programs.
+\item It should avoid global locks, on resources shared across all threads, as much as possible.
+\item It's performance (FIX ME: cite performance benchmarks) should be comparable to the commonly used allocators (FIX ME: cite common allocators).
+\item It should be a lightweight memory allocator.
+\end{itemize}
+This chapter presents a new stand-alone concurrent low-latency memory-allocator ($\approx$1,200 lines of code), called llheap (low-latency heap), for C/\CC programs using kernel threads (1:1 threading), and specialized versions of the allocator for the programming languages \uC and \CFA using user-level threads running over multiple kernel threads (M:N threading).
+The new allocator fulfills the GNU C Library allocator API~\cite{GNUallocAPI}.
+\section{llheap}
+The primary design objective for llheap is low-latency across all allocator calls independent of application access-patterns and/or number of threads, \ie very seldom does the allocator have a delay during an allocator call.
+(Large allocations requiring initialization, \eg zero fill, and/or copying are not covered by the low-latency objective.)
+A direct consequence of this objective is very simple or no storage coalescing;
+hence, llheap's design is willing to use more storage to lower latency.
+This objective is apropos because systems research and industrial applications are striving for low latency and computers have huge amounts of RAM memory.
+Finally, llheap's performance should be comparable with the current best allocators (see performance comparison in \VRef[Chapter]{c:Performance}).
+% The objective of llheap's new design was to fulfill following requirements:
+% \begin{itemize}
+% \item It should be concurrent and thread-safe for multi-threaded programs.
+% \item It should avoid global locks, on resources shared across all threads, as much as possible.
+% \item It's performance (FIX ME: cite performance benchmarks) should be comparable to the commonly used allocators (FIX ME: cite common allocators).
+% \item It should be a lightweight memory allocator.
+% \end{itemize}
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+<<<<<<< HEAD
 \section{Design choices for uHeap}\label{sec:allocatorSec}
 uHeap's design was reviewed and changed to fulfill new requirements (FIX ME: cite allocator philosophy). For this purpose, following two designs of uHeapLmm were proposed:
+\paragraph{Design 1: Centralized}
+One heap, but lower bucket sizes are N-shared across KTs.
+This design leverages the fact that 95\% of allocation requests are less than 512 bytes and there are only 3--5 different request sizes.
+When KTs $\le$ N, the important bucket sizes are uncontented.
+When KTs $>$ N, the free buckets are contented.
+Therefore, threads are only contending for a small number of buckets, which are distributed among them to reduce contention.
+\begin{cquote}
+=======
+\section{Design Choices}
+>>>>>>> bb7c77dc425e289ed60aa638529b3e5c7c3e4961
+llheap's design was reviewed and changed multiple times throughout the thesis.
+Some of the rejected designs are discussed because they show the path to the final design (see discussion in \VRef{s:MultipleHeaps}).
+Note, a few simples tests for a design choice were compared with the current best allocators to determine the viability of a design.
+\subsection{Allocation Fastpath}
+\label{s:AllocationFastpath}
+These designs look at the allocation/free \newterm{fastpath}, \ie when an allocation can immediately return free storage or returned storage is not coalesced.
+\paragraph{T:1 model}
+\VRef[Figure]{f:T1SharedBuckets} shows one heap accessed by multiple kernel threads (KTs) using a bucket array, where smaller bucket sizes are N-shared across KTs.
+This design leverages the fact that 95\% of allocation requests are less than 1024 bytes and there are only 3--5 different request sizes.
+When KTs $\le$ N, the common bucket sizes are uncontented;
+when KTs $>$ N, the free buckets are contented and latency increases significantly.
+In all cases, a KT must acquire/release a lock, contented or uncontented, along the fast allocation path because a bucket is shared.
+Therefore, while threads are contending for a small number of buckets sizes, the buckets are distributed among them to reduce contention, which lowers latency;
+however, picking N is workload specific.
+\begin{figure}
+\centering
+\input{AllocDS1}
+\caption{T:1 with Shared Buckets}
+\label{f:T1SharedBuckets}
+\end{figure}
+Problems:
+\begin{itemize}
+\item
+Need to know when a KT is created/destroyed to assign/unassign a shared bucket-number from the memory allocator.
+\item
+When no thread is assigned a bucket number, its free storage is unavailable.
+\item
+All KTs contend for the global-pool lock for initial allocations, before free-lists get populated.
+\end{itemize}
+Tests showed having locks along the allocation fast-path produced a significant increase in allocation costs and any contention among KTs produces a significant spike in latency.
+\paragraph{T:H model}
+\VRef[Figure]{f:THSharedHeaps} shows a fixed number of heaps (N), each a local free pool, where the heaps are sharded across the KTs.
+A KT can point directly to its assigned heap or indirectly through the corresponding heap bucket.
+When KT $\le$ N, the heaps are uncontented;
+when KTs $>$ N, the heaps are contented.
+In all cases, a KT must acquire/release a lock, contented or uncontented along the fast allocation path because a heap is shared.
+By adjusting N upwards, this approach reduces contention but increases storage (time versus space);
+however, picking N is workload specific.
+\begin{figure}
 \centering
 \input{AllocDS2}
+\end{cquote}
+Problems: need to know when a kernel thread (KT) is created and destroyed to know when to assign a shared bucket-number.
+When no thread is assigned a bucket number, its free storage is unavailable. All KTs will be contended for one lock on sbrk for their initial allocations (before free-lists gets populated).
+\paragraph{Design 2: Decentralized N Heaps}
+Fixed number of heaps: shard the heap into N heaps each with a bump-area allocated from the @sbrk@ area.
+Kernel threads (KT) are assigned to the N heaps.
+When KTs $\le$ N, the heaps are uncontented.
+When KTs $>$ N, the heaps are contented.
+By adjusting N, this approach reduces storage at the cost of speed due to contention.
+In all cases, a thread acquires/releases a lock, contented or uncontented.
+\begin{cquote}
+\centering
+\input{AllocDS1}
+\end{cquote}
+Problems: need to know when a KT is created and destroyed to know when to assign/un-assign a heap to the KT.
+\paragraph{Design 3: Decentralized Per-thread Heaps}
+Design 3 is similar to design 2 but instead of having an M:N model, it uses a 1:1 model. So, instead of having N heaos and sharing them among M KTs, Design 3 has one heap for each KT.
+Dynamic number of heaps: create a thread-local heap for each kernel thread (KT) with a bump-area allocated from the @sbrk@ area.
+Each KT will have its own exclusive thread-local heap. Heap will be uncontended between KTs regardless how many KTs have been created.
+Operations on @sbrk@ area will still be protected by locks.
+%\begin{cquote}
+%\centering
+%\input{AllocDS3} FIXME add figs
+%\end{cquote}
+Problems: We cannot destroy the heap when a KT exits because our dynamic objects have ownership and they are returned to the heap that created them when the program frees a dynamic object. All dynamic objects point back to their owner heap. If a thread A creates an object O, passes it to another thread B, and A itself exits. When B will free object O, O should return to A's heap so A's heap should be preserved for the lifetime of the whole program as their might be objects in-use of other threads that were allocated by A. Also, we need to know when a KT is created and destroyed to know when to create/destroy a heap for the KT.
+\paragraph{Design 4: Decentralized Per-CPU Heaps}
+Design 4 is similar to Design 3 but instead of having a heap for each thread, it creates a heap for each CPU.
+Fixed number of heaps for a machine: create a heap for each CPU with a bump-area allocated from the @sbrk@ area.
+Each CPU will have its own CPU-local heap. When the program does a dynamic memory operation, it will be entertained by the heap of the CPU where the process is currently running on.
+Each CPU will have its own exclusive heap. Just like Design 3(FIXME cite), heap will be uncontended between KTs regardless how many KTs have been created.
+Operations on @sbrk@ area will still be protected by locks.
+To deal with preemtion during a dynamic memory operation, librseq(FIXME cite) will be used to make sure that the whole dynamic memory operation completes on one CPU. librseq's restartable sequences can make it possible to re-run a critical section and undo the current writes if a preemption happened during the critical section's execution.
+%\begin{cquote}
+%\centering
+%\input{AllocDS4} FIXME add figs
+%\end{cquote}
+Problems: This approach was slower than the per-thread model. Also, librseq does not provide such restartable sequences to detect preemtions in user-level threading system which is important to us as CFA(FIXME cite) has its own threading system that we want to support.
+Out of the four designs, Design 3 was chosen because of the following reasons.
+\begin{itemize}
+\item
+Decentralized designes are better in general as compared to centralized design because their concurrency is better across all bucket-sizes as design 1 shards a few buckets of selected sizes while other designs shards all the buckets. Decentralized designes shard the whole heap which has all the buckets with the addition of sharding sbrk area. So Design 1 was eliminated.
+\item
+Design 2 was eliminated because it has a possibility of contention in-case of KT > N while Design 3 and 4 have no contention in any scenerio.
+\item
+Design 4 was eliminated because it was slower than Design 3 and it provided no way to achieve user-threading safety using librseq. We had to use CFA interruption handling to achive user-threading safety which has some cost to it. Desing 4 was already slower than Design 3, adding cost of interruption handling on top of that would have made it even slower.
+\end{itemize}
+\subsection{Advantages of distributed design}
+The distributed design of uHeap is concurrent to work in multi-threaded applications.
+Some key benefits of the distributed design of uHeap are as follows:
+\begin{itemize}
+\item
+The bump allocation is concurrent as memory taken from sbrk is sharded across all heaps as bump allocation reserve. The call to sbrk will be protected using locks but bump allocation (on memory taken from sbrk) will not be contended once the sbrk call has returned.
+\item
+Low or almost no contention on heap resources.
+\item
+It is possible to use sharing and stealing techniques to share/find unused storage, when a free list is unused or empty.
+\item
+Distributed design avoids unnecassry locks on resources shared across all KTs.
+\end{itemize}
+\caption{T:H with Shared Heaps}
+\label{f:THSharedHeaps}
+\end{figure}
+Problems:
+\begin{itemize}
+\item
+Need to know when a KT is created/destroyed to assign/unassign a heap from the memory allocator.
+\item
+When no thread is assigned to a heap, its free storage is unavailable.
+\item
+Ownership issues arise (see \VRef{s:Ownership}).
+\item
+All KTs contend for the local/global-pool lock for initial allocations, before free-lists get populated.
+\end{itemize}
+Tests showed having locks along the allocation fast-path produced a significant increase in allocation costs and any contention among KTs produces a significant spike in latency.
+\paragraph{T:H model, H = number of CPUs}
+This design is the T:H model but H is set to the number of CPUs on the computer or the number restricted to an application, \eg via @taskset@.
+(See \VRef[Figure]{f:THSharedHeaps} but with a heap bucket per CPU.)
+Hence, each CPU logically has its own private heap and local pool.
+A memory operation is serviced from the heap associated with the CPU executing the operation.
+This approach removes fastpath locking and contention, regardless of the number of KTs mapped across the CPUs, because only one KT is running on each CPU at a time (modulo operations on the global pool and ownership).
+This approach is essentially an M:N approach where M is the number if KTs and N is the number of CPUs.
+Problems:
+\begin{itemize}
+\item
+Need to know when a CPU is added/removed from the @taskset@.
+\item
+Need a fast way to determine the CPU a KT is executing on to access the appropriate heap.
+\item
+Need to prevent preemption during a dynamic memory operation because of the \newterm{serially-reusable problem}.
+\begin{quote}
+A sequence of code that is guaranteed to run to completion before being invoked to accept another input is called serially-reusable code.~\cite{SeriallyReusable}
+\end{quote}
+If a KT is preempted during an allocation operation, the operating system can schedule another KT on the same CPU, which can begin an allocation operation before the previous operation associated with this CPU has completed, invalidating heap correctness.
+Note, the serially-reusable problem can occur in sequential programs with preemption, if the signal handler calls the preempted function, unless the function is serially reusable.
+Essentially, the serially-reusable problem is a race condition on an unprotected critical section, where the operating system is providing the second thread via the signal handler.
+Library @librseq@~\cite{librseq} was used to perform a fast determination of the CPU and to ensure all memory operations complete on one CPU using @librseq@'s restartable sequences, which restart the critical section after undoing its writes, if the critical section is preempted.
+\end{itemize}
+Tests showed that @librseq@ can determine the particular CPU quickly but setting up the restartable critical-section along the allocation fast-path produced a significant increase in allocation costs.
+Also, the number of undoable writes in @librseq@ is limited and restartable sequences cannot deal with user-level thread (UT) migration across KTs.
+For example, UT$_1$ is executing a memory operation by KT$_1$ on CPU$_1$ and a time-slice preemption occurs.
+The signal handler context switches UT$_1$ onto the user-level ready-queue and starts running UT$_2$ on KT$_1$, which immediately calls a memory operation.
+Since KT$_1$ is still executing on CPU$_1$, @librseq@ takes no action because it assumes KT$_1$ is still executing the same critical section.
+Then UT$_1$ is scheduled onto KT$_2$ by the user-level scheduler, and its memory operation continues in parallel with UT$_2$ using references into the heap associated with CPU$_1$, which corrupts CPU$_1$'s heap.
+If @librseq@ had an @rseq_abort@ which:
+\begin{enumerate}
+\item
+Marked the current restartable critical-section as cancelled so it restarts when attempting to commit.
+\item
+Do nothing if there is no current restartable critical section in progress.
+\end{enumerate}
+Then @rseq_abort@ could be called on the backside of a  user-level context-switching.
+A feature similar to this idea might exist for hardware transactional-memory.
+A significant effort was made to make this approach work but its complexity, lack of robustness, and performance costs resulted in its rejection.
+\paragraph{1:1 model}
+This design is the T:H model with T = H, where there is one thread-local heap for each KT.
+(See \VRef[Figure]{f:THSharedHeaps} but with a heap bucket per KT and no bucket or local-pool lock.)
+Hence, immediately after a KT starts, its heap is created and just before a KT terminates, its heap is (logically) deleted.
+Heaps are uncontended for a KTs memory operations to its heap (modulo operations on the global pool and ownership).
+Problems:
+\begin{itemize}
+\item
+Need to know when a KT is starts/terminates to create/delete its heap.
+\noindent
+It is possible to leverage constructors/destructors for thread-local objects to get a general handle on when a KT starts/terminates.
+\item
+There is a classic \newterm{memory-reclamation} problem for ownership because storage passed to another thread can be returned to a terminated heap.
+\noindent
+The classic solution only deletes a heap after all referents are returned, which is complex.
+The cheap alternative is for heaps to persist for program duration to handle outstanding referent frees.
+If old referents return storage to a terminated heap, it is handled in the same way as an active heap.
+To prevent heap blowup, terminated heaps can be reused by new KTs, where a reused heap may be populated with free storage from a prior KT (external fragmentation).
+In most cases, heap blowup is not a problem because programs have a small allocation set-size, so the free storage from a prior KT is apropos for a new KT.
+\item
+There can be significant external fragmentation as the number of KTs increases.
+\noindent
+In many concurrent applications, good performance is achieved with the number of KTs proportional to the number of CPUs.
+Since the number of CPUs is relatively small, >~1024, and a heap relatively small, $\approx$10K bytes (not including any associated freed storage), the worst-case external fragmentation is still small compared to the RAM available on large servers with many CPUs.
+\item
+There is the same serially-reusable problem with UTs migrating across KTs.
+\end{itemize}
+Tests showed this design produced the closest performance match with the best current allocators, and code inspection showed most of these allocators use different variations of this approach.
+\vspace{5pt}
+\noindent
+The conclusion from this design exercise is: any atomic fence, atomic instruction (lock free), or lock along the allocation fastpath produces significant slowdown.
+For the T:1 and T:H models, locking must exist along the allocation fastpath because the buckets or heaps maybe shared by multiple threads, even when KTs $\le$ N.
+For the T:H=CPU and 1:1 models, locking is eliminated along the allocation fastpath.
+However, T:H=CPU has poor operating-system support to determine the CPU id (heap id) and prevent the serially-reusable problem for KTs.
+More operating system support is required to make this model viable, but there is still the serially-reusable problem with user-level threading.
+Leaving the 1:1 model with no atomic actions along the fastpath and no special operating-system support required.
+The 1:1 model still has the serially-reusable problem with user-level threading, which is addressed in \VRef{s:UserlevelThreadingSupport}, and the greatest potential for heap blowup for certain allocation patterns.
+% \begin{itemize}
+% \item
+% A decentralized design is better to centralized design because their concurrency is better across all bucket-sizes as design 1 shards a few buckets of selected sizes while other designs shards all the buckets. Decentralized designs shard the whole heap which has all the buckets with the addition of sharding @sbrk@ area. So Design 1 was eliminated.
+% \item
+% Design 2 was eliminated because it has a possibility of contention in-case of KT > N while Design 3 and 4 have no contention in any scenario.
+% \item
+% Design 3 was eliminated because it was slower than Design 4 and it provided no way to achieve user-threading safety using librseq. We had to use CFA interruption handling to achieve user-threading safety which has some cost to it.
+% that  because of 4 was already slower than Design 3, adding cost of interruption handling on top of that would have made it even slower.
+% \end{itemize}
+% Of the four designs for a low-latency memory allocator, the 1:1 model was chosen for the following reasons:
+% \subsection{Advantages of distributed design}
+%
+% The distributed design of llheap is concurrent to work in multi-threaded applications.
+% Some key benefits of the distributed design of llheap are as follows:
+% \begin{itemize}
+% \item
+% The bump allocation is concurrent as memory taken from @sbrk@ is sharded across all heaps as bump allocation reserve. The call to @sbrk@ will be protected using locks but bump allocation (on memory taken from @sbrk@) will not be contended once the @sbrk@ call has returned.
+% \item
+% Low or almost no contention on heap resources.
+% \item
+% It is possible to use sharing and stealing techniques to share/find unused storage, when a free list is unused or empty.
+% \item
+% Distributed design avoids unnecessary locks on resources shared across all KTs.
+% \end{itemize}
+\subsection{Allocation Latency}
+A primary goal of llheap is low latency.
+Two forms of latency are internal and external.
+Internal latency is the time to perform an allocation, while external latency is time to obtain/return storage from/to the operating system.
+Ideally latency is $O(1)$ with a small constant.
+To obtain $O(1)$ internal latency means no searching on the allocation fastpath, largely prohibits coalescing, which leads to external fragmentation.
+The mitigating factor is that most programs have well behaved allocation patterns, where the majority of allocation operations can be $O(1)$, and heap blowup does not occur without coalescing (although the allocation footprint may be slightly larger).
+To obtain $O(1)$ external latency means obtaining one large storage area from the operating system and subdividing it across all program allocations, which requires a good guess at the program storage high-watermark and potential large external fragmentation.
+Excluding real-time operating-systems, operating-system operations are unbounded, and hence some external latency is unavoidable.
+The mitigating factor is that operating-system calls can often be reduced if a programmer has a sense of the storage high-watermark and the allocator is capable of using this information (see @malloc_expansion@ \VPageref{p:malloc_expansion}).
+Furthermore, while operating-system calls are unbounded, many are now reasonably fast, so their latency is tolerable and infrequent.
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{uHeap Structure}
+As described in (FIXME cite 2.4) uHeap uses following features of multi-threaded memory allocators.
+\begin{itemize}
+\item
+uHeap has multiple heaps without a global heap and uses 1:1 model. (FIXME cite 2.5 1:1 model)
+\item
+uHeap uses object ownership. (FIXME cite 2.5.2)
+\item
+uHeap does not use object containers (FIXME cite 2.6) or any coalescing technique. Instead each dynamic object allocated by uHeap has a header than contains bookkeeping information.
+\item
+Each thread-local heap in uHeap has its own allocation buffer that is taken from the system using sbrk() call. (FIXME cite 2.7)
+\item
+Unless a heap is freeing an object that is owned by another thread's heap or heap is using sbrk() system call, uHeap is mostly lock-free which eliminates most of the contention on shared resources. (FIXME cite 2.8)
+\end{itemize}
+As uHeap uses a heap per-thread model to reduce contention on heap resources, we manage a list of heaps (heap-list) that can be used by threads. The list is empty at the start of the program. When a kernel thread (KT) is created, we check if heap-list is empty. If no then a heap is removed from the heap-list and is given to this new KT to use exclusively. If yes then a new heap object is created in dynamic memory and is given to this new KT to use exclusively. When a KT exits, its heap is not destroyed but instead its heap is put on the heap-list and is ready to be reused by new KTs.
+This reduces the memory footprint as the objects on free-lists of a KT that has exited can be reused by a new KT. Also, we preserve all the heaps that were created during the lifetime of the program till the end of the program. uHeap uses object ownership where an object is freed to the free-buckets of the heap that allocated it. Even after a KT A has exited, its heap has to be preserved as there might be objects in-use of other threads that were initially allocated by A and the passed to other threads.
+\section{llheap Structure}
+\VRef[Figure]{f:llheapStructure} shows the design of llheap, which uses the following features:
+\begin{itemize}
+\item
+:1 multiple-heap model to minimize the fastpath,
+\item
+can be built with or without heap ownership,
+\item
+headers per allocation versus containers,
+\item
+no coalescing to minimize latency,
+\item
+global heap memory (pool) obtained from the operating system using @mmap@ to create and reuse heaps needed by threads,
+\item
+local reserved memory (pool) per heap obtained from global pool,
+\item
+global reserved memory (pool) obtained from the operating system using @sbrk@ call,
+\item
+optional fast-lookup table for converting allocation requests into bucket sizes,
+\item
+optional statistic-counters table for accumulating counts of allocation operations.
+\end{itemize}
 \begin{figure}
 \centering
+<<<<<<< HEAD
 \includegraphics[width=0.65\textwidth]{figures/NewHeapStructure.eps}
 \caption{uHeap Structure}
 \label{fig:heapStructureFig}
+=======
+% \includegraphics[width=0.65\textwidth]{figures/NewHeapStructure.eps}
+\input{llheap}
+\caption{llheap Structure}
+\label{f:llheapStructure}
+>>>>>>> bb7c77dc425e289ed60aa638529b3e5c7c3e4961
 \end{figure}
+Each heap uses seggregated free-buckets that have free objects of a specific size. Each free-bucket of a specific size has following 2 lists in it:
+\begin{itemize}
+\item
+Free list is used when a thread is freeing an object that is owned by its own heap so free list does not use any locks/atomic-operations as it is only used by the owner KT.
+\item
+Away list is used when a thread A is freeing an object that is owned by another KT B's heap. This object should be freed to the owner heap (B's heap) so A will place the object on the away list of B. Away list is lock protected as it is shared by all other threads.
+\end{itemize}
+When a dynamic object of a size S is requested. The thread-local heap will check if S is greater than or equal to the mmap threshhold. Any request larger than the mmap threshhold is fulfilled by allocating an mmap area of that size and such requests are not allocated on sbrk area. The value of this threshhold can be changed using mallopt routine but the new value should not be larger than our biggest free-bucket size.
+Algorithm~\ref{alg:heapObjectAlloc} briefly shows how an allocation request is fulfilled.
+\begin{algorithm}
+\caption{Dynamic object allocation of size S}\label{alg:heapObjectAlloc}
+llheap starts by creating an array of $N$ global heaps from storage obtained using @mmap@, where $N$ is the number of computer cores, that persists for program duration.
+There is a global bump-pointer to the next free heap in the array.
+When this array is exhausted, another array is allocated.
+There is a global top pointer for a heap intrusive link to chain free heaps from terminated threads.
+When statistics are turned on, there is a global top pointer for a heap intrusive link to chain \emph{all} the heaps, which is traversed to accumulate statistics counters across heaps using @malloc_stats@.
+When a KT starts, a heap is allocated from the current array for exclusive use by the KT.
+When a KT terminates, its heap is chained onto the heap free-list for reuse by a new KT, which prevents unbounded growth of heaps.
+The free heaps is a stack so hot storage is reused first.
+Preserving all heaps created during the program lifetime, solves the storage lifetime problem, when ownership is used.
+This approach wastes storage if a large number of KTs are created/terminated at program start and then the program continues sequentially.
+llheap can be configured with object ownership, where an object is freed to the heap from which it is allocated, or object no-ownership, where an object is freed to the KT's current heap.
+Each heap uses segregated free-buckets that have free objects distributed across 91 different sizes from 16 to 4M.
+The number of buckets used is determined dynamically depending on the crossover point from @sbrk@ to @mmap@ allocation using @mallopt( M_MMAP_THRESHOLD )@, \ie small objects managed by the program and large objects managed by the operating system.
+Each free bucket of a specific size has the following two lists:
+\begin{itemize}
+\item
+A free stack used solely by the KT heap-owner, so push/pop operations do not require locking.
+The free objects are a stack so hot storage is reused first.
+\item
+For ownership, a shared away-stack for KTs to return storage allocated by other KTs, so push/pop operations require locking.
+When the free stack is empty, the entire ownership stack is removed and becomes the head of the corresponding free stack.
+\end{itemize}
+Algorithm~\ref{alg:heapObjectAlloc} shows the allocation outline for an object of size $S$.
+First, the allocation is divided into small (@sbrk@) or large (@mmap@).
+For large allocations, the storage is mapped directly from the operating system.
+For small allocations, $S$ is quantized into a bucket size.
+Quantizing is performed using a binary search over the ordered bucket array.
+An optional optimization is fast lookup $O(1)$ for sizes < 64K from a 64K array of type @char@, where each element has an index to the corresponding bucket.
+(Type @char@ restricts the number of bucket sizes to 256.)
+For $S$ > 64K, a binary search is used.
+Then, the allocation storage is obtained from the following locations (in order), with increasing latency.
+\begin{enumerate}[topsep=0pt,itemsep=0pt,parsep=0pt]
+\item
+bucket's free stack,
+\item
+bucket's away stack,
+\item
+heap's local pool
+\item
+global pool
+\item
+operating system (@sbrk@)
+\end{enumerate}
+\begin{figure}
+\vspace*{-10pt}
+\begin{algorithm}[H]
+\small
+\caption{Dynamic object allocation of size $S$}\label{alg:heapObjectAlloc}
 \begin{algorithmic}[1]
 \State $\textit{O} \gets \text{NULL}$
+\If {$S < \textit{mmap-threshhold}$}
+        \State $\textit{B} \gets (\text{smallest free-bucket} \geq S)$
+\If {$S >= \textit{mmap-threshhold}$}
+        \State $\textit{O} \gets \text{allocate dynamic memory using system call mmap with size S}$
+\Else
+        \State $\textit{B} \gets \text{smallest free-bucket} \geq S$
         \If {$\textit{B's free-list is empty}$}
                 \If {$\textit{B's away-list is empty}$}
                         \If {$\textit{heap's allocation buffer} < S$}
                                 \State $\text{get allocation buffer using system call sbrk()}$
+                                \State $\text{get allocation from global pool (which might call \lstinline{sbrk})}$
                         \EndIf
                         \State $\textit{O} \gets \text{bump allocate an object of size S from allocation buffer}$
 …
         \EndIf
         \State $\textit{O's owner} \gets \text{B}$
-\Else
-        \State $\textit{O} \gets \text{allocate dynamic memory using system call mmap with size S}$
 \EndIf
 \State $\Return \textit{ O}$
 …
 \end{algorithm}
+<<<<<<< HEAD
 Algorithm~\ref{alg:heapObjectFreeOwn} shows how a free request is fulfilled if object ownership is turned on. Algorithm~\ref{alg:heapObjectFreeNoOwn} shows how the same free request is fulfilled without object ownership.
 …
 \If {$\textit{A was mmap-ed}$}
         \State $\text{return A's dynamic memory to system using system call munmap}$
+=======
+\vspace*{-15pt}
+\begin{algorithm}[H]
+\small
+\caption{Dynamic object free at address $A$ with object ownership}\label{alg:heapObjectFreeOwn}
+\begin{algorithmic}[1]
+\If {$\textit{A mapped allocation}$}
+        \State $\text{return A's dynamic memory to system using system call \lstinline{munmap}}$
+>>>>>>> bb7c77dc425e289ed60aa638529b3e5c7c3e4961
 \Else
         \State $\text{B} \gets \textit{O's owner}$
 …
 \end{algorithmic}
 \end{algorithm}
+<<<<<<< HEAD
 \begin{algorithm}
 …
 \end{algorithm}
+=======
+>>>>>>> bb7c77dc425e289ed60aa638529b3e5c7c3e4961
+\vspace*{-15pt}
+\begin{algorithm}[H]
+\small
+\caption{Dynamic object free at address $A$ without object ownership}\label{alg:heapObjectFreeNoOwn}
+\begin{algorithmic}[1]
+\If {$\textit{A mapped allocation}$}
+        \State $\text{return A's dynamic memory to system using system call \lstinline{munmap}}$
+\Else
+        \State $\text{B} \gets \textit{O's owner}$
+        \If {$\textit{B is thread-local heap's bucket}$}
+                \State $\text{push A to B's free-list}$
+        \Else
+                \State $\text{C} \gets \textit{thread local heap's bucket with same size as B}$
+                \State $\text{push A to C's free-list}$
+        \EndIf
+\EndIf
+\end{algorithmic}
+\end{algorithm}
+\end{figure}
+Algorithm~\ref{alg:heapObjectFreeOwn} shows the de-allocation (free) outline for an object at address $A$ with ownership.
+First, the address is divided into small (@sbrk@) or large (@mmap@).
+For large allocations, the storage is unmapped back to the operating system.
+For small allocations, the bucket associated with the request size is retrieved.
+If the bucket is local to the thread, the allocation is pushed onto the thread's associated bucket.
+If the bucket is not local to the thread, the allocation is pushed onto the owning thread's associated away stack.
+Algorithm~\ref{alg:heapObjectFreeNoOwn} shows the de-allocation (free) outline for an object at address $A$ without ownership.
+The algorithm is the same as for ownership except if the bucket is not local to the thread.
+Then the corresponding bucket of the owner thread is computed for the deallocating thread, and the allocation is pushed onto the deallocating thread's bucket.
+Finally, the llheap design funnels \label{p:FunnelRoutine} all allocation/deallocation operations through routines @malloc@/@free@, which are the only routines to directly access and manage the internal data structures of the heap.
+Other allocation operations, \eg @calloc@, @memalign@, and @realloc@, are composed of calls to @malloc@ and possibly @free@, and may manipulate header information after storage is allocated.
+This design simplifies heap-management code during development and maintenance.
+\subsection{Alignment}
+All dynamic memory allocations must have a minimum storage alignment for the contained object(s).
+Often the minimum memory alignment, M, is the bus width (32 or 64-bit) or the largest register (double, long double) or largest atomic instruction (DCAS) or vector data (MMMX).
+In general, the minimum storage alignment is 8/16-byte boundary on 32/64-bit computers.
+For consistency, the object header is normally aligned at this same boundary.
+Larger alignments must be a power of 2, such page alignment (4/8K).
+Any alignment request, N, $\le$ the minimum alignment is handled as a normal allocation with minimal alignment.
+For alignments greater than the minimum, the obvious approach for aligning to address @A@ is: compute the next address that is a multiple of @N@ after the current end of the heap, @E@, plus room for the header before @A@ and the size of the allocation after @A@, moving the end of the heap to @E'@.
+\begin{center}
+\input{Alignment1}
+\end{center}
+The storage between @E@ and @H@ is chained onto the appropriate free list for future allocations.
+This approach is also valid within any sufficiently large free block, where @E@ is the start of the free block, and any unused storage before @H@ or after the allocated object becomes free storage.
+In this approach, the aligned address @A@ is the same as the allocated storage address @P@, \ie @P@ $=$ @A@ for all allocation routines, which simplifies deallocation.
+However, if there are a large number of aligned requests, this approach leads to memory fragmentation from the small free areas around the aligned object.
+As well, it does not work for large allocations, where many memory allocators switch from program @sbrk@ to operating-system @mmap@.
+The reason is that @mmap@ only starts on a page boundary, and it is difficult to reuse the storage before the alignment boundary for other requests.
+Finally, this approach is incompatible with allocator designs that funnel allocation requests through @malloc@ as it directly manipulates management information within the allocator to optimize the space/time of a request.
+Instead, llheap alignment is accomplished by making a \emph{pessimistically} allocation request for sufficient storage to ensure that \emph{both} the alignment and size request are satisfied, \eg:
+\begin{center}
+\input{Alignment2}
+\end{center}
+The amount of storage necessary is @alignment - M + size@, which ensures there is an address, @A@, after the storage returned from @malloc@, @P@, that is a multiple of @alignment@ followed by sufficient storage for the data object.
+The approach is pessimistic because if @P@ already has the correct alignment @N@, the initial allocation has already requested sufficient space to move to the next multiple of @N@.
+For this special case, there is @alignment - M@ bytes of unused storage after the data object, which subsequently can be used by @realloc@.
+Note, the address returned is @A@, which is subsequently returned to @free@.
+However, to correctly free the allocated object, the value @P@ must be computable, since that is the value generated by @malloc@ and returned within @memalign@.
+Hence, there must be a mechanism to detect when @P@ $\neq$ @A@ and how to compute @P@ from @A@.
+The llheap approach uses two headers:
+the \emph{original} header associated with a memory allocation from @malloc@, and a \emph{fake} header within this storage before the alignment boundary @A@, which is returned from @memalign@, e.g.:
+\begin{center}
+\input{Alignment2Impl}
+\end{center}
+Since @malloc@ has a minimum alignment of @M@, @P@ $\neq$ @A@ only holds for alignments of @M@ or greater.
+When @P@ $\neq$ @A@, the minimum distance between @P@ and @A@ is @M@ bytes, due to the pessimistic storage-allocation.
+Therefore, there is always room for an @M@-byte fake header before @A@.
+The fake header must supply an indicator to distinguish it from a normal header and the location of address @P@ generated by @malloc@.
+This information is encoded as an offset from A to P and the initialize alignment (discussed in \VRef{s:ReallocStickyProperties}).
+To distinguish a fake header from a normal header, the least-significant bit of the alignment is used because the offset participates in multiple calculations, while the alignment is just remembered data.
+\begin{center}
+\input{FakeHeader}
+\end{center}
+\subsection{\lstinline{realloc} and Sticky Properties}
+\label{s:ReallocStickyProperties}
+Allocation routine @realloc@ provides a memory-management pattern for shrinking/enlarging an existing allocation, while maintaining some or all of the object data, rather than performing the following steps manually.
+\begin{flushleft}
+\begin{tabular}{ll}
+\multicolumn{1}{c}{\textbf{realloc pattern}} & \multicolumn{1}{c}{\textbf{manually}} \\
+\begin{lstlisting}
+T * naddr = realloc( oaddr, newSize );
+\end{lstlisting}
+&
+\begin{lstlisting}
+T * naddr = (T *)malloc( newSize ); $\C[2.4in]{// new storage}$
+memcpy( naddr, addr, oldSize );  $\C{// copy old bytes}$
+free( addr );                           $\C{// free old storage}$
+addr = naddr;                           $\C{// change pointer}\CRT$
+\end{lstlisting}
+\end{tabular}
+\end{flushleft}
+The realloc pattern leverages available storage at the end of an allocation due to bucket sizes, possibly eliminating a new allocation and copying.
+This pattern is not used enough to reduce storage management costs.
+In fact, if @oaddr@ is @nullptr@, @realloc@ does a @malloc@, so even the initial @malloc@ can be a @realloc@ for consistency in the pattern.
+The hidden problem for this pattern is the effect of zero fill and alignment with respect to reallocation.
+Are these properties transient or persistent (``sticky'')?
+For example, when memory is initially allocated by @calloc@ or @memalign@ with zero fill or alignment properties, respectively, what happens when those allocations are given to @realloc@ to change size.
+That is, if @realloc@ logically extends storage into unused bucket space or allocates new storage to satisfy a size change, are initial allocation properties preserve?
+Currently, allocation properties are not preserved, so subsequent use of @realloc@ storage may cause inefficient execution or errors due to lack of zero fill or alignment.
+This silent problem is unintuitive to programmers and difficult to locate because it is transient.
+To prevent these problems, llheap preserves initial allocation properties for the lifetime of an allocation and the semantics of @realloc@ are augmented to preserve these properties, with additional query routines.
+This change makes the realloc pattern efficient and safe.
+\subsection{Header}
+To preserve allocation properties requires storing additional information with an allocation,
+The only available location is the header, where \VRef[Figure]{f:llheapNormalHeader} shows the llheap storage layout.
+The header has two data field sized appropriately for 32/64-bit alignment requirements.
+The first field is a union of three values:
+\begin{description}
+\item[bucket pointer]
+is for allocated storage and points back to the bucket associated with this storage requests (see \VRef[Figure]{f:llheapStructure} for the fields accessible in a bucket).
+\item[mapped size]
+is for mapped storage and is the storage size for use in unmapping.
+\item[next free block]
+is for free storage and is an intrusive pointer chaining same-size free blocks onto a bucket's free stack.
+\end{description}
+The second field remembers the request size versus the allocation (bucket) size, \eg request 42 bytes which is rounded up to 64 bytes.
+Since programmers think in request sizes rather than allocation sizes, the request size allows better generation of statistics or errors.
+\begin{figure}
+\centering
+\input{Header}
+\caption{llheap Normal Header}
+\label{f:llheapNormalHeader}
+\end{figure}
+The low-order 3-bits of the first field are \emph{unused} for any stored values, whereas the second field may use all of its bits.
+The 3 unused bits are used to represent mapped allocation, zero filled, and alignment, respectively.
+Note, the alignment bit is not used in the normal header and the zero-filled/mapped bits are not used in the fake header.
+This implementation allows a fast test if any of the lower 3-bits are on (@&@ and compare).
+If no bits are on, it implies a basic allocation, which is handled quickly;
+otherwise, the bits are analysed and appropriate actions are taken for the complex cases.
+Since most allocations are basic, this implementation results in a significant performance gain along the allocation and free fastpath.
+\section{Statistics and Debugging}
+llheap can be built to accumulate fast and largely contention-free allocation statistics to help understand allocation behaviour.
+Incrementing statistic counters must appear on the allocation fastpath.
+As noted, any atomic operation along the fastpath produces a significant increase in allocation costs.
+To make statistics performant enough for use on running systems, each heap has its own set of statistic counters, so heap operations do not require atomic operations.
+To locate all statistic counters, heaps are linked together in statistics mode, and this list is locked and traversed to sum all counters across heaps.
+Note, the list is locked to prevent errors traversing an active list;
+the statistics counters are not locked and can flicker during accumulation, which is not an issue with atomic read/write.
+\VRef[Figure]{f:StatiticsOutput} shows an example of statistics output, which covers all allocation operations and information about deallocating storage not owned by a thread.
+No other memory allocator studied provides as comprehensive statistical information.
+Finally, these statistics were invaluable during the development of this thesis for debugging and verifying correctness, and hence, should be equally valuable to application developers.
+\begin{figure}
+\begin{lstlisting}
+Heap statistics: (storage request / allocation)
+  malloc >0 calls 2,766; 0 calls 2,064; storage 12,715 / 13,367 bytes
+  aalloc >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  calloc >0 calls 6; 0 calls 0; storage 1,008 / 1,104 bytes
+  memalign >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  amemalign >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  cmemalign >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  resize >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  realloc >0 calls 0; 0 calls 0; storage 0 / 0 bytes
+  free !null calls 2,766; null calls 4,064; storage 12,715 / 13,367 bytes
+  away pulls 0; pushes 0; storage 0 / 0 bytes
+  sbrk calls 1; storage 10,485,760 bytes
+  mmap calls 10,000; storage 10,000 / 10,035 bytes
+  munmap calls 10,000; storage 10,000 / 10,035 bytes
+  threads started 4; exited 3
+  heaps new 4; reused 0
+\end{lstlisting}
+\caption{Statistics Output}
+\label{f:StatiticsOutput}
+\end{figure}
+llheap can also be built with debug checking, which inserts many asserts along all allocation paths.
+These assertions detect incorrect allocation usage, like double frees, unfreed storage, or memory corruptions because internal values (like header fields) are overwritten.
+These checks are best effort as opposed to complete allocation checking as in @valgrind@.
+Nevertheless, the checks detect many allocation problems.
+There is an unfortunate problem in detecting unfreed storage because some library routines assume their allocations have life-time duration, and hence, do not free their storage.
+For example, @printf@ allocates a 1024 buffer on first call and never deletes this buffer.
+To prevent a false positive for unfreed storage, it is possible to specify an amount of storage that is never freed (see @malloc_unfreed@ \VPageref{p:malloc_unfreed}), and it is subtracted from the total allocate/free difference.
+Determining the amount of never-freed storage is annoying, but once done, any warnings of unfreed storage are application related.
+Tests indicate only a 30\% performance increase when statistics \emph{and} debugging are enabled, and the latency cost for accumulating statistic is mitigated by limited calls, often only one at the end of the program.
+\section{User-level Threading Support}
+\label{s:UserlevelThreadingSupport}
+The serially-reusable problem (see \VRef{s:AllocationFastpath}) occurs for kernel threads in the ``T:H model, H = number of CPUs'' model and for user threads in the ``1:1'' model, where llheap uses the ``1:1'' model.
+The solution is to prevent interrupts that can result in CPU or KT change during operations that are logically critical sections.
+Locking these critical sections negates any attempt for a quick fastpath and results in high contention.
+For user-level threading, the serially-reusable problem appears with time slicing for preemptable scheduling, as the signal handler context switches to another user-level thread.
+Without time slicing, a user thread performing a long computation can prevent execution (starve) other threads.
+To prevent starvation for an allocation-active thread, \ie the time slice always triggers in an allocation critical-section for one thread, a thread-local \newterm{rollforward} flag is set in the signal handler when it aborts a time slice.
+The rollforward flag is tested at the end of each allocation funnel routine (see \VPageref{p:FunnelRoutine}), and if set, it is reset and a volunteer yield (context switch) is performed to allow other threads to execute.
+llheap uses two techniques to detect when execution is in a allocation operation or routine called from allocation operation, to abort any time slice during this period.
+On the slowpath when executing expensive operations, like @sbrk@ or @mmap@, interrupts are disabled/enabled by setting thread-local flags so the signal handler aborts immediately.
+On the fastpath, disabling/enabling interrupts is too expensive as accessing thread-local storage can be expensive and not thread-safe.
+For example, the ARM processor stores the thread-local pointer in a coprocessor register that cannot perform atomic base-displacement addressing.
+Hence, there is a window between loading the thread-local pointer from the coprocessor register into a normal register and adding the displacement when a time slice can move a thread.
+The fast technique defines a special code section and places all non-interruptible routines in this section.
+The linker places all code in this section into a contiguous block of memory, but the order of routines within the block is unspecified.
+Then, the signal handler compares the program counter at the point of interrupt with the the start and end address of the non-interruptible section, and aborts if executing within this section and sets the rollforward flag.
+This technique is fragile because any calls in the non-interruptible code outside of the non-interruptible section (like @sbrk@) must be bracketed with disable/enable interrupts and these calls must be along the slowpath.
+Hence, for correctness, this approach requires inspection of generated assembler code for routines placed in the non-interruptible section.
+This issue is mitigated by the llheap funnel design so only funnel routines and a few statistics routines are placed in the non-interruptible section and their assembler code examined.
+These techniques are used in both the \uC and \CFA versions of llheap, where both of these systems have user-level threading.
+\section{Bootstrapping}
+There are problems bootstrapping a memory allocator.
+\begin{enumerate}
+\item
+Programs can be statically or dynamically linked.
+\item
+The order the linker schedules startup code is poorly supported.
+\item
+Knowing a KT's start and end independently from the KT code is difficult.
+\end{enumerate}
+For static linking, the allocator is loaded with the program.
+Hence, allocation calls immediately invoke the allocator operation defined by the loaded allocation library and there is only one memory allocator used in the program.
+This approach allows allocator substitution by placing an allocation library before any other in the linked/load path.
+Allocator substitution is similar for dynamic linking, but the problem is that the dynamic loader starts first and needs to perform dynamic allocations \emph{before} the substitution allocator is loaded.
+As a result, the dynamic loader uses a default allocator until the substitution allocator is loaded, after which all allocation operations are handled by the substitution allocator, including from the dynamic loader.
+Hence, some part of the @sbrk@ area may be used by the default allocator and statistics about allocation operations cannot be correct.
+Furthermore, dynamic linking goes through trampolines, so there is an additional cost along the allocator fastpath for all allocation operations.
+Testing showed up to a 5\% performance increase for dynamic linking over static linking, even when using @tls_model("initial-exec")@ so the dynamic loader can obtain tighter binding.
+All allocator libraries need to perform startup code to initialize data structures, such as the heap array for llheap.
+The problem is getting initialized done before the first allocator call.
+However, there does not seem to be mechanism to tell either the static or dynamic loader to first perform initialization code before any calls to a loaded library.
+As a result, calls to allocation routines occur without initialization.
+To deal with this problem, it is necessary to put a conditional initialization check along the allocation fastpath to trigger initialization (singleton pattern).
+Two other important execution points are program startup and termination, which include prologue or epilogue code to bootstrap a program, which programmers are unaware of.
+For example, dynamic-memory allocations before/after the application starts should not be considered in statistics because the application does not make these calls.
+llheap establishes these two points using routines:
+\begin{lstlisting}
+__attribute__(( constructor( 100 ) )) static void startup( void ) {
+        // clear statistic counters
+        // reset allocUnfreed counter
+}
+__attribute__(( destructor( 100 ) )) static void shutdown( void ) {
+        // sum allocUnfreed for all heaps
+        // subtract global unfreed storage
+        // if allocUnfreed > 0 then print warning message
+}
+\end{lstlisting}
+which use global constructor/destructor priority 100, where the linker calls these routines at program prologue/epilogue in increasing/decreasing order of priority.
+Application programs may only use global constructor/destructor priorities greater than 100.
+Hence, @startup@ is called after the program prologue but before the application starts, and @shutdown@ is called after the program terminates but before the program epilogue.
+By resetting counters in @startup@, prologue allocations are ignored, and checking unfreed storage in @shutdown@ checks only application memory management, ignoring the program epilogue.
+While @startup@/@shutdown@ apply to the program KT, a concurrent program creates additional KTs that do not trigger these routines.
+However, it is essential for the allocator to know when each KT is started/terminated.
+One approach is to create a thread-local object with a construct/destructor, which is triggered after a new KT starts and before it terminates, respectively.
+\begin{lstlisting}
+struct ThreadManager {
+        volatile bool pgm_thread;
+        ThreadManager() {} // unusable
+        ~ThreadManager() { if ( pgm_thread ) heapManagerDtor(); }
+};
+static thread_local ThreadManager threadManager;
+\end{lstlisting}
+Unfortunately, thread-local variables are created lazily, \ie on the first dereference of @threadManager@, which then triggers its constructor.
+Therefore, the constructor is useless for knowing when a KT starts because the KT must reference it, and the allocator does not control the application KT.
+Fortunately, the singleton pattern needed for initializing the program KT also triggers KT allocator initialization, which can then reference @pgm_thread@ to call @threadManager@'s constructor, otherwise its destructor is not called.
+Now when a KT terminates, @~ThreadManager@ is called to chained it onto the global-heap free-stack, where @pgm_thread@ is set to true only for the program KT.
+The conditional destructor call prevents closing down the program heap, which must remain available because epilogue code may free more storage.
+Finally, there is a recursive problem when the singleton pattern dereferences @pgm_thread@ to initialize the thread-local object, because its initialization calls @atExit@, which immediately calls @malloc@ to obtain storage.
+This recursion is handled with another thread-local flag to prevent double initialization.
+A similar problem exists when the KT terminates and calls member @~ThreadManager@, because immediately afterwards, the terminating KT calls @free@ to deallocate the storage obtained from the @atExit@.
+In the meantime, the terminated heap has been put on the global-heap free-stack, and may be active by a new KT, so the @atExit@ free is handled as a free to another heap and put onto the away list using locking.
+For user threading systems, the KTs are controlled by the runtime, and hence, start/end pointers are known and interact directly with the llheap allocator for \uC and \CFA, which eliminates or simplifies several of these problems.
+The following API was created to provide interaction between the language runtime and the allocator.
+\begin{lstlisting}
+void startTask();                       $\C{// KT starts}$
+void finishTask();                      $\C{// KT ends}$
+void startup();                         $\C{// when application code starts}$
+void shutdown();                        $\C{// when application code ends}$
+bool traceHeap();                       $\C{// enable allocation/free printing for debugging}$
+bool traceHeapOn();                     $\C{// start printing allocation/free calls}$
+bool traceHeapOff();                    $\C{// stop printing allocation/free calls}$
+\end{lstlisting}
+This kind of API is necessary to allow concurrent runtime systems to interact with difference memory allocators in a consistent way.
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 \section{Added Features and Methods}
+To improve the uHeap allocator (FIX ME: cite uHeap) interface and make it more user friendly, we added a few more routines to the C allocator. Also, we built a \CFA (FIX ME: cite cforall) interface on top of C interface to increase the usability of the allocator.
+\subsection{C Interface}
+We added a few more features and routines to the allocator's C interface that can make the allocator more usable to the programmers. THese features will programmer more control on the dynamic memory allocation.
+The C dynamic-allocation API (see \VRef[Figure]{f:CDynamicAllocationAPI}) is neither orthogonal nor complete.
+For example,
+\begin{itemize}
+\item
+It is possible to zero fill or align an allocation but not both.
+\item
+It is \emph{only} possible to zero fill an array allocation.
+\item
+It is not possible to resize a memory allocation without data copying.
+\item
+@realloc@ does not preserve initial allocation properties.
+\end{itemize}
+As a result, programmers must provide these options, which is error prone, resulting in blaming the entire programming language for a poor dynamic-allocation API.
+Furthermore, newer programming languages have better type systems that can provide safer and more powerful APIs for memory allocation.
+\begin{figure}
+\begin{lstlisting}
+void * malloc( size_t size );
+void * calloc( size_t nmemb, size_t size );
+void * realloc( void * ptr, size_t size );
+void * reallocarray( void * ptr, size_t nmemb, size_t size );
+void free( void * ptr );
+void * memalign( size_t alignment, size_t size );
+void * aligned_alloc( size_t alignment, size_t size );
+int posix_memalign( void ** memptr, size_t alignment, size_t size );
+void * valloc( size_t size );
+void * pvalloc( size_t size );
+struct mallinfo mallinfo( void );
+int mallopt( int param, int val );
+int malloc_trim( size_t pad );
+size_t malloc_usable_size( void * ptr );
+void malloc_stats( void );
+int malloc_info( int options, FILE * fp );
+\end{lstlisting}
+\caption{C Dynamic-Allocation API}
+\label{f:CDynamicAllocationAPI}
+\end{figure}
+The following presents design and API changes for C, \CC (\uC), and \CFA, all of which are implemented in llheap.
 \subsection{Out of Memory}
 …
 Most allocators use @nullptr@ to indicate an allocation failure, specifically out of memory;
 hence the need to return an alternate value for a zero-sized allocation.
+The alternative is to abort a program when out of memory.
+In theory, notifying the programmer allows recovery;
+in practice, it is almost impossible to gracefully when out of memory, so the cheaper approach of returning @nullptr@ for a zero-sized allocation is chosen.
+\subsection{\lstinline{void * aalloc( size_t dim, size_t elemSize )}}
+@aalloc@ is an extension of malloc. It allows programmer to allocate a dynamic array of objects without calculating the total size of array explicitly. The only alternate of this routine in the other allocators is calloc but calloc also fills the dynamic memory with 0 which makes it slower for a programmer who only wants to dynamically allocate an array of objects without filling it with 0.
+\paragraph{Usage}
+A different approach allowed by the C API is to abort a program when out of memory and return @nullptr@ for a zero-sized allocation.
+In theory, notifying the programmer of memory failure allows recovery;
+in practice, it is almost impossible to gracefully recover when out of memory.
+Hence, the cheaper approach of returning @nullptr@ for a zero-sized allocation is chosen because no pseudo allocation is necessary.
+\subsection{C Interface}
+For C, it is possible to increase functionality and orthogonality of the dynamic-memory API to make allocation better for programmers.
+For existing C allocation routines:
+\begin{itemize}
+\item
+@calloc@ sets the sticky zero-fill property.
+\item
+@memalign@, @aligned_alloc@, @posix_memalign@, @valloc@ and @pvalloc@ set the sticky alignment property.
+\item
+@realloc@ and @reallocarray@ preserve sticky properties.
+\end{itemize}
+The C dynamic-memory API is extended with the following routines:
+\paragraph{\lstinline{void * aalloc( size_t dim, size_t elemSize )}}
+extends @calloc@ for allocating a dynamic array of objects without calculating the total size of array explicitly but \emph{without} zero-filling the memory.
+@aalloc@ is significantly faster than @calloc@, which is the only alternative.
+\noindent\textbf{Usage}
 @aalloc@ takes two parameters.
+\begin{itemize}
+\item
+@dim@: number of objects in the array
+\item
+@elemSize@: size of the object in the array.
+\end{itemize}
+It returns address of dynamic object allocatoed on heap that can contain dim number of objects of the size elemSize. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{void * resize( void * oaddr, size_t size )}}
+@resize@ is an extension of relloc. It allows programmer to reuse a cuurently allocated dynamic object with a new size requirement. Its alternate in the other allocators is @realloc@ but relloc also copy the data in old object to the new object which makes it slower for the programmer who only wants to reuse an old dynamic object for a new size requirement but does not want to preserve the data in the old object to the new object.
+\paragraph{Usage}
+\begin{itemize}
+\item
+@dim@: number of array objects
+\item
+@elemSize@: size of array object
+\end{itemize}
+It returns the address of the dynamic array or @NULL@ if either @dim@ or @elemSize@ are zero.
+\paragraph{\lstinline{void * resize( void * oaddr, size_t size )}}
+extends @realloc@ for resizing an existing allocation \emph{without} copying previous data into the new allocation or preserving sticky properties.
+@resize@ is significantly faster than @realloc@, which is the only alternative.
+\noindent\textbf{Usage}
 @resize@ takes two parameters.
+\begin{itemize}
+\item
+@oaddr@: the address of the old object that needs to be resized.
+\item
+@size@: the new size requirement of the to which the old object needs to be resized.
+\end{itemize}
+It returns an object that is of the size given but it does not preserve the data in the old object. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{void * resize( void * oaddr, size_t nalign, size_t size )}}
+This @resize@ is an extension of the above @resize@ (FIX ME: cite above resize). In addition to resizing the size of of an old object, it can also realign the old object to a new alignment requirement.
+\paragraph{Usage}
+This resize takes three parameters. It takes an additional parameter of nalign as compared to the above resize (FIX ME: cite above resize).
+\begin{itemize}
+\item
+@oaddr@: the address of the old object that needs to be resized.
+\item
+@nalign@: the new alignment to which the old object needs to be realigned.
+\item
+@size@: the new size requirement of the to which the old object needs to be resized.
+\end{itemize}
+It returns an object with the size and alignment given in the parameters. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{void * amemalign( size_t alignment, size_t dim, size_t elemSize )}}
+amemalign is a hybrid of memalign and aalloc. It allows programmer to allocate an aligned dynamic array of objects without calculating the total size of the array explicitly. It frees the programmer from calculating the total size of the array.
+\paragraph{Usage}
+amemalign takes three parameters.
+\begin{itemize}
+\item
+@alignment@: the alignment to which the dynamic array needs to be aligned.
+\item
+@dim@: number of objects in the array
+\item
+@elemSize@: size of the object in the array.
+\end{itemize}
+It returns a dynamic array of objects that has the capacity to contain dim number of objects of the size of elemSize. The returned dynamic array is aligned to the given alignment. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{void * cmemalign( size_t alignment, size_t dim, size_t elemSize )}}
+cmemalign is a hybrid of amemalign and calloc. It allows programmer to allocate an aligned dynamic array of objects that is 0 filled. The current way to do this in other allocators is to allocate an aligned object with memalign and then fill it with 0 explicitly. This routine provides both features of aligning and 0 filling, implicitly.
+\paragraph{Usage}
+cmemalign takes three parameters.
+\begin{itemize}
+\item
+@alignment@: the alignment to which the dynamic array needs to be aligned.
+\item
+@dim@: number of objects in the array
+\item
+@elemSize@: size of the object in the array.
+\end{itemize}
+It returns a dynamic array of objects that has the capacity to contain dim number of objects of the size of elemSize. The returned dynamic array is aligned to the given alignment and is 0 filled. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{size_t malloc_alignment( void * addr )}}
+@malloc_alignment@ returns the alignment of a currently allocated dynamic object. It allows the programmer in memory management and personal bookkeeping. It helps the programmer in verofying the alignment of a dynamic object especially in a scenerio similar to prudcer-consumer where a producer allocates a dynamic object and the consumer needs to assure that the dynamic object was allocated with the required alignment.
+\paragraph{Usage}
+@malloc_alignment@ takes one parameters.
+\begin{itemize}
+\item
+@addr@: the address of the currently allocated dynamic object.
+\end{itemize}
+@malloc_alignment@ returns the alignment of the given dynamic object. On failure, it return the value of default alignment of the uHeap allocator.
+\subsection{\lstinline{bool malloc_zero_fill( void * addr )}}
+@malloc_zero_fill@ returns whether a currently allocated dynamic object was initially zero filled at the time of allocation. It allows the programmer in memory management and personal bookkeeping. It helps the programmer in verifying the zero filled property of a dynamic object especially in a scenerio similar to prudcer-consumer where a producer allocates a dynamic object and the consumer needs to assure that the dynamic object was zero filled at the time of allocation.
+\paragraph{Usage}
+\begin{itemize}
+\item
+@oaddr@: address to be resized
+\item
+@size@: new allocation size (smaller or larger than previous)
+\end{itemize}
+It returns the address of the old or new storage with the specified new size or @NULL@ if @size@ is zero.
+\paragraph{\lstinline{void * amemalign( size_t alignment, size_t dim, size_t elemSize )}}
+extends @aalloc@ and @memalign@ for allocating an aligned dynamic array of objects.
+Sets sticky alignment property.
+\noindent\textbf{Usage}
+@amemalign@ takes three parameters.
+\begin{itemize}
+\item
+@alignment@: alignment requirement
+\item
+@dim@: number of array objects
+\item
+@elemSize@: size of array object
+\end{itemize}
+It returns the address of the aligned dynamic-array or @NULL@ if either @dim@ or @elemSize@ are zero.
+\paragraph{\lstinline{void * cmemalign( size_t alignment, size_t dim, size_t elemSize )}}
+extends @amemalign@ with zero fill and has the same usage as @amemalign@.
+Sets sticky zero-fill and alignment property.
+It returns the address of the aligned, zero-filled dynamic-array or @NULL@ if either @dim@ or @elemSize@ are zero.
+\paragraph{\lstinline{size_t malloc_alignment( void * addr )}}
+returns the alignment of the dynamic object for use in aligning similar allocations.
+\noindent\textbf{Usage}
+@malloc_alignment@ takes one parameter.
+\begin{itemize}
+\item
+@addr@: address of an allocated object.
+\end{itemize}
+It returns the alignment of the given object, where objects not allocated with alignment return the minimal allocation alignment.
+\paragraph{\lstinline{bool malloc_zero_fill( void * addr )}}
+returns true if the object has the zero-fill sticky property for use in zero filling similar allocations.
+\noindent\textbf{Usage}
 @malloc_zero_fill@ takes one parameters.
 \begin{itemize}
 \item
+@addr@: the address of the currently allocated dynamic object.
+\end{itemize}
+@malloc_zero_fill@ returns true if the dynamic object was initially zero filled and return false otherwise. On failure, it returns false.
+\subsection{\lstinline{size_t malloc_size( void * addr )}}
+@malloc_size@ returns the allocation size of a currently allocated dynamic object. It allows the programmer in memory management and personal bookkeeping. It helps the programmer in verofying the alignment of a dynamic object especially in a scenerio similar to prudcer-consumer where a producer allocates a dynamic object and the consumer needs to assure that the dynamic object was allocated with the required size. Its current alternate in the other allocators is @malloc_usable_size@. But, @malloc_size@ is different from @malloc_usable_size@ as @malloc_usabe_size@ returns the total data capacity of dynamic object including the extra space at the end of the dynamic object. On the other hand, @malloc_size@ returns the size that was given to the allocator at the allocation of the dynamic object. This size is updated when an object is realloced, resized, or passed through a similar allocator routine.
+\paragraph{Usage}
+@addr@: address of an allocated object.
+\end{itemize}
+It returns true if the zero-fill sticky property is set and false otherwise.
+\paragraph{\lstinline{size_t malloc_size( void * addr )}}
+returns the request size of the dynamic object (updated when an object is resized) for use in similar allocations.
+See also @malloc_usable_size@.
+\noindent\textbf{Usage}
 @malloc_size@ takes one parameters.
+\begin{itemize}
+\item
+@addr@: the address of the currently allocated dynamic object.
+\end{itemize}
+@malloc_size@ returns the allocation size of the given dynamic object. On failure, it return zero.
+\subsection{\lstinline{void * realloc( void * oaddr, size_t nalign, size_t size )}}
+This @realloc@ is an extension of the default @realloc@ (FIX ME: cite default @realloc@). In addition to reallocating an old object and preserving the data in old object, it can also realign the old object to a new alignment requirement.
+\paragraph{Usage}
+This @realloc@ takes three parameters. It takes an additional parameter of nalign as compared to the default @realloc@.
+\begin{itemize}
+\item
+@oaddr@: the address of the old object that needs to be reallocated.
+\item
+@nalign@: the new alignment to which the old object needs to be realigned.
+\item
+@size@: the new size requirement of the to which the old object needs to be resized.
+\end{itemize}
+It returns an object with the size and alignment given in the parameters that preserves the data in the old object. On failure, it returns a @NULL@ pointer.
+\subsection{\CFA Malloc Interface}
+We added some routines to the malloc interface of \CFA. These routines can only be used in \CFA and not in our standalone uHeap allocator as these routines use some features that are only provided by \CFA and not by C. It makes the allocator even more usable to the programmers.
+\CFA provides the liberty to know the returned type of a call to the allocator. So, mainly in these added routines, we removed the object size parameter from the routine as allocator can calculate the size of the object from the returned type.
+\subsection{\lstinline{T * malloc( void )}}
+This malloc is a simplified polymorphic form of defualt malloc (FIX ME: cite malloc). It does not take any parameter as compared to default malloc that takes one parameter.
+\paragraph{Usage}
+This malloc takes no parameters.
+It returns a dynamic object of the size of type @T@. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * aalloc( size_t dim )}}
+This aalloc is a simplified polymorphic form of above aalloc (FIX ME: cite aalloc). It takes one parameter as compared to the above aalloc that takes two parameters.
+\paragraph{Usage}
+aalloc takes one parameters.
+\begin{itemize}
+\item
+@dim@: required number of objects in the array.
+\end{itemize}
+It returns a dynamic object that has the capacity to contain dim number of objects, each of the size of type @T@. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * calloc( size_t dim )}}
+This calloc is a simplified polymorphic form of defualt calloc (FIX ME: cite calloc). It takes one parameter as compared to the default calloc that takes two parameters.
+\paragraph{Usage}
+This calloc takes one parameter.
+\begin{itemize}
+\item
+@dim@: required number of objects in the array.
+\end{itemize}
+It returns a dynamic object that has the capacity to contain dim number of objects, each of the size of type @T@. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * resize( T * ptr, size_t size )}}
+This resize is a simplified polymorphic form of above resize (FIX ME: cite resize with alignment). It takes two parameters as compared to the above resize that takes three parameters. It frees the programmer from explicitly mentioning the alignment of the allocation as \CFA provides gives allocator the liberty to get the alignment of the returned type.
+\paragraph{Usage}
+This resize takes two parameters.
+\begin{itemize}
+\item
+@ptr@: address of the old object.
+\item
+@size@: the required size of the new object.
+\end{itemize}
+It returns a dynamic object of the size given in paramters. The returned object is aligned to the alignemtn of type @T@. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * realloc( T * ptr, size_t size )}}
+This @realloc@ is a simplified polymorphic form of defualt @realloc@ (FIX ME: cite @realloc@ with align). It takes two parameters as compared to the above @realloc@ that takes three parameters. It frees the programmer from explicitly mentioning the alignment of the allocation as \CFA provides gives allocator the liberty to get the alignment of the returned type.
+\paragraph{Usage}
+This @realloc@ takes two parameters.
+\begin{itemize}
+\item
+@ptr@: address of the old object.
+\item
+@size@: the required size of the new object.
+\end{itemize}
+It returns a dynamic object of the size given in paramters that preserves the data in the given object. The returned object is aligned to the alignemtn of type @T@. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * memalign( size_t align )}}
+This memalign is a simplified polymorphic form of defualt memalign (FIX ME: cite memalign). It takes one parameters as compared to the default memalign that takes two parameters.
+\paragraph{Usage}
+memalign takes one parameters.
+\begin{itemize}
+\item
+@align@: the required alignment of the dynamic object.
+\end{itemize}
+It returns a dynamic object of the size of type @T@ that is aligned to given parameter align. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * amemalign( size_t align, size_t dim )}}
+This amemalign is a simplified polymorphic form of above amemalign (FIX ME: cite amemalign). It takes two parameter as compared to the above amemalign that takes three parameters.
+\paragraph{Usage}
+amemalign takes two parameters.
+\begin{itemize}
+\item
+@align@: required alignment of the dynamic array.
+\item
+@dim@: required number of objects in the array.
+\end{itemize}
+It returns a dynamic object that has the capacity to contain dim number of objects, each of the size of type @T@. The returned object is aligned to the given parameter align. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * cmemalign( size_t align, size_t dim  )}}
+This cmemalign is a simplified polymorphic form of above cmemalign (FIX ME: cite cmemalign). It takes two parameter as compared to the above cmemalign that takes three parameters.
+\paragraph{Usage}
+cmemalign takes two parameters.
+\begin{itemize}
+\item
+@align@: required alignment of the dynamic array.
+\item
+@dim@: required number of objects in the array.
+\end{itemize}
+It returns a dynamic object that has the capacity to contain dim number of objects, each of the size of type @T@. The returned object is aligned to the given parameter align and is zero filled. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * aligned_alloc( size_t align )}}
+This @aligned_alloc@ is a simplified polymorphic form of defualt @aligned_alloc@ (FIX ME: cite @aligned_alloc@). It takes one parameter as compared to the default @aligned_alloc@ that takes two parameters.
+\paragraph{Usage}
+This @aligned_alloc@ takes one parameter.
+\begin{itemize}
+\item
+@align@: required alignment of the dynamic object.
+\end{itemize}
+It returns a dynamic object of the size of type @T@ that is aligned to the given parameter. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{int posix_memalign( T ** ptr, size_t align )}}
+This @posix_memalign@ is a simplified polymorphic form of defualt @posix_memalign@ (FIX ME: cite @posix_memalign@). It takes two parameters as compared to the default @posix_memalign@ that takes three parameters.
+\paragraph{Usage}
+This @posix_memalign@ takes two parameter.
+\begin{itemize}
+\item
+@ptr@: variable address to store the address of the allocated object.
+\item
+@align@: required alignment of the dynamic object.
+\end{itemize}
+It stores address of the dynamic object of the size of type @T@ in given parameter ptr. This object is aligned to the given parameter. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * valloc( void )}}
+This @valloc@ is a simplified polymorphic form of defualt @valloc@ (FIX ME: cite @valloc@). It takes no parameters as compared to the default @valloc@ that takes one parameter.
+\paragraph{Usage}
+@valloc@ takes no parameters.
+It returns a dynamic object of the size of type @T@ that is aligned to the page size. On failure, it returns a @NULL@ pointer.
+\subsection{\lstinline{T * pvalloc( void )}}
+\paragraph{Usage}
+@pvalloc@ takes no parameters.
+It returns a dynamic object of the size that is calcutaed by rouding the size of type @T@. The returned object is also aligned to the page size. On failure, it returns a @NULL@ pointer.
+\subsection{Alloc Interface}
+In addition to improve allocator interface both for \CFA and our standalone allocator uHeap in C. We also added a new alloc interface in \CFA that increases usability of dynamic memory allocation.
+This interface helps programmers in three major ways.
+\begin{itemize}
+\item
+Routine Name: alloc interfce frees programmers from remmebring different routine names for different kind of dynamic allocations.
+\item
+Parametre Positions: alloc interface frees programmers from remembering parameter postions in call to routines.
+\item
+Object Size: alloc interface does not require programmer to mention the object size as \CFA allows allocator to determince the object size from returned type of alloc call.
+\end{itemize}
+Alloc interface uses polymorphism, backtick routines (FIX ME: cite backtick) and ttype parameters of \CFA (FIX ME: cite ttype) to provide a very simple dynamic memory allocation interface to the programmers. The new interfece has just one routine name alloc that can be used to perform a wide range of dynamic allocations. The parameters use backtick functions to provide a similar-to named parameters feature for our alloc interface so that programmers do not have to remember parameter positions in alloc call except the position of dimension (dim) parameter.
+\subsection{Routine: \lstinline{T * alloc( ... )}}
+Call to alloc wihout any parameter returns one object of size of type @T@ allocated dynamically.
+Only the dimension (dim) parameter for array allocation has the fixed position in the alloc routine. If programmer wants to allocate an array of objects that the required number of members in the array has to be given as the first parameter to the alloc routine.
+alocc routine accepts six kinds of arguments. Using different combinations of tha parameters, different kind of allocations can be performed. Any combincation of parameters can be used together except @`realloc@ and @`resize@ that should not be used simultanously in one call to routine as it creates ambiguity about whether to reallocate or resize a currently allocated dynamic object. If both @`resize@ and @`realloc@ are used in a call to alloc then the latter one will take effect or unexpected resulted might be produced.
+\paragraph{Dim}
+This is the only parameter in the alloc routine that has a fixed-position and it is also the only parameter that does not use a backtick function. It has to be passed at the first position to alloc call in-case of an array allocation of objects of type @T@.
+It represents the required number of members in the array allocation as in \CFA's aalloc (FIX ME: cite aalloc).
+This parameter should be of type @size_t@.
+Example: @int a = alloc( 5 )@
+This call will return a dynamic array of five integers.
+\paragraph{Align}
+This parameter is position-free and uses a backtick routine align (@`align@). The parameter passed with @`align@ should be of type @size_t@. If the alignment parameter is not a power of two or is less than the default alignment of the allocator (that can be found out using routine libAlign in \CFA) then the passed alignment parameter will be rejected and the default alignment will be used.
+Example: @int b = alloc( 5 , 64`align )@
+This call will return a dynamic array of five integers. It will align the allocated object to 64.
+\paragraph{Fill}
+This parameter is position-free and uses a backtick routine fill (@`fill@). In case of @realloc@, only the extra space after copying the data in the old object will be filled with given parameter.
+Three types of parameters can be passed using `fill.
+\begin{itemize}
+\item
+@char@: A char can be passed with @`fill@ to fill the whole dynamic allocation with the given char recursively till the end of required allocation.
+\item
+Object of returned type: An object of type of returned type can be passed with @`fill@ to fill the whole dynamic allocation with the given object recursively till the end of required allocation.
+\item
+Dynamic object of returned type: A dynamic object of type of returned type can be passed with @`fill@ to fill the dynamic allocation with the given dynamic object. In this case, the allocated memory is not filled recursively till the end of allocation. The filling happen untill the end object passed to @`fill@ or the end of requested allocation reaches.
+\end{itemize}
+Example: @int b = alloc( 5 , 'a'`fill )@
+This call will return a dynamic array of five integers. It will fill the allocated object with character 'a' recursively till the end of requested allocation size.
+Example: @int b = alloc( 5 , 4`fill )@
+This call will return a dynamic array of five integers. It will fill the allocated object with integer 4 recursively till the end of requested allocation size.
+Example: @int b = alloc( 5 , a`fill )@ where @a@ is a pointer of int type
+This call will return a dynamic array of five integers. It will copy data in a to the returned object non-recursively untill end of a or the newly allocated object is reached.
+\paragraph{Resize}
+This parameter is position-free and uses a backtick routine resize (@`resize@). It represents the old dynamic object (oaddr) that the programmer wants to
+\begin{itemize}
+\item
+resize to a new size.
+\item
+realign to a new alignment
+\item
+fill with something.
+\end{itemize}
+The data in old dynamic object will not be preserved in the new object. The type of object passed to @`resize@ and the returned type of alloc call can be different.
+Example: @int b = alloc( 5 , a`resize )@
+This call will resize object a to a dynamic array that can contain 5 integers.
+Example: @int b = alloc( 5 , a`resize , 32`align )@
+This call will resize object a to a dynamic array that can contain 5 integers. The returned object will also be aligned to 32.
+Example: @int b = alloc( 5 , a`resize , 32`align , 2`fill )@
+This call will resize object a to a dynamic array that can contain 5 integers. The returned object will also be aligned to 32 and will be filled with 2.
+\paragraph{Realloc}
+This parameter is position-free and uses a backtick routine @realloc@ (@`realloc@). It represents the old dynamic object (oaddr) that the programmer wants to
+\begin{itemize}
+\item
+realloc to a new size.
+\item
+realign to a new alignment
+\item
+fill with something.
+\end{itemize}
+The data in old dynamic object will be preserved in the new object. The type of object passed to @`realloc@ and the returned type of alloc call cannot be different.
+Example: @int b = alloc( 5 , a`realloc )@
+This call will realloc object a to a dynamic array that can contain 5 integers.
+Example: @int b = alloc( 5 , a`realloc , 32`align )@
+This call will realloc object a to a dynamic array that can contain 5 integers. The returned object will also be aligned to 32.
+Example: @int b = alloc( 5 , a`realloc , 32`align , 2`fill )@
+This call will resize object a to a dynamic array that can contain 5 integers. The returned object will also be aligned to 32. The extra space after copying data of a to the returned object will be filled with 2.
+\begin{itemize}
+\item
+@addr@: address of an allocated object.
+\end{itemize}
+It returns the request size or zero if @addr@ is @NULL@.
+\paragraph{\lstinline{int malloc_stats_fd( int fd )}}
+changes the file descriptor where @malloc_stats@ writes statistics (default @stdout@).
+\noindent\textbf{Usage}
+@malloc_stats_fd@ takes one parameters.
+\begin{itemize}
+\item
+@fd@: files description.
+\end{itemize}
+It returns the previous file descriptor.
+\paragraph{\lstinline{size_t malloc_expansion()}}
+\label{p:malloc_expansion}
+set the amount (bytes) to extend the heap when there is insufficient free storage to service an allocation request.
+It returns the heap extension size used throughout a program, \ie called once at heap initialization.
+\paragraph{\lstinline{size_t malloc_mmap_start()}}
+set the crossover between allocations occurring in the @sbrk@ area or separately mapped.
+It returns the crossover point used throughout a program, \ie called once at heap initialization.
+\paragraph{\lstinline{size_t malloc_unfreed()}}
+\label{p:malloc_unfreed}
+amount subtracted to adjust for unfreed program storage (debug only).
+It returns the new subtraction amount and called by @malloc_stats@.
+\subsection{\CC Interface}
+The following extensions take advantage of overload polymorphism in the \CC type-system.
+\paragraph{\lstinline{void * resize( void * oaddr, size_t nalign, size_t size )}}
+extends @resize@ with an alignment re\-quirement.
+\noindent\textbf{Usage}
+takes three parameters.
+\begin{itemize}
+\item
+@oaddr@: address to be resized
+\item
+@nalign@: alignment requirement
+\item
+@size@: new allocation size (smaller or larger than previous)
+\end{itemize}
+It returns the address of the old or new storage with the specified new size and alignment, or @NULL@ if @size@ is zero.
+\paragraph{\lstinline{void * realloc( void * oaddr, size_t nalign, size_t size )}}
+extends @realloc@ with an alignment re\-quirement and has the same usage as aligned @resize@.
+\subsection{\CFA Interface}
+The following extensions take advantage of overload polymorphism in the \CFA type-system.
+The key safety advantage of the \CFA type system is using the return type to select overloads;
+hence, a polymorphic routine knows the returned type and its size.
+This capability is used to remove the object size parameter and correctly cast the return storage to match the result type.
+For example, the following is the \CFA wrapper for C @malloc@:
+\begin{cfa}
+forall( T & | sized(T) ) {
+        T * malloc( void ) {
+                if ( _Alignof(T) <= libAlign() ) return @(T *)@malloc( @sizeof(T)@ ); // C allocation
+                else return @(T *)@memalign( @_Alignof(T)@, @sizeof(T)@ ); // C allocation
+        } // malloc
+\end{cfa}
+and is used as follows:
+\begin{lstlisting}
+int * i = malloc();
+double * d = malloc();
+struct Spinlock { ... } __attribute__(( aligned(128) ));
+Spinlock * sl = malloc();
+\end{lstlisting}
+where each @malloc@ call provides the return type as @T@, which is used with @sizeof@, @_Alignof@, and casting the storage to the correct type.
+This interface removes many of the common allocation errors in C programs.
+\VRef[Figure]{f:CFADynamicAllocationAPI} show the \CFA wrappers for the equivalent C/\CC allocation routines with same semantic behaviour.
+\begin{figure}
+\begin{lstlisting}
+T * malloc( void );
+T * aalloc( size_t dim );
+T * calloc( size_t dim );
+T * resize( T * ptr, size_t size );
+T * realloc( T * ptr, size_t size );
+T * memalign( size_t align );
+T * amemalign( size_t align, size_t dim );
+T * cmemalign( size_t align, size_t dim  );
+T * aligned_alloc( size_t align );
+int posix_memalign( T ** ptr, size_t align );
+T * valloc( void );
+T * pvalloc( void );
+\end{lstlisting}
+\caption{\CFA C-Style Dynamic-Allocation API}
+\label{f:CFADynamicAllocationAPI}
+\end{figure}
+In addition to the \CFA C-style allocator interface, a new allocator interface is provided to further increase orthogonality and usability of dynamic-memory allocation.
+This interface helps programmers in three ways.
+\begin{itemize}
+\item
+naming: \CFA regular and @ttype@ polymorphism is used to encapsulate a wide range of allocation functionality into a single routine name, so programmers do not have to remember multiple routine names for different kinds of dynamic allocations.
+\item
+named arguments: individual allocation properties are specified using postfix function call, so programmers do have to remember parameter positions in allocation calls.
+\item
+object size: like the \CFA C-style interface, programmers do not have to specify object size or cast allocation results.
+\end{itemize}
+Note, postfix function call is an alternative call syntax, using backtick @`@, where the argument appears before the function name, \eg
+\begin{cfa}
+duration ?@`@h( int h );                // ? denote the position of the function operand
+duration ?@`@m( int m );
+duration ?@`@s( int s );
+duration dur = 3@`@h + 42@`@m + 17@`@s;
+\end{cfa}
+@ttype@ polymorphism is similar to \CC variadic templates.
+\paragraph{\lstinline{T * alloc( ... )} or \lstinline{T * alloc( size_t dim, ... )}}
+is overloaded with a variable number of specific allocation routines, or an integer dimension parameter followed by a variable number specific allocation routines.
+A call without parameters returns a dynamically allocated object of type @T@ (@malloc@).
+A call with only the dimension (dim) parameter returns a dynamically allocated array of objects of type @T@ (@aalloc@).
+The variable number of arguments consist of allocation properties, which can be combined to produce different kinds of allocations.
+The only restriction is for properties @realloc@ and @resize@, which cannot be combined.
+The allocation property functions are:
+\subparagraph{\lstinline{T_align ?`align( size_t alignment )}}
+to align the allocation.
+The alignment parameter must be $\ge$ the default alignment (@libAlign()@ in \CFA) and a power of two, \eg:
+\begin{cfa}
+int * i0 = alloc( @4096`align@ );  sout | i0 | nl;
+int * i1 = alloc( 3, @4096`align@ );  sout | i1; for (i; 3 ) sout | &i1[i]; sout | nl;
+x555555572000
+x555555574000 0x555555574000 0x555555574004 0x555555574008
+\end{cfa}
+returns a dynamic object and object array aligned on a 4096-byte boundary.
+\subparagraph{\lstinline{S_fill(T) ?`fill ( /* various types */ )}}
+to initialize storage.
+There are three ways to fill storage:
+\begin{enumerate}
+\item
+A char fills each byte of each object.
+\item
+An object of the returned type fills each object.
+\item
+An object array pointer fills some or all of the corresponding object array.
+\end{enumerate}
+For example:
+\begin{cfa}[numbers=left]
+int * i0 = alloc( @0n`fill@ );  sout | *i0 | nl;  // disambiguate 0
+int * i1 = alloc( @5`fill@ );  sout | *i1 | nl;
+int * i2 = alloc( @'\xfe'`fill@ ); sout | hex( *i2 ) | nl;
+int * i3 = alloc( 5, @5`fill@ );  for ( i; 5 ) sout | i3[i]; sout | nl;
+int * i4 = alloc( 5, @0xdeadbeefN`fill@ );  for ( i; 5 ) sout | hex( i4[i] ); sout | nl;
+int * i5 = alloc( 5, @i3`fill@ );  for ( i; 5 ) sout | i5[i]; sout | nl;
+int * i6 = alloc( 5, @[i3, 3]`fill@ );  for ( i; 5 ) sout | i6[i]; sout | nl;
+\end{cfa}
+\begin{lstlisting}[numbers=left]
+xfefefefe
+5 5 5 5
+xdeadbeef 0xdeadbeef 0xdeadbeef 0xdeadbeef 0xdeadbeef
+5 5 5 5
+5 5 -555819298 -555819298  // two undefined values
+\end{lstlisting}
+Examples 1 to 3, fill an object with a value or characters.
+Examples 4 to 7, fill an array of objects with values, another array, or part of an array.
+\subparagraph{\lstinline{S_resize(T) ?`resize( void * oaddr )}}
+used to resize, realign, and fill, where the old object data is not copied to the new object.
+The old object type may be different from the new object type, since the values are not used.
+For example:
+\begin{cfa}[numbers=left]
+int * i = alloc( @5`fill@ );  sout | i | *i;
+i = alloc( @i`resize@, @256`align@, @7`fill@ );  sout | i | *i;
+double * d = alloc( @i`resize@, @4096`align@, @13.5`fill@ );  sout | d | *d;
+\end{cfa}
+\begin{lstlisting}[numbers=left]
+x55555556d5c0 5
+x555555570000 7
+x555555571000 13.5
+\end{lstlisting}
+Examples 2 to 3 change the alignment, fill, and size for the initial storage of @i@.
+\begin{cfa}[numbers=left]
+int * ia = alloc( 5, @5`fill@ );  for ( i; 5 ) sout | ia[i]; sout | nl;
+ia = alloc( 10, @ia`resize@, @7`fill@ ); for ( i; 10 ) sout | ia[i]; sout | nl;
+sout | ia; ia = alloc( 5, @ia`resize@, @512`align@, @13`fill@ ); sout | ia; for ( i; 5 ) sout | ia[i]; sout | nl;;
+ia = alloc( 3, @ia`resize@, @4096`align@, @2`fill@ );  sout | ia; for ( i; 3 ) sout | &ia[i] | ia[i]; sout | nl;
+\end{cfa}
+\begin{lstlisting}[numbers=left]
+5 5 5 5
+7 7 7 7 7 7 7 7 7
+x55555556d560 0x555555571a00 13 13 13 13 13
+x555555572000 0x555555572000 2 0x555555572004 2 0x555555572008 2
+\end{lstlisting}
+Examples 2 to 4 change the array size, alignment and fill for the initial storage of @ia@.
+\subparagraph{\lstinline{S_realloc(T) ?`realloc( T * a ))}}
+used to resize, realign, and fill, where the old object data is copied to the new object.
+The old object type must be the same as the new object type, since the values used.
+Note, for @fill@, only the extra space after copying the data from the old object is filled with the given parameter.
+For example:
+\begin{cfa}[numbers=left]
+int * i = alloc( @5`fill@ );  sout | i | *i;
+i = alloc( @i`realloc@, @256`align@ );  sout | i | *i;
+i = alloc( @i`realloc@, @4096`align@, @13`fill@ );  sout | i | *i;
+\end{cfa}
+\begin{lstlisting}[numbers=left]
+x55555556d5c0 5
+x555555570000 5
+x555555571000 5
+\end{lstlisting}
+Examples 2 to 3 change the alignment for the initial storage of @i@.
+The @13`fill@ for example 3 does nothing because no extra space is added.
+\begin{cfa}[numbers=left]
+int * ia = alloc( 5, @5`fill@ );  for ( i; 5 ) sout | ia[i]; sout | nl;
+ia = alloc( 10, @ia`realloc@, @7`fill@ ); for ( i; 10 ) sout | ia[i]; sout | nl;
+sout | ia; ia = alloc( 1, @ia`realloc@, @512`align@, @13`fill@ ); sout | ia; for ( i; 1 ) sout | ia[i]; sout | nl;;
+ia = alloc( 3, @ia`realloc@, @4096`align@, @2`fill@ );  sout | ia; for ( i; 3 ) sout | &ia[i] | ia[i]; sout | nl;
+\end{cfa}
+\begin{lstlisting}[numbers=left]
+5 5 5 5
+5 5 5 5 7 7 7 7 7
+x55555556c560 0x555555570a00 5
+x555555571000 0x555555571000 5 0x555555571004 2 0x555555571008 2
+\end{lstlisting}
+Examples 2 to 4 change the array size, alignment and fill for the initial storage of @ia@.
+The @13`fill@ for example 3 does nothing because no extra space is added.
+These \CFA allocation features are used extensively in the development of the \CFA runtime.

doc/theses/mubeen_zulfiqar_MMath/background.tex

-              rba897d21
+              r2e9b59b
 \VRef[Figure]{f:AllocatorComponents} shows the two important data components for a memory allocator, management and storage, collectively called the \newterm{heap}.
 The \newterm{management data} is a data structure located at a known memory address and contains all information necessary to manage the storage data.
 The management data starts with fixed-sized information in the static-data memory that flows into the dynamic-allocation memory.
+The management data starts with fixed-sized information in the static-data memory that references components in the dynamic-allocation memory.
 The \newterm{storage data} is composed of allocated and freed objects, and \newterm{reserved memory}.
 Allocated objects (white) are variable sized, and allocated and maintained by the program;
+Allocated objects (light grey) are variable sized, and allocated and maintained by the program;
 \ie only the program knows the location of allocated storage, not the memory allocator.
 \begin{figure}[h]
 …
 \label{f:AllocatorComponents}
 \end{figure}
 Freed objects (light grey) are memory deallocated by the program, which are linked into one or more lists facilitating easy location for new allocations.
+Freed objects (white) represent memory deallocated by the program, which are linked into one or more lists facilitating easy location of new allocations.
 Often the free list is chained internally so it does not consume additional storage, \ie the link fields are placed at known locations in the unused memory blocks.
 Reserved memory (dark grey) is one or more blocks of memory obtained from the operating system but not yet allocated to the program;
 …
 The trailer may be used to simplify an allocation implementation, \eg coalescing, and/or for security purposes to mark the end of an object.
 An object may be preceded by padding to ensure proper alignment.
+Some algorithms quantize allocation requests into distinct sizes resulting in additional spacing after objects less than the quantized value.
+Some algorithms quantize allocation requests into distinct sizes, called \newterm{buckets}, resulting in additional spacing after objects less than the quantized value.
+(Note, the buckets are often organized as an array of ascending bucket sizes for fast searching, \eg binary search, and the array is stored in the heap management-area, where each bucket is a top point to the freed objects of that size.)
 When padding and spacing are necessary, neither can be used to satisfy a future allocation request while the current allocation exists.
 A free object also contains management data, \eg size, chaining, etc.
 …
 Fragmentation is memory requested from the operating system but not used by the program;
 hence, allocated objects are not fragmentation.
 \VRef[Figure]{f:InternalExternalFragmentation}) shows fragmentation is divided into two forms: internal or external.
+\VRef[Figure]{f:InternalExternalFragmentation} shows fragmentation is divided into two forms: internal or external.
 \begin{figure}
 …
 An allocator should strive to keep internal management information to a minimum.
 \newterm{External fragmentation} is all memory space reserved from the operating system but not allocated to the program~\cite{Wilson95,Lim98,Siebert00}, which includes freed objects, all external management data, and reserved memory.
+\newterm{External fragmentation} is all memory space reserved from the operating system but not allocated to the program~\cite{Wilson95,Lim98,Siebert00}, which includes all external management data, freed objects, and reserved memory.
 This memory is problematic in two ways: heap blowup and highly fragmented memory.
 \newterm{Heap blowup} occurs when memory freed by the program is not reused for future allocations leading to potentially unbounded external fragmentation growth~\cite{Berger00}.
 …
 \end{figure}
 For a single-threaded memory allocator, three basic approaches for controlling fragmentation have been identified~\cite{Johnstone99}.
+For a single-threaded memory allocator, three basic approaches for controlling fragmentation are identified~\cite{Johnstone99}.
 The first approach is a \newterm{sequential-fit algorithm} with one list of free objects that is searched for a block large enough to fit a requested object size.
 Different search policies determine the free object selected, \eg the first free object large enough or closest to the requested size.
 …
 The second approach is a \newterm{segregated} or \newterm{binning algorithm} with a set of lists for different sized freed objects.
 When an object is allocated, the requested size is rounded up to the nearest bin-size, possibly with spacing after the object.
+When an object is allocated, the requested size is rounded up to the nearest bin-size, often leading to spacing after the object.
 A binning algorithm is fast at finding free memory of the appropriate size and allocating it, since the first free object on the free list is used.
 The fewer bin-sizes, the fewer lists need to be searched and maintained;
 …
 Temporal locality commonly occurs during an iterative computation with a fix set of disjoint variables, while spatial locality commonly occurs when traversing an array.
 Hardware takes advantage of temporal and spatial locality through multiple levels of caching (\ie memory hierarchy).
+Hardware takes advantage of temporal and spatial locality through multiple levels of caching, \ie memory hierarchy.
 When an object is accessed, the memory physically located around the object is also cached with the expectation that the current and nearby objects will be referenced within a short period of time.
 For example, entire cache lines are transferred between memory and cache and entire virtual-memory pages are transferred between disk and memory.
 …
 There are a number of ways a memory allocator can degrade locality by increasing the working set.
 For example, a memory allocator may access multiple free objects before finding one to satisfy an allocation request (\eg sequential-fit algorithm).
+For example, a memory allocator may access multiple free objects before finding one to satisfy an allocation request, \eg sequential-fit algorithm.
 If there are a (large) number of objects accessed in very different areas of memory, the allocator may perturb the program's memory hierarchy causing multiple cache or page misses~\cite{Grunwald93}.
 Another way locality can be degraded is by spatially separating related data.
 …
 A multi-threaded memory-allocator does not run any threads itself, but is used by a multi-threaded program.
 In addition to single-threaded design issues of locality and fragmentation, a multi-threaded allocator may be simultaneously accessed by multiple threads, and hence, must deal with concurrency issues such as mutual exclusion, false sharing, and additional forms of heap blowup.
+In addition to single-threaded design issues of fragmentation and locality, a multi-threaded allocator is simultaneously accessed by multiple threads, and hence, must deal with concurrency issues such as mutual exclusion, false sharing, and additional forms of heap blowup.
 …
 Second is when multiple threads contend for a shared resource simultaneously, and hence, some threads must wait until the resource is released.
 Contention can be reduced in a number of ways:
+\begin{itemize}[itemsep=0pt]
+\item
 using multiple fine-grained locks versus a single lock, spreading the contention across a number of locks;
+\item
 using trylock and generating new storage if the lock is busy, yielding a classic space versus time tradeoff;
+\item
 using one of the many lock-free approaches for reducing contention on basic data-structure operations~\cite{Oyama99}.
+However, all of these approaches have degenerate cases where contention occurs.
+\end{itemize}
+However, all of these approaches have degenerate cases where program contention is high, which occurs outside of the allocator.
 …
 \label{s:MultipleHeaps}
 A single-threaded allocator has at most one thread and heap, while a multi-threaded allocator has potentially multiple threads and heaps.
+A multi-threaded allocator has potentially multiple threads and heaps.
 The multiple threads cause complexity, and multiple heaps are a mechanism for dealing with the complexity.
 The spectrum ranges from multiple threads using a single heap, denoted as T:1 (see \VRef[Figure]{f:SingleHeap}), to multiple threads sharing multiple heaps, denoted as T:H (see \VRef[Figure]{f:SharedHeaps}), to one thread per heap, denoted as 1:1 (see \VRef[Figure]{f:PerThreadHeap}), which is almost back to a single-threaded allocator.
 …
 An alternative implementation is for all heaps to share one reserved memory, which requires a separate lock for the reserved storage to ensure mutual exclusion when acquiring new memory.
 Because multiple threads can allocate/free/reallocate adjacent storage, all forms of false sharing may occur.
 Other storage-management options are to use @mmap@ to set aside (large) areas of virtual memory for each heap and suballocate each heap's storage within that area.
+Other storage-management options are to use @mmap@ to set aside (large) areas of virtual memory for each heap and suballocate each heap's storage within that area, pushing part of the storage management complexity back to the operating system.
 \begin{figure}
 …
 \paragraph{1:1 model (thread heaps)} where each thread has its own heap, which eliminates most contention and locking because threads seldom accesses another thread's heap (see ownership in \VRef{s:Ownership}).
+\paragraph{1:1 model (thread heaps)} where each thread has its own heap eliminating most contention and locking because threads seldom access another thread's heap (see ownership in \VRef{s:Ownership}).
 An additional benefit of thread heaps is improved locality due to better memory layout.
 As each thread only allocates from its heap, all objects for a thread are consolidated in the storage area for that heap, better utilizing each CPUs cache and accessing fewer pages.
 …
 Second is to place the thread heap on a list of available heaps and reuse it for a new thread in the future.
 Destroying the thread heap immediately may reduce external fragmentation sooner, since all free objects are freed to the global heap and may be reused by other threads.
 Alternatively, reusing thread heaps may improve performance if the inheriting thread makes similar allocation requests as the thread that previously held the thread heap.
+Alternatively, reusing thread heaps may improve performance if the inheriting thread makes similar allocation requests as the thread that previously held the thread heap because any unfreed storage is immediately accessible..
 …
 However, an important goal of user-level threading is for fast operations (creation/termination/context-switching) by not interacting with the operating system, which allows the ability to create large numbers of high-performance interacting threads ($>$ 10,000).
 It is difficult to retain this goal, if the user-threading model is directly involved with the heap model.
 \VRef[Figure]{f:UserLevelKernelHeaps} shows that virtually all user-level threading systems use whatever kernel-level heap-model provided by the language runtime.
+\VRef[Figure]{f:UserLevelKernelHeaps} shows that virtually all user-level threading systems use whatever kernel-level heap-model is provided by the language runtime.
 Hence, a user thread allocates/deallocates from/to the heap of the kernel thread on which it is currently executing.
 …
 Adopting this model results in a subtle problem with shared heaps.
 With kernel threading, an operation that is started by a kernel thread is always completed by that thread.
+For example, if a kernel thread starts an allocation/deallocation on a shared heap, it always completes that operation with that heap even if preempted.
+Any correctness locking associated with the shared heap is preserved across preemption.
+For example, if a kernel thread starts an allocation/deallocation on a shared heap, it always completes that operation with that heap even if preempted, \ie any locking correctness associated with the shared heap is preserved across preemption.
 However, this correctness property is not preserved for user-level threading.
 …
 However, eagerly disabling/enabling time-slicing on the allocation/deallocation fast path is expensive, because preemption is rare (10--100 milliseconds).
 Instead, techniques exist to lazily detect this case in the interrupt handler, abort the preemption, and return to the operation so it can complete atomically.
 Occasionally ignoring a preemption should be benign.
+Occasionally ignoring a preemption should be benign, but a persistent lack of preemption can result in both short and long term starvation.
 …
 \newterm{Ownership} defines which heap an object is returned-to on deallocation.
 If a thread returns an object to the heap it was originally allocated from, the heap has ownership of its objects.
 Alternatively, a thread can return an object to the heap it is currently allocating from, which can be any heap accessible during a thread's lifetime.
+If a thread returns an object to the heap it was originally allocated from, a heap has ownership of its objects.
+Alternatively, a thread can return an object to the heap it is currently associated with, which can be any heap accessible during a thread's lifetime.
 \VRef[Figure]{f:HeapsOwnership} shows an example of multiple heaps (minus the global heap) with and without ownership.
 Again, the arrows indicate the direction memory conceptually moves for each kind of operation.
 …
 Only with the 1:1 model and ownership is active and passive false-sharing avoided (see \VRef{s:Ownership}).
 Passive false-sharing may still occur, if delayed ownership is used.
+Finally, a completely free container can become reserved storage and be reset to allocate objects of a new size or freed to the global heap.
 \begin{figure}
 …
 \caption{Free-list Structure with Container Ownership}
 \end{figure}
-A fragmented heap has multiple containers that may be partially or completely free.
-A completely free container can become reserved storage and be reset to allocate objects of a new size.
-When a heap reaches a threshold of free objects, it moves some free storage to the global heap for reuse to prevent heap blowup.
-Without ownership, when a heap frees objects to the global heap, individual objects must be passed, and placed on the global-heap's free-list.
-Containers cannot be freed to the global heap unless completely free because
 When a container changes ownership, the ownership of all objects within it change as well.
 …
 Note, once the object is freed by Task$_1$, no more false sharing can occur until the container changes ownership again.
 To prevent this form of false sharing, container movement may be restricted to when all objects in the container are free.
 One implementation approach that increases the freedom to return a free container to the operating system involves allocating containers using a call like @mmap@, which allows memory at an arbitrary address to be returned versus only storage at the end of the contiguous @sbrk@ area.
+One implementation approach that increases the freedom to return a free container to the operating system involves allocating containers using a call like @mmap@, which allows memory at an arbitrary address to be returned versus only storage at the end of the contiguous @sbrk@ area, again pushing storage management complexity back to the operating system.
 \begin{figure}
 …
 \end{figure}
 As mentioned, an implementation may have only one heap deal with the global heap, so the other heap can be simplified.
+As mentioned, an implementation may have only one heap interact with the global heap, so the other heap can be simplified.
 For example, if only the private heap interacts with the global heap, the public heap can be reduced to a lock-protected free-list of objects deallocated by other threads due to ownership, called a \newterm{remote free-list}.
 To avoid heap blowup, the private heap allocates from the remote free-list when it reaches some threshold or it has no free storage.
 …
 An allocation buffer is reserved memory (see~\VRef{s:AllocatorComponents}) not yet allocated to the program, and is used for allocating objects when the free list is empty.
 That is, rather than requesting new storage for a single object, an entire buffer is requested from which multiple objects are allocated later.
 Both any heap may use an allocation buffer, resulting in allocation from the buffer before requesting objects (containers) from the global heap or operating system, respectively.
+Any heap may use an allocation buffer, resulting in allocation from the buffer before requesting objects (containers) from the global heap or operating system, respectively.
 The allocation buffer reduces contention and the number of global/operating-system calls.
 For coalescing, a buffer is split into smaller objects by allocations, and recomposed into larger buffer areas during deallocations.
 Allocation buffers are useful initially when there are no freed objects in a heap because many allocations usually occur when a thread starts.
+Allocation buffers are useful initially when there are no freed objects in a heap because many allocations usually occur when a thread starts (simple bump allocation).
 Furthermore, to prevent heap blowup, objects should be reused before allocating a new allocation buffer.
 Thus, allocation buffers are often allocated more frequently at program/thread start, and then their use often diminishes.
+Thus, allocation buffers are often allocated more frequently at program/thread start, and then allocations often diminish.
 Using an allocation buffer with a thread heap avoids active false-sharing, since all objects in the allocation buffer are allocated to the same thread.
 …
 \label{s:LockFreeOperations}
 A lock-free algorithm guarantees safe concurrent-access to a data structure, so that at least one thread can make progress in the system, but an individual task has no bound to execution, and hence, may starve~\cite[pp.~745--746]{Herlihy93}.
+% A wait-free algorithm puts a finite bound on the number of steps any thread takes to complete an operation, so an individual task cannot starve
+A \newterm{lock-free algorithm} guarantees safe concurrent-access to a data structure, so that at least one thread makes progress, but an individual task has no execution bound and may starve~\cite[pp.~745--746]{Herlihy93}.
+(A \newterm{wait-free algorithm} puts a bound on the number of steps any thread takes to complete an operation to prevent starvation.)
 Lock-free operations can be used in an allocator to reduce or eliminate the use of locks.
 Locks are a problem for high contention or if the thread holding the lock is preempted and other threads attempt to use that lock.
 With respect to the heap, these situations are unlikely unless all threads makes extremely high use of dynamic-memory allocation, which can be an indication of poor design.
+While locks and lock-free data-structures often have equal performance, lock-free has the advantage of not holding a lock across preemption so other threads can continue to make progress.
+With respect to the heap, these situations are unlikely unless all threads make extremely high use of dynamic-memory allocation, which can be an indication of poor design.
 Nevertheless, lock-free algorithms can reduce the number of context switches, since a thread does not yield/block while waiting for a lock;
 on the other hand, a thread may busy-wait for an unbounded period.
+on the other hand, a thread may busy-wait for an unbounded period holding a processor.
 Finally, lock-free implementations have greater complexity and hardware dependency.
 Lock-free algorithms can be applied most easily to simple free-lists, \eg remote free-list, to allow lock-free insertion and removal from the head of a stack.
 Implementing lock-free operations for more complex data-structures (queue~\cite{Valois94}/deque~\cite{Sundell08}) is more complex.
+Implementing lock-free operations for more complex data-structures (queue~\cite{Valois94}/deque~\cite{Sundell08}) is correspondingly more complex.
 Michael~\cite{Michael04} and Gidenstam \etal \cite{Gidenstam05} have created lock-free variations of the Hoard allocator.

doc/theses/mubeen_zulfiqar_MMath/figures/AllocDS1.fig

-              rba897d21
+              r2e9b59b
 -2
 2
 4200 1575 4500 1725
 3 0 1 0 0 50 -1 20 0.000 1 0.0000 4275 1650 20 20 4275 1650 4295 1650
 3 0 1 0 0 50 -1 20 0.000 1 0.0000 4350 1650 20 20 4350 1650 4370 1650
 3 0 1 0 0 50 -1 20 0.000 1 0.0000 4425 1650 20 20 4425 1650 4445 1650
+2850 2100 3150 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 2925 2175 20 20 2925 2175 2945 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3000 2175 20 20 3000 2175 3020 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3075 2175 20 20 3075 2175 3095 2175
 -6
+2850 2475 3150 2850
+4050 2100 4350 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4125 2175 20 20 4125 2175 4145 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4200 2175 20 20 4200 2175 4220 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4275 2175 20 20 4275 2175 4295 2175
+-6
+4650 2100 4950 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4725 2175 20 20 4725 2175 4745 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4800 2175 20 20 4800 2175 4820 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4875 2175 20 20 4875 2175 4895 2175
+-6
+3450 2100 3750 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3525 2175 20 20 3525 2175 3545 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3600 2175 20 20 3600 2175 3620 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3675 2175 20 20 3675 2175 3695 2175
+-6
+3300 2175 3600 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
 2475 2925 2700
+2175 3375 2400
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
 2700 3150 2700 3150 2850 2850 2850 2850 2700
+2400 3600 2400 3600 2550 3300 2550 3300 2400
 -6
+4350 2475 4650 2850
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 3150 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 2850 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 4650 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 4950 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 4500 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 5100 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 3450 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 3750 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 3300 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 3900 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5250 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5400 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5550 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5700 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5850 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 2700 2250
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+2475 4425 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 4650 2700 4650 2850 4350 2850 4350 2700
+-6
+3600 2475 3825 3150
+1275 3375 1575
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+2475 3675 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 3825 2700 3825 2850 3600 2850 3600 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3000 3825 3000 3825 3150 3600 3150 3600 3000
+1275 2700 1575
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 2775 1575
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+2775 3675 3000
+-6
+4875 3600 5175 3750
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4950 3675 20 20 4950 3675 4970 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5025 3675 20 20 5025 3675 5045 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5100 3675 20 20 5100 3675 5120 3675
+-6
+4875 2325 5175 2475
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4950 2400 20 20 4950 2400 4970 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5025 2400 20 20 5025 2400 5045 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5100 2400 20 20 5100 2400 5120 2400
+-6
+5625 2325 5925 2475
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5700 2400 20 20 5700 2400 5720 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5775 2400 20 20 5775 2400 5795 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5850 2400 20 20 5850 2400 5870 2400
+-6
+5625 3600 5925 3750
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5700 3675 20 20 5700 3675 5720 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5775 3675 20 20 5775 3675 5795 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5850 3675 20 20 5850 3675 5870 3675
+-6
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2400 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2550 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2700 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2850 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 3000 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 3600 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 3900 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 4050 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 4200 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 4350 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 4500 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1500 3300 1800
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1500 3600 1800
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1500 3900 1800
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+1500 4800 1500 4800 1800 3000 1800 3000 1500
+1275 5175 1575
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 5625 1575
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 3750 1575
 1 1 1 0 7 50 -1 -1 4.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+1650 2625 2100
+1275 3825 1575
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1950 6000 1950
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 6000 2100
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+1800 6000 1800 6000 2250 2700 2250 2700 1800
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
 1650 2550 2100
+2175 2775 2400
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+1650 4050 2100
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1650 3975 2100
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2100 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2250 3150 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2250 4650 2250
+2475 2775 2700
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
 2100 3150 2100 3150 2550 1950 2550 1950 2100
+2700 2850 2700 2850 2850 2700 2850 2700 2700
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2100 4650 2100 4650 2550 3450 2550 3450 2100
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2250 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 3750 2550
+2400 2850 2400 2850 2550 2700 2550 2700 2400
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+2475 2025 2700
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2775 2025 3000
+2175 4575 2400
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3000 2100 3000 2100 3150 1950 3150 1950 3000
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 2100 2700 2100 2850 1950 2850 1950 2700
+2400 5025 2400 5025 2550 4500 2550 4500 2400
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
 1 1.00 45.00 90.00
 3750 2700 3750 2700 3525
+3375 4350 3375 4350 3150
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3525 3150 3525 3150 3900 1950 3900 1950 3525
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
+1 1.00 45.00 90.00
+3750 4200 3750 4200 3525
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3525 4650 3525 4650 3900 3450 3900 3450 3525
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
+1 1.00 45.00 90.00
+4650 4200 4650 4200 4275
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+4275 4650 4275 4650 4875 3150 4875 3150 4275
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2400 3150 2400
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2400 4650 2400
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 5400 3900
+2 0 50 -1 0 11 0.0000 2 120 300 1875 2250 lock\001
+1 0 50 -1 0 12 0.0000 2 135 1935 3900 1425 N kernel-thread buckets\001
+1 0 50 -1 0 12 0.0000 2 195 810 4425 2025 heap$_2$\001
+1 0 50 -1 0 12 0.0000 2 195 810 2175 2025 heap$_1$\001
+2 0 50 -1 0 11 0.0000 2 120 270 1875 2400 size\001
+2 0 50 -1 0 11 0.0000 2 120 270 1875 2550 free\001
+1 0 50 -1 0 12 0.0000 2 180 825 2550 3450 local pool\001
+0 0 50 -1 0 12 0.0000 2 135 360 3525 3700 lock\001
+0 0 50 -1 0 12 0.0000 2 135 360 3225 4450 lock\001
+2 0 50 -1 0 12 0.0000 2 135 600 1875 3000 free list\001
+1 0 50 -1 0 12 0.0000 2 180 825 4050 3450 local pool\001
+1 0 50 -1 0 12 0.0000 2 180 1455 3900 4200 global pool (sbrk)\001
+0 0 50 -1 0 12 0.0000 2 135 360 2025 3700 lock\001
+1 0 50 -1 0 12 0.0000 2 180 720 6450 3150 free pool\001
+1 0 50 -1 0 12 0.0000 2 180 390 6450 2925 heap\001
+3150 5100 3150 5100 3525 3600 3525 3600 3150
+2 0 50 -1 0 11 0.0000 2 135 300 2625 1950 lock\001
+1 0 50 -1 0 11 0.0000 2 150 1155 3000 1725 N$\\times$S$_1$\001
+1 0 50 -1 0 11 0.0000 2 150 1155 3600 1725 N$\\times$S$_2$\001
+1 0 50 -1 0 12 0.0000 2 180 390 4425 1500 heap\001
+2 0 50 -1 0 12 0.0000 2 135 1140 2550 1425 kernel threads\001
+2 0 50 -1 0 11 0.0000 2 120 270 2625 2100 size\001
+2 0 50 -1 0 11 0.0000 2 120 270 2625 2250 free\001
+2 0 50 -1 0 12 0.0000 2 135 600 2625 2700 free list\001
+0 0 50 -1 0 12 0.0000 2 135 360 3675 3325 lock\001
+1 0 50 -1 0 12 0.0000 2 180 1455 4350 3075 global pool (sbrk)\001
+1 0 50 -1 0 11 0.0000 2 150 1110 4800 1725 N$\\times$S$_t$\001

doc/theses/mubeen_zulfiqar_MMath/figures/AllocDS2.fig

-              rba897d21
+              r2e9b59b
 -2
 2
+2850 2100 3150 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 2925 2175 20 20 2925 2175 2945 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3000 2175 20 20 3000 2175 3020 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3075 2175 20 20 3075 2175 3095 2175
+-6
+4050 2100 4350 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4125 2175 20 20 4125 2175 4145 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4200 2175 20 20 4200 2175 4220 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4275 2175 20 20 4275 2175 4295 2175
+-6
+4650 2100 4950 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4725 2175 20 20 4725 2175 4745 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4800 2175 20 20 4800 2175 4820 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4875 2175 20 20 4875 2175 4895 2175
+-6
+3450 2100 3750 2250
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3525 2175 20 20 3525 2175 3545 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3600 2175 20 20 3600 2175 3620 2175
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 3675 2175 20 20 3675 2175 3695 2175
+-6
+3300 2175 3600 2550
+2850 2475 3150 2850
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
 2175 3375 2400
+2475 2925 2700
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2400 3600 2400 3600 2550 3300 2550 3300 2400
+2700 3150 2700 3150 2850 2850 2850 2850 2700
+-6
+4350 2475 4650 2850
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2475 4425 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 4650 2700 4650 2850 4350 2850 4350 2700
+-6
+3600 2475 3825 3150
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2475 3675 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 3825 2700 3825 2850 3600 2850 3600 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3000 3825 3000 3825 3150 3600 3150 3600 3000
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2775 3675 3000
+-6
+1950 3525 3150 3900
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
+1 1.00 45.00 90.00
+3750 2700 3750 2700 3525
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3525 3150 3525 3150 3900 1950 3900 1950 3525
+0 0 50 -1 0 12 0.0000 2 135 360 2025 3700 lock\001
+-6
+4050 1575 4350 1725
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4125 1650 20 20 4125 1650 4145 1650
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4200 1650 20 20 4200 1650 4220 1650
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4275 1650 20 20 4275 1650 4295 1650
+-6
+4875 2325 6150 3750
+4875 2325 5175 2475
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4950 2400 20 20 4950 2400 4970 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5025 2400 20 20 5025 2400 5045 2400
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5100 2400 20 20 5100 2400 5120 2400
+-6
+4875 3600 5175 3750
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 4950 3675 20 20 4950 3675 4970 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5025 3675 20 20 5025 3675 5045 3675
+3 0 1 0 0 50 -1 20 0.000 1 0.0000 5100 3675 20 20 5100 3675 5120 3675
+-6
+1 0 50 -1 0 12 0.0000 2 180 900 5700 3150 local pools\001
+1 0 50 -1 0 12 0.0000 2 180 465 5700 2925 heaps\001
+-6
+3600 4050 5100 4650
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
+1 1.00 45.00 90.00
+4500 4350 4500 4350 4275
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+4275 5100 4275 5100 4650 3600 4650 3600 4275
+1 0 50 -1 0 12 0.0000 2 180 1455 4350 4200 global pool (sbrk)\001
+0 0 50 -1 0 12 0.0000 2 135 360 3675 4450 lock\001
 -6
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 3150 2250
+2100 2400 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 2850 2250
+2100 2550 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 4650 2250
+2100 2700 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 4950 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 4500 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 5100 2250
+2100 2850 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 3450 2250
+2100 3000 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 3750 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 3300 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 3900 2250
+2100 3600 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 5250 2250
+2100 3900 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 5400 2250
+2100 4050 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 5550 2250
+2100 4200 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
 1800 5700 2250
+2100 4350 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1800 5850 2250
+1 0 3 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1725 2700 2250
+2100 4500 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1500 3300 1800
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1500 3600 1800
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+1500 4800 1500 4800 1800 3000 1800 3000 1500
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
 1275 3375 1575
+1650 2550 2100
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+1275 2700 1575
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 2775 1575
+1650 4050 2100
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2100 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2250 3150 2250
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2250 4650 2250
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2100 3150 2100 3150 2550 1950 2550 1950 2100
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2100 4650 2100 4650 2550 3450 2550 3450 2100
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 2250 2550
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 3750 2550
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
 1275 5175 1575
+2475 2025 2700
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
 1 1.00 45.00 90.00
+1275 5625 1575
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 3750 1575
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+1275 3825 1575
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+1950 6000 1950
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2100 6000 2100
+2775 2025 3000
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+1800 6000 1800 6000 2250 2700 2250 2700 1800
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2175 2775 2400
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2475 2775 2700
+3000 2100 3000 2100 3150 1950 3150 1950 3000
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2700 2850 2700 2850 2850 2700 2850 2700 2700
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2400 2850 2400 2850 2550 2700 2550 2700 2400
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
+1 1.00 45.00 90.00
+2175 4575 2400
+2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+2400 5025 2400 5025 2550 4500 2550 4500 2400
+2700 2100 2700 2100 2850 1950 2850 1950 2700
 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 3
 1 1.00 45.00 90.00
 3525 4650 3525 4650 3150
+3750 4200 3750 4200 3525
 2 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 5
+3150 5100 3150 5100 3750 3600 3750 3600 3150
+2 0 50 -1 0 11 0.0000 2 120 300 2625 1950 lock\001
+1 0 50 -1 0 10 0.0000 2 150 1155 3000 1725 N$\\times$S$_1$\001
+1 0 50 -1 0 10 0.0000 2 150 1155 3600 1725 N$\\times$S$_2$\001
+1 0 50 -1 0 12 0.0000 2 180 390 4425 1500 heap\001
+2 0 50 -1 0 12 0.0000 2 135 1140 2550 1425 kernel threads\001
+2 0 50 -1 0 11 0.0000 2 120 270 2625 2100 size\001
+2 0 50 -1 0 11 0.0000 2 120 270 2625 2250 free\001
+2 0 50 -1 0 12 0.0000 2 135 600 2625 2700 free list\001
+0 0 50 -1 0 12 0.0000 2 135 360 3675 3325 lock\001
+1 0 50 -1 0 12 0.0000 2 180 1455 4350 3075 global pool (sbrk)\001
+1 0 50 -1 0 10 0.0000 2 150 1110 4800 1725 N$\\times$S$_t$\001
+3525 4650 3525 4650 3900 3450 3900 3450 3525
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2400 3150 2400
+1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
+2400 4650 2400
+2 0 50 -1 0 11 0.0000 2 135 300 1875 2250 lock\001
+1 0 50 -1 0 12 0.0000 2 180 1245 3900 1425 H heap buckets\001
+1 0 50 -1 0 12 0.0000 2 180 810 4425 2025 heap$_2$\001
+1 0 50 -1 0 12 0.0000 2 180 810 2175 2025 heap$_1$\001
+2 0 50 -1 0 11 0.0000 2 120 270 1875 2400 size\001
+2 0 50 -1 0 11 0.0000 2 120 270 1875 2550 free\001
+1 0 50 -1 0 12 0.0000 2 180 825 2550 3450 local pool\001
+0 0 50 -1 0 12 0.0000 2 135 360 3525 3700 lock\001
+2 0 50 -1 0 12 0.0000 2 135 600 1875 3000 free list\001
+1 0 50 -1 0 12 0.0000 2 180 825 4050 3450 local pool\001

doc/theses/mubeen_zulfiqar_MMath/intro.tex

-              rba897d21
+              r2e9b59b
 Attempts have been made to perform quasi garbage collection in C/\CC~\cite{Boehm88}, but it is a compromise.
 This thesis only examines dynamic memory-management with \emph{explicit} deallocation.
 While garbage collection and compaction are not part this work, many of the results are applicable to the allocation phase in any memory-management approach.
+While garbage collection and compaction are not part this work, many of the work's results are applicable to the allocation phase in any memory-management approach.
 Most programs use a general-purpose allocator, often the one provided implicitly by the programming-language's runtime.
 …
 \begin{enumerate}[leftmargin=*]
 \item
+Implementation of a new stand-lone concurrent low-latency memory-allocator ($\approx$1,200 lines of code) for C/\CC programs using kernel threads (1:1 threading), and specialized versions of the allocator for programming languages \uC and \CFA using user-level threads running over multiple kernel threads (M:N threading).
+\item
+Adopt returning of @nullptr@ for a zero-sized allocation, rather than an actual memory address, both of which can be passed to @free@.
+\item
+Extended the standard C heap functionality by preserving with each allocation its original request size versus the amount allocated, if an allocation is zero fill, and the allocation alignment.
+\item
+Use the zero fill and alignment as \emph{sticky} properties for @realloc@, to realign existing storage, or preserve existing zero-fill and alignment when storage is copied.
+Implementation of a new stand-lone concurrent low-latency memory-allocator ($\approx$1,200 lines of code) for C/\CC programs using kernel threads (1:1 threading), and specialized versions of the allocator for the programming languages \uC and \CFA using user-level threads running over multiple kernel threads (M:N threading).
+\item
+Adopt @nullptr@ return for a zero-sized allocation, rather than an actual memory address, which can be passed to @free@.
+\item
+Extend the standard C heap functionality by preserving with each allocation:
+\begin{itemize}[itemsep=0pt]
+\item
+its request size plus the amount allocated,
+\item
+whether an allocation is zero fill,
+\item
+and allocation alignment.
+\end{itemize}
+\item
+Use the preserved zero fill and alignment as \emph{sticky} properties for @realloc@ to zero-fill and align when storage is extended or copied.
 Without this extension, it is unsafe to @realloc@ storage initially allocated with zero-fill/alignment as these properties are not preserved when copying.
 This silent generation of a problem is unintuitive to programmers and difficult to locate because it is transient.
 …
 @resize( oaddr, alignment, size )@ re-purpose an old allocation with new alignment but \emph{without} preserving fill.
 \item
 @realloc( oaddr, alignment, size )@ same as previous @realloc@ but adding or changing alignment.
+@realloc( oaddr, alignment, size )@ same as @realloc@ but adding or changing alignment.
 \item
 @aalloc( dim, elemSize )@ same as @calloc@ except memory is \emph{not} zero filled.
 …
 \item
 Provide additional heap wrapper functions in \CFA to provide a complete orthogonal set of allocation operations and properties.
+Provide additional heap wrapper functions in \CFA creating an orthogonal set of allocation operations and properties.
 \item
 …
 @malloc_size( addr )@ returns the size of the memory allocation pointed-to by @addr@.
 \item
 @malloc_usable_size( addr )@ returns the usable size of the memory pointed-to by @addr@, i.e., the bin size containing the allocation, where @malloc_size( addr )@ $\le$ @malloc_usable_size( addr )@.
+@malloc_usable_size( addr )@ returns the usable (total) size of the memory pointed-to by @addr@, i.e., the bin size containing the allocation, where @malloc_size( addr )@ $\le$ @malloc_usable_size( addr )@.
 \end{itemize}
 …
 \item
 Provide complete, fast, and contention-free allocation statistics to help understand program behaviour:
+Provide complete, fast, and contention-free allocation statistics to help understand allocation behaviour:
 \begin{itemize}
 \item

doc/theses/mubeen_zulfiqar_MMath/performance.tex

rba897d21	r2e9b59b
1	1	\chapter{Performance}
	2	\label{c:Performance}
2	3
3	4	\section{Machine Specification}

doc/theses/mubeen_zulfiqar_MMath/uw-ethesis.bib

-              rba897d21
+              r2e9b59b
+}
+@misc{nedmalloc,
+    author      = {Niall Douglas},
+    title       = {nedmalloc version 1.06 Beta},
+    month       = jan,
+    year        = 2010,
+    note        = {\textsf{http://\-prdownloads.\-sourceforge.\-net/\-nedmalloc/\-nedmalloc\_v1.06beta1\_svn1151.zip}},
+@misc{ptmalloc2,
+    author      = {Wolfram Gloger},
+    title       = {ptmalloc version 2},
+    month       = jun,
+    year        = 2006,
+    note        = {\href{http://www.malloc.de/malloc/ptmalloc2-current.tar.gz}{http://www.malloc.de/\-malloc/\-ptmalloc2-current.tar.gz}},
+}
+@misc{GNUallocAPI,
+    author      = {GNU},
+    title       = {Summary of malloc-Related Functions},
+    year        = 2020,
+    note        = {\href{https://www.gnu.org/software/libc/manual/html\_node/Summary-of-Malloc.html}{https://www.gnu.org/\-software/\-libc/\-manual/\-html\_node/\-Summary-of-Malloc.html}},
+}
+@misc{SeriallyReusable,
+    author      = {IBM},
+    title       = {Serially reusable programs},
+    month       = mar,
+    year        = 2021,
+    note        = {\href{https://www.ibm.com/docs/en/ztpf/1.1.0.15?topic=structures-serially-reusable-programs}{https://www.ibm.com/\-docs/\-en/\-ztpf/\-1.1.0.15?\-topic=structures-serially-reusable-programs}},
+}
+@misc{librseq,
+    author      = {Mathieu Desnoyers},
+    title       = {Library for Restartable Sequences},
+    month       = mar,
+    year        = 2022,
+    note        = {\href{https://github.com/compudj/librseq}{https://github.com/compudj/librseq}},
+}

doc/theses/mubeen_zulfiqar_MMath/uw-ethesis.tex

-              rba897d21
+              r2e9b59b
 % For hyperlinked PDF, suitable for viewing on a computer, use this:
 \documentclass[letterpaper,12pt,titlepage,oneside,final]{book}
+\usepackage[T1]{fontenc}        % Latin-1 => 256-bit characters, => | not dash, <> not Spanish question marks
 % For PDF, suitable for double-sided printing, change the PrintVersion variable below to "true" and use this \documentclass line instead of the one above:
 …
 % Use the "hyperref" package
 % N.B. HYPERREF MUST BE THE LAST PACKAGE LOADED; ADD ADDITIONAL PKGS ABOVE
+\usepackage[pagebackref=true]{hyperref} % with basic options
+\usepackage{url}
+\usepackage[dvips,pagebackref=true]{hyperref} % with basic options
 %\usepackage[pdftex,pagebackref=true]{hyperref}
 % N.B. pagebackref=true provides links back from the References to the body text. This can cause trouble for printing.
 …
     citecolor=blue,        % color of links to bibliography
     filecolor=magenta,      % color of file links
+    urlcolor=blue           % color of external links
+    urlcolor=blue,           % color of external links
+    breaklinks=true
+}
 \ifthenelse{\boolean{PrintVersion}}{   % for improved print quality, change some hyperref options
 …
     urlcolor=black
 }}{} % end of ifthenelse (no else)
+%\usepackage[dvips,plainpages=false,pdfpagelabels,pdfpagemode=UseNone,pagebackref=true,breaklinks=true,colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue]{hyperref}
+\usepackage{breakurl}
+\urlstyle{sf}
 %\usepackage[automake,toc,abbreviations]{glossaries-extra} % Exception to the rule of hyperref being the last add-on package
 …
 \input{common}
 %\usepackageinput{common}
+\CFAStyle                                               % CFA code-style for all languages
+\CFAStyle                                               % CFA code-style
+\lstset{language=CFA}                                   % default language
 \lstset{basicstyle=\linespread{0.9}\sf}                 % CFA typewriter font
 \newcommand{\uC}{$\mu$\CC}

doc/theses/thierry_delisle_PhD/thesis/Makefile

-              rba897d21
+              r2e9b59b
 PICTURES = ${addsuffix .pstex, \
         base \
+        base_avg \
+        cache-share \
+        cache-noshare \
         empty \
         emptybit \
 …
         system \
         cycle \
+        result.cycle.jax.ops \
+}
 …
         python3 $< $@
+build/result.%.ns.svg : data/% | ${Build}
+        ../../../../benchmark/plot.py -f $< -o $@ -y "ns per ops"
+build/result.%.ops.svg : data/% | ${Build}
+        ../../../../benchmark/plot.py -f $< -o $@ -y "Ops per second"
 ## pstex with inverted colors
 %.dark.pstex : fig/%.fig Makefile | ${Build}

doc/theses/thierry_delisle_PhD/thesis/fig/base.fig

-              rba897d21
+              r2e9b59b
 5210 5550 4950 5250 4950 5100 5210 5250 5470 5550 5470
 5210
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 0 0 2
+5700 3600 1200
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 0 0 2
+5700 4800 1200
+1 1 1 0 7 50 -1 -1 4.000 0 0 -1 0 0 2
+5700 6000 1200
 2 -1 50 -1 0 12 0.0000 2 135 630 2100 3075 Threads\001
 2 -1 50 -1 0 12 0.0000 2 165 450 2100 2850 Ready\001

doc/theses/thierry_delisle_PhD/thesis/glossary.tex

rba897d21	r2e9b59b
101	101
102	102	\longnewglossaryentry{at}
103		{name={~~fred~~}}
	103	{name={task}}
104	104	{
105	105	Abstract object representing an unit of work. Systems will offer one or more concrete implementations of this concept (\eg \gls{kthrd}, \gls{job}), however, most of the concept of schedulings are independent of the particular implementations of the work representation. For this reason, this document use the term \Gls{at} to mean any representation and not one in particular.

doc/theses/thierry_delisle_PhD/thesis/local.bib

-              rba897d21
+              r2e9b59b
   note = "[Online; accessed 9-February-2021]"
+}
+@misc{wiki:rcu,
+  author = "{Wikipedia contributors}",
+  title = "Read-copy-update --- {W}ikipedia{,} The Free Encyclopedia",
+  year = "2022",
+  url = "https://en.wikipedia.org/wiki/Linear_congruential_generator",
+  note = "[Online; accessed 12-April-2022]"
+}
+@misc{wiki:rwlock,
+  author = "{Wikipedia contributors}",
+  title = "Readers-writer lock --- {W}ikipedia{,} The Free Encyclopedia",
+  year = "2021",
+  url = "https://en.wikipedia.org/wiki/Readers%E2%80%93writer_lock",
+  note = "[Online; accessed 12-April-2022]"
+}

doc/theses/thierry_delisle_PhD/thesis/text/core.tex

-              rba897d21
+              r2e9b59b
 Before discussing scheduling in general, where it is important to address systems that are changing states, this document discusses scheduling in a somewhat ideal scenario, where the system has reached a steady state. For this purpose, a steady state is loosely defined as a state where there are always \glspl{thrd} ready to run and the system has the resources necessary to accomplish the work, \eg, enough workers. In short, the system is neither overloaded nor underloaded.
 I believe it is important to discuss the steady state first because it is the easiest case to handle and, relatedly, the case in which the best performance is to be expected. As such, when the system is either overloaded or underloaded, a common approach is to try to adapt the system to this new load and return to the steady state, \eg, by adding or removing workers. Therefore, flaws in scheduling the steady state can to be pervasive in all states.
+It is important to discuss the steady state first because it is the easiest case to handle and, relatedly, the case in which the best performance is to be expected. As such, when the system is either overloaded or underloaded, a common approach is to try to adapt the system to this new load and return to the steady state, \eg, by adding or removing workers. Therefore, flaws in scheduling the steady state tend to be pervasive in all states.
 \section{Design Goals}
 …
 It is important to note that these guarantees are expected only up to a point. \Glspl{thrd} that are ready to run should not be prevented to do so, but they still share the limited hardware resources. Therefore, the guarantee is considered respected if a \gls{thrd} gets access to a \emph{fair share} of the hardware resources, even if that share is very small.
 Similarly the performance guarantee, the lack of interference among threads, is only relevant up to a point. Ideally, the cost of running and blocking should be constant regardless of contention, but the guarantee is considered satisfied if the cost is not \emph{too high} with or without contention. How much is an acceptable cost is obviously highly variable. For this document, the performance experimentation attempts to show the cost of scheduling is at worst equivalent to existing algorithms used in popular languages. This demonstration can be made by comparing applications built in \CFA to applications built with other languages or other models. Recall programmer expectation is that the impact of the scheduler can be ignored. Therefore, if the cost of scheduling is equivalent to or lower than other popular languages, I consider the guarantee achieved.
+Similarly the performance guarantee, the lack of interference among threads, is only relevant up to a point. Ideally, the cost of running and blocking should be constant regardless of contention, but the guarantee is considered satisfied if the cost is not \emph{too high} with or without contention. How much is an acceptable cost is obviously highly variable. For this document, the performance experimentation attempts to show the cost of scheduling is at worst equivalent to existing algorithms used in popular languages. This demonstration can be made by comparing applications built in \CFA to applications built with other languages or other models. Recall programmer expectation is that the impact of the scheduler can be ignored. Therefore, if the cost of scheduling is compatitive to other popular languages, the guarantee will be consider achieved.
 More precisely the scheduler should be:
 …
 \end{itemize}
+\subsection{Fairness vs Scheduler Locality}
+\subsection{Fairness Goals}
+For this work fairness will be considered as having two strongly related requirements: true starvation freedom and ``fast'' load balancing.
+\paragraph{True starvation freedom} is more easily defined: As long as at least one \proc continues to dequeue \ats, all read \ats should be able to run eventually.
+In any running system, \procs can stop dequeing \ats if they start running a \at that will simply never park.
+Traditional workstealing schedulers do not have starvation freedom in these cases.
+Now this requirement begs the question, what about preemption?
+Generally speaking preemption happens on the timescale of several milliseconds, which brings us to the next requirement: ``fast'' load balancing.
+\paragraph{Fast load balancing} means that load balancing should happen faster than preemption would normally allow.
+For interactive applications that need to run at 60, 90, 120 frames per second, \ats having to wait for several millseconds to run are effectively starved.
+Therefore load-balancing should be done at a faster pace, one that can detect starvation at the microsecond scale.
+With that said, this is a much fuzzier requirement since it depends on the number of \procs, the number of \ats and the general load of the system.
+\subsection{Fairness vs Scheduler Locality} \label{fairnessvlocal}
 An important performance factor in modern architectures is cache locality. Waiting for data at lower levels or not present in the cache can have a major impact on performance. Having multiple \glspl{hthrd} writing to the same cache lines also leads to cache lines that must be waited on. It is therefore preferable to divide data among each \gls{hthrd}\footnote{This partitioning can be an explicit division up front or using data structures where different \glspl{hthrd} are naturally routed to different cache lines.}.
 For a scheduler, having good locality\footnote{This section discusses \emph{internal locality}, \ie, the locality of the data used by the scheduler versus \emph{external locality}, \ie, how the data used by the application is affected by scheduling. External locality is a much more complicated subject and is discussed in part~\ref{Evaluation} on evaluation.}, \ie, having the data local to each \gls{hthrd}, generally conflicts with fairness. Indeed, good locality often requires avoiding the movement of cache lines, while fairness requires dynamically moving a \gls{thrd}, and as consequence cache lines, to a \gls{hthrd} that is currently available.
+For a scheduler, having good locality\footnote{This section discusses \emph{internal locality}, \ie, the locality of the data used by the scheduler versus \emph{external locality}, \ie, how the data used by the application is affected by scheduling. External locality is a much more complicated subject and is discussed in the next section.}, \ie, having the data local to each \gls{hthrd}, generally conflicts with fairness. Indeed, good locality often requires avoiding the movement of cache lines, while fairness requires dynamically moving a \gls{thrd}, and as consequence cache lines, to a \gls{hthrd} that is currently available.
 However, I claim that in practice it is possible to strike a balance between fairness and performance because these goals do not necessarily overlap temporally, where Figure~\ref{fig:fair} shows a visual representation of this behaviour. As mentioned, some unfairness is acceptable; therefore it is desirable to have an algorithm that prioritizes cache locality as long as thread delay does not exceed the execution mental-model.
 …
 \end{figure}
+\section{Design}
+\subsection{Performance Challenges}\label{pref:challenge}
+While there exists a multitude of potential scheduling algorithms, they generally always have to contend with the same performance challenges. Since these challenges are recurring themes in the design of a scheduler it is relevant to describe the central ones here before looking at the design.
+\subsubsection{Scalability}
+The most basic performance challenge of a scheduler is scalability.
+Given a large number of \procs and an even larger number of \ats, scalability measures how fast \procs can enqueue and dequeues \ats.
+One could expect that doubling the number of \procs would double the rate at which \ats are dequeued, but contention on the internal data structure of the scheduler can lead to worst improvements.
+While the ready-queue itself can be sharded to alleviate the main source of contention, auxillary scheduling features, \eg counting ready \ats, can also be sources of contention.
+\subsubsection{Migration Cost}
+Another important source of latency in scheduling is migration.
+An \at is said to have migrated if it is executed by two different \proc consecutively, which is the process discussed in \ref{fairnessvlocal}.
+Migrations can have many different causes, but it certain programs it can be all but impossible to limit migrations.
+Chapter~\ref{microbench} for example, has a benchmark where any \at can potentially unblock any other \at, which can leat to \ats migrating more often than not.
+Because of this it is important to design the internal data structures of the scheduler to limit the latency penalty from migrations.
+\section{Inspirations}
 In general, a na\"{i}ve \glsxtrshort{fifo} ready-queue does not scale with increased parallelism from \glspl{hthrd}, resulting in decreased performance. The problem is adding/removing \glspl{thrd} is a single point of contention. As shown in the evaluation sections, most production schedulers do scale when adding \glspl{hthrd}. The solution to this problem is to shard the ready-queue : create multiple sub-ready-queues that multiple \glspl{hthrd} can access and modify without interfering.
 Before going into the design of \CFA's scheduler proper, I want to discuss two sharding solutions which served as the inspiration scheduler in this thesis.
+Before going into the design of \CFA's scheduler proper, it is relevant to discuss two sharding solutions which served as the inspiration scheduler in this thesis.
 \subsection{Work-Stealing}
+As I mentioned in \ref{existing:workstealing}, a popular pattern shard the ready-queue is work-stealing. As mentionned, in this pattern each \gls{proc} has its own ready-queue and \glspl{proc} only access each other's ready-queue if they run out of work.
+The interesting aspect of workstealing happen in easier scheduling cases, \ie enough work for everyone but no more and no load balancing needed. In these cases, work-stealing is close to optimal scheduling: it can achieve perfect locality and have no contention.
+As mentioned in \ref{existing:workstealing}, a popular pattern shard the ready-queue is work-stealing.
+In this pattern each \gls{proc} has its own local ready-queue and \glspl{proc} only access each other's ready-queue if they run out of work on their local ready-queue.
+The interesting aspect of workstealing happen in easier scheduling cases, \ie enough work for everyone but no more and no load balancing needed.
+In these cases, work-stealing is close to optimal scheduling: it can achieve perfect locality and have no contention.
 On the other hand, work-stealing schedulers only attempt to do load-balancing when a \gls{proc} runs out of work.
 This means that the scheduler may never balance unfairness that does not result in a \gls{proc} running out of work.
+This means that the scheduler never balances unfair loads unless they result in a \gls{proc} running out of work.
 Chapter~\ref{microbench} shows that in pathological cases this problem can lead to indefinite starvation.
 Based on these observation, I conclude that \emph{perfect} scheduler should behave very similarly to work-stealing in the easy cases, but should have more proactive load-balancing if the need arises.
+Based on these observation, the conclusion is that a \emph{perfect} scheduler should behave very similarly to work-stealing in the easy cases, but should have more proactive load-balancing if the need arises.
 \subsection{Relaxed-Fifo}
 An entirely different scheme is to create a ``relaxed-FIFO'' queue as in \todo{cite Trevor's paper}. This approach forgos any ownership between \gls{proc} and ready-queue, and simply creates a pool of ready-queues from which the \glspl{proc} can pick from.
 \Glspl{proc} choose ready-queus at random, but timestamps are added to all elements of the queue and dequeues are done by picking two queues and dequeing the oldest element.
+All subqueues are protected by TryLocks and \procs simply pick a different subqueue if they fail to acquire the TryLock.
 The result is a queue that has both decent scalability and sufficient fairness.
 The lack of ownership means that as long as one \gls{proc} is still able to repeatedly dequeue elements, it is unlikely that any element will stay on the queue for much longer than any other element.
 …
 While the fairness, of this scheme is good, it does suffer in terms of performance.
+It requires very wide sharding, \eg at least 4 queues per \gls{hthrd}, and the randomness means locality can suffer significantly and finding non-empty queues can be difficult.
+\section{\CFA}
+The \CFA is effectively attempting to merge these two approaches, keeping the best of both.
+It is based on the
+It requires very wide sharding, \eg at least 4 queues per \gls{hthrd}, and finding non-empty queues can be difficult if there are too few ready \ats.
+\section{Relaxed-FIFO++}
+Since it has inherent fairness quelities and decent performance in the presence of many \ats, the relaxed-FIFO queue appears as a good candidate to form the basis of a scheduler.
+The most obvious problems is for workloads where the number of \ats is barely greater than the number of \procs.
+In these situations, the wide sharding means most of the sub-queues from which the relaxed queue is formed will be empty.
+The consequence is that when a dequeue operations attempts to pick a sub-queue at random, it is likely that it picks an empty sub-queue and will have to pick again.
+This problem can repeat an unbounded number of times.
+As this is the most obvious challenge, it is worth addressing first.
+The obvious solution is to supplement each subqueue with some sharded data structure that keeps track of which subqueues are empty.
+This data structure can take many forms, for example simple bitmask or a binary tree that tracks which branch are empty.
+Following a binary tree on each pick has fairly good Big O complexity and many modern architectures have powerful bitmask manipulation instructions.
+However, precisely tracking which sub-queues are empty is actually fundamentally problematic.
+The reason is that each subqueues are already a form of sharding and the sharding width has presumably already chosen to avoid contention.
+However, tracking which ready queue is empty is only useful if the tracking mechanism uses denser sharding than the sub queues, then it will invariably create a new source of contention.
+But if the tracking mechanism is not denser than the sub-queues, then it will generally not provide useful because reading this new data structure risks being as costly as simply picking a sub-queue at random.
+Early experiments with this approach have shown that even with low success rates, randomly picking a sub-queue can be faster than a simple tree walk.
+The exception to this rule is using local tracking.
+If each \proc keeps track locally of which sub-queue is empty, then this can be done with a very dense data structure without introducing a new source of contention.
+The consequence of local tracking however, is that the information is not complete.
+Each \proc is only aware of the last state it saw each subqueues but does not have any information about freshness.
+Even on systems with low \gls{hthrd} count, \eg 4 or 8, this can quickly lead to the local information being no better than the random pick.
+This is due in part to the cost of this maintaining this information and its poor quality.
+However, using a very low cost approach to local tracking may actually be beneficial.
+If the local tracking is no more costly than the random pick, than \emph{any} improvement to the succes rate, however low it is, would lead to a performance benefits.
+This leads to the following approach:
+\subsection{Dynamic Entropy}\cit{https://xkcd.com/2318/}
+The Relaxed-FIFO approach can be made to handle the case of mostly empty sub-queues by tweaking the \glsxtrlong{prng}.
+The \glsxtrshort{prng} state can be seen as containing a list of all the future sub-queues that will be accessed.
+While this is not particularly useful on its own, the consequence is that if the \glsxtrshort{prng} algorithm can be run \emph{backwards}, then the state also contains a list of all the subqueues that were accessed.
+Luckily, bidirectional \glsxtrshort{prng} algorithms do exist, for example some Linear Congruential Generators\cit{https://en.wikipedia.org/wiki/Linear\_congruential\_generator} support running the algorithm backwards while offering good quality and performance.
+This particular \glsxtrshort{prng} can be used as follows:
+Each \proc maintains two \glsxtrshort{prng} states, which whill be refered to as \texttt{F} and \texttt{B}.
+When a \proc attempts to dequeue a \at, it picks the subqueues by running the \texttt{B} backwards.
+When a \proc attempts to enqueue a \at, it runs \texttt{F} forward to pick to subqueue to enqueue to.
+If the enqueue is successful, the state \texttt{B} is overwritten with the content of \texttt{F}.
+The result is that each \proc will tend to dequeue \ats that it has itself enqueued.
+When most sub-queues are empty, this technique increases the odds of finding \ats at very low cost, while also offering an improvement on locality in many cases.
+However, while this approach does notably improve performance in many cases, this algorithm is still not competitive with work-stealing algorithms.
+The fundamental problem is that the constant randomness limits how much locality the scheduler offers.
+This becomes problematic both because the scheduler is likely to get cache misses on internal data-structures and because migration become very frequent.
+Therefore since the approach of modifying to relaxed-FIFO algorithm to behave more like work stealing does not seem to pan out, the alternative is to do it the other way around.
+\section{Work Stealing++}
+To add stronger fairness guarantees to workstealing a few changes.
+First, the relaxed-FIFO algorithm has fundamentally better fairness because each \proc always monitors all subqueues.
+Therefore the workstealing algorithm must be prepended with some monitoring.
+Before attempting to dequeue from a \proc's local queue, the \proc must make some effort to make sure remote queues are not being neglected.
+To make this possible, \procs must be able to determie which \at has been on the ready-queue the longest.
+Which is the second aspect that much be added.
+The relaxed-FIFO approach uses timestamps for each \at and this is also what is done here.
 \begin{figure}
         \centering
         \input{base.pstex_t}
         \caption[Base \CFA design]{Base \CFA design \smallskip\newline A list of sub-ready queues offers the sharding, two per \glspl{proc}. However, \glspl{proc} can access any of the sub-queues.}
+        \caption[Base \CFA design]{Base \CFA design \smallskip\newline A Pool of sub-ready queues offers the sharding, two per \glspl{proc}. Each \gls{proc} have local subqueues, however \glspl{proc} can access any of the sub-queues. Each \at is timestamped when enqueued.}
         \label{fig:base}
 \end{figure}
+% The common solution to the single point of contention is to shard the ready-queue so each \gls{hthrd} can access the ready-queue without contention, increasing performance.
+% \subsection{Sharding} \label{sec:sharding}
+% An interesting approach to sharding a queue is presented in \cit{Trevors paper}. This algorithm presents a queue with a relaxed \glsxtrshort{fifo} guarantee using an array of strictly \glsxtrshort{fifo} sublists as shown in Figure~\ref{fig:base}. Each \emph{cell} of the array has a timestamp for the last operation and a pointer to a linked-list with a lock. Each node in the list is marked with a timestamp indicating when it is added to the list. A push operation is done by picking a random cell, acquiring the list lock, and pushing to the list. If the cell is locked, the operation is simply retried on another random cell until a lock is acquired. A pop operation is done in a similar fashion except two random cells are picked. If both cells are unlocked with non-empty lists, the operation pops the node with the oldest timestamp. If one of the cells is unlocked and non-empty, the operation pops from that cell. If both cells are either locked or empty, the operation picks two new random cells and tries again.
+% \begin{figure}
+%       \centering
+%       \input{base.pstex_t}
+%       \caption[Relaxed FIFO list]{Relaxed FIFO list \smallskip\newline List at the base of the scheduler: an array of strictly FIFO lists. The timestamp is in all nodes and cell arrays.}
+%       \label{fig:base}
+% \end{figure}
+% \subsection{Finding threads}
+% Once threads have been distributed onto multiple queues, identifying empty queues becomes a problem. Indeed, if the number of \glspl{thrd} does not far exceed the number of queues, it is probable that several of the cell queues are empty. Figure~\ref{fig:empty} shows an example with 2 \glspl{thrd} running on 8 queues, where the chances of getting an empty queue is 75\% per pick, meaning two random picks yield a \gls{thrd} only half the time. This scenario leads to performance problems since picks that do not yield a \gls{thrd} are not useful and do not necessarily help make more informed guesses.
+% \begin{figure}
+%       \centering
+%       \input{empty.pstex_t}
+%       \caption[``More empty'' Relaxed FIFO list]{``More empty'' Relaxed FIFO list \smallskip\newline Emptier state of the queue: the array contains many empty cells, that is strictly FIFO lists containing no elements.}
+%       \label{fig:empty}
+% \end{figure}
+% There are several solutions to this problem, but they ultimately all have to encode if a cell has an empty list. My results show the density and locality of this encoding is generally the dominating factor in these scheme. Classic solutions to this problem use one of three techniques to encode the information:
+% \paragraph{Dense Information} Figure~\ref{fig:emptybit} shows a dense bitmask to identify the cell queues currently in use. This approach means processors can often find \glspl{thrd} in constant time, regardless of how many underlying queues are empty. Furthermore, modern x86 CPUs have extended bit manipulation instructions (BMI2) that allow searching the bitmask with very little overhead compared to the randomized selection approach for a filled ready queue, offering good performance even in cases with many empty inner queues. However, this technique has its limits: with a single word\footnote{Word refers here to however many bits can be written atomically.} bitmask, the total amount of ready-queue sharding is limited to the number of bits in the word. With a multi-word bitmask, this maximum limit can be increased arbitrarily, but the look-up time increases. Finally, a dense bitmap, either single or multi-word, causes additional contention problems that reduces performance because of cache misses after updates. This central update bottleneck also means the information in the bitmask is more often stale before a processor can use it to find an item, \ie mask read says there are available \glspl{thrd} but none on queue when the subsequent atomic check is done.
+% \begin{figure}
+%       \centering
+%       \vspace*{-5pt}
+%       {\resizebox{0.75\textwidth}{!}{\input{emptybit.pstex_t}}}
+%       \vspace*{-5pt}
+%       \caption[Underloaded queue with bitmask]{Underloaded queue with bitmask indicating array cells with items.}
+%       \label{fig:emptybit}
+%       \vspace*{10pt}
+%       {\resizebox{0.75\textwidth}{!}{\input{emptytree.pstex_t}}}
+%       \vspace*{-5pt}
+%       \caption[Underloaded queue with binary search-tree]{Underloaded queue with binary search-tree indicating array cells with items.}
+%       \label{fig:emptytree}
+%       \vspace*{10pt}
+%       {\resizebox{0.95\textwidth}{!}{\input{emptytls.pstex_t}}}
+%       \vspace*{-5pt}
+%       \caption[Underloaded queue with per processor bitmask]{Underloaded queue with per processor bitmask indicating array cells with items.}
+%       \label{fig:emptytls}
+% \end{figure}
+% \paragraph{Sparse Information} Figure~\ref{fig:emptytree} shows an approach using a hierarchical tree data-structure to reduce contention and has been shown to work in similar cases~\cite{ellen2007snzi}. However, this approach may lead to poorer performance due to the inherent pointer chasing cost while still allowing significant contention on the nodes of the tree if the tree is shallow.
+% \paragraph{Local Information} Figure~\ref{fig:emptytls} shows an approach using dense information, similar to the bitmap, but each \gls{hthrd} keeps its own independent copy. While this approach can offer good scalability \emph{and} low latency, the liveliness and discovery of the information can become a problem. This case is made worst in systems with few processors where even blind random picks can find \glspl{thrd} in a few tries.
+% I built a prototype of these approaches and none of these techniques offer satisfying performance when few threads are present. All of these approach hit the same 2 problems. First, randomly picking sub-queues is very fast. That speed means any improvement to the hit rate can easily be countered by a slow-down in look-up speed, whether or not there are empty lists. Second, the array is already sharded to avoid contention bottlenecks, so any denser data structure tends to become a bottleneck. In all cases, these factors meant the best cases scenario, \ie many threads, would get worst throughput, and the worst-case scenario, few threads, would get a better hit rate, but an equivalent poor throughput. As a result I tried an entirely different approach.
+% \subsection{Dynamic Entropy}\cit{https://xkcd.com/2318/}
+% In the worst-case scenario there are only few \glspl{thrd} ready to run, or more precisely given $P$ \glspl{proc}\footnote{For simplicity, this assumes there is a one-to-one match between \glspl{proc} and \glspl{hthrd}.}, $T$ \glspl{thrd} and $\epsilon$ a very small number, than the worst case scenario can be represented by $T = P + \epsilon$, with $\epsilon \ll P$. It is important to note in this case that fairness is effectively irrelevant. Indeed, this case is close to \emph{actually matching} the model of the ``Ideal multi-tasking CPU'' on page \pageref{q:LinuxCFS}. In this context, it is possible to use a purely internal-locality based approach and still meet the fairness requirements. This approach simply has each \gls{proc} running a single \gls{thrd} repeatedly. Or from the shared ready-queue viewpoint, each \gls{proc} pushes to a given sub-queue and then pops from the \emph{same} subqueue. The challenge is for the the scheduler to achieve good performance in both the $T = P + \epsilon$ case and the $T \gg P$ case, without affecting the fairness guarantees in the later.
+% To handle this case, I use a \glsxtrshort{prng}\todo{Fix missing long form} in a novel way. There exist \glsxtrshort{prng}s that are fast, compact and can be run forward \emph{and} backwards.  Linear congruential generators~\cite{wiki:lcg} are an example of \glsxtrshort{prng}s of such \glsxtrshort{prng}s. The novel approach is to use the ability to run backwards to ``replay'' the \glsxtrshort{prng}. The scheduler uses an exclusive \glsxtrshort{prng} instance per \gls{proc}, the random-number seed effectively starts an encoding that produces a list of all accessed subqueues, from latest to oldest. Replaying the \glsxtrshort{prng} to identify cells accessed recently and which probably have data still cached.
+% The algorithm works as follows:
+% \begin{itemize}
+%       \item Each \gls{proc} has two \glsxtrshort{prng} instances, $F$ and $B$.
+%       \item Push and Pop operations occur as discussed in Section~\ref{sec:sharding} with the following exceptions:
+%       \begin{itemize}
+%               \item Push operations use $F$ going forward on each try and on success $F$ is copied into $B$.
+%               \item Pop operations use $B$ going backwards on each try.
+%       \end{itemize}
+% \end{itemize}
+% The main benefit of this technique is that it basically respects the desired properties of Figure~\ref{fig:fair}. When looking for work, a \gls{proc} first looks at the last cell they pushed to, if any, and then move backwards through its accessed cells. As the \gls{proc} continues looking for work, $F$ moves backwards and $B$ stays in place. As a result, the relation between the two becomes weaker, which means that the probablisitic fairness of the algorithm reverts to normal. Chapter~\ref{proofs} discusses more formally the fairness guarantees of this algorithm.
+% \section{Details}
+The algorithm is structure as shown in Figure~\ref{fig:base}.
+This is very similar to classic workstealing except the local queues are placed in an array so \procs can access eachother's queue in constant time.
+Sharding width can be adjusted based on need.
+When a \proc attempts to dequeue a \at, it first picks a random remote queue and compares its timestamp to the timestamps of the local queue(s), dequeue from the remote queue if needed.
+Implemented as as naively state above, this approach has some obvious performance problems.
+First, it is necessary to have some damping effect on helping.
+Random effects like cache misses and preemption can add spurious but short bursts of latency for which helping is not helpful, pun intended.
+The effect of these bursts would be to cause more migrations than needed and make this workstealing approach slowdown to the match the relaxed-FIFO approach.
+\begin{figure}
+        \centering
+        \input{base_avg.pstex_t}
+        \caption[\CFA design with Moving Average]{\CFA design with Moving Average \smallskip\newline A moving average is added to each subqueue.}
+        \label{fig:base-ma}
+\end{figure}
+A simple solution to this problem is to compare an exponential moving average\cit{https://en.wikipedia.org/wiki/Moving\_average\#Exponential\_moving\_average} instead if the raw timestamps, shown in Figure~\ref{fig:base-ma}.
+Note that this is slightly more complex than it sounds because since the \at at the head of a subqueue is still waiting, its wait time has not ended.
+Therefore the exponential moving average is actually an exponential moving average of how long each already dequeued \at have waited.
+To compare subqueues, the timestamp at the head must be compared to the current time, yielding the bestcase wait time for the \at at the head of the queue.
+This new waiting is averaged with the stored average.
+To limit even more the amount of unnecessary migration, a bias can be added to the local queue, where a remote queue is helped only if its moving average is more than \emph{X} times the local queue's average.
+None of the experimentation that I have run with these scheduler seem to indicate that the choice of the weight for the moving average or the choice of bis is particularly important.
+Weigths and biases of similar \emph{magnitudes} have similar effects.
+With these additions to workstealing, scheduling can be made as fair as the relaxed-FIFO approach, well avoiding the majority of unnecessary migrations.
+Unfortunately, the performance of this approach does suffer in the cases with no risks of starvation.
+The problem is that the constant polling of remote subqueues generally entail a cache miss.
+To make things worst, remote subqueues that are very active, \ie \ats are frequently enqueued and dequeued from them, the higher the chances are that polling will incurr a cache-miss.
+Conversly, the active subqueues do not benefit much from helping since starvation is already a non-issue.
+This puts this algorithm in an akward situation where it is paying for a cost, but the cost itself suggests the operation was unnecessary.
+The good news is that this problem can be mitigated
+\subsection{Redundant Timestamps}
+The problem with polling remote queues is due to a tension between the consistency requirement on the subqueue.
+For the subqueues, correctness is critical. There must be a consensus among \procs on which subqueues hold which \ats.
+Since the timestamps are use for fairness, it is alco important to have consensus and which \at is the oldest.
+However, when deciding if a remote subqueue is worth polling, correctness is much less of a problem.
+Since the only need is that a subqueue will eventually be polled, some data staleness can be acceptable.
+This leads to a tension where stale timestamps are only problematic in some cases.
+Furthermore, stale timestamps can be somewhat desirable since lower freshness requirements means less tension on the cache coherence protocol.
+\begin{figure}
+        \centering
+        % \input{base_ts2.pstex_t}
+        \caption[\CFA design with Redundant Timestamps]{\CFA design with Redundant Timestamps \smallskip\newline A array is added containing a copy of the timestamps. These timestamps are written to with relaxed atomics, without fencing, leading to fewer cache invalidations.}
+        \label{fig:base-ts2}
+\end{figure}
+A solution to this is to create a second array containing a copy of the timestamps and average.
+This copy is updated \emph{after} the subqueue's critical sections using relaxed atomics.
+\Glspl{proc} now check if polling is needed by comparing the copy of the remote timestamp instead of the actual timestamp.
+The result is that since there is no fencing, the writes can be buffered and cause fewer cache invalidations.
+The correctness argument here is somewhat subtle.
+The data used for deciding whether or not to poll a queue can be stale as long as it does not cause starvation.
+Therefore, it is acceptable if stale data make queues appear older than they really are but not fresher.
+For the timestamps, this means that missing writes to the timestamp is acceptable since they will make the head \at look older.
+For the moving average, as long as the operation are RW-safe, the average is guaranteed to yield a value that is between the oldest and newest values written.
+Therefore this unprotected read of the timestamp and average satisfy the limited correctness that is required.
+\begin{figure}
+        \centering
+        \input{cache-share.pstex_t}
+        \caption[CPU design with wide L3 sharing]{CPU design with wide L3 sharing \smallskip\newline A very simple CPU with 4 \glspl{hthrd}. L1 and L2 are private to each \gls{hthrd} but the L3 is shared across to entire core.}
+        \label{fig:cache-share}
+\end{figure}
+\begin{figure}
+        \centering
+        \input{cache-noshare.pstex_t}
+        \caption[CPU design with a narrower L3 sharing]{CPU design with a narrower L3 sharing \smallskip\newline A different CPU design, still with 4 \glspl{hthrd}. L1 and L2 are still private to each \gls{hthrd} but the L3 is shared some of the CPU but there is still two distinct L3 instances.}
+        \label{fig:cache-noshare}
+\end{figure}
+With redundant tiemstamps this scheduling algorithm achieves both the fairness and performance requirements, on some machines.
+The problem is that the cost of polling and helping is not necessarily consistent across each \gls{hthrd}.
+For example, on machines where the motherboard holds multiple CPU, cache misses can be satisfied from a cache that belongs to the CPU that missed, the \emph{local} CPU, or by a different CPU, a \emph{remote} one.
+Cache misses that are satisfied by a remote CPU will have higher latency than if it is satisfied by the local CPU.
+However, this is not specific to systems with multiple CPUs.
+Depending on the cache structure, cache-misses can have different latency for the same CPU.
+The AMD EPYC 7662 CPUs that is described in Chapter~\ref{microbench} is an example of that.
+Figure~\ref{fig:cache-share} and Figure~\ref{fig:cache-noshare} show two different cache topologies with highlight this difference.
+In Figure~\ref{fig:cache-share}, all cache instances are either private to a \gls{hthrd} or shared to the entire system, this means latency due to cache-misses are likely fairly consistent.
+By comparison, in Figure~\ref{fig:cache-noshare} misses in the L2 cache can be satisfied by a hit in either instance of the L3.
+However, the memory access latency to the remote L3 instance will be notably higher than the memory access latency to the local L3.
+The impact of these different design on this algorithm is that scheduling will scale very well on architectures similar to Figure~\ref{fig:cache-share}, both will have notably worst scalling with many narrower L3 instances.
+This is simply because as the number of L3 instances grow, so two does the chances that the random helping will cause significant latency.
+The solution is to have the scheduler be aware of the cache topology.
+\subsection{Per CPU Sharding}
+Building a scheduler that is aware of cache topology poses two main challenges: discovering cache topology and matching \procs to cache instance.
+Sadly, there is no standard portable way to discover cache topology in C.
+Therefore, while this is a significant portability challenge, it is outside the scope of this thesis to design a cross-platform cache discovery mechanisms.
+The rest of this work assumes discovering the cache topology based on Linux's \texttt{/sys/devices/system/cpu} directory.
+This leaves the challenge of matching \procs to cache instance, or more precisely identifying which subqueues of the ready queue are local to which cache instance.
+Once this matching is available, the helping algorithm can be changed to add bias so that \procs more often help subqueues local to the same cache instance
+\footnote{Note that like other biases mentioned in this section, the actual bias value does not appear to need precise tuinng.}.
+The obvious approach to mapping cache instances to subqueues is to statically tie subqueues to CPUs.
+Instead of having each subqueue local to a specific \proc, the system is initialized with subqueues for each \glspl{hthrd} up front.
+Then \procs dequeue and enqueue by first asking which CPU id they are local to, in order to identify which subqueues are the local ones.
+\Glspl{proc} can get the CPU id from \texttt{sched\_getcpu} or \texttt{librseq}.
+This approach solves the performance problems on systems with topologies similar to Figure~\ref{fig:cache-noshare}.
+However, it actually causes some subtle fairness problems in some systems, specifically systems with few \procs and many \glspl{hthrd}.
+In these cases, the large number of subqueues and the bias agains subqueues tied to different cache instances make it so it is very unlikely any single subqueue is picked.
+To make things worst, the small number of \procs mean that few helping attempts will be made.
+This combination of few attempts and low chances make it so a \at stranded on a subqueue that is not actively dequeued from may wait very long before it gets randomly helped.
+On a system with 2 \procs, 256 \glspl{hthrd} with narrow cache sharing, and a 100:1 bias, it can actually take multiple seconds for a \at to get dequeued from a remote queue.
+Therefore, a more dynamic matching of subqueues to cache instance is needed.
+\subsection{Topological Work Stealing}
+The approach that is used in the \CFA scheduler is to have per-\proc subqueue, but have an excplicit data-structure track which cache instance each subqueue is tied to.
+This is requires some finess because reading this data structure must lead to fewer cache misses than not having the data structure in the first place.
+A key element however is that, like the timestamps for helping, reading the cache instance mapping only needs to give the correct result \emph{often enough}.
+Therefore the algorithm can be built as follows: Before enqueuing or dequeing a \at, each \proc queries the CPU id and the corresponding cache instance.
+Since subqueues are tied to \procs, each \proc can then update the cache instance mapped to the local subqueue(s).
+To avoid unnecessary cache line invalidation, the map is only written to if the mapping changes.

doc/theses/thierry_delisle_PhD/thesis/text/eval_micro.tex

-              rba897d21
+              r2e9b59b
 The first step of evaluation is always to test-out small controlled cases, to ensure that the basics are working properly.
 This sections presents five different experimental setup, evaluating some of the basic features of \CFA's scheduler.
+\section{Benchmark Environment}
+All of these benchmarks are run on two distinct hardware environment, an AMD and an INTEL machine.
+\paragraph{AMD} The AMD machine is a server with two AMD EPYC 7662 CPUs and 256GB of DDR4 RAM.
+The server runs Ubuntu 20.04.2 LTS on top of Linux Kernel 5.8.0-55.
+These EPYCs have 64 cores per CPUs and 2 \glspl{hthrd} per core, for a total of 256 \glspl{hthrd}.
+The cpus each have 4 MB, 64 MB and 512 MB of L1, L2 and L3 caches respectively.
+Each L1 and L2 instance are only shared by \glspl{hthrd} on a given core, but each L3 instance is shared by 4 cores, therefore 8 \glspl{hthrd}.
+\paragraph{Intel} The Intel machine is a server with four Intel Xeon Platinum 8160 CPUs and 384GB of DDR4 RAM.
+The server runs Ubuntu 20.04.2 LTS on top of Linux Kernel 5.8.0-55.
+These Xeon Platinums have 24 cores per CPUs and 2 \glspl{hthrd} per core, for a total of 192 \glspl{hthrd}.
+The cpus each have 3 MB, 96 MB and 132 MB of L1, L2 and L3 caches respectively.
+Each L1 and L2 instance are only shared by \glspl{hthrd} on a given core, but each L3 instance is shared across the entire CPU, therefore 48 \glspl{hthrd}.
+This limited sharing of the last level cache on the AMD machine is markedly different than the Intel machine. Indeed, while on both architectures L2 cache misses that are served by L3 caches on a different cpu incurr a significant latency, on AMD it is also the case that cache misses served by a different L3 instance on the same cpu still incur high latency.
 \section{Cycling latency}
 …
 \end{figure}
-\todo{check term ``idle sleep handling''}
 To avoid this benchmark from being dominated by the idle sleep handling, the number of rings is kept at least as high as the number of \glspl{proc} available.
 Beyond this point, adding more rings serves to mitigate even more the idle sleep handling.
 This is to avoid the case where one of the worker \glspl{at} runs out of work because of the variation on the number of ready \glspl{at} mentionned above.
+This is to avoid the case where one of the \glspl{proc} runs out of work because of the variation on the number of ready \glspl{at} mentionned above.
 The actual benchmark is more complicated to handle termination, but that simply requires using a binary semphore or a channel instead of raw \texttt{park}/\texttt{unpark} and carefully picking the order of the \texttt{P} and \texttt{V} with respect to the loop condition.
-\todo{code, setup, results}
 \begin{lstlisting}
         Thread.main() {
 …
 \end{lstlisting}
+\begin{figure}
+        \centering
+        \input{result.cycle.jax.ops.pstex_t}
+        \vspace*{-10pt}
+        \label{fig:cycle:ns:jax}
+\end{figure}
 \section{Yield}

doc/theses/thierry_delisle_PhD/thesis/text/existing.tex

-              rba897d21
+              r2e9b59b
 Scheduling is the process of assigning resources to incomming requests.
 A very common form of this is assigning available workers to work-requests.
 The need for scheduling is very common in Computer Science, \eg Operating Systems and Hypervisors schedule available CPUs, NICs schedule available bamdwith, but it is also common in other fields.
 For example, assmebly lines are an example of scheduling where parts needed assembly are assigned to line workers.
+The need for scheduling is very common in Computer Science, \eg Operating Systems and Hypervisors schedule available CPUs, NICs schedule available bamdwith, but scheduling is also common in other fields.
+For example, in assmebly lines assigning parts in need of assembly to line workers is a form of scheduling.
 In all these cases, the choice of a scheduling algorithm generally depends first and formost on how much information is available to the scheduler.
 …
 \section{Naming Convention}
 Scheduling has been studied by various different communities concentrating on different incarnation of the same problems. As a result, their is no real naming convention for scheduling that is respected across these communities. For this document, I will use the term \newterm{task} to refer to the abstract objects being scheduled and the term \newterm{worker} to refer to the objects which will execute these tasks.
+Scheduling has been studied by various different communities concentrating on different incarnation of the same problems. As a result, their is no real naming convention for scheduling that is respected across these communities. For this document, I will use the term \newterm{\Gls{at}} to refer to the abstract objects being scheduled and the term \newterm{\Gls{proc}} to refer to the objects which will execute these \glspl{at}.
 \section{Static Scheduling}
 Static schedulers require that tasks have their dependencies and costs explicitly and exhaustively specified prior schedule.
+Static schedulers require that \glspl{at} have their dependencies and costs explicitly and exhaustively specified prior schedule.
 The scheduler then processes this input ahead of time and producess a \newterm{schedule} to which the system can later adhere.
 This approach is generally popular in real-time systems since the need for strong guarantees justifies the cost of supplying this information.
 …
 \section{Dynamic Scheduling}
 It may be difficult to fulfill the requirements of static scheduler if dependencies are conditionnal. In this case, it may be preferable to detect dependencies at runtime. This detection effectively takes the form of halting or suspending a task with unfulfilled dependencies and adding one or more new task(s) to the system. The new task(s) have the responsability of adding the dependent task back in the system once completed. As a consequence, the scheduler may have an incomplete view of the system, seeing only tasks we no pending dependencies. Schedulers that support this detection at runtime are referred to as \newterm{Dynamic Schedulers}.
+It may be difficult to fulfill the requirements of static scheduler if dependencies are conditionnal. In this case, it may be preferable to detect dependencies at runtime. This detection effectively takes the form of adding one or more new \gls{at}(s) to the system as their dependencies are resolved. As well as potentially halting or suspending a \gls{at} that dynamically detect unfulfilled dependencies. Each \gls{at} has the responsability of adding the dependent \glspl{at} back in the system once completed. As a consequence, the scheduler may have an incomplete view of the system, seeing only \glspl{at} we no pending dependencies. Schedulers that support this detection at runtime are referred to as \newterm{Dynamic Schedulers}.
 \subsection{Explicitly Informed Dynamic Schedulers}
 While dynamic schedulers do not have access to an exhaustive list of dependencies for a task, they may require to provide more or less information about each task, including for example: expected duration, required ressources, relative importance, etc. The scheduler can then use this information to direct the scheduling decisions. \cit{Examples of schedulers with more information} Precisely providing this information can be difficult for programmers, especially \emph{predicted} behaviour, and the scheduler may need to support some amount of imprecision in the provided information. For example, specifying that a tasks takes approximately 5 seconds to complete, rather than exactly 5 seconds. User provided information can also become a significant burden depending how the effort to provide the information scales with the number of tasks and there complexity. For example, providing an exhaustive list of files read by 5 tasks is an easier requirement the providing an exhaustive list of memory addresses accessed by 10'000 distinct tasks.
+While dynamic schedulers do not have access to an exhaustive list of dependencies for a \gls{at}, they may require to provide more or less information about each \gls{at}, including for example: expected duration, required ressources, relative importance, etc. The scheduler can then use this information to direct the scheduling decisions. \cit{Examples of schedulers with more information} Precisely providing this information can be difficult for programmers, especially \emph{predicted} behaviour, and the scheduler may need to support some amount of imprecision in the provided information. For example, specifying that a \glspl{at} takes approximately 5 seconds to complete, rather than exactly 5 seconds. User provided information can also become a significant burden depending how the effort to provide the information scales with the number of \glspl{at} and there complexity. For example, providing an exhaustive list of files read by 5 \glspl{at} is an easier requirement the providing an exhaustive list of memory addresses accessed by 10'000 distinct \glspl{at}.
 Since the goal of this thesis is to provide a scheduler as a replacement for \CFA's existing \emph{uninformed} scheduler, Explicitly Informed schedulers are less relevant to this project. Nevertheless, some strategies are worth mentionnding.
 \subsubsection{Prority Scheduling}
+A commonly used information that schedulers used to direct the algorithm is priorities. Each Task is given a priority and higher-priority tasks are preferred to lower-priority ones. The simplest priority scheduling algorithm is to simply require that every task have a distinct pre-established priority and always run the available task with the highest priority. Asking programmers to provide an exhaustive set of unique priorities can be prohibitive when the system has a large number of tasks. It can therefore be diserable for schedulers to support tasks with identical priorities and/or automatically setting and adjusting priorites for tasks.
+A commonly used information that schedulers used to direct the algorithm is priorities. Each Task is given a priority and higher-priority \glspl{at} are preferred to lower-priority ones. The simplest priority scheduling algorithm is to simply require that every \gls{at} have a distinct pre-established priority and always run the available \gls{at} with the highest priority. Asking programmers to provide an exhaustive set of unique priorities can be prohibitive when the system has a large number of \glspl{at}. It can therefore be diserable for schedulers to support \glspl{at} with identical priorities and/or automatically setting and adjusting priorites for \glspl{at}. The most common operating some variation on priorities with overlaps and dynamic priority adjustments. For example, Microsoft Windows uses a pair of priorities
+\cit{https://docs.microsoft.com/en-us/windows/win32/procthread/scheduling-priorities,https://docs.microsoft.com/en-us/windows/win32/taskschd/taskschedulerschema-priority-settingstype-element}, one specified by users out of ten possible options and one adjusted by the system.
 \subsection{Uninformed and Self-Informed Dynamic Schedulers}
 Several scheduling algorithms do not require programmers to provide additionnal information on each task, and instead make scheduling decisions based solely on internal state and/or information implicitly gathered by the scheduler.
+Several scheduling algorithms do not require programmers to provide additionnal information on each \gls{at}, and instead make scheduling decisions based solely on internal state and/or information implicitly gathered by the scheduler.
 \subsubsection{Feedback Scheduling}
 As mentionned, Schedulers may also gather information about each tasks to direct their decisions. This design effectively moves the scheduler to some extent into the realm of \newterm{Control Theory}\cite{wiki:controltheory}. This gathering does not generally involve programmers and as such does not increase programmer burden the same way explicitly provided information may. However, some feedback schedulers do offer the option to programmers to offer additionnal information on certain tasks, in order to direct scheduling decision. The important distinction being whether or not the scheduler can function without this additionnal information.
+As mentionned, Schedulers may also gather information about each \glspl{at} to direct their decisions. This design effectively moves the scheduler to some extent into the realm of \newterm{Control Theory}\cite{wiki:controltheory}. This gathering does not generally involve programmers and as such does not increase programmer burden the same way explicitly provided information may. However, some feedback schedulers do offer the option to programmers to offer additionnal information on certain \glspl{at}, in order to direct scheduling decision. The important distinction being whether or not the scheduler can function without this additionnal information.
 \section{Work Stealing}\label{existing:workstealing}
 One of the most popular scheduling algorithm in practice (see~\ref{existing:prod}) is work-stealing. This idea, introduce by \cite{DBLP:conf/fpca/BurtonS81}, effectively has each worker work on its local tasks first, but allows the possibility for other workers to steal local tasks if they run out of tasks. \cite{DBLP:conf/focs/Blumofe94} introduced the more familiar incarnation of this, where each workers has queue of tasks to accomplish and workers without tasks steal tasks from random workers. (The Burton and Sleep algorithm had trees of tasks and stole only among neighbours). Blumofe and Leiserson also prove worst case space and time requirements for well-structured computations.
+One of the most popular scheduling algorithm in practice (see~\ref{existing:prod}) is work-stealing. This idea, introduce by \cite{DBLP:conf/fpca/BurtonS81}, effectively has each worker work on its local \glspl{at} first, but allows the possibility for other workers to steal local \glspl{at} if they run out of \glspl{at}. \cite{DBLP:conf/focs/Blumofe94} introduced the more familiar incarnation of this, where each workers has queue of \glspl{at} to accomplish and workers without \glspl{at} steal \glspl{at} from random workers. (The Burton and Sleep algorithm had trees of \glspl{at} and stole only among neighbours). Blumofe and Leiserson also prove worst case space and time requirements for well-structured computations.
 Many variations of this algorithm have been proposed over the years\cite{DBLP:journals/ijpp/YangH18}, both optmizations of existing implementations and approaches that account for new metrics.
 …
 \paragraph{Granularity} A significant portion of early Work Stealing research was concentrating on \newterm{Implicit Parellelism}\cite{wiki:implicitpar}. Since the system was responsible to split the work, granularity is a challenge that cannot be left to the programmers (as opposed to \newterm{Explicit Parellelism}\cite{wiki:explicitpar} where the burden can be left to programmers). In general, fine granularity is better for load balancing and coarse granularity reduces communication overhead. The best performance generally means finding a middle ground between the two. Several methods can be employed, but I believe these are less relevant for threads, which are generally explicit and more coarse grained.
 \paragraph{Task Placement} Since modern computers rely heavily on cache hierarchies\cit{Do I need a citation for this}, migrating tasks from one core to another can be .  \cite{DBLP:journals/tpds/SquillanteL93}
+\paragraph{Task Placement} Since modern computers rely heavily on cache hierarchies\cit{Do I need a citation for this}, migrating \glspl{at} from one core to another can be .  \cite{DBLP:journals/tpds/SquillanteL93}
 \todo{The survey is not great on this subject}
 …
 \subsection{Theoretical Results}
 There is also a large body of research on the theoretical aspects of work stealing. These evaluate, for example, the cost of migration\cite{DBLP:conf/sigmetrics/SquillanteN91,DBLP:journals/pe/EagerLZ86}, how affinity affects performance\cite{DBLP:journals/tpds/SquillanteL93,DBLP:journals/mst/AcarBB02,DBLP:journals/ipl/SuksompongLS16} and theoretical models for heterogenous systems\cite{DBLP:journals/jpdc/MirchandaneyTS90,DBLP:journals/mst/BenderR02,DBLP:conf/sigmetrics/GastG10}. \cite{DBLP:journals/jacm/BlellochGM99} examine the space bounds of Work Stealing and \cite{DBLP:journals/siamcomp/BerenbrinkFG03} show that for underloaded systems, the scheduler will complete computations in finite time, \ie is \newterm{stable}. Others show that Work-Stealing is applicable to various scheduling contexts\cite{DBLP:journals/mst/AroraBP01,DBLP:journals/anor/TchiboukdjianGT13,DBLP:conf/isaac/TchiboukdjianGTRB10,DBLP:conf/ppopp/AgrawalLS10,DBLP:conf/spaa/AgrawalFLSSU14}. \cite{DBLP:conf/ipps/ColeR13} also studied how Randomized Work Stealing affects false sharing among tasks.
+There is also a large body of research on the theoretical aspects of work stealing. These evaluate, for example, the cost of migration\cite{DBLP:conf/sigmetrics/SquillanteN91,DBLP:journals/pe/EagerLZ86}, how affinity affects performance\cite{DBLP:journals/tpds/SquillanteL93,DBLP:journals/mst/AcarBB02,DBLP:journals/ipl/SuksompongLS16} and theoretical models for heterogenous systems\cite{DBLP:journals/jpdc/MirchandaneyTS90,DBLP:journals/mst/BenderR02,DBLP:conf/sigmetrics/GastG10}. \cite{DBLP:journals/jacm/BlellochGM99} examine the space bounds of Work Stealing and \cite{DBLP:journals/siamcomp/BerenbrinkFG03} show that for underloaded systems, the scheduler will complete computations in finite time, \ie is \newterm{stable}. Others show that Work-Stealing is applicable to various scheduling contexts\cite{DBLP:journals/mst/AroraBP01,DBLP:journals/anor/TchiboukdjianGT13,DBLP:conf/isaac/TchiboukdjianGTRB10,DBLP:conf/ppopp/AgrawalLS10,DBLP:conf/spaa/AgrawalFLSSU14}. \cite{DBLP:conf/ipps/ColeR13} also studied how Randomized Work Stealing affects false sharing among \glspl{at}.
 However, as \cite{DBLP:journals/ijpp/YangH18} highlights, it is worth mentionning that this theoretical research has mainly focused on ``fully-strict'' computations, \ie workloads that can be fully represented with a Direct Acyclic Graph. It is unclear how well these distributions represent workloads in real world scenarios.
 \section{Preemption}
 One last aspect of scheduling worth mentionning is preemption since many schedulers rely on it for some of their guarantees. Preemption is the idea of interrupting tasks that have been running for too long, effectively injecting suspend points in the applications. There are multiple techniques to achieve this but they all aim to have the effect of guaranteeing that suspend points in a task are never further apart than some fixed duration. While this helps schedulers guarantee that no tasks will unfairly monopolize a worker, preemption can effectively added to any scheduler. Therefore, the only interesting aspect of preemption for the design of scheduling is whether or not to require it.
+One last aspect of scheduling worth mentionning is preemption since many schedulers rely on it for some of their guarantees. Preemption is the idea of interrupting \glspl{at} that have been running for too long, effectively injecting suspend points in the applications. There are multiple techniques to achieve this but they all aim to have the effect of guaranteeing that suspend points in a \gls{at} are never further apart than some fixed duration. While this helps schedulers guarantee that no \glspl{at} will unfairly monopolize a worker, preemption can effectively added to any scheduler. Therefore, the only interesting aspect of preemption for the design of scheduling is whether or not to require it.
 \section{Schedulers in Production}\label{existing:prod}
 …
 \subsection{Operating System Schedulers}
 Operating System Schedulers tend to be fairly complex schedulers, they generally support some amount of real-time, aim to balance interactive and non-interactive tasks and support for multiple users sharing hardware without requiring these users to cooperate. Here are more details on a few schedulers used in the common operating systems: Linux, FreeBsd, Microsoft Windows and Apple's OS X. The information is less complete for operating systems behind closed source.
+Operating System Schedulers tend to be fairly complex schedulers, they generally support some amount of real-time, aim to balance interactive and non-interactive \glspl{at} and support for multiple users sharing hardware without requiring these users to cooperate. Here are more details on a few schedulers used in the common operating systems: Linux, FreeBsd, Microsoft Windows and Apple's OS X. The information is less complete for operating systems behind closed source.
 \paragraph{Linux's CFS}
 The default scheduler used by Linux (the Completely Fair Scheduler)\cite{MAN:linux/cfs,MAN:linux/cfs2} is a feedback scheduler based on CPU time. For each processor, it constructs a Red-Black tree of tasks waiting to run, ordering them by amount of CPU time spent. The scheduler schedules the task that has spent the least CPU time. It also supports the concept of \newterm{Nice values}, which are effectively multiplicative factors on the CPU time spent. The ordering of tasks is also impacted by a group based notion of fairness, where tasks belonging to groups having spent less CPU time are preferred to tasks beloning to groups having spent more CPU time. Linux achieves load-balancing by regularly monitoring the system state\cite{MAN:linux/cfs/balancing} and using some heuristic on the load (currently CPU time spent in the last millisecond plus decayed version of the previous time slots\cite{MAN:linux/cfs/pelt}.).
+The default scheduler used by Linux (the Completely Fair Scheduler)\cite{MAN:linux/cfs,MAN:linux/cfs2} is a feedback scheduler based on CPU time. For each processor, it constructs a Red-Black tree of \glspl{at} waiting to run, ordering them by amount of CPU time spent. The scheduler schedules the \gls{at} that has spent the least CPU time. It also supports the concept of \newterm{Nice values}, which are effectively multiplicative factors on the CPU time spent. The ordering of \glspl{at} is also impacted by a group based notion of fairness, where \glspl{at} belonging to groups having spent less CPU time are preferred to \glspl{at} beloning to groups having spent more CPU time. Linux achieves load-balancing by regularly monitoring the system state\cite{MAN:linux/cfs/balancing} and using some heuristic on the load (currently CPU time spent in the last millisecond plus decayed version of the previous time slots\cite{MAN:linux/cfs/pelt}.).
 \cite{DBLP:conf/eurosys/LoziLFGQF16} shows that Linux's CFS also does work-stealing to balance the workload of each processors, but the paper argues this aspect can be improved significantly. The issues highlighted sem to stem from Linux's need to support fairness across tasks \emph{and} across users\footnote{Enforcing fairness across users means, for example, that given two users: one with a single task and the other with one thousand tasks, the user with a single task does not receive one one thousandth of the CPU time.}, increasing the complexity.
+\cite{DBLP:conf/eurosys/LoziLFGQF16} shows that Linux's CFS also does work-stealing to balance the workload of each processors, but the paper argues this aspect can be improved significantly. The issues highlighted sem to stem from Linux's need to support fairness across \glspl{at} \emph{and} across users\footnote{Enforcing fairness across users means, for example, that given two users: one with a single \gls{at} and the other with one thousand \glspl{at}, the user with a single \gls{at} does not receive one one thousandth of the CPU time.}, increasing the complexity.
 Linux also offers a FIFO scheduler, a real-time schedulerwhich runs the highest-priority task, and a round-robin scheduler, which is an extension of the fifo-scheduler that adds fixed time slices. \cite{MAN:linux/sched}
+Linux also offers a FIFO scheduler, a real-time schedulerwhich runs the highest-priority \gls{at}, and a round-robin scheduler, which is an extension of the fifo-scheduler that adds fixed time slices. \cite{MAN:linux/sched}
 \paragraph{FreeBSD}
 …
 \paragraph{Windows(OS)}
 Microsoft's Operating System's Scheduler\cite{MAN:windows/scheduler} is a feedback scheduler with priorities. It supports 32 levels of priorities, some of which are reserved for real-time and prviliged applications. It schedules tasks based on the highest priorities (lowest number) and how much cpu time each tasks have used. The scheduler may also temporarily adjust priorities after certain effects like the completion of I/O requests.
+Microsoft's Operating System's Scheduler\cite{MAN:windows/scheduler} is a feedback scheduler with priorities. It supports 32 levels of priorities, some of which are reserved for real-time and prviliged applications. It schedules \glspl{at} based on the highest priorities (lowest number) and how much cpu time each \glspl{at} have used. The scheduler may also temporarily adjust priorities after certain effects like the completion of I/O requests.
 \todo{load balancing}
 …
 \subsection{User-Level Schedulers}
 By comparison, user level schedulers tend to be simpler, gathering fewer metrics and avoid complex notions of fairness. Part of the simplicity is due to the fact that all tasks have the same user, and therefore cooperation is both feasible and probable.
+By comparison, user level schedulers tend to be simpler, gathering fewer metrics and avoid complex notions of fairness. Part of the simplicity is due to the fact that all \glspl{at} have the same user, and therefore cooperation is both feasible and probable.
 \paragraph{Go}
 Go's scheduler uses a Randomized Work Stealing algorithm that has a global runqueue(\emph{GRQ}) and each processor(\emph{P}) has both a fixed-size runqueue(\emph{LRQ}) and a high-priority next ``chair'' holding a single element.\cite{GITHUB:go,YTUBE:go} Preemption is present, but only at function call boundaries.
 …
 \paragraph{Intel\textregistered ~Threading Building Blocks}
 \newterm{Thread Building Blocks}(TBB) is Intel's task parellelism\cite{wiki:taskparallel} framework. It runs tasks or \newterm{jobs}, schedulable objects that must always run to completion, on a pool of worker threads. TBB's scheduler is a variation of Randomized Work Stealing that also supports higher-priority graph-like dependencies\cite{MAN:tbb/scheduler}. It schedules tasks as follows (where \textit{t} is the last task completed):
+\newterm{Thread Building Blocks}(TBB) is Intel's task parellelism\cite{wiki:taskparallel} framework. It runs \newterm{jobs}, uninterruptable \glspl{at}, schedulable objects that must always run to completion, on a pool of worker threads. TBB's scheduler is a variation of Randomized Work Stealing that also supports higher-priority graph-like dependencies\cite{MAN:tbb/scheduler}. It schedules \glspl{at} as follows (where \textit{t} is the last \gls{at} completed):
 \begin{displayquote}
         \begin{enumerate}
 …
 \paragraph{Grand Central Dispatch}
 This is an API produce by Apple\cit{Official GCD source} that offers task parellelism\cite{wiki:taskparallel}. Its distinctive aspect is that it uses multiple ``Dispatch Queues'', some of which are created by programmers. These queues each have their own local ordering guarantees, \eg tasks on queue $A$ are executed in \emph{FIFO} order.
+This is an API produce by Apple\cit{Official GCD source} that offers task parellelism\cite{wiki:taskparallel}. Its distinctive aspect is that it uses multiple ``Dispatch Queues'', some of which are created by programmers. These queues each have their own local ordering guarantees, \eg \glspl{at} on queue $A$ are executed in \emph{FIFO} order.
 \todo{load balancing and scheduling}

doc/theses/thierry_delisle_PhD/thesis/text/io.tex

-              rba897d21
+              r2e9b59b
 The consequence is that the amount of parallelism used to prepare submissions for the next system call is limited.
 Beyond this limit, the length of the system call is the throughput limiting factor.
 I concluded from early experiments that preparing submissions seems to take about as long as the system call itself, which means that with a single @io_uring@ instance, there is no benefit in terms of \io throughput to having more than two \glspl{hthrd}.
+I concluded from early experiments that preparing submissions seems to take at most as long as the system call itself, which means that with a single @io_uring@ instance, there is no benefit in terms of \io throughput to having more than two \glspl{hthrd}.
 Therefore the design of the submission engine must manage multiple instances of @io_uring@ running in parallel, effectively sharding @io_uring@ instances.
 Similarly to scheduling, this sharding can be done privately, \ie, one instance per \glspl{proc}, in decoupled pools, \ie, a pool of \glspl{proc} use a pool of @io_uring@ instances without one-to-one coupling between any given instance and any given \gls{proc}, or some mix of the two.
 …
 The only added complexity is that the number of SQEs is fixed, which means allocation can fail.
 Allocation failures need to be pushed up to the routing algorithm: \glspl{thrd} attempting \io operations must not be directed to @io_uring@ instances without sufficient SQEs available.
+Allocation failures need to be pushed up to a routing algorithm: \glspl{thrd} attempting \io operations must not be directed to @io_uring@ instances without sufficient SQEs available.
 Furthermore, the routing algorithm should block operations up-front if none of the instances have available SQEs.
 …
 In the case of designating a \gls{thrd}, ideally, when multiple \glspl{thrd} attempt to submit operations to the same @io_uring@ instance, all requests would be batched together and one of the \glspl{thrd} would do the system call on behalf of the others, referred to as the \newterm{submitter}.
 In practice however, it is important that the \io requests are not left pending indefinitely and as such, it may be required to have a current submitter and a next submitter.
+In practice however, it is important that the \io requests are not left pending indefinitely and as such, it may be required to have a ``next submitter'' that guarentees everything that is missed by the current submitter is seen by the next one.
 Indeed, as long as there is a ``next'' submitter, \glspl{thrd} submitting new \io requests can move on, knowing that some future system call will include their request.
 Once the system call is done, the submitter must also free SQEs so that the allocator can reused them.
 …
 If the submission side does not designate submitters, polling can also submit all SQEs as it is polling events.
 A simple approach to polling is to allocate a \gls{thrd} per @io_uring@ instance and simply let the poller \glspl{thrd} poll their respective instances when scheduled.
-This design is especially convenient for reasons explained in Chapter~\ref{practice}.
 With this pool of instances approach, the big advantage is that it is fairly flexible.
 It does not impose restrictions on what \glspl{thrd} submitting \io operations can and cannot do between allocations and submissions.
 It also can gracefully handles running out of ressources, SQEs or the kernel returning @EBUSY@.
+It also can gracefully handle running out of ressources, SQEs or the kernel returning @EBUSY@.
 The down side to this is that many of the steps used for submitting need complex synchronization to work properly.
 The routing and allocation algorithm needs to keep track of which ring instances have available SQEs, block incoming requests if no instance is available, prevent barging if \glspl{thrd} are already queued up waiting for SQEs and handle SQEs being freed.
 The submission side needs to safely append SQEs to the ring buffer, correctly handle chains, make sure no SQE is dropped or left pending forever, notify the allocation side when SQEs can be reused and handle the kernel returning @EBUSY@.
 All this synchronization may have a significant cost and, compare to the next approach presented, this synchronization is entirely overhead.
+All this synchronization may have a significant cost and, compared to the next approach presented, this synchronization is entirely overhead.
 \subsubsection{Private Instances}
 Another approach is to simply create one ring instance per \gls{proc}.
 This alleviate the need for synchronization on the submissions, requiring only that \glspl{thrd} are not interrupted in between two submission steps.
+This alleviates the need for synchronization on the submissions, requiring only that \glspl{thrd} are not interrupted in between two submission steps.
 This is effectively the same requirement as using @thread_local@ variables.
 Since SQEs that are allocated must be submitted to the same ring, on the same \gls{proc}, this effectively forces the application to submit SQEs in allocation order
 …
 \paragraph{Pending Allocations} can be more complicated to handle.
 If the arbiter has available instances, the arbiter can attempt to directly hand over the instance and satisfy the request.
+Otherwise
+Otherwise it must hold onto the list of threads until SQEs are made available again.
+This handling becomes that much more complex if pending allocation require more than one SQE, since the arbiter must make a decision between statisfying requests in FIFO ordering or satisfy requests for fewer SQEs first.
+While this arbiter has the potential to solve many of the problems mentionned in above, it also introduces a significant amount of complexity.
+Tracking which processors are borrowing which instances and which instances have SQEs available ends-up adding a significant synchronization prelude to any I/O operation.
+Any submission must start with a handshake that pins the currently borrowed instance, if available.
+An attempt to allocate is then made, but the arbiter can concurrently be attempting to allocate from the same instance from a different \gls{hthrd}.
+Once the allocation is completed, the submission must still check that the instance is still burrowed before attempt to flush.
+These extra synchronization steps end-up having a similar cost to the multiple shared instances approach.
+Furthermore, if the number of instances does not match the number of processors actively submitting I/O, the system can fall into a state where instances are constantly being revoked and end-up cycling the processors, which leads to significant cache deterioration.
+Because of these reasons, this approach, which sounds promising on paper, does not improve on the private instance approach in practice.
+\subsubsection{Private Instances V2}
 …
 Finally, the last important part of the \io subsystem is it's interface. There are multiple approaches that can be offered to programmers, each with advantages and disadvantages. The new \io subsystem can replace the C runtime's API or extend it. And in the later case the interface can go from very similar to vastly different. The following sections discuss some useful options using @read@ as an example. The standard Linux interface for C is :
 @ssize_t read(int fd, void *buf, size_t count);@.
+@ssize_t read(int fd, void *buf, size_t count);@
 \subsection{Replacement}
+Replacing the C \glsxtrshort{api}
+Replacing the C \glsxtrshort{api} is the more intrusive and draconian approach.
+The goal is to convince the compiler and linker to replace any calls to @read@ to direct them to the \CFA implementation instead of glibc's.
+This has the advantage of potentially working transparently and supporting existing binaries without needing recompilation.
+It also offers a, presumably, well known and familiar API that C programmers can simply continue to work with.
+However, this approach also entails a plethora of subtle technical challenges which generally boils down to making a perfect replacement.
+If the \CFA interface replaces only \emph{some} of the calls to glibc, then this can easily lead to esoteric concurrency bugs.
+Since the gcc ecosystems does not offer a scheme for such perfect replacement, this approach was rejected as being laudable but infeasible.
 \subsection{Synchronous Extension}
+An other interface option is to simply offer an interface that is different in name only. For example:
+@ssize_t cfa_read(int fd, void *buf, size_t count);@
+\noindent This is much more feasible but still familiar to C programmers.
+It comes with the caveat that any code attempting to use it must be recompiled, which can be a big problem considering the amount of existing legacy C binaries.
+However, it has the advantage of implementation simplicity.
 \subsection{Asynchronous Extension}
+It is important to mention that there is a certain irony to using only synchronous, therefore blocking, interfaces for a feature often referred to as ``non-blocking'' \io.
+A fairly traditional way of doing this is using futures\cit{wikipedia futures}.
+As simple way of doing so is as follows:
+@future(ssize_t) read(int fd, void *buf, size_t count);@
+\noindent Note that this approach is not necessarily the most idiomatic usage of futures.
+The definition of read above ``returns'' the read content through an output parameter which cannot be synchronized on.
+A more classical asynchronous API could look more like:
+@future([ssize_t, void *]) read(int fd, size_t count);@
+\noindent However, this interface immediately introduces memory lifetime challenges since the call must effectively allocate a buffer to be returned.
+Because of the performance implications of this, the first approach is considered preferable as it is more familiar to C programmers.
 \subsection{Interface directly to \lstinline{io_uring}}
+Finally, an other interface that can be relevant is to simply expose directly the underlying \texttt{io\_uring} interface. For example:
+@array(SQE, want) cfa_io_allocate(int want);@
+@void cfa_io_submit( const array(SQE, have) & );@
+\noindent This offers more flexibility to users wanting to fully use all of the \texttt{io\_uring} features.
+However, it is not the most user-friendly option.
+It obviously imposes a strong dependency between user code and \texttt{io\_uring} but at the same time restricting users to usages that are compatible with how \CFA internally uses \texttt{io\_uring}.

doc/theses/thierry_delisle_PhD/thesis/text/practice.tex

-              rba897d21
+              r2e9b59b
 The scheduling algorithm discribed in Chapter~\ref{core} addresses scheduling in a stable state.
 However, it does not address problems that occur when the system changes state.
 Indeed the \CFA runtime, supports expanding and shrinking the number of KTHREAD\_place \todo{add kthrd to glossary}, both manually and, to some extent automatically.
+Indeed the \CFA runtime, supports expanding and shrinking the number of \procs, both manually and, to some extent, automatically.
 This entails that the scheduling algorithm must support these transitions.
+\section{Resizing}
+More precise \CFA supports adding \procs using the RAII object @processor@.
+These objects can be created at any time and can be destroyed at any time.
+They are normally create as automatic stack variables, but this is not a requirement.
+The consequence is that the scheduler and \io subsystems must support \procs comming in and out of existence.
+\section{Manual Resizing}
+The consequence of dynamically changing the number of \procs is that all internal arrays that are sized based on the number of \procs neede to be \texttt{realloc}ed.
+This also means that any references into these arrays, pointers or indexes, may need to be fixed when shrinking\footnote{Indexes may still need fixing because there is no guarantee the \proc causing the shrink had the highest index. Therefore indexes need to be reassigned to preserve contiguous indexes.}.
+There are no performance requirements, within reason, for resizing since this is usually considered as part of setup and teardown.
+However, this operation has strict correctness requirements since shrinking and idle sleep can easily lead to deadlocks.
+It should also avoid as much as possible any effect on performance when the number of \procs remain constant.
+This later requirement prehibits simple solutions, like simply adding a global lock to these arrays.
+\subsection{Read-Copy-Update}
+One solution is to use the Read-Copy-Update\cite{wiki:rcu} pattern.
+In this pattern, resizing is done by creating a copy of the internal data strucures, updating the copy with the desired changes, and then attempt an Idiana Jones Switch to replace the original witht the copy.
+This approach potentially has the advantage that it may not need any synchronization to do the switch.
+The switch definitely implies a race where \procs could still use the previous, original, data structure after the copy was switched in.
+The important question then becomes whether or not this race can be recovered from.
+If the changes that arrived late can be transferred from the original to the copy then this solution works.
+For linked-lists, dequeing is somewhat of a problem.
+Dequeing from the original will not necessarily update the copy which could lead to multiple \procs dequeing the same \at.
+Fixing this requires making the array contain pointers to subqueues rather than the subqueues themselves.
+Another challenge is that the original must be kept until all \procs have witnessed the change.
+This is a straight forward memory reclamation challenge but it does mean that every operation will need \emph{some} form of synchronization.
+If each of these operation does need synchronization then it is possible a simpler solution achieves the same performance.
+Because in addition to the classic challenge of memory reclamation, transferring the original data to the copy before reclaiming it poses additional challenges.
+Especially merging subqueues while having a minimal impact on fairness and locality.
+\subsection{Read-Writer Lock}
+A simpler approach would be to use a \newterm{Readers-Writer Lock}\cite{wiki:rwlock} where the resizing requires acquiring the lock as a writer while simply enqueing/dequeing \ats requires acquiring the lock as a reader.
+Using a Readers-Writer lock solves the problem of dynamically resizing and leaves the challenge of finding or building a lock with sufficient good read-side performance.
+Since this is not a very complex challenge and an ad-hoc solution is perfectly acceptable, building a Readers-Writer lock was the path taken.
+To maximize reader scalability, the readers should not contend with eachother when attempting to acquire and release the critical sections.
+This effectively requires that each reader have its own piece of memory to mark as locked and unlocked.
+Reades then acquire the lock wait for writers to finish the critical section and then acquire their local spinlocks.
+Writers acquire the global lock, so writers have mutual exclusion among themselves, and then acquires each of the local reader locks.
+Acquiring all the local locks guarantees mutual exclusion between the readers and the writer, while the wait on the read side prevents readers from continously starving the writer.
+\todo{reference listings}
+\begin{lstlisting}
+void read_lock() {
+        // Step 1 : make sure no writers in
+        while write_lock { Pause(); }
+        // May need fence here
+        // Step 2 : acquire our local lock
+        while atomic_xchg( tls.lock ) {
+                Pause();
+        }
+}
+void read_unlock() {
+        tls.lock = false;
+}
+\end{lstlisting}
+\begin{lstlisting}
+void write_lock()  {
+        // Step 1 : lock global lock
+        while atomic_xchg( write_lock ) {
+                Pause();
+        }
+        // Step 2 : lock per-proc locks
+        for t in all_tls {
+                while atomic_xchg( t.lock ) {
+                        Pause();
+                }
+        }
+}
+void write_unlock() {
+        // Step 1 : release local locks
+        for t in all_tls {
+                t.lock = false;
+        }
+        // Step 2 : release global lock
+        write_lock = false;
+}
+\end{lstlisting}
 \section{Idle-Sleep}
+In addition to users manually changing the number of \procs, it is desireable to support ``removing'' \procs when there is not enough \ats for all the \procs to be useful.
+While manual resizing is expected to be rare, the number of \ats is expected to vary much more which means \procs may need to be ``removed'' for only short periods of time.
+Furthermore, race conditions that spuriously lead to the impression no \ats are ready are actually common in practice.
+Therefore \procs should not be actually \emph{removed} but simply put into an idle state where the \gls{kthrd} is blocked until more \ats become ready.
+This state is referred to as \newterm{Idle-Sleep}.
+Idle sleep effectively encompasses several challenges.
+First some data structure needs to keep track of all \procs that are in idle sleep.
+Because of idle sleep can be spurious, this data structure has strict performance requirements in addition to the strict correctness requirements.
+Next, some tool must be used to block kernel threads \glspl{kthrd}, \eg \texttt{pthread\_cond\_wait}, pthread semaphores.
+The complexity here is to support \at parking and unparking, timers, \io operations and all other \CFA features with minimal complexity.
+Finally, idle sleep also includes a heuristic to determine the appropriate number of \procs to be in idle sleep an any given time.
+This third challenge is however outside the scope of this thesis because developping a general heuristic is involved enough to justify its own work.
+The \CFA scheduler simply follows the ``Race-to-Idle'\cit{https://doi.org/10.1137/1.9781611973099.100}' approach where a sleeping \proc is woken any time an \at becomes ready and \procs go to idle sleep anytime they run out of work.
+\section{Tracking Sleepers}
+Tracking which \procs are in idle sleep requires a data structure holding all the sleeping \procs, but more importantly it requires a concurrent \emph{handshake} so that no \at is stranded on a ready-queue with no active \proc.
+The classic challenge is when a \at is made ready while a \proc is going to sleep, there is a race where the new \at may not see the sleeping \proc and the sleeping \proc may not see the ready \at.
+Furthermore, the ``Race-to-Idle'' approach means that there is some
+\section{Sleeping}
+\subsection{Event FDs}
+\subsection{Epoll}
+\subsection{\texttt{io\_uring}}
+\section{Reducing Latency}

doc/theses/thierry_delisle_PhD/thesis/thesis.tex

-              rba897d21
+              r2e9b59b
 \newcommand\io{\glsxtrshort{io}\xspace}%
+\newcommand\at{\gls{at}\xspace}%
+\newcommand\ats{\glspl{at}\xspace}%
+\newcommand\proc{\gls{proc}\xspace}%
+\newcommand\procs{\glspl{proc}\xspace}%
 %======================================================================

libcfa/src/Makefile.am

-              rba897d21
+              r2e9b59b
         bits/queue.hfa \
         bits/sequence.hfa \
+        concurrency/iofwd.hfa \
+        concurrency/barrier.hfa \
         containers/array.hfa \
-        concurrency/iofwd.hfa \
         containers/list.hfa \
         containers/queueLockFree.hfa \
 …
         concurrency/exception.hfa \
         concurrency/kernel.hfa \
+        concurrency/kernel/cluster.hfa \
         concurrency/locks.hfa \
         concurrency/monitor.hfa \
 …
         concurrency/io/call.cfa \
         concurrency/iofwd.hfa \
         concurrency/kernel_private.hfa \
+        concurrency/kernel/private.hfa \
         concurrency/kernel/startup.cfa \
         concurrency/preemption.cfa \

libcfa/src/concurrency/coroutine.cfa

rba897d21	r2e9b59b
27	27	#include <unwind.h>
28	28
29		#include "kernel_private.hfa"
	29	#include "kernel/private.hfa"
30	30	#include "exception.hfa"
31	31	#include "math.hfa"

libcfa/src/concurrency/io.cfa

-              rba897d21
+              r2e9b59b
         #include "kernel.hfa"
         #include "kernel/fwd.hfa"
+        #include "kernel_private.hfa"
+        #include "kernel/private.hfa"
+        #include "kernel/cluster.hfa"
         #include "io/types.hfa"
 …
         extern void __kernel_unpark( thread$ * thrd, unpark_hint );
+        bool __cfa_io_drain( processor * proc ) {
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( ready_schedule_islocked() );
+                /* paranoid */ verify( proc );
+                /* paranoid */ verify( proc->io.ctx );
+                // Drain the queue
+                $io_context * ctx = proc->io.ctx;
+                unsigned head = *ctx->cq.head;
+                unsigned tail = *ctx->cq.tail;
+                const __u32 mask = *ctx->cq.mask;
+                __u32 count = tail - head;
+                __STATS__( false, io.calls.drain++; io.calls.completed += count; )
+                if(count == 0) return false;
+                for(i; count) {
+                        unsigned idx = (head + i) & mask;
+                        volatile struct io_uring_cqe & cqe = ctx->cq.cqes[idx];
+                        /* paranoid */ verify(&cqe);
+                        struct io_future_t * future = (struct io_future_t *)(uintptr_t)cqe.user_data;
+                        __cfadbg_print_safe( io, "Kernel I/O : Syscall completed : cqe %p, result %d for %p\n", &cqe, cqe.res, future );
+                        __kernel_unpark( fulfil( *future, cqe.res, false ), UNPARK_LOCAL );
+                }
+                __cfadbg_print_safe(io, "Kernel I/O : %u completed\n", count);
+                // Mark to the kernel that the cqe has been seen
+                // Ensure that the kernel only sees the new value of the head index after the CQEs have been read.
+                __atomic_store_n( ctx->cq.head, head + count, __ATOMIC_SEQ_CST );
+                /* paranoid */ verify( ready_schedule_islocked() );
+                /* paranoid */ verify( ! __preemption_enabled() );
+                return true;
+        }
+        bool __cfa_io_flush( processor * proc, int min_comp ) {
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( proc );
+                /* paranoid */ verify( proc->io.ctx );
+                __attribute__((unused)) cluster * cltr = proc->cltr;
+                $io_context & ctx = *proc->io.ctx;
+                __ioarbiter_flush( ctx );
+                if(ctx.sq.to_submit != 0 || min_comp > 0) {
+                        __STATS__( true, io.calls.flush++; )
+                        int ret = syscall( __NR_io_uring_enter, ctx.fd, ctx.sq.to_submit, min_comp, min_comp > 0 ? IORING_ENTER_GETEVENTS : 0, (sigset_t *)0p, _NSIG / 8);
+        static void ioring_syscsll( struct $io_context & ctx, unsigned int min_comp, unsigned int flags ) {
+                __STATS__( true, io.calls.flush++; )
+                int ret;
+                for() {
+                        ret = syscall( __NR_io_uring_enter, ctx.fd, ctx.sq.to_submit, min_comp, flags, (sigset_t *)0p, _NSIG / 8);
                         if( ret < 0 ) {
                                 switch((int)errno) {
+                                case EINTR:
+                                        continue;
                                 case EAGAIN:
-                                case EINTR:
                                 case EBUSY:
                                         // Update statistics
 …
+                                }
+                        }
+                        __cfadbg_print_safe(io, "Kernel I/O : %u submitted to io_uring %d\n", ret, ctx.fd);
+                        __STATS__( true, io.calls.submitted += ret; )
+                        /* paranoid */ verify( ctx.sq.to_submit <= *ctx.sq.num );
+                        /* paranoid */ verify( ctx.sq.to_submit >= ret );
+                        ctx.sq.to_submit -= ret;
+                        /* paranoid */ verify( ctx.sq.to_submit <= *ctx.sq.num );
+                        // Release the consumed SQEs
+                        __release_sqes( ctx );
+                        break;
+                }
+                __cfadbg_print_safe(io, "Kernel I/O : %u submitted to io_uring %d\n", ret, ctx.fd);
+                __STATS__( true, io.calls.submitted += ret; )
+                /* paranoid */ verify( ctx.sq.to_submit <= *ctx.sq.num );
+                /* paranoid */ verify( ctx.sq.to_submit >= ret );
+                ctx.sq.to_submit -= ret;
+                /* paranoid */ verify( ctx.sq.to_submit <= *ctx.sq.num );
+                // Release the consumed SQEs
+                __release_sqes( ctx );
+                /* paranoid */ verify( ! __preemption_enabled() );
+                __atomic_store_n(&ctx.proc->io.pending, false, __ATOMIC_RELAXED);
+        }
+        static bool try_acquire( $io_context * ctx ) __attribute__((nonnull(1))) {
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( ready_schedule_islocked() );
+                {
+                        const __u32 head = *ctx->cq.head;
+                        const __u32 tail = *ctx->cq.tail;
+                        if(head == tail) return false;
+                }
+                // Drain the queue
+                if(!__atomic_try_acquire(&ctx->cq.lock)) {
+                        __STATS__( false, io.calls.locked++; )
+                        return false;
+                }
+                return true;
+        }
+        static bool __cfa_do_drain( $io_context * ctx, cluster * cltr ) __attribute__((nonnull(1, 2))) {
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( ready_schedule_islocked() );
+                /* paranoid */ verify( ctx->cq.lock == true );
+                const __u32 mask = *ctx->cq.mask;
+                unsigned long long ts_prev = ctx->cq.ts;
+                // re-read the head and tail in case it already changed.
+                const __u32 head = *ctx->cq.head;
+                const __u32 tail = *ctx->cq.tail;
+                const __u32 count = tail - head;
+                __STATS__( false, io.calls.drain++; io.calls.completed += count; )
+                for(i; count) {
+                        unsigned idx = (head + i) & mask;
+                        volatile struct io_uring_cqe & cqe = ctx->cq.cqes[idx];
+                        /* paranoid */ verify(&cqe);
+                        struct io_future_t * future = (struct io_future_t *)(uintptr_t)cqe.user_data;
+                        // __cfadbg_print_safe( io, "Kernel I/O : Syscall completed : cqe %p, result %d for %p\n", &cqe, cqe.res, future );
+                        __kernel_unpark( fulfil( *future, cqe.res, false ), UNPARK_LOCAL );
+                }
+                unsigned long long ts_next = ctx->cq.ts = rdtscl();
+                // Mark to the kernel that the cqe has been seen
+                // Ensure that the kernel only sees the new value of the head index after the CQEs have been read.
+                __atomic_store_n( ctx->cq.head, head + count, __ATOMIC_SEQ_CST );
+                ctx->proc->idle_wctx.drain_time = ts_next;
+                __cfadbg_print_safe(io, "Kernel I/O : %u completed age %llu\n", count, ts_next);
+                /* paranoid */ verify( ready_schedule_islocked() );
+                /* paranoid */ verify( ! __preemption_enabled() );
+                __atomic_unlock(&ctx->cq.lock);
+                touch_tsc( cltr->sched.io.tscs, ctx->cq.id, ts_prev, ts_next );
+                return true;
+        }
+        bool __cfa_io_drain( processor * proc ) {
+                bool local = false;
+                bool remote = false;
+                ready_schedule_lock();
+                cluster * const cltr = proc->cltr;
+                $io_context * const ctx = proc->io.ctx;
+                /* paranoid */ verify( cltr );
+                /* paranoid */ verify( ctx );
+                with(cltr->sched) {
+                        const size_t ctxs_count = io.count;
+                        /* paranoid */ verify( ready_schedule_islocked() );
                         /* paranoid */ verify( ! __preemption_enabled() );
+                        ctx.proc->io.pending = false;
+                }
+                ready_schedule_lock();
+                bool ret = __cfa_io_drain( proc );
+                        /* paranoid */ verify( active_processor() == proc );
+                        /* paranoid */ verify( __shard_factor.io > 0 );
+                        /* paranoid */ verify( ctxs_count > 0 );
+                        /* paranoid */ verify( ctx->cq.id < ctxs_count );
+                        const unsigned this_cache = cache_id(cltr, ctx->cq.id / __shard_factor.io);
+                        const unsigned long long ctsc = rdtscl();
+                        if(proc->io.target == MAX) {
+                                uint64_t chaos = __tls_rand();
+                                unsigned ext = chaos & 0xff;
+                                unsigned other  = (chaos >> 8) % (ctxs_count);
+                                if(ext < 3 || __atomic_load_n(&caches[other / __shard_factor.io].id, __ATOMIC_RELAXED) == this_cache) {
+                                        proc->io.target = other;
+                                }
+                        }
+                        else {
+                                const unsigned target = proc->io.target;
+                                /* paranoid */ verify( io.tscs[target].tv != MAX );
+                                HELP: if(target < ctxs_count) {
+                                        const unsigned long long cutoff = calc_cutoff(ctsc, ctx->cq.id, ctxs_count, io.data, io.tscs, __shard_factor.io);
+                                        const unsigned long long age = moving_average(ctsc, io.tscs[target].tv, io.tscs[target].ma);
+                                        __cfadbg_print_safe(io, "Kernel I/O: Help attempt on %u from %u, age %'llu vs cutoff %'llu, %s\n", target, ctx->cq.id, age, cutoff, age > cutoff ? "yes" : "no");
+                                        if(age <= cutoff) break HELP;
+                                        if(!try_acquire(io.data[target])) break HELP;
+                                        if(!__cfa_do_drain( io.data[target], cltr )) break HELP;
+                                        remote = true;
+                                        __STATS__( false, io.calls.helped++; )
+                                }
+                                proc->io.target = MAX;
+                        }
+                }
+                // Drain the local queue
+                if(try_acquire( proc->io.ctx )) {
+                        local = __cfa_do_drain( proc->io.ctx, cltr );
+                }
+                /* paranoid */ verify( ready_schedule_islocked() );
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( active_processor() == proc );
                 ready_schedule_unlock();
+                return ret;
+                return local || remote;
+        }
+        bool __cfa_io_flush( processor * proc ) {
+                /* paranoid */ verify( ! __preemption_enabled() );
+                /* paranoid */ verify( proc );
+                /* paranoid */ verify( proc->io.ctx );
+                $io_context & ctx = *proc->io.ctx;
+                __ioarbiter_flush( ctx );
+                if(ctx.sq.to_submit != 0) {
+                        ioring_syscsll(ctx, 0, 0);
+                }
+                return __cfa_io_drain( proc );
+        }
 …
                 struct io_uring_sqe * sqes = ctx->sq.sqes;
                 for(i; want) {
                         __cfadbg_print_safe(io, "Kernel I/O : filling loop\n");
+                        // __cfadbg_print_safe(io, "Kernel I/O : filling loop\n");
                         out_sqes[i] = &sqes[idxs[i]];
+                }
 …
                 // copy all the indexes we want from the available list
                 for(i; want) {
                         __cfadbg_print_safe(io, "Kernel I/O : allocating loop\n");
+                        // __cfadbg_print_safe(io, "Kernel I/O : allocating loop\n");
                         idxs[i] = sq.free_ring.array[(fhead + i) & mask];
+                }
 …
         // sqe == &sqes[idx]
         struct $io_context * cfa_io_allocate(struct io_uring_sqe * sqes[], __u32 idxs[], __u32 want) {
                 __cfadbg_print_safe(io, "Kernel I/O : attempting to allocate %u\n", want);
+                // __cfadbg_print_safe(io, "Kernel I/O : attempting to allocate %u\n", want);
                 disable_interrupts();
 …
                 /* paranoid */ verify( ctx );
                 __cfadbg_print_safe(io, "Kernel I/O : attempting to fast allocation\n");
+                // __cfadbg_print_safe(io, "Kernel I/O : attempting to fast allocation\n");
                 // We can proceed to the fast path
 …
                         enable_interrupts();
                         __cfadbg_print_safe(io, "Kernel I/O : fast allocation successful from ring %d\n", ctx->fd);
+                        // __cfadbg_print_safe(io, "Kernel I/O : fast allocation successful from ring %d\n", ctx->fd);
                         __fill( sqes, want, idxs, ctx );
 …
                 /* paranoid */ verify( ioarb );
                 __cfadbg_print_safe(io, "Kernel I/O : falling back on arbiter for allocation\n");
+                // __cfadbg_print_safe(io, "Kernel I/O : falling back on arbiter for allocation\n");
                 struct $io_context * ret = __ioarbiter_allocate(*ioarb, idxs, want);
                 __cfadbg_print_safe(io, "Kernel I/O : slow allocation completed from ring %d\n", ret->fd);
+                // __cfadbg_print_safe(io, "Kernel I/O : slow allocation completed from ring %d\n", ret->fd);
                 __fill( sqes, want, idxs,ret );
 …
                 // Add the sqes to the array
                 for( i; have ) {
                         __cfadbg_print_safe(io, "Kernel I/O : __submit loop\n");
+                        // __cfadbg_print_safe(io, "Kernel I/O : __submit loop\n");
                         sq.kring.array[ (tail + i) & mask ] = idxs[i];
+                }
 …
                 sq.to_submit += have;
                 ctx->proc->io.pending = true;
                 ctx->proc->io.dirty   = true;
+                __atomic_store_n(&ctx->proc->io.pending, true, __ATOMIC_RELAXED);
+                __atomic_store_n(&ctx->proc->io.dirty  , true, __ATOMIC_RELAXED);
+        }
 …
                 if(sq.to_submit > 30) {
                         __tls_stats()->io.flush.full++;
                         __cfa_io_flush( ctx->proc, 0 );
+                        __cfa_io_flush( ctx->proc );
+                }
                 if(!lazy) {
                         __tls_stats()->io.flush.eager++;
                         __cfa_io_flush( ctx->proc, 0 );
+                        __cfa_io_flush( ctx->proc );
+                }
+        }
         void cfa_io_submit( struct $io_context * inctx, __u32 idxs[], __u32 have, bool lazy ) __attribute__((nonnull (1))) {
                 __cfadbg_print_safe(io, "Kernel I/O : attempting to submit %u (%s)\n", have, lazy ? "lazy" : "eager");
+                // __cfadbg_print_safe(io, "Kernel I/O : attempting to submit %u (%s)\n", have, lazy ? "lazy" : "eager");
                 disable_interrupts();
 …
                         enable_interrupts();
                         __cfadbg_print_safe(io, "Kernel I/O : submitted on fast path\n");
+                        // __cfadbg_print_safe(io, "Kernel I/O : submitted on fast path\n");
                         return;
+                }
 …
                 enable_interrupts();
                 __cfadbg_print_safe(io, "Kernel I/O : falling back on arbiter for submission\n");
+                // __cfadbg_print_safe(io, "Kernel I/O : falling back on arbiter for submission\n");
                 __ioarbiter_submit(inctx, idxs, have, lazy);
 …
                 // go through the range and release the sqes
                 for( i; count ) {
                         __cfadbg_print_safe(io, "Kernel I/O : release loop\n");
+                        // __cfadbg_print_safe(io, "Kernel I/O : release loop\n");
                         __u32 idx = ctx.sq.kring.array[ (phead + i) & mask ];
                         ctx.sq.free_ring.array[ (ftail + i) & mask ] = idx;
 …
         static $io_context * __ioarbiter_allocate( $io_arbiter & this, __u32 idxs[], __u32 want ) {
                 __cfadbg_print_safe(io, "Kernel I/O : arbiter allocating\n");
+                // __cfadbg_print_safe(io, "Kernel I/O : arbiter allocating\n");
                 __STATS__( false, io.alloc.block += 1; )
 …
                 bool we = enqueue(ctx->ext_sq, (__outstanding_io&)ei);
                 ctx->proc->io.pending = true;
+                __atomic_store_n(&ctx->proc->io.pending, true, __ATOMIC_SEQ_CST);
                 if( we ) {
 …
                         // We can proceed to the fast path
+                        if( !__alloc(ctx, &idx, 1) ) return false;
+                        if( !__alloc(ctx, &idx, 1) ) {
+                                /* paranoid */ verify( false ); // for now check if this happens, next time just abort the sleep.
+                                return false;
+                        }
                         // Allocation was successful
 …
                         /* paranoid */ verify( sqe->user_data == (uintptr_t)&future );
                         __submit( ctx, &idx, 1, true );
+                        __submit_only( ctx, &idx, 1 );
                         /* paranoid */ verify( proc == __cfaabi_tls.this_processor );
 …
                         return true;
+                }
+                void __cfa_io_idle( processor * proc ) {
+                        iovec iov;
+                        __atomic_acquire( &proc->io.ctx->cq.lock );
+                        __attribute__((used)) volatile bool was_reset = false;
+                        with( proc->idle_wctx) {
+                                // Do we already have a pending read
+                                if(available(*ftr)) {
+                                        // There is no pending read, we need to add one
+                                        reset(*ftr);
+                                        iov.iov_base = rdbuf;
+                                        iov.iov_len  = sizeof(eventfd_t);
+                                        __kernel_read(proc, *ftr, iov, evfd );
+                                        ftr->result = 0xDEADDEAD;
+                                        *((eventfd_t *)rdbuf) = 0xDEADDEADDEADDEAD;
+                                        was_reset = true;
+                                }
+                        }
+                        if( !__atomic_load_n( &proc->do_terminate, __ATOMIC_SEQ_CST ) ) {
+                                __ioarbiter_flush( *proc->io.ctx );
+                                proc->idle_wctx.sleep_time = rdtscl();
+                                ioring_syscsll( *proc->io.ctx, 1, IORING_ENTER_GETEVENTS);
+                        }
+                        ready_schedule_lock();
+                        __cfa_do_drain( proc->io.ctx, proc->cltr );
+                        ready_schedule_unlock();
+                        asm volatile ("" :: "m" (was_reset));
+                }
         #endif
 #endif

libcfa/src/concurrency/io/setup.cfa

-              rba897d21
+              r2e9b59b
         void __cfa_io_start( processor * proc ) {}
+        bool __cfa_io_flush( processor * proc, int ) { return false; }
+        bool __cfa_io_flush( processor * proc ) { return false; }
+        bool __cfa_io_drain( processor * proc ) __attribute__((nonnull (1))) { return false; }
+        void __cfa_io_idle ( processor * ) __attribute__((nonnull (1))) {}
         void __cfa_io_stop ( processor * proc ) {}
 …
 #else
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Waddress-of-packed-member"
         #include <errno.h>
         #include <stdint.h>
 …
         #include "bitmanip.hfa"
         #include "fstream.hfa"
+        #include "kernel_private.hfa"
+        #include "kernel/private.hfa"
+        #include "limits.hfa"
         #include "thread.hfa"
+#pragma GCC diagnostic pop
         void ?{}(io_context_params & this) {
 …
                 this.ext_sq.empty = true;
                 (this.ext_sq.queue){};
                 __io_uring_setup( this, cl.io.params, proc->idle_fd );
+                __io_uring_setup( this, cl.io.params, proc->idle_wctx.evfd );
                 __cfadbg_print_safe(io_core, "Kernel I/O : Created ring for io_context %u (%p)\n", this.fd, &this);
+        }
 …
                 __cfadbg_print_safe(io_core, "Kernel I/O : Destroyed ring for io_context %u\n", this.fd);
+        }
-        extern void __disable_interrupts_hard();
-        extern void __enable_interrupts_hard();
         static void __io_uring_setup( $io_context & this, const io_context_params & params_in, int procfd ) {
 …
                 // completion queue
+                cq.lock      = false;
+                cq.id        = MAX;
+                cq.ts        = rdtscl();
                 cq.head      = (volatile __u32 *)(((intptr_t)cq.ring_ptr) + params.cq_off.head);
                 cq.tail      = (volatile __u32 *)(((intptr_t)cq.ring_ptr) + params.cq_off.tail);
 …
                         __cfadbg_print_safe(io_core, "Kernel I/O : registering %d for completion with ring %d\n", procfd, fd);
-                        __disable_interrupts_hard();
                         int ret = syscall( __NR_io_uring_register, fd, IORING_REGISTER_EVENTFD, &procfd, 1);
                         if (ret < 0) {
                                 abort("KERNEL ERROR: IO_URING EVENTFD REGISTER - %s\n", strerror(errno));
+                        }
-                        __enable_interrupts_hard();
                         __cfadbg_print_safe(io_core, "Kernel I/O : registered %d for completion with ring %d\n", procfd, fd);

libcfa/src/concurrency/io/types.hfa

-              rba897d21
+              r2e9b59b
 #include "bits/locks.hfa"
 #include "bits/queue.hfa"
+#include "iofwd.hfa"
 #include "kernel/fwd.hfa"
+#include "limits.hfa"
 #if defined(CFA_HAVE_LINUX_IO_URING_H)
 …
         struct __cmp_ring_t {
+                volatile bool lock;
+                unsigned id;
+                unsigned long long ts;
                 // Head and tail of the ring
                 volatile __u32 * head;
 …
         };
+        static inline unsigned long long ts($io_context *& this) {
+                const __u32 head = *this->cq.head;
+                const __u32 tail = *this->cq.tail;
+                if(head == tail) return MAX;
+                return this->cq.ts;
+        }
         struct __pending_alloc {
                 inline __outstanding_io;
 …
         // void __ioctx_prepare_block($io_context & ctx);
 #endif
-//-----------------------------------------------------------------------
-// IO user data
-struct io_future_t {
-        future_t self;
-        __s32 result;
-};
-static inline {
-        thread$ * fulfil( io_future_t & this, __s32 result, bool do_unpark = true ) {
-                this.result = result;
-                return fulfil(this.self, do_unpark);
+        }
-        // Wait for the future to be fulfilled
-        bool wait     ( io_future_t & this ) { return wait     (this.self); }
-        void reset    ( io_future_t & this ) { return reset    (this.self); }
-        bool available( io_future_t & this ) { return available(this.self); }
+}

libcfa/src/concurrency/iofwd.hfa

-              rba897d21
+              r2e9b59b
+}
 #include "bits/defs.hfa"
+#include "kernel/fwd.hfa"
 #include "time.hfa"
 …
 struct cluster;
-struct io_future_t;
 struct $io_context;
 …
 struct io_uring_sqe;
+//-----------------------------------------------------------------------
+// IO user data
+struct io_future_t {
+        future_t self;
+        __s32 result;
+};
+static inline {
+        thread$ * fulfil( io_future_t & this, __s32 result, bool do_unpark = true ) {
+                this.result = result;
+                return fulfil(this.self, do_unpark);
+        }
+        // Wait for the future to be fulfilled
+        bool wait     ( io_future_t & this ) { return wait     (this.self); }
+        void reset    ( io_future_t & this ) { return reset    (this.self); }
+        bool available( io_future_t & this ) { return available(this.self); }
+}
 //----------

libcfa/src/concurrency/kernel.cfa

-              rba897d21
+              r2e9b59b
 // #define __CFA_DEBUG_PRINT_RUNTIME_CORE__
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Waddress-of-packed-member"
 //C Includes
 #include <errno.h>
 …
 #include <signal.h>
 #include <unistd.h>
 extern "C" {
         #include <sys/eventfd.h>
 …
 //CFA Includes
 #include "kernel_private.hfa"
+#include "kernel/private.hfa"
 #include "preemption.hfa"
 #include "strstream.hfa"
 …
 #define __CFA_INVOKE_PRIVATE__
 #include "invoke.h"
+#pragma GCC diagnostic pop
 #if !defined(__CFA_NO_STATISTICS__)
 …
 static void __wake_one(cluster * cltr);
 static void idle_sleep(processor * proc, io_future_t & future, iovec & iov);
+static void idle_sleep(processor * proc);
 static bool mark_idle (__cluster_proc_list & idles, processor & proc);
 static void mark_awake(__cluster_proc_list & idles, processor & proc);
+extern void __cfa_io_start( processor * );
+extern bool __cfa_io_drain( processor * );
+extern bool __cfa_io_flush( processor *, int min_comp );
+extern void __cfa_io_stop ( processor * );
+static inline bool __maybe_io_drain( processor * );
+extern bool __cfa_io_drain( processor * proc ) __attribute__((nonnull (1)));
+extern bool __cfa_io_flush( processor * ) __attribute__((nonnull (1)));
+extern void __cfa_io_idle( processor * ) __attribute__((nonnull (1)));
 #if defined(CFA_WITH_IO_URING_IDLE)
 …
         verify(this);
         io_future_t future; // used for idle sleep when io_uring is present
         future.self.ptr = 1p;  // mark it as already fulfilled so we know if there is a pending request or not
+        eventfd_t idle_val;
         iovec idle_iovec = { &idle_val, sizeof(idle_val) };
         __cfa_io_start( this );
+        /* paranoid */ verify( this->idle_wctx.ftr   != 0p );
+        /* paranoid */ verify( this->idle_wctx.rdbuf != 0p );
+        // used for idle sleep when io_uring is present
+        // mark it as already fulfilled so we know if there is a pending request or not
+        this->idle_wctx.ftr->self.ptr = 1p;
         __cfadbg_print_safe(runtime_core, "Kernel : core %p starting\n", this);
 …
                 for() {
                         // Check if there is pending io
                         __maybe_io_drain( this );
+                        __cfa_io_drain( this );
                         // Try to get the next thread
 …
                         if( !readyThread ) {
+                                // there is no point in holding submissions if we are idle
                                 __IO_STATS__(true, io.flush.idle++; )
+                                __cfa_io_flush( this, 0 );
+                                __cfa_io_flush( this );
+                                // drain again in case something showed up
+                                __cfa_io_drain( this );
                                 readyThread = __next_thread( this->cltr );
 …
                         if( !readyThread ) for(5) {
+                                readyThread = __next_thread_slow( this->cltr );
+                                if( readyThread ) break;
+                                // It's unlikely we still I/O to submit, but the arbiter could
                                 __IO_STATS__(true, io.flush.idle++; )
+                                readyThread = __next_thread_slow( this->cltr );
+                                if( readyThread ) break;
+                                __cfa_io_flush( this, 0 );
+                                __cfa_io_flush( this );
+                                // drain again in case something showed up
+                                __cfa_io_drain( this );
+                        }
 …
+                                }
                                 idle_sleep( this, future, idle_iovec );
+                                idle_sleep( this );
                                 // We were woken up, remove self from idle
 …
                         if( __atomic_load_n(&this->do_terminate, __ATOMIC_SEQ_CST) ) break MAIN_LOOP;
                         if(this->io.pending && !this->io.dirty) {
+                        if(__atomic_load_n(&this->io.pending, __ATOMIC_RELAXED) && !__atomic_load_n(&this->io.dirty, __ATOMIC_RELAXED)) {
                                 __IO_STATS__(true, io.flush.dirty++; )
                                 __cfa_io_flush( this, 0 );
+                                __cfa_io_flush( this );
+                        }
+                }
 …
                 __cfadbg_print_safe(runtime_core, "Kernel : core %p stopping\n", this);
+        }
-        for(int i = 0; !available(future); i++) {
-                if(i > 1000) __cfaabi_dbg_write( "ERROR: kernel has bin spinning on a flush after exit loop.\n", 60);
-                __cfa_io_flush( this, 1 );
+        }
-        __cfa_io_stop( this );
         post( this->terminated );
 …
         int fd = 1;
         if( __atomic_load_n(&fdp->fd, __ATOMIC_SEQ_CST) != 1 ) {
                 fd = __atomic_exchange_n(&fdp->fd, 1, __ATOMIC_RELAXED);
+        if( __atomic_load_n(&fdp->sem, __ATOMIC_SEQ_CST) != 1 ) {
+                fd = __atomic_exchange_n(&fdp->sem, 1, __ATOMIC_RELAXED);
+        }
         switch(fd) {
+                __attribute__((unused)) int ret;
         case 0:
                 // If the processor isn't ready to sleep then the exchange will already wake it up
 …
                 // If the processor was ready to sleep, we need to wake it up with an actual write
                 val = 1;
+                eventfd_write( fd, val );
+                ret = eventfd_write( fd, val );
+                /* paranoid */ verifyf( ret == 0, "Expected return to be 0, was %d\n", ret );
                 #if !defined(__CFA_NO_STATISTICS__)
 …
         __cfadbg_print_safe(runtime_core, "Kernel : waking Processor %p\n", this);
+        this->idle_wctx.fd = 1;
+        this->idle_wctx.sem = 1;
+        this->idle_wctx.wake__time = rdtscl();
         eventfd_t val;
         val = 1;
+        eventfd_write( this->idle_fd, val );
+        /* paranoid */ verify( ! __preemption_enabled() );
+}
+static void idle_sleep(processor * this, io_future_t & future, iovec & iov) {
+        __attribute__((unused)) int ret = eventfd_write( this->idle_wctx.evfd, val );
+        /* paranoid */ verifyf( ret == 0, "Expected return to be 0, was %d\n", ret );
+        /* paranoid */ verify( ! __preemption_enabled() );
+}
+static void idle_sleep(processor * this) {
+        /* paranoid */ verify( this->idle_wctx.evfd != 1 );
+        /* paranoid */ verify( this->idle_wctx.evfd != 2 );
         // Tell everyone we are ready to go do sleep
         for() {
                 int expected = this->idle_wctx.fd;
+                int expected = this->idle_wctx.sem;
                 // Someone already told us to wake-up! No time for a nap.
 …
                 // Try to mark that we are going to sleep
                 if(__atomic_compare_exchange_n(&this->idle_wctx.fd, &expected, this->idle_fd, false,  __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST) ) {
+                if(__atomic_compare_exchange_n(&this->idle_wctx.sem, &expected, this->idle_wctx.evfd, false,  __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST) ) {
                         // Every one agreed, taking a nap
                         break;
 …
+                {
                         eventfd_t val;
                         ssize_t ret = read( this->idle_fd, &val, sizeof(val) );
+                        ssize_t ret = read( this->idle_wctx.evfd, &val, sizeof(val) );
                         if(ret < 0) {
                                 switch((int)errno) {
 …
                 #endif
         #else
+                // Do we already have a pending read
+                if(available(future)) {
+                        // There is no pending read, we need to add one
+                        reset(future);
+                        __kernel_read(this, future, iov, this->idle_fd );
+                }
+                __cfa_io_flush( this, 1 );
+                __cfa_io_idle( this );
         #endif
+}
 …
         __STATS__(true, ready.sleep.halts++; )
         proc.idle_wctx.fd = 0;
+        proc.idle_wctx.sem = 0;
         /* paranoid */ verify( ! __preemption_enabled() );
 …
 #endif
+static inline bool __maybe_io_drain( processor * proc ) {
+        bool ret = false;
+        #if defined(CFA_HAVE_LINUX_IO_URING_H)
+                __cfadbg_print_safe(runtime_core, "Kernel : core %p checking io for ring %d\n", proc, proc->io.ctx->fd);
+                // Check if we should drain the queue
+                $io_context * ctx = proc->io.ctx;
+                unsigned head = *ctx->cq.head;
+                unsigned tail = *ctx->cq.tail;
+                if(head == tail) return false;
+                ready_schedule_lock();
+                ret = __cfa_io_drain( proc );
+                ready_schedule_unlock();
+        #endif
+        return ret;
+}
 //-----------------------------------------------------------------------------
 …
         void print_stats_now( cluster & this, int flags ) {
                 crawl_cluster_stats( this );
                 __print_stats( this.stats, this.print_stats, "Cluster", this.name, (void*)&this );
+                __print_stats( this.stats, flags, "Cluster", this.name, (void*)&this );
+        }
 #endif

libcfa/src/concurrency/kernel.hfa

-              rba897d21
+              r2e9b59b
 extern struct cluster * mainCluster;
+// Processor id, required for scheduling threads
+// Coroutine used py processors for the 2-step context switch
 coroutine processorCtx_t {
         struct processor * proc;
 };
+struct io_future_t;
+// Information needed for idle sleep
 struct __fd_waitctx {
+        volatile int fd;
+        // semaphore/future like object
+        // values can be 0, 1 or some file descriptor.
+        // 0 - is the default state
+        // 1 - means the proc should wake-up immediately
+        // FD - means the proc is going asleep and should be woken by writing to the FD.
+        volatile int sem;
+        // The event FD that corresponds to this processor
+        int evfd;
+        // buffer into which the proc will read from evfd
+        // unused if not using io_uring for idle sleep
+        void * rdbuf;
+        // future use to track the read of the eventfd
+        // unused if not using io_uring for idle sleep
+        io_future_t * ftr;
+        volatile unsigned long long wake__time;
+        volatile unsigned long long sleep_time;
+        volatile unsigned long long drain_time;
 };
 …
         struct {
                 $io_context * ctx;
+                bool pending;
+                bool dirty;
+                unsigned target;
+                volatile bool pending;
+                volatile bool dirty;
         } io;
 …
         bool pending_preemption;
+        // Idle lock (kernel semaphore)
+        int idle_fd;
+        // Idle waitctx
+        // context for idle sleep
         struct __fd_waitctx idle_wctx;
 …
 void ^?{}(__intrusive_lane_t & this);
 // Aligned timestamps which are used by the relaxed ready queue
+// Aligned timestamps which are used by the ready queue and io subsystem
 struct __attribute__((aligned(128))) __timestamp_t {
         volatile unsigned long long tv;
 …
 };
+static inline void  ?{}(__timestamp_t & this) { this.tv = 0; this.ma = 0; }
+static inline void ^?{}(__timestamp_t &) {}
 struct __attribute__((aligned(16))) __cache_id_t {
         volatile unsigned id;
 };
-// Aligned timestamps which are used by the relaxed ready queue
-struct __attribute__((aligned(128))) __help_cnts_t {
-        volatile unsigned long long src;
-        volatile unsigned long long dst;
-        volatile unsigned long long tri;
-};
-static inline void  ?{}(__timestamp_t & this) { this.tv = 0; this.ma = 0; }
-static inline void ^?{}(__timestamp_t &) {}
-struct __attribute__((aligned(128))) __ready_queue_caches_t;
-void  ?{}(__ready_queue_caches_t & this);
-void ^?{}(__ready_queue_caches_t & this);
-//TODO adjust cache size to ARCHITECTURE
-// Structure holding the ready queue
-struct __ready_queue_t {
-        // Data tracking the actual lanes
-        // On a seperate cacheline from the used struct since
-        // used can change on each push/pop but this data
-        // only changes on shrink/grow
-        struct {
-                // Arary of lanes
-                __intrusive_lane_t * volatile data;
-                // Array of times
-                __timestamp_t * volatile tscs;
-                __cache_id_t * volatile caches;
-                // Array of stats
-                __help_cnts_t * volatile help;
-                // Number of lanes (empty or not)
-                volatile size_t count;
-        } lanes;
-};
-void  ?{}(__ready_queue_t & this);
-void ^?{}(__ready_queue_t & this);
-#if !defined(__CFA_NO_STATISTICS__)
-        unsigned cnt(const __ready_queue_t & this, unsigned idx);
-#endif
 // Idle Sleep
 …
 // Cluster
 struct __attribute__((aligned(128))) cluster {
+        // Ready queue for threads
+        __ready_queue_t ready_queue;
+        struct {
+                struct {
+                        // Arary of subqueues
+                        __intrusive_lane_t * data;
+                        // Time since subqueues were processed
+                        __timestamp_t * tscs;
+                        // Number of subqueue / timestamps
+                        size_t count;
+                } readyQ;
+                struct {
+                        // Array of $io_
+                        $io_context ** data;
+                        // Time since subqueues were processed
+                        __timestamp_t * tscs;
+                        // Number of I/O subqueues
+                        size_t count;
+                } io;
+                // Cache each kernel thread belongs to
+                __cache_id_t * caches;
+        } sched;
+        // // Ready queue for threads
+        // __ready_queue_t ready_queue;
         // Name of the cluster

libcfa/src/concurrency/kernel/fwd.hfa

rba897d21	r2e9b59b
248	248	// check if the future is available
249	249	bool available( future_t & this ) {
	250	while( this.ptr == 2p ) Pause();
250	251	return this.ptr == 1p;
251	252	}

libcfa/src/concurrency/kernel/private.hfa

-              rba897d21
+              r2e9b59b
 // file "LICENCE" distributed with Cforall.
 //
 // kernel_private.hfa --
+// kernel/private.hfa --
 //
 // Author           : Thierry Delisle
 …
 #if !defined(__cforall_thread__)
         #error kernel_private.hfa should only be included in libcfathread source
+        #error kernel/private.hfa should only be included in libcfathread source
 #endif
 …
 #else
         #ifndef _GNU_SOURCE
         #error kernel_private requires gnu_source
+        #error kernel/private requires gnu_source
         #endif
         #include <sched.h>
 …
 // Defines whether or not we *want* to use io_uring_enter as the idle_sleep blocking call
 #define CFA_WANT_IO_URING_IDLE
+// #define CFA_WANT_IO_URING_IDLE
 // Defines whether or not we *can* use io_uring_enter as the idle_sleep blocking call
 …
 void ready_queue_shrink(struct cluster * cltr);
+//-----------------------------------------------------------------------
+// Decrease the width of the ready queue (number of lanes) by 4
+void ready_queue_close(struct cluster * cltr);
 // Local Variables: //

libcfa/src/concurrency/kernel/startup.cfa

-              rba897d21
+              r2e9b59b
 // CFA Includes
+#include "kernel_private.hfa"
+#include "kernel/private.hfa"
+#include "iofwd.hfa"
 #include "startup.hfa"                                  // STARTUP_PRIORITY_XXX
 #include "limits.hfa"
 …
 extern void __kernel_alarm_startup(void);
 extern void __kernel_alarm_shutdown(void);
+extern void __cfa_io_start( processor * );
+extern void __cfa_io_stop ( processor * );
 //-----------------------------------------------------------------------------
 …
 KERNEL_STORAGE(__stack_t,            mainThreadCtx);
 KERNEL_STORAGE(__scheduler_RWLock_t, __scheduler_lock);
+KERNEL_STORAGE(eventfd_t,            mainIdleEventFd);
+KERNEL_STORAGE(io_future_t,          mainIdleFuture);
 #if !defined(__CFA_NO_STATISTICS__)
 KERNEL_STORAGE(__stats_t, mainProcStats);
 …
         (*mainProcessor){};
+        mainProcessor->idle_wctx.rdbuf = &storage_mainIdleEventFd;
+        mainProcessor->idle_wctx.ftr   = (io_future_t*)&storage_mainIdleFuture;
+        /* paranoid */ verify( sizeof(storage_mainIdleEventFd) == sizeof(eventfd_t) );
+        __cfa_io_start( mainProcessor );
         register_tls( mainProcessor );
 …
         unregister_tls( mainProcessor );
+        __cfa_io_stop( mainProcessor );
         // Destroy the main processor and its context in reverse order of construction
 …
         proc->local_data = &__cfaabi_tls;
+        __cfa_io_start( proc );
         register_tls( proc );
+        // used for idle sleep when io_uring is present
+        io_future_t future;
+        eventfd_t idle_buf;
+        proc->idle_wctx.ftr = &future;
+        proc->idle_wctx.rdbuf = &idle_buf;
         // SKULLDUGGERY: We want to create a context for the processor coroutine
 …
         unregister_tls( proc );
+        __cfa_io_stop( proc );
         return 0p;
 …
         this.rdq.its = 0;
         this.rdq.itr = 0;
         this.rdq.id  = MAX;
+        this.rdq.id  = 0;
         this.rdq.target = MAX;
         this.rdq.last = MAX;
 …
         this.local_data = 0p;
         this.idle_fd = eventfd(0, 0);
         if (idle_fd < 0) {
+        idle_wctx.evfd = eventfd(0, 0);
+        if (idle_wctx.evfd < 0) {
                 abort("KERNEL ERROR: PROCESSOR EVENTFD - %s\n", strerror(errno));
+        }
+        this.idle_wctx.fd = 0;
+        idle_wctx.sem = 0;
+        idle_wctx.wake__time = 0;
         // I'm assuming these two are reserved for standard input and output
         // so I'm using them as sentinels with idle_wctx.
         /* paranoid */ verify( this.idle_fd != 0 );
         /* paranoid */ verify( this.idle_fd != 1 );
+        /* paranoid */ verify( idle_wctx.evfd != 0 );
+        /* paranoid */ verify( idle_wctx.evfd != 1 );
         #if !defined(__CFA_NO_STATISTICS__)
 …
 // Not a ctor, it just preps the destruction but should not destroy members
 static void deinit(processor & this) {
         close(this.idle_fd);
+        close(this.idle_wctx.evfd);
+}
 …
         this.name = name;
         this.preemption_rate = preemption_rate;
+        ready_queue{};
+        this.sched.readyQ.data = 0p;
+        this.sched.readyQ.tscs = 0p;
+        this.sched.readyQ.count = 0;
+        this.sched.io.tscs = 0p;
+        this.sched.io.data = 0p;
+        this.sched.caches = 0p;
         #if !defined(__CFA_NO_STATISTICS__)
 …
         // Unlock the RWlock
         ready_mutate_unlock( last_size );
+        ready_queue_close( &this );
+        /* paranoid */ verify( this.sched.readyQ.data == 0p );
+        /* paranoid */ verify( this.sched.readyQ.tscs == 0p );
+        /* paranoid */ verify( this.sched.readyQ.count == 0 );
+        /* paranoid */ verify( this.sched.io.tscs == 0p );
+        /* paranoid */ verify( this.sched.caches == 0p );
         enable_interrupts( false ); // Don't poll, could be in main cluster
         #if !defined(__CFA_NO_STATISTICS__)

libcfa/src/concurrency/locks.cfa

rba897d21	r2e9b59b
19	19
20	20	#include "locks.hfa"
21		#include "kernel_private.hfa"
	21	#include "kernel/private.hfa"
22	22
23	23	#include <kernel.hfa>

libcfa/src/concurrency/locks.hfa

-              rba897d21
+              r2e9b59b
+}
 static inline bool lock(linear_backoff_then_block_lock & this) with(this) {
+static inline void lock(linear_backoff_then_block_lock & this) with(this) {
         // if owner just return
         if (active_thread() == owner) return true;
+        if (active_thread() == owner) return;
         size_t compare_val = 0;
         int spin = spin_start;
 …
         for( ;; ) {
                 compare_val = 0;
                 if (internal_try_lock(this, compare_val)) return true;
+                if (internal_try_lock(this, compare_val)) return;
                 if (2 == compare_val) break;
                 for (int i = 0; i < spin; i++) Pause();
 …
+        }
         if(2 != compare_val && try_lock_contention(this)) return true;
+        if(2 != compare_val && try_lock_contention(this)) return;
         // block until signalled
+        while (block(this)) if(try_lock_contention(this)) return true;
+        // this should never be reached as block(this) always returns true
+        return false;
+        while (block(this)) if(try_lock_contention(this)) return;
+}

libcfa/src/concurrency/monitor.cfa

rba897d21	r2e9b59b
22	22	#include <inttypes.h>
23	23
24		#include "kernel_private.hfa"
	24	#include "kernel/private.hfa"
25	25
26	26	#include "bits/algorithm.hfa"

libcfa/src/concurrency/mutex.cfa

rba897d21	r2e9b59b
21	21	#include "mutex.hfa"
22	22
23		#include "kernel_private.hfa"
	23	#include "kernel/private.hfa"
24	24
25	25	//-----------------------------------------------------------------------------

libcfa/src/concurrency/mutex_stmt.hfa

-              rba897d21
+              r2e9b59b
 };
+struct __mutex_stmt_lock_guard {
+    void ** lockarr;
+    __lock_size_t count;
+};
+static inline void ?{}( __mutex_stmt_lock_guard & this, void * lockarr [], __lock_size_t count  ) {
+    this.lockarr = lockarr;
+    this.count = count;
+    // Sort locks based on address
+    __libcfa_small_sort(this.lockarr, count);
+    // acquire locks in order
+    // for ( size_t i = 0; i < count; i++ ) {
+    //     lock(*this.lockarr[i]);
+    // }
+}
+static inline void ^?{}( __mutex_stmt_lock_guard & this ) with(this) {
+    // for ( size_t i = count; i > 0; i-- ) {
+    //     unlock(*lockarr[i - 1]);
+    // }
+}
 forall(L & | is_lock(L)) {
-    struct __mutex_stmt_lock_guard {
-        L ** lockarr;
-        __lock_size_t count;
-    };
-    static inline void ?{}( __mutex_stmt_lock_guard(L) & this, L * lockarr [], __lock_size_t count  ) {
-        this.lockarr = lockarr;
-        this.count = count;
-        // Sort locks based on address
-        __libcfa_small_sort(this.lockarr, count);
-        // acquire locks in order
-        for ( size_t i = 0; i < count; i++ ) {
-            lock(*this.lockarr[i]);
+        }
+    }
-    static inline void ^?{}( __mutex_stmt_lock_guard(L) & this ) with(this) {
-        for ( size_t i = count; i > 0; i-- ) {
-            unlock(*lockarr[i - 1]);
+        }
+    }
     struct scoped_lock {
 …
+    }
     static inline L * __get_ptr( L & this ) {
+    static inline void * __get_mutexstmt_lock_ptr( L & this ) {
         return &this;
+    }
     static inline L __get_type( L & this );
+    static inline L __get_mutexstmt_lock_type( L & this );
     static inline L __get_type( L * this );
+    static inline L __get_mutexstmt_lock_type( L * this );
+}

libcfa/src/concurrency/preemption.cfa

rba897d21	r2e9b59b
31	31	#include "bits/debug.hfa"
32	32	#include "bits/signal.hfa"
33		#include "kernel_private.hfa"
	33	#include "kernel/private.hfa"
34	34
35	35

libcfa/src/concurrency/ready_queue.cfa

-              rba897d21
+              r2e9b59b
-// #define USE_RELAXED_FIFO
-// #define USE_WORK_STEALING
-// #define USE_CPU_WORK_STEALING
 #define USE_AWARE_STEALING
 #include "bits/defs.hfa"
 #include "device/cpu.hfa"
+#include "kernel_private.hfa"
+#include "stdlib.hfa"
+#include "limits.hfa"
+#include "math.hfa"
+#include <errno.h>
+#include <unistd.h>
+extern "C" {
+        #include <sys/syscall.h>  // __NR_xxx
+}
+#include "kernel/cluster.hfa"
+#include "kernel/private.hfa"
+// #include <errno.h>
+// #include <unistd.h>
 #include "ready_subqueue.hfa"
 …
 #endif
-// No overriden function, no environment variable, no define
-// fall back to a magic number
-#ifndef __CFA_MAX_PROCESSORS__
-        #define __CFA_MAX_PROCESSORS__ 1024
-#endif
-#if   defined(USE_AWARE_STEALING)
-        #define READYQ_SHARD_FACTOR 2
-        #define SEQUENTIAL_SHARD 2
-#elif defined(USE_CPU_WORK_STEALING)
-        #define READYQ_SHARD_FACTOR 2
-#elif defined(USE_RELAXED_FIFO)
-        #define BIAS 4
-        #define READYQ_SHARD_FACTOR 4
-        #define SEQUENTIAL_SHARD 1
-#elif defined(USE_WORK_STEALING)
-        #define READYQ_SHARD_FACTOR 2
-        #define SEQUENTIAL_SHARD 2
-#else
-        #error no scheduling strategy selected
-#endif
 static inline struct thread$ * try_pop(struct cluster * cltr, unsigned w __STATS(, __stats_readyQ_pop_t & stats));
 static inline struct thread$ * try_pop(struct cluster * cltr, unsigned i, unsigned j __STATS(, __stats_readyQ_pop_t & stats));
 static inline struct thread$ * search(struct cluster * cltr);
-static inline [unsigned, bool] idx_from_r(unsigned r, unsigned preferred);
-// returns the maximum number of processors the RWLock support
-__attribute__((weak)) unsigned __max_processors() {
-        const char * max_cores_s = getenv("CFA_MAX_PROCESSORS");
-        if(!max_cores_s) {
-                __cfadbg_print_nolock(ready_queue, "No CFA_MAX_PROCESSORS in ENV\n");
-                return __CFA_MAX_PROCESSORS__;
+        }
-        char * endptr = 0p;
-        long int max_cores_l = strtol(max_cores_s, &endptr, 10);
-        if(max_cores_l < 1 || max_cores_l > 65535) {
-                __cfadbg_print_nolock(ready_queue, "CFA_MAX_PROCESSORS out of range : %ld\n", max_cores_l);
-                return __CFA_MAX_PROCESSORS__;
+        }
-        if('\0' != *endptr) {
-                __cfadbg_print_nolock(ready_queue, "CFA_MAX_PROCESSORS not a decimal number : %s\n", max_cores_s);
-                return __CFA_MAX_PROCESSORS__;
+        }
-        return max_cores_l;
+}
-#if   defined(CFA_HAVE_LINUX_LIBRSEQ)
-        // No forward declaration needed
-        #define __kernel_rseq_register rseq_register_current_thread
-        #define __kernel_rseq_unregister rseq_unregister_current_thread
-#elif defined(CFA_HAVE_LINUX_RSEQ_H)
-        static void __kernel_raw_rseq_register  (void);
-        static void __kernel_raw_rseq_unregister(void);
-        #define __kernel_rseq_register __kernel_raw_rseq_register
-        #define __kernel_rseq_unregister __kernel_raw_rseq_unregister
-#else
-        // No forward declaration needed
-        // No initialization needed
-        static inline void noop(void) {}
-        #define __kernel_rseq_register noop
-        #define __kernel_rseq_unregister noop
-#endif
-//=======================================================================
-// Cluster wide reader-writer lock
-//=======================================================================
-void  ?{}(__scheduler_RWLock_t & this) {
-        this.max   = __max_processors();
-        this.alloc = 0;
-        this.ready = 0;
-        this.data  = alloc(this.max);
-        this.write_lock  = false;
-        /*paranoid*/ verify(__atomic_is_lock_free(sizeof(this.alloc), &this.alloc));
-        /*paranoid*/ verify(__atomic_is_lock_free(sizeof(this.ready), &this.ready));
+}
-void ^?{}(__scheduler_RWLock_t & this) {
-        free(this.data);
+}
-//=======================================================================
-// Lock-Free registering/unregistering of threads
-unsigned register_proc_id( void ) with(*__scheduler_lock) {
-        __kernel_rseq_register();
-        bool * handle = (bool *)&kernelTLS().sched_lock;
-        // Step - 1 : check if there is already space in the data
-        uint_fast32_t s = ready;
-        // Check among all the ready
-        for(uint_fast32_t i = 0; i < s; i++) {
-                bool * volatile * cell = (bool * volatile *)&data[i]; // Cforall is bugged and the double volatiles causes problems
-                /* paranoid */ verify( handle != *cell );
-                bool * null = 0p; // Re-write every loop since compare thrashes it
-                if( __atomic_load_n(cell, (int)__ATOMIC_RELAXED) == null
-                        && __atomic_compare_exchange_n( cell, &null, handle, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) {
-                        /* paranoid */ verify(i < ready);
-                        /* paranoid */ verify( (kernelTLS().sched_id = i, true) );
-                        return i;
+                }
+        }
-        if(max <= alloc) abort("Trying to create more than %ud processors", __scheduler_lock->max);
-        // Step - 2 : F&A to get a new spot in the array.
-        uint_fast32_t n = __atomic_fetch_add(&alloc, 1, __ATOMIC_SEQ_CST);
-        if(max <= n) abort("Trying to create more than %ud processors", __scheduler_lock->max);
-        // Step - 3 : Mark space as used and then publish it.
-        data[n] = handle;
-        while() {
-                unsigned copy = n;
-                if( __atomic_load_n(&ready, __ATOMIC_RELAXED) == n
-                        && __atomic_compare_exchange_n(&ready, &copy, n + 1, true, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST))
-                        break;
-                Pause();
+        }
-        // Return new spot.
-        /* paranoid */ verify(n < ready);
-        /* paranoid */ verify( (kernelTLS().sched_id = n, true) );
-        return n;
+}
-void unregister_proc_id( unsigned id ) with(*__scheduler_lock) {
-        /* paranoid */ verify(id < ready);
-        /* paranoid */ verify(id == kernelTLS().sched_id);
-        /* paranoid */ verify(data[id] == &kernelTLS().sched_lock);
-        bool * volatile * cell = (bool * volatile *)&data[id]; // Cforall is bugged and the double volatiles causes problems
-        __atomic_store_n(cell, 0p, __ATOMIC_RELEASE);
-        __kernel_rseq_unregister();
+}
-//-----------------------------------------------------------------------
-// Writer side : acquire when changing the ready queue, e.g. adding more
-//  queues or removing them.
-uint_fast32_t ready_mutate_lock( void ) with(*__scheduler_lock) {
-        /* paranoid */ verify( ! __preemption_enabled() );
-        // Step 1 : lock global lock
-        // It is needed to avoid processors that register mid Critical-Section
-        //   to simply lock their own lock and enter.
-        __atomic_acquire( &write_lock );
-        // Make sure we won't deadlock ourself
-        // Checking before acquiring the writer lock isn't safe
-        // because someone else could have locked us.
-        /* paranoid */ verify( ! kernelTLS().sched_lock );
-        // Step 2 : lock per-proc lock
-        // Processors that are currently being registered aren't counted
-        //   but can't be in read_lock or in the critical section.
-        // All other processors are counted
-        uint_fast32_t s = ready;
-        for(uint_fast32_t i = 0; i < s; i++) {
-                volatile bool * llock = data[i];
-                if(llock) __atomic_acquire( llock );
+        }
-        /* paranoid */ verify( ! __preemption_enabled() );
-        return s;
+}
-void ready_mutate_unlock( uint_fast32_t last_s ) with(*__scheduler_lock) {
-        /* paranoid */ verify( ! __preemption_enabled() );
-        // Step 1 : release local locks
-        // This must be done while the global lock is held to avoid
-        //   threads that where created mid critical section
-        //   to race to lock their local locks and have the writer
-        //   immidiately unlock them
-        // Alternative solution : return s in write_lock and pass it to write_unlock
-        for(uint_fast32_t i = 0; i < last_s; i++) {
-                volatile bool * llock = data[i];
-                if(llock) __atomic_store_n(llock, (bool)false, __ATOMIC_RELEASE);
+        }
-        // Step 2 : release global lock
-        /*paranoid*/ assert(true == write_lock);
-        __atomic_store_n(&write_lock, (bool)false, __ATOMIC_RELEASE);
-        /* paranoid */ verify( ! __preemption_enabled() );
+}
-//=======================================================================
-// caches handling
-struct __attribute__((aligned(128))) __ready_queue_caches_t {
-        // Count States:
-        // - 0  : No one is looking after this cache
-        // - 1  : No one is looking after this cache, BUT it's not empty
-        // - 2+ : At least one processor is looking after this cache
-        volatile unsigned count;
-};
-void  ?{}(__ready_queue_caches_t & this) { this.count = 0; }
-void ^?{}(__ready_queue_caches_t & this) {}
-static inline void depart(__ready_queue_caches_t & cache) {
-        /* paranoid */ verify( cache.count > 1);
-        __atomic_fetch_add(&cache.count, -1, __ATOMIC_SEQ_CST);
-        /* paranoid */ verify( cache.count != 0);
-        /* paranoid */ verify( cache.count < 65536 ); // This verify assumes no cluster will have more than 65000 kernel threads mapped to a single cache, which could be correct but is super weird.
+}
-static inline void arrive(__ready_queue_caches_t & cache) {
-        // for() {
-        //      unsigned expected = cache.count;
-        //      unsigned desired  = 0 == expected ? 2 : expected + 1;
-        // }
+}
 //=======================================================================
 // Cforall Ready Queue used for scheduling
 //=======================================================================
+unsigned long long moving_average(unsigned long long currtsc, unsigned long long instsc, unsigned long long old_avg) {
+        /* paranoid */ verifyf( currtsc < 45000000000000000, "Suspiciously large current time: %'llu (%llx)\n", currtsc, currtsc );
+        /* paranoid */ verifyf( instsc  < 45000000000000000, "Suspiciously large insert time: %'llu (%llx)\n", instsc, instsc );
+        /* paranoid */ verifyf( old_avg < 15000000000000, "Suspiciously large previous average: %'llu (%llx)\n", old_avg, old_avg );
+        const unsigned long long new_val = currtsc > instsc ? currtsc - instsc : 0;
+        const unsigned long long total_weight = 16;
+        const unsigned long long new_weight   = 4;
+        const unsigned long long old_weight = total_weight - new_weight;
+        const unsigned long long ret = ((new_weight * new_val) + (old_weight * old_avg)) / total_weight;
+        return ret;
+}
+void ?{}(__ready_queue_t & this) with (this) {
+        #if defined(USE_CPU_WORK_STEALING)
+                lanes.count = cpu_info.hthrd_count * READYQ_SHARD_FACTOR;
+                lanes.data = alloc( lanes.count );
+                lanes.tscs = alloc( lanes.count );
+                lanes.help = alloc( cpu_info.hthrd_count );
+                for( idx; (size_t)lanes.count ) {
+                        (lanes.data[idx]){};
+                        lanes.tscs[idx].tv = rdtscl();
+                        lanes.tscs[idx].ma = rdtscl();
+                }
+                for( idx; (size_t)cpu_info.hthrd_count ) {
+                        lanes.help[idx].src = 0;
+                        lanes.help[idx].dst = 0;
+                        lanes.help[idx].tri = 0;
+                }
+        #else
+                lanes.data   = 0p;
+                lanes.tscs   = 0p;
+                lanes.caches = 0p;
+                lanes.help   = 0p;
+                lanes.count  = 0;
+        #endif
+}
+void ^?{}(__ready_queue_t & this) with (this) {
+        #if !defined(USE_CPU_WORK_STEALING)
+                verify( SEQUENTIAL_SHARD == lanes.count );
+        #endif
+        free(lanes.data);
+        free(lanes.tscs);
+        free(lanes.caches);
+        free(lanes.help);
+}
+//-----------------------------------------------------------------------
+#if defined(USE_AWARE_STEALING)
+        __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) {
+                processor * const proc = kernelTLS().this_processor;
+                const bool external = (!proc) || (cltr != proc->cltr);
+                const bool remote   = hint == UNPARK_REMOTE;
+                unsigned i;
+                if( external || remote ) {
+                        // Figure out where thread was last time and make sure it's valid
+                        /* paranoid */ verify(thrd->preferred >= 0);
+                        if(thrd->preferred * READYQ_SHARD_FACTOR < lanes.count) {
+                                /* paranoid */ verify(thrd->preferred * READYQ_SHARD_FACTOR < lanes.count);
+                                unsigned start = thrd->preferred * READYQ_SHARD_FACTOR;
+                                do {
+                                        unsigned r = __tls_rand();
+                                        i = start + (r % READYQ_SHARD_FACTOR);
+                                        /* paranoid */ verify( i < lanes.count );
+                                        // If we can't lock it retry
+                                } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                        } else {
+                                do {
+                                        i = __tls_rand() % lanes.count;
+                                } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                        }
+// void ?{}(__ready_queue_t & this) with (this) {
+//      lanes.data   = 0p;
+//      lanes.tscs   = 0p;
+//      lanes.caches = 0p;
+//      lanes.count  = 0;
+// }
+// void ^?{}(__ready_queue_t & this) with (this) {
+//      free(lanes.data);
+//      free(lanes.tscs);
+//      free(lanes.caches);
+// }
+//-----------------------------------------------------------------------
+__attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->sched) {
+        processor * const proc = kernelTLS().this_processor;
+        const bool external = (!proc) || (cltr != proc->cltr);
+        const bool remote   = hint == UNPARK_REMOTE;
+        const size_t lanes_count = readyQ.count;
+        /* paranoid */ verify( __shard_factor.readyq > 0 );
+        /* paranoid */ verify( lanes_count > 0 );
+        unsigned i;
+        if( external || remote ) {
+                // Figure out where thread was last time and make sure it's valid
+                /* paranoid */ verify(thrd->preferred >= 0);
+                unsigned start = thrd->preferred * __shard_factor.readyq;
+                if(start < lanes_count) {
+                        do {
+                                unsigned r = __tls_rand();
+                                i = start + (r % __shard_factor.readyq);
+                                /* paranoid */ verify( i < lanes_count );
+                                // If we can't lock it retry
+                        } while( !__atomic_try_acquire( &readyQ.data[i].lock ) );
                 } else {
                         do {
+                                unsigned r = proc->rdq.its++;
+                                i = proc->rdq.id + (r % READYQ_SHARD_FACTOR);
+                                /* paranoid */ verify( i < lanes.count );
+                                // If we can't lock it retry
+                        } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                }
+                // Actually push it
+                push(lanes.data[i], thrd);
+                // Unlock and return
+                __atomic_unlock( &lanes.data[i].lock );
+                #if !defined(__CFA_NO_STATISTICS__)
+                        if(unlikely(external || remote)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);
+                        else __tls_stats()->ready.push.local.success++;
+                #endif
+        }
+        static inline unsigned long long calc_cutoff(const unsigned long long ctsc, const processor * proc, __ready_queue_t & rdq) {
+                unsigned start = proc->rdq.id;
+                unsigned long long max = 0;
+                for(i; READYQ_SHARD_FACTOR) {
+                        unsigned long long ptsc = ts(rdq.lanes.data[start + i]);
+                        if(ptsc != -1ull) {
+                                /* paranoid */ verify( start + i < rdq.lanes.count );
+                                unsigned long long tsc = moving_average(ctsc, ptsc, rdq.lanes.tscs[start + i].ma);
+                                if(tsc > max) max = tsc;
+                        }
+                }
+                return (max + 2 * max) / 2;
+        }
+        __attribute__((hot)) struct thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {
+                /* paranoid */ verify( lanes.count > 0 );
+                /* paranoid */ verify( kernelTLS().this_processor );
+                /* paranoid */ verify( kernelTLS().this_processor->rdq.id < lanes.count );
+                processor * const proc = kernelTLS().this_processor;
+                unsigned this = proc->rdq.id;
+                /* paranoid */ verify( this < lanes.count );
+                __cfadbg_print_safe(ready_queue, "Kernel : pop from %u\n", this);
+                // Figure out the current cpu and make sure it is valid
+                const int cpu = __kernel_getcpu();
+                /* paranoid */ verify(cpu >= 0);
+                /* paranoid */ verify(cpu < cpu_info.hthrd_count);
+                unsigned this_cache = cpu_info.llc_map[cpu].cache;
+                // Super important: don't write the same value over and over again
+                // We want to maximise our chances that his particular values stays in cache
+                if(lanes.caches[this / READYQ_SHARD_FACTOR].id != this_cache)
+                        __atomic_store_n(&lanes.caches[this / READYQ_SHARD_FACTOR].id, this_cache, __ATOMIC_RELAXED);
+                const unsigned long long ctsc = rdtscl();
+                if(proc->rdq.target == MAX) {
+                        uint64_t chaos = __tls_rand();
+                        unsigned ext = chaos & 0xff;
+                        unsigned other  = (chaos >> 8) % (lanes.count);
+                        if(ext < 3 || __atomic_load_n(&lanes.caches[other / READYQ_SHARD_FACTOR].id, __ATOMIC_RELAXED) == this_cache) {
+                                proc->rdq.target = other;
+                        }
+                }
+                else {
+                        const unsigned target = proc->rdq.target;
+                        __cfadbg_print_safe(ready_queue, "Kernel : %u considering helping %u, tcsc %llu\n", this, target, lanes.tscs[target].tv);
+                        /* paranoid */ verify( lanes.tscs[target].tv != MAX );
+                        if(target < lanes.count) {
+                                const unsigned long long cutoff = calc_cutoff(ctsc, proc, cltr->ready_queue);
+                                const unsigned long long age = moving_average(ctsc, lanes.tscs[target].tv, lanes.tscs[target].ma);
+                                __cfadbg_print_safe(ready_queue, "Kernel : Help attempt on %u from %u, age %'llu vs cutoff %'llu, %s\n", target, this, age, cutoff, age > cutoff ? "yes" : "no");
+                                if(age > cutoff) {
+                                        thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help));
+                                        if(t) return t;
+                                }
+                        }
+                        proc->rdq.target = MAX;
+                }
+                for(READYQ_SHARD_FACTOR) {
+                        unsigned i = this + (proc->rdq.itr++ % READYQ_SHARD_FACTOR);
+                        if(thread$ * t = try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.local))) return t;
+                }
+                // All lanes where empty return 0p
+                return 0p;
+        }
+        __attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) with (cltr->ready_queue) {
+                unsigned i = __tls_rand() % lanes.count;
+                return try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.steal));
+        }
+        __attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) {
+                return search(cltr);
+        }
+#endif
+#if defined(USE_CPU_WORK_STEALING)
+        __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) {
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushing %p on cluster %p\n", thrd, cltr);
+                processor * const proc = kernelTLS().this_processor;
+                const bool external = (!proc) || (cltr != proc->cltr);
+                // Figure out the current cpu and make sure it is valid
+                const int cpu = __kernel_getcpu();
+                /* paranoid */ verify(cpu >= 0);
+                /* paranoid */ verify(cpu < cpu_info.hthrd_count);
+                /* paranoid */ verify(cpu * READYQ_SHARD_FACTOR < lanes.count);
+                // Figure out where thread was last time and make sure it's
+                /* paranoid */ verify(thrd->preferred >= 0);
+                /* paranoid */ verify(thrd->preferred < cpu_info.hthrd_count);
+                /* paranoid */ verify(thrd->preferred * READYQ_SHARD_FACTOR < lanes.count);
+                const int prf = thrd->preferred * READYQ_SHARD_FACTOR;
+                const cpu_map_entry_t & map;
+                choose(hint) {
+                        case UNPARK_LOCAL : &map = &cpu_info.llc_map[cpu];
+                        case UNPARK_REMOTE: &map = &cpu_info.llc_map[prf];
+                }
+                /* paranoid */ verify(map.start * READYQ_SHARD_FACTOR < lanes.count);
+                /* paranoid */ verify(map.self * READYQ_SHARD_FACTOR < lanes.count);
+                /* paranoid */ verifyf((map.start + map.count) * READYQ_SHARD_FACTOR <= lanes.count, "have %zu lanes but map can go up to %u", lanes.count, (map.start + map.count) * READYQ_SHARD_FACTOR);
+                const int start = map.self * READYQ_SHARD_FACTOR;
+                unsigned i;
+                                i = __tls_rand() % lanes_count;
+                        } while( !__atomic_try_acquire( &readyQ.data[i].lock ) );
+                }
+        } else {
                 do {
+                        unsigned r;
+                        if(unlikely(external)) { r = __tls_rand(); }
+                        else { r = proc->rdq.its++; }
+                        choose(hint) {
+                                case UNPARK_LOCAL : i = start + (r % READYQ_SHARD_FACTOR);
+                                case UNPARK_REMOTE: i = prf   + (r % READYQ_SHARD_FACTOR);
+                        }
+                        unsigned r = proc->rdq.its++;
+                        i = proc->rdq.id + (r % __shard_factor.readyq);
+                        /* paranoid */ verify( i < lanes_count );
                         // If we can't lock it retry
+                } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                // Actually push it
+                push(lanes.data[i], thrd);
+                // Unlock and return
+                __atomic_unlock( &lanes.data[i].lock );
+                #if !defined(__CFA_NO_STATISTICS__)
+                        if(unlikely(external)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);
+                        else __tls_stats()->ready.push.local.success++;
+                #endif
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushed %p on cluster %p (idx: %u, mask %llu, first %d)\n", thrd, cltr, i, used.mask[0], lane_first);
+        }
+        // Pop from the ready queue from a given cluster
+        __attribute__((hot)) thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {
+                /* paranoid */ verify( lanes.count > 0 );
+                /* paranoid */ verify( kernelTLS().this_processor );
+                processor * const proc = kernelTLS().this_processor;
+                const int cpu = __kernel_getcpu();
+                /* paranoid */ verify(cpu >= 0);
+                /* paranoid */ verify(cpu < cpu_info.hthrd_count);
+                /* paranoid */ verify(cpu * READYQ_SHARD_FACTOR < lanes.count);
+                const cpu_map_entry_t & map = cpu_info.llc_map[cpu];
+                /* paranoid */ verify(map.start * READYQ_SHARD_FACTOR < lanes.count);
+                /* paranoid */ verify(map.self * READYQ_SHARD_FACTOR < lanes.count);
+                /* paranoid */ verifyf((map.start + map.count) * READYQ_SHARD_FACTOR <= lanes.count, "have %zu lanes but map can go up to %u", lanes.count, (map.start + map.count) * READYQ_SHARD_FACTOR);
+                const int start = map.self * READYQ_SHARD_FACTOR;
+                const unsigned long long ctsc = rdtscl();
+                // Did we already have a help target
+                if(proc->rdq.target == MAX) {
+                        unsigned long long max = 0;
+                        for(i; READYQ_SHARD_FACTOR) {
+                                unsigned long long tsc = moving_average(ctsc, ts(lanes.data[start + i]), lanes.tscs[start + i].ma);
+                                if(tsc > max) max = tsc;
+                        }
+                        //  proc->rdq.cutoff = (max + 2 * max) / 2;
+                        /* paranoid */ verify(lanes.count < 65536); // The following code assumes max 65536 cores.
+                        /* paranoid */ verify(map.count < 65536); // The following code assumes max 65536 cores.
+                        if(0 == (__tls_rand() % 100)) {
+                                proc->rdq.target = __tls_rand() % lanes.count;
+                        } else {
+                                unsigned cpu_chaos = map.start + (__tls_rand() % map.count);
+                                proc->rdq.target = (cpu_chaos * READYQ_SHARD_FACTOR) + (__tls_rand() % READYQ_SHARD_FACTOR);
+                                /* paranoid */ verify(proc->rdq.target >= (map.start * READYQ_SHARD_FACTOR));
+                                /* paranoid */ verify(proc->rdq.target <  ((map.start + map.count) * READYQ_SHARD_FACTOR));
+                        }
+                        /* paranoid */ verify(proc->rdq.target != MAX);
+                }
+                else {
+                        unsigned long long max = 0;
+                        for(i; READYQ_SHARD_FACTOR) {
+                                unsigned long long tsc = moving_average(ctsc, ts(lanes.data[start + i]), lanes.tscs[start + i].ma);
+                                if(tsc > max) max = tsc;
+                        }
+                        const unsigned long long cutoff = (max + 2 * max) / 2;
+                        {
+                                unsigned target = proc->rdq.target;
+                                proc->rdq.target = MAX;
+                                lanes.help[target / READYQ_SHARD_FACTOR].tri++;
+                                if(moving_average(ctsc, lanes.tscs[target].tv, lanes.tscs[target].ma) > cutoff) {
+                                        thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help));
+                                        proc->rdq.last = target;
+                                        if(t) return t;
+                                }
+                                proc->rdq.target = MAX;
+                        }
+                        unsigned last = proc->rdq.last;
+                        if(last != MAX && moving_average(ctsc, lanes.tscs[last].tv, lanes.tscs[last].ma) > cutoff) {
+                                thread$ * t = try_pop(cltr, last __STATS(, __tls_stats()->ready.pop.help));
+                                if(t) return t;
+                        }
+                        else {
+                                proc->rdq.last = MAX;
+                        }
+                }
+                for(READYQ_SHARD_FACTOR) {
+                        unsigned i = start + (proc->rdq.itr++ % READYQ_SHARD_FACTOR);
+                        if(thread$ * t = try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.local))) return t;
+                }
+                // All lanes where empty return 0p
+                return 0p;
+        }
+        __attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) with (cltr->ready_queue) {
+                processor * const proc = kernelTLS().this_processor;
+                unsigned last = proc->rdq.last;
+                if(last != MAX) {
+                        struct thread$ * t = try_pop(cltr, last __STATS(, __tls_stats()->ready.pop.steal));
+                        if(t) return t;
+                        proc->rdq.last = MAX;
+                }
+                unsigned i = __tls_rand() % lanes.count;
+                return try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.steal));
+        }
+        __attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) {
+                return search(cltr);
+        }
+#endif
+#if defined(USE_RELAXED_FIFO)
+        //-----------------------------------------------------------------------
+        // get index from random number with or without bias towards queues
+        static inline [unsigned, bool] idx_from_r(unsigned r, unsigned preferred) {
+                unsigned i;
+                bool local;
+                unsigned rlow  = r % BIAS;
+                unsigned rhigh = r / BIAS;
+                if((0 != rlow) && preferred >= 0) {
+                        // (BIAS - 1) out of BIAS chances
+                        // Use perferred queues
+                        i = preferred + (rhigh % READYQ_SHARD_FACTOR);
+                        local = true;
+                }
+                else {
+                        // 1 out of BIAS chances
+                        // Use all queues
+                        i = rhigh;
+                        local = false;
+                }
+                return [i, local];
+        }
+        __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) {
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushing %p on cluster %p\n", thrd, cltr);
+                const bool external = (hint != UNPARK_LOCAL) || (!kernelTLS().this_processor) || (cltr != kernelTLS().this_processor->cltr);
+                /* paranoid */ verify(external || kernelTLS().this_processor->rdq.id < lanes.count );
+                bool local;
+                int preferred = external ? -1 : kernelTLS().this_processor->rdq.id;
+                // Try to pick a lane and lock it
+                unsigned i;
+                do {
+                        // Pick the index of a lane
+                        unsigned r = __tls_rand_fwd();
+                        [i, local] = idx_from_r(r, preferred);
+                        i %= __atomic_load_n( &lanes.count, __ATOMIC_RELAXED );
+                        #if !defined(__CFA_NO_STATISTICS__)
+                                if(unlikely(external)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.attempt, 1, __ATOMIC_RELAXED);
+                                else if(local) __tls_stats()->ready.push.local.attempt++;
+                                else __tls_stats()->ready.push.share.attempt++;
+                        #endif
+                        // If we can't lock it retry
+                } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                // Actually push it
+                push(lanes.data[i], thrd);
+                // Unlock and return
+                __atomic_unlock( &lanes.data[i].lock );
+                // Mark the current index in the tls rng instance as having an item
+                __tls_rand_advance_bck();
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushed %p on cluster %p (idx: %u, mask %llu, first %d)\n", thrd, cltr, i, used.mask[0], lane_first);
+                // Update statistics
+                #if !defined(__CFA_NO_STATISTICS__)
+                        if(unlikely(external)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);
+                        else if(local) __tls_stats()->ready.push.local.success++;
+                        else __tls_stats()->ready.push.share.success++;
+                #endif
+        }
+        // Pop from the ready queue from a given cluster
+        __attribute__((hot)) thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {
+                /* paranoid */ verify( lanes.count > 0 );
+                /* paranoid */ verify( kernelTLS().this_processor );
+                /* paranoid */ verify( kernelTLS().this_processor->rdq.id < lanes.count );
+                unsigned count = __atomic_load_n( &lanes.count, __ATOMIC_RELAXED );
+                int preferred = kernelTLS().this_processor->rdq.id;
+                // As long as the list is not empty, try finding a lane that isn't empty and pop from it
+                for(25) {
+                        // Pick two lists at random
+                        unsigned ri = __tls_rand_bck();
+                        unsigned rj = __tls_rand_bck();
+                        unsigned i, j;
+                        __attribute__((unused)) bool locali, localj;
+                        [i, locali] = idx_from_r(ri, preferred);
+                        [j, localj] = idx_from_r(rj, preferred);
+                        i %= count;
+                        j %= count;
+                        // try popping from the 2 picked lists
+                        struct thread$ * thrd = try_pop(cltr, i, j __STATS(, *(locali || localj ? &__tls_stats()->ready.pop.local : &__tls_stats()->ready.pop.help)));
+                        if(thrd) {
+                                return thrd;
+                        }
+                }
+                // All lanes where empty return 0p
+                return 0p;
+        }
+        __attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) { return pop_fast(cltr); }
+        __attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) {
+                return search(cltr);
+        }
+#endif
+#if defined(USE_WORK_STEALING)
+        __attribute__((hot)) void push(struct cluster * cltr, struct thread$ * thrd, unpark_hint hint) with (cltr->ready_queue) {
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushing %p on cluster %p\n", thrd, cltr);
+                // #define USE_PREFERRED
+                #if !defined(USE_PREFERRED)
+                const bool external = (hint != UNPARK_LOCAL) || (!kernelTLS().this_processor) || (cltr != kernelTLS().this_processor->cltr);
+                /* paranoid */ verify(external || kernelTLS().this_processor->rdq.id < lanes.count );
+                #else
+                        unsigned preferred = thrd->preferred;
+                        const bool external = (hint != UNPARK_LOCAL) || (!kernelTLS().this_processor) || preferred == MAX || thrd->curr_cluster != cltr;
+                        /* paranoid */ verifyf(external || preferred < lanes.count, "Invalid preferred queue %u for %u lanes", preferred, lanes.count );
+                        unsigned r = preferred % READYQ_SHARD_FACTOR;
+                        const unsigned start = preferred - r;
+                #endif
+                // Try to pick a lane and lock it
+                unsigned i;
+                do {
+                        #if !defined(__CFA_NO_STATISTICS__)
+                                if(unlikely(external)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.attempt, 1, __ATOMIC_RELAXED);
+                                else __tls_stats()->ready.push.local.attempt++;
+                        #endif
+                        if(unlikely(external)) {
+                                i = __tls_rand() % lanes.count;
+                        }
+                        else {
+                                #if !defined(USE_PREFERRED)
+                                        processor * proc = kernelTLS().this_processor;
+                                        unsigned r = proc->rdq.its++;
+                                        i =  proc->rdq.id + (r % READYQ_SHARD_FACTOR);
+                                #else
+                                        i = start + (r++ % READYQ_SHARD_FACTOR);
+                                #endif
+                        }
+                        // If we can't lock it retry
+                } while( !__atomic_try_acquire( &lanes.data[i].lock ) );
+                // Actually push it
+                push(lanes.data[i], thrd);
+                // Unlock and return
+                __atomic_unlock( &lanes.data[i].lock );
+                #if !defined(__CFA_NO_STATISTICS__)
+                        if(unlikely(external)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);
+                        else __tls_stats()->ready.push.local.success++;
+                #endif
+                __cfadbg_print_safe(ready_queue, "Kernel : Pushed %p on cluster %p (idx: %u, mask %llu, first %d)\n", thrd, cltr, i, used.mask[0], lane_first);
+        }
+        // Pop from the ready queue from a given cluster
+        __attribute__((hot)) thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {
+                /* paranoid */ verify( lanes.count > 0 );
+                /* paranoid */ verify( kernelTLS().this_processor );
+                /* paranoid */ verify( kernelTLS().this_processor->rdq.id < lanes.count );
+                processor * proc = kernelTLS().this_processor;
+                if(proc->rdq.target == MAX) {
+                        unsigned long long min = ts(lanes.data[proc->rdq.id]);
+                        for(int i = 0; i < READYQ_SHARD_FACTOR; i++) {
+                                unsigned long long tsc = ts(lanes.data[proc->rdq.id + i]);
+                                if(tsc < min) min = tsc;
+                        }
+                        proc->rdq.cutoff = min;
+                        proc->rdq.target = __tls_rand() % lanes.count;
+                }
+                else {
+                        unsigned target = proc->rdq.target;
+                        proc->rdq.target = MAX;
+                        const unsigned long long bias = 0; //2_500_000_000;
+                        const unsigned long long cutoff = proc->rdq.cutoff > bias ? proc->rdq.cutoff - bias : proc->rdq.cutoff;
+                        if(lanes.tscs[target].tv < cutoff && ts(lanes.data[target]) < cutoff) {
+                } while( !__atomic_try_acquire( &readyQ.data[i].lock ) );
+        }
+        // Actually push it
+        push(readyQ.data[i], thrd);
+        // Unlock and return
+        __atomic_unlock( &readyQ.data[i].lock );
+        #if !defined(__CFA_NO_STATISTICS__)
+                if(unlikely(external || remote)) __atomic_fetch_add(&cltr->stats->ready.push.extrn.success, 1, __ATOMIC_RELAXED);
+                else __tls_stats()->ready.push.local.success++;
+        #endif
+}
+__attribute__((hot)) struct thread$ * pop_fast(struct cluster * cltr) with (cltr->sched) {
+        const size_t lanes_count = readyQ.count;
+        /* paranoid */ verify( __shard_factor.readyq > 0 );
+        /* paranoid */ verify( lanes_count > 0 );
+        /* paranoid */ verify( kernelTLS().this_processor );
+        /* paranoid */ verify( kernelTLS().this_processor->rdq.id < lanes_count );
+        processor * const proc = kernelTLS().this_processor;
+        unsigned this = proc->rdq.id;
+        /* paranoid */ verify( this < lanes_count );
+        __cfadbg_print_safe(ready_queue, "Kernel : pop from %u\n", this);
+        // Figure out the current cache is
+        const unsigned this_cache = cache_id(cltr, this / __shard_factor.readyq);
+        const unsigned long long ctsc = rdtscl();
+        if(proc->rdq.target == MAX) {
+                uint64_t chaos = __tls_rand();
+                unsigned ext = chaos & 0xff;
+                unsigned other  = (chaos >> 8) % (lanes_count);
+                if(ext < 3 || __atomic_load_n(&caches[other / __shard_factor.readyq].id, __ATOMIC_RELAXED) == this_cache) {
+                        proc->rdq.target = other;
+                }
+        }
+        else {
+                const unsigned target = proc->rdq.target;
+                __cfadbg_print_safe(ready_queue, "Kernel : %u considering helping %u, tcsc %llu\n", this, target, readyQ.tscs[target].tv);
+                /* paranoid */ verify( readyQ.tscs[target].tv != MAX );
+                if(target < lanes_count) {
+                        const unsigned long long cutoff = calc_cutoff(ctsc, proc->rdq.id, lanes_count, cltr->sched.readyQ.data, cltr->sched.readyQ.tscs, __shard_factor.readyq);
+                        const unsigned long long age = moving_average(ctsc, readyQ.tscs[target].tv, readyQ.tscs[target].ma);
+                        __cfadbg_print_safe(ready_queue, "Kernel : Help attempt on %u from %u, age %'llu vs cutoff %'llu, %s\n", target, this, age, cutoff, age > cutoff ? "yes" : "no");
+                        if(age > cutoff) {
                                 thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help));
                                 if(t) return t;
+                        }
+                }
+                for(READYQ_SHARD_FACTOR) {
+                        unsigned i = proc->rdq.id + (proc->rdq.itr++ % READYQ_SHARD_FACTOR);
+                        if(thread$ * t = try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.local))) return t;
+                }
+                return 0p;
+        }
+        __attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) with (cltr->ready_queue) {
+                unsigned i = __tls_rand() % lanes.count;
+                return try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.steal));
+        }
+        __attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) with (cltr->ready_queue) {
+                return search(cltr);
+        }
+#endif
+                proc->rdq.target = MAX;
+        }
+        for(__shard_factor.readyq) {
+                unsigned i = this + (proc->rdq.itr++ % __shard_factor.readyq);
+                if(thread$ * t = try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.local))) return t;
+        }
+        // All lanes where empty return 0p
+        return 0p;
+}
+__attribute__((hot)) struct thread$ * pop_slow(struct cluster * cltr) {
+        unsigned i = __tls_rand() % (cltr->sched.readyQ.count);
+        return try_pop(cltr, i __STATS(, __tls_stats()->ready.pop.steal));
+}
+__attribute__((hot)) struct thread$ * pop_search(struct cluster * cltr) {
+        return search(cltr);
+}
 //=======================================================================
 …
 //-----------------------------------------------------------------------
 // try to pop from a lane given by index w
 static inline struct thread$ * try_pop(struct cluster * cltr, unsigned w __STATS(, __stats_readyQ_pop_t & stats)) with (cltr->ready_queue) {
         /* paranoid */ verify( w < lanes.count );
+static inline struct thread$ * try_pop(struct cluster * cltr, unsigned w __STATS(, __stats_readyQ_pop_t & stats)) with (cltr->sched) {
+        /* paranoid */ verify( w < readyQ.count );
         __STATS( stats.attempt++; )
         // Get relevant elements locally
         __intrusive_lane_t & lane = lanes.data[w];
+        __intrusive_lane_t & lane = readyQ.data[w];
         // If list looks empty retry
 …
         // Actually pop the list
         struct thread$ * thrd;
+        #if defined(USE_AWARE_STEALING) || defined(USE_WORK_STEALING) || defined(USE_CPU_WORK_STEALING)
+                unsigned long long tsc_before = ts(lane);
+        #endif
+        unsigned long long tsv;
+        [thrd, tsv] = pop(lane);
+        unsigned long long ts_prev = ts(lane);
+        unsigned long long ts_next;
+        [thrd, ts_next] = pop(lane);
         /* paranoid */ verify(thrd);
         /* paranoid */ verify(tsv);
+        /* paranoid */ verify(ts_next);
         /* paranoid */ verify(lane.lock);
 …
         __STATS( stats.success++; )
+        #if defined(USE_AWARE_STEALING) || defined(USE_WORK_STEALING) || defined(USE_CPU_WORK_STEALING)
+                if (tsv != MAX) {
+                        unsigned long long now = rdtscl();
+                        unsigned long long pma = __atomic_load_n(&lanes.tscs[w].ma, __ATOMIC_RELAXED);
+                        __atomic_store_n(&lanes.tscs[w].tv, tsv, __ATOMIC_RELAXED);
+                        __atomic_store_n(&lanes.tscs[w].ma, moving_average(now, tsc_before, pma), __ATOMIC_RELAXED);
+                }
+        #endif
+        #if defined(USE_AWARE_STEALING) || defined(USE_CPU_WORK_STEALING)
+                thrd->preferred = w / READYQ_SHARD_FACTOR;
+        #else
+                thrd->preferred = w;
+        #endif
+        touch_tsc(readyQ.tscs, w, ts_prev, ts_next);
+        thrd->preferred = w / __shard_factor.readyq;
         // return the popped thread
 …
 // try to pop from any lanes making sure you don't miss any threads push
 // before the start of the function
+static inline struct thread$ * search(struct cluster * cltr) with (cltr->ready_queue) {
+        /* paranoid */ verify( lanes.count > 0 );
+        unsigned count = __atomic_load_n( &lanes.count, __ATOMIC_RELAXED );
+static inline struct thread$ * search(struct cluster * cltr) {
+        const size_t lanes_count = cltr->sched.readyQ.count;
+        /* paranoid */ verify( lanes_count > 0 );
+        unsigned count = __atomic_load_n( &lanes_count, __ATOMIC_RELAXED );
         unsigned offset = __tls_rand();
         for(i; count) {
 …
 // get preferred ready for new thread
 unsigned ready_queue_new_preferred() {
         unsigned pref = 0;
+        unsigned pref = MAX;
         if(struct thread$ * thrd = publicTLS_get( this_thread )) {
                 pref = thrd->preferred;
+        }
-        else {
-                #if defined(USE_CPU_WORK_STEALING)
-                        pref = __kernel_getcpu();
-                #endif
+        }
-        #if defined(USE_CPU_WORK_STEALING)
-                /* paranoid */ verify(pref >= 0);
-                /* paranoid */ verify(pref < cpu_info.hthrd_count);
-        #endif
         return pref;
 …
 //-----------------------------------------------------------------------
-// Check that all the intrusive queues in the data structure are still consistent
-static void check( __ready_queue_t & q ) with (q) {
-        #if defined(__CFA_WITH_VERIFY__)
+                {
-                        for( idx ; lanes.count ) {
-                                __intrusive_lane_t & sl = lanes.data[idx];
-                                assert(!lanes.data[idx].lock);
-                                        if(is_empty(sl)) {
-                                                assert( sl.anchor.next == 0p );
-                                                assert( sl.anchor.ts   == -1llu );
-                                                assert( mock_head(sl)  == sl.prev );
-                                        } else {
-                                                assert( sl.anchor.next != 0p );
-                                                assert( sl.anchor.ts   != -1llu );
-                                                assert( mock_head(sl)  != sl.prev );
+                                        }
+                        }
+                }
-        #endif
+}
-//-----------------------------------------------------------------------
 // Given 2 indexes, pick the list with the oldest push an try to pop from it
 static inline struct thread$ * try_pop(struct cluster * cltr, unsigned i, unsigned j __STATS(, __stats_readyQ_pop_t & stats)) with (cltr->ready_queue) {
+static inline struct thread$ * try_pop(struct cluster * cltr, unsigned i, unsigned j __STATS(, __stats_readyQ_pop_t & stats)) with (cltr->sched) {
         // Pick the bet list
         int w = i;
         if( __builtin_expect(!is_empty(lanes.data[j]), true) ) {
                 w = (ts(lanes.data[i]) < ts(lanes.data[j])) ? i : j;
+        if( __builtin_expect(!is_empty(readyQ.data[j]), true) ) {
+                w = (ts(readyQ.data[i]) < ts(readyQ.data[j])) ? i : j;
+        }
         return try_pop(cltr, w __STATS(, stats));
+}
-// Call this function of the intrusive list was moved using memcpy
-// fixes the list so that the pointers back to anchors aren't left dangling
-static inline void fix(__intrusive_lane_t & ll) {
-                        if(is_empty(ll)) {
-                                verify(ll.anchor.next == 0p);
-                                ll.prev = mock_head(ll);
+                        }
+}
-static void assign_list(unsigned & value, dlist(processor) & list, unsigned count) {
-        processor * it = &list`first;
-        for(unsigned i = 0; i < count; i++) {
-                /* paranoid */ verifyf( it, "Unexpected null iterator, at index %u of %u\n", i, count);
-                it->rdq.id = value;
-                it->rdq.target = MAX;
-                value += READYQ_SHARD_FACTOR;
-                it = &(*it)`next;
+        }
+}
-static void reassign_cltr_id(struct cluster * cltr) {
-        unsigned preferred = 0;
-        assign_list(preferred, cltr->procs.actives, cltr->procs.total - cltr->procs.idle);
-        assign_list(preferred, cltr->procs.idles  , cltr->procs.idle );
+}
-static void fix_times( struct cluster * cltr ) with( cltr->ready_queue ) {
-        #if defined(USE_AWARE_STEALING) || defined(USE_WORK_STEALING)
-                lanes.tscs = alloc(lanes.count, lanes.tscs`realloc);
-                for(i; lanes.count) {
-                        lanes.tscs[i].tv = rdtscl();
-                        lanes.tscs[i].ma = 0;
+                }
-        #endif
+}
-#if defined(USE_CPU_WORK_STEALING)
-        // ready_queue size is fixed in this case
-        void ready_queue_grow(struct cluster * cltr) {}
-        void ready_queue_shrink(struct cluster * cltr) {}
-#else
-        // Grow the ready queue
-        void ready_queue_grow(struct cluster * cltr) {
-                size_t ncount;
-                int target = cltr->procs.total;
-                /* paranoid */ verify( ready_mutate_islocked() );
-                __cfadbg_print_safe(ready_queue, "Kernel : Growing ready queue\n");
-                // Make sure that everything is consistent
-                /* paranoid */ check( cltr->ready_queue );
-                // grow the ready queue
-                with( cltr->ready_queue ) {
-                        // Find new count
-                        // Make sure we always have atleast 1 list
-                        if(target >= 2) {
-                                ncount = target * READYQ_SHARD_FACTOR;
-                        } else {
-                                ncount = SEQUENTIAL_SHARD;
+                        }
-                        // Allocate new array (uses realloc and memcpies the data)
-                        lanes.data = alloc( ncount, lanes.data`realloc );
-                        // Fix the moved data
-                        for( idx; (size_t)lanes.count ) {
-                                fix(lanes.data[idx]);
+                        }
-                        // Construct new data
-                        for( idx; (size_t)lanes.count ~ ncount) {
-                                (lanes.data[idx]){};
+                        }
-                        // Update original
-                        lanes.count = ncount;
-                        lanes.caches = alloc( target, lanes.caches`realloc );
+                }
-                fix_times(cltr);
-                reassign_cltr_id(cltr);
-                // Make sure that everything is consistent
-                /* paranoid */ check( cltr->ready_queue );
-                __cfadbg_print_safe(ready_queue, "Kernel : Growing ready queue done\n");
-                /* paranoid */ verify( ready_mutate_islocked() );
+        }
-        // Shrink the ready queue
-        void ready_queue_shrink(struct cluster * cltr) {
-                /* paranoid */ verify( ready_mutate_islocked() );
-                __cfadbg_print_safe(ready_queue, "Kernel : Shrinking ready queue\n");
-                // Make sure that everything is consistent
-                /* paranoid */ check( cltr->ready_queue );
-                int target = cltr->procs.total;
-                with( cltr->ready_queue ) {
-                        // Remember old count
-                        size_t ocount = lanes.count;
-                        // Find new count
-                        // Make sure we always have atleast 1 list
-                        lanes.count = target >= 2 ? target * READYQ_SHARD_FACTOR: SEQUENTIAL_SHARD;
-                        /* paranoid */ verify( ocount >= lanes.count );
-                        /* paranoid */ verify( lanes.count == target * READYQ_SHARD_FACTOR || target < 2 );
-                        // for printing count the number of displaced threads
-                        #if defined(__CFA_DEBUG_PRINT__) || defined(__CFA_DEBUG_PRINT_READY_QUEUE__)
-                                __attribute__((unused)) size_t displaced = 0;
-                        #endif
-                        // redistribute old data
-                        for( idx; (size_t)lanes.count ~ ocount) {
-                                // Lock is not strictly needed but makes checking invariants much easier
-                                __attribute__((unused)) bool locked = __atomic_try_acquire(&lanes.data[idx].lock);
-                                verify(locked);
-                                // As long as we can pop from this lane to push the threads somewhere else in the queue
-                                while(!is_empty(lanes.data[idx])) {
-                                        struct thread$ * thrd;
-                                        unsigned long long _;
-                                        [thrd, _] = pop(lanes.data[idx]);
-                                        push(cltr, thrd, true);
-                                        // for printing count the number of displaced threads
-                                        #if defined(__CFA_DEBUG_PRINT__) || defined(__CFA_DEBUG_PRINT_READY_QUEUE__)
-                                                displaced++;
-                                        #endif
+                                }
-                                // Unlock the lane
-                                __atomic_unlock(&lanes.data[idx].lock);
-                                // TODO print the queue statistics here
-                                ^(lanes.data[idx]){};
+                        }
-                        __cfadbg_print_safe(ready_queue, "Kernel : Shrinking ready queue displaced %zu threads\n", displaced);
-                        // Allocate new array (uses realloc and memcpies the data)
-                        lanes.data = alloc( lanes.count, lanes.data`realloc );
-                        // Fix the moved data
-                        for( idx; (size_t)lanes.count ) {
-                                fix(lanes.data[idx]);
+                        }
-                        lanes.caches = alloc( target, lanes.caches`realloc );
+                }
-                fix_times(cltr);
-                reassign_cltr_id(cltr);
-                // Make sure that everything is consistent
-                /* paranoid */ check( cltr->ready_queue );
-                __cfadbg_print_safe(ready_queue, "Kernel : Shrinking ready queue done\n");
-                /* paranoid */ verify( ready_mutate_islocked() );
+        }
-#endif
-#if !defined(__CFA_NO_STATISTICS__)
-        unsigned cnt(const __ready_queue_t & this, unsigned idx) {
-                /* paranoid */ verify(this.lanes.count > idx);
-                return this.lanes.data[idx].cnt;
+        }
-#endif
-#if   defined(CFA_HAVE_LINUX_LIBRSEQ)
-        // No definition needed
-#elif defined(CFA_HAVE_LINUX_RSEQ_H)
-        #if defined( __x86_64 ) || defined( __i386 )
-                #define RSEQ_SIG        0x53053053
-        #elif defined( __ARM_ARCH )
-                #ifdef __ARMEB__
-                #define RSEQ_SIG    0xf3def5e7      /* udf    #24035    ; 0x5de3 (ARMv6+) */
-                #else
-                #define RSEQ_SIG    0xe7f5def3      /* udf    #24035    ; 0x5de3 */
-                #endif
-        #endif
-        extern void __disable_interrupts_hard();
-        extern void __enable_interrupts_hard();
-        static void __kernel_raw_rseq_register  (void) {
-                /* paranoid */ verify( __cfaabi_rseq.cpu_id == RSEQ_CPU_ID_UNINITIALIZED );
-                // int ret = syscall(__NR_rseq, &__cfaabi_rseq, sizeof(struct rseq), 0, (sigset_t *)0p, _NSIG / 8);
-                int ret = syscall(__NR_rseq, &__cfaabi_rseq, sizeof(struct rseq), 0, RSEQ_SIG);
-                if(ret != 0) {
-                        int e = errno;
-                        switch(e) {
-                        case EINVAL: abort("KERNEL ERROR: rseq register invalid argument");
-                        case ENOSYS: abort("KERNEL ERROR: rseq register no supported");
-                        case EFAULT: abort("KERNEL ERROR: rseq register with invalid argument");
-                        case EBUSY : abort("KERNEL ERROR: rseq register already registered");
-                        case EPERM : abort("KERNEL ERROR: rseq register sig  argument  on unregistration does not match the signature received on registration");
-                        default: abort("KERNEL ERROR: rseq register unexpected return %d", e);
+                        }
+                }
+        }
-        static void __kernel_raw_rseq_unregister(void) {
-                /* paranoid */ verify( __cfaabi_rseq.cpu_id >= 0 );
-                // int ret = syscall(__NR_rseq, &__cfaabi_rseq, sizeof(struct rseq), RSEQ_FLAG_UNREGISTER, (sigset_t *)0p, _NSIG / 8);
-                int ret = syscall(__NR_rseq, &__cfaabi_rseq, sizeof(struct rseq), RSEQ_FLAG_UNREGISTER, RSEQ_SIG);
-                if(ret != 0) {
-                        int e = errno;
-                        switch(e) {
-                        case EINVAL: abort("KERNEL ERROR: rseq unregister invalid argument");
-                        case ENOSYS: abort("KERNEL ERROR: rseq unregister no supported");
-                        case EFAULT: abort("KERNEL ERROR: rseq unregister with invalid argument");
-                        case EBUSY : abort("KERNEL ERROR: rseq unregister already registered");
-                        case EPERM : abort("KERNEL ERROR: rseq unregister sig  argument  on unregistration does not match the signature received on registration");
-                        default: abort("KERNEL ERROR: rseq unregisteunexpected return %d", e);
+                        }
+                }
+        }
-#else
-        // No definition needed
-#endif

libcfa/src/concurrency/ready_subqueue.hfa

-              rba897d21
+              r2e9b59b
 #define __CFA_NO_SCHED_STATS__
 #include "containers/queueLockFree.hfa"
+#include "limits.hfa"
 // Intrusives lanes which are used by the relaxed ready queue
 …
+}
-// Ctor
-void ?{}( __intrusive_lane_t & this ) {
-        this.lock = false;
-        this.prev = mock_head(this);
-        this.anchor.next = 0p;
-        this.anchor.ts   = -1llu;
-        #if !defined(__CFA_NO_STATISTICS__)
-                this.cnt  = 0;
-        #endif
-        // We add a boat-load of assertions here because the anchor code is very fragile
-        /* paranoid */ _Static_assert( offsetof( thread$, link ) == offsetof(__intrusive_lane_t, anchor) );
-        /* paranoid */ verify( offsetof( thread$, link ) == offsetof(__intrusive_lane_t, anchor) );
-        /* paranoid */ verify( ((uintptr_t)( mock_head(this) ) + offsetof( thread$, link )) == (uintptr_t)(&this.anchor) );
-        /* paranoid */ verify( &mock_head(this)->link.next == &this.anchor.next );
-        /* paranoid */ verify( &mock_head(this)->link.ts   == &this.anchor.ts   );
-        /* paranoid */ verify( mock_head(this)->link.next == 0p );
-        /* paranoid */ verify( mock_head(this)->link.ts   == -1llu  );
-        /* paranoid */ verify( mock_head(this) == this.prev );
-        /* paranoid */ verify( __alignof__(__intrusive_lane_t) == 128 );
-        /* paranoid */ verify( __alignof__(this) == 128 );
-        /* paranoid */ verifyf( ((intptr_t)(&this) % 128) == 0, "Expected address to be aligned %p %% 128 == %zd", &this, ((intptr_t)(&this) % 128) );
+}
-// Dtor is trivial
-void ^?{}( __intrusive_lane_t & this ) {
-        // Make sure the list is empty
-        /* paranoid */ verify( this.anchor.next == 0p );
-        /* paranoid */ verify( this.anchor.ts   == -1llu );
-        /* paranoid */ verify( mock_head(this)  == this.prev );
+}
 // Push a thread onto this lane
 // returns true of lane was empty before push, false otherwise
 …
         /* paranoid */ verify( this.lock );
         /* paranoid */ verify( node->link.next == 0p );
         /* paranoid */ verify( node->link.ts   == -1llu  );
+        /* paranoid */ verify( node->link.ts   == MAX  );
         /* paranoid */ verify( this.prev->link.next == 0p );
         /* paranoid */ verify( this.prev->link.ts   == -1llu  );
+        /* paranoid */ verify( this.prev->link.ts   == MAX  );
         if( this.anchor.next == 0p ) {
                 /* paranoid */ verify( this.anchor.next == 0p );
                 /* paranoid */ verify( this.anchor.ts   == -1llu );
+                /* paranoid */ verify( this.anchor.ts   == MAX );
                 /* paranoid */ verify( this.anchor.ts   != 0  );
                 /* paranoid */ verify( this.prev == mock_head( this ) );
         } else {
                 /* paranoid */ verify( this.anchor.next != 0p );
                 /* paranoid */ verify( this.anchor.ts   != -1llu );
+                /* paranoid */ verify( this.anchor.ts   != MAX );
                 /* paranoid */ verify( this.anchor.ts   != 0  );
                 /* paranoid */ verify( this.prev != mock_head( this ) );
 …
         /* paranoid */ verify( this.lock );
         /* paranoid */ verify( this.anchor.next != 0p );
         /* paranoid */ verify( this.anchor.ts   != -1llu );
+        /* paranoid */ verify( this.anchor.ts   != MAX );
         /* paranoid */ verify( this.anchor.ts   != 0  );
 …
         bool is_empty = this.anchor.next == 0p;
         node->link.next = 0p;
         node->link.ts   = -1llu;
+        node->link.ts   = MAX;
         #if !defined(__CFA_NO_STATISTICS__)
                 this.cnt--;
 …
         /* paranoid */ verify( node->link.next == 0p );
         /* paranoid */ verify( node->link.ts   == -1llu  );
+        /* paranoid */ verify( node->link.ts   == MAX  );
         /* paranoid */ verify( node->link.ts   != 0  );
         /* paranoid */ verify( this.anchor.ts  != 0  );

libcfa/src/concurrency/stats.cfa

-              rba897d21
+              r2e9b59b
                         stats->io.calls.drain       = 0;
                         stats->io.calls.completed   = 0;
+                        stats->io.calls.locked      = 0;
+                        stats->io.calls.helped      = 0;
                         stats->io.calls.errors.busy = 0;
                         stats->io.ops.sockread      = 0;
 …
                         tally_one( &cltr->io.calls.drain      , &proc->io.calls.drain       );
                         tally_one( &cltr->io.calls.completed  , &proc->io.calls.completed   );
+                        tally_one( &cltr->io.calls.locked     , &proc->io.calls.locked      );
+                        tally_one( &cltr->io.calls.helped     , &proc->io.calls.helped      );
                         tally_one( &cltr->io.calls.errors.busy, &proc->io.calls.errors.busy );
                         tally_one( &cltr->io.ops.sockread     , &proc->io.ops.sockread      );
 …
                                      |   " sub " | eng3(io.calls.submitted) | "/" | eng3(io.calls.flush) | "(" | ws(3, 3, avgsubs) | "/flush)"
                                      | " - cmp " | eng3(io.calls.completed) | "/" | eng3(io.calls.drain) | "(" | ws(3, 3, avgcomp) | "/drain)"
+                                     | " - cmp " | eng3(io.calls.locked) | "locked, " | eng3(io.calls.helped) | "helped"
                                      | " - " | eng3(io.calls.errors.busy) | " EBUSY";
                                 sstr | " - sub: " | eng3(io.flush.full) | "full, " | eng3(io.flush.dirty) | "drty, " | eng3(io.flush.idle) | "idle, " | eng3(io.flush.eager) | "eagr, " | eng3(io.flush.external) | "ext";

libcfa/src/concurrency/stats.hfa

rba897d21	r2e9b59b
103	103	volatile uint64_t drain;
104	104	volatile uint64_t completed;
	105	volatile uint64_t locked;
	106	volatile uint64_t helped;
105	107	volatile uint64_t flush;
106	108	volatile uint64_t submitted;

libcfa/src/concurrency/thread.cfa

rba897d21	r2e9b59b
19	19	#include "thread.hfa"
20	20
21		#include "kernel_private.hfa"
	21	#include "kernel/private.hfa"
22	22	#include "exception.hfa"
23	23

libcfa/src/containers/array.hfa

rba897d21	r2e9b59b
	1	#include <assert.h>
1	2
2	3
…	…
34	35
35	36	static inline Timmed & ?[?]( arpk(N, S, Timmed, Tbase) & a, int i ) {
	37	assert( i < N );
36	38	return (Timmed &) a.strides[i];
37	39	}
38	40
39	41	static inline Timmed & ?[?]( arpk(N, S, Timmed, Tbase) & a, unsigned int i ) {
	42	assert( i < N );
40	43	return (Timmed &) a.strides[i];
41	44	}
42	45
43	46	static inline Timmed & ?[?]( arpk(N, S, Timmed, Tbase) & a, long int i ) {
	47	assert( i < N );
44	48	return (Timmed &) a.strides[i];
45	49	}
46	50
47	51	static inline Timmed & ?[?]( arpk(N, S, Timmed, Tbase) & a, unsigned long int i ) {
	52	assert( i < N );
48	53	return (Timmed &) a.strides[i];
49	54	}

libcfa/src/device/cpu.hfa

rba897d21	r2e9b59b
13	13	// Update Count :
14	14	//
	15
	16	#pragma once
15	17
16	18	#include <stddef.h>

libcfa/src/fstream.cfa

-              rba897d21
+              r2e9b59b
 // Created On       : Wed May 27 17:56:53 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Mon Jan 10 08:45:05 2022
 // Update Count     : 513
+// Last Modified On : Sat Apr  9 14:55:54 2022
+// Update Count     : 515
 //
 …
     for ( cnt; 10 ) {
                 errno = 0;
+                disable_interrupts();
                 len = vfprintf( (FILE *)(os.file$), format, args );
+                enable_interrupts();
           if ( len != EOF || errno != EINTR ) break;            // timer interrupt ?
           if ( cnt == 9 ) abort( "ofstream fmt EINTR spinning exceeded" );
 …
     for () {                                                                                    // no check for EINTR limit waiting for keyboard input
                 errno = 0;
+                disable_interrupts();
                 len = vfscanf( (FILE *)(is.file$), format, args );
+                enable_interrupts();
           if ( len != EOF || errno != EINTR ) break;            // timer interrupt ?
     } // for

src/AST/Convert.cpp

-              rba897d21
+              r2e9b59b
 // Author           : Thierry Delisle
 // Created On       : Thu May 09 15::37::05 2019
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Wed Feb  2 13:19:22 2022
 // Update Count     : 41
+// Last Modified By : Andrew Beach
+// Last Modified On : Wed Mar 16 15:01:00 2022
+// Update Count     : 42
 //
 …
 //================================================================================================
 namespace ast {
+// This is to preserve the FindSpecialDecls hack. It does not (and perhaps should not)
+// allow us to use the same stratagy in the new ast.
+// xxx - since convert back pass works, this concern seems to be unnecessary.
+// these need to be accessed in new FixInit now
+ast::ptr<ast::Type> sizeType = nullptr;
+const ast::FunctionDecl * dereferenceOperator = nullptr;
+const ast::StructDecl   * dtorStruct = nullptr;
+const ast::FunctionDecl * dtorStructDestroy = nullptr;
+// These are the shared local information used by ConverterNewToOld and
+// ConverterOldToNew to update the global information in the two versions.
+static ast::ptr<ast::Type> sizeType = nullptr;
+static const ast::FunctionDecl * dereferenceOperator = nullptr;
+static const ast::StructDecl   * dtorStruct = nullptr;
+static const ast::FunctionDecl * dtorStructDestroy = nullptr;
+}
 …
                 decl->parent = get<AggregateDecl>().accept1( node->parent );
                 declPostamble( decl, node );
                 return nullptr;
+                return nullptr; // ??
+        }
 …
                         node->name,
                         get<Attribute>().acceptL( node->attributes ),
+                        LinkageSpec::Spec( node->linkage.val )
+                );
+                return aggregatePostamble( decl, node );
+                        LinkageSpec::Spec( node->linkage.val ),
+                        get<Type>().accept1(node->base)
+                );
+                return aggregatePostamble( decl, node ); // Node info, including members, processed in aggregatePostamble
+        }
 …
                 this->node = stmt;
                 return nullptr;
+        }
+        void clausePostamble( Statement * stmt, const ast::StmtClause * node ) {
+                stmt->location = node->location;
+                this->node = stmt;
+        }
 …
                 auto stmt = new SwitchStmt(
                         get<Expression>().accept1( node->cond ),
                         get<Statement>().acceptL( node->stmts )
+                        get<Statement>().acceptL( node->cases )
                 );
                 return stmtPostamble( stmt, node );
+        }
         const ast::Stmt * visit( const ast::CaseStmt * node ) override final {
+        const ast::CaseClause * visit( const ast::CaseClause * node ) override final {
                 if ( inCache( node ) ) return nullptr;
                 auto stmt = new CaseStmt(
 …
                         node->isDefault()
                 );
+                return stmtPostamble( stmt, node );
+                clausePostamble( stmt, node );
+                return nullptr;
+        }
 …
+        }
         const ast::Stmt * visit( const ast::CatchStmt * node ) override final {
+        const ast::CatchClause * visit( const ast::CatchClause * node ) override final {
                 if ( inCache( node ) ) return nullptr;
                 CatchStmt::Kind kind;
 …
                         break;
                 default:
                         assertf(false, "Invalid ast::CatchStmt::Kind: %d\n", node->kind);
+                        assertf(false, "Invalid ast::ExceptionKind: %d\n", node->kind);
+                }
                 auto stmt = new CatchStmt(
 …
                         get<Statement>().accept1( node->body )
                 );
                 return stmtPostamble( stmt, node );
+        }
         const ast::Stmt * visit( const ast::FinallyStmt * node ) override final {
+                return clausePostamble( stmt, node ), nullptr;
+        }
+        const ast::FinallyClause * visit( const ast::FinallyClause * node ) override final {
                 if ( inCache( node ) ) return nullptr;
                 auto stmt = new FinallyStmt( get<CompoundStmt>().accept1( node->body ) );
                 return stmtPostamble( stmt, node );
+                return clausePostamble( stmt, node ), nullptr;
+        }
 …
+        }
+        const ast::Expr * visit( const ast::DimensionExpr * node ) override final {
+                auto expr = visitBaseExpr( node, new DimensionExpr( node->name ) );
+                this->node = expr;
+                return nullptr;
+        }
         const ast::Expr * visit( const ast::AsmExpr * node ) override final {
                 auto expr = visitBaseExpr( node,
 …
                 return strict_dynamic_cast< ast::Decl * >( node );
+        }
         ConverterOldToNew() = default;
         ConverterOldToNew(const ConverterOldToNew &) = delete;
 …
                 getAccept1< ast::type, decltype( old->child ) >( old->child )
         template<typename NewT, typename OldC>
         std::vector< ast::ptr<NewT> > getAcceptV( const OldC& old ) {
 …
 #       define GET_ACCEPT_V(child, type) \
                 getAcceptV< ast::type, decltype( old->child ) >( old->child )
+#       define GET_ACCEPT_E(child, type) \
+                getAccept1< ast::type, decltype( old->base ) >( old->base )
         template<typename NewT, typename OldC>
 …
+        }
+        // Convert SynTree::EnumDecl to AST::EnumDecl
         virtual void visit( const EnumDecl * old ) override final {
                 if ( inCache( old ) ) return;
 …
                         old->name,
                         GET_ACCEPT_V(attributes, Attribute),
+                        { old->linkage.val }
+                        { old->linkage.val },
+                        GET_ACCEPT_1(base, Type),
+                        old->enumValues
                 );
                 cache.emplace( old, decl );
 …
                 decl->uniqueId   = old->uniqueId;
                 decl->storage    = { old->storageClasses.val };
                 this->node = decl;
+        }
 …
                         old->location,
                         GET_ACCEPT_1(condition, Expr),
                         GET_ACCEPT_V(statements, Stmt),
+                        GET_ACCEPT_V(statements, CaseClause),
                         GET_LABELS_V(old->labels)
                 );
 …
         virtual void visit( const CaseStmt * old ) override final {
                 if ( inCache( old ) ) return;
                 this->node = new ast::CaseStmt(
+                this->node = new ast::CaseClause(
                         old->location,
                         GET_ACCEPT_1(condition, Expr),
+                        GET_ACCEPT_V(stmts, Stmt),
+                        GET_LABELS_V(old->labels)
+                );
+                        GET_ACCEPT_V(stmts, Stmt)
+                );
+                auto labels = GET_LABELS_V(old->labels);
+                assertf(labels.empty(), "Labels found on CaseStmt.");
                 cache.emplace( old, this->node );
+        }
 …
                         old->location,
                         GET_ACCEPT_1(block, CompoundStmt),
                         GET_ACCEPT_V(handlers, CatchStmt),
                         GET_ACCEPT_1(finallyBlock, FinallyStmt),
+                        GET_ACCEPT_V(handlers, CatchClause),
+                        GET_ACCEPT_1(finallyBlock, FinallyClause),
                         GET_LABELS_V(old->labels)
                 );
 …
+                }
                 this->node = new ast::CatchStmt(
+                this->node = new ast::CatchClause(
                         old->location,
                         kind,
                         GET_ACCEPT_1(decl, Decl),
                         GET_ACCEPT_1(cond, Expr),
+                        GET_ACCEPT_1(body, Stmt),
+                        GET_LABELS_V(old->labels)
+                );
+                        GET_ACCEPT_1(body, Stmt)
+                );
+                auto labels = GET_LABELS_V(old->labels);
+                assertf(labels.empty(), "Labels found on CatchStmt.");
                 cache.emplace( old, this->node );
+        }
 …
         virtual void visit( const FinallyStmt * old ) override final {
                 if ( inCache( old ) ) return;
+                this->node = new ast::FinallyStmt(
+                        old->location,
+                        GET_ACCEPT_1(block, CompoundStmt),
+                        GET_LABELS_V(old->labels)
+                );
+                this->node = new ast::FinallyClause(
+                        old->location,
+                        GET_ACCEPT_1(block, CompoundStmt)
+                );
+                auto labels = GET_LABELS_V(old->labels);
+                assertf(labels.empty(), "Labels found on FinallyStmt.");
                 cache.emplace( old, this->node );
+        }
 …
         virtual void visit( const DimensionExpr * old ) override final {
+                // DimensionExpr gets desugared away in Validate.
+                // As long as new-AST passes don't use it, this cheap-cheerful error
+                // detection helps ensure that these occurrences have been compiled
+                // away, as expected.  To move the DimensionExpr boundary downstream
+                // or move the new-AST translation boundary upstream, implement
+                // DimensionExpr in the new AST and implement a conversion.
+                (void) old;
+                assert(false && "DimensionExpr should not be present at new-AST boundary");
+                this->node = visitBaseExpr( old,
+                        new ast::DimensionExpr( old->location, old->name )
+                );
+        }
 …
                 for (auto & param : foralls) {
                         ty->forall.emplace_back(new ast::TypeInstType(param->name, param));
+                        ty->forall.emplace_back(new ast::TypeInstType(param));
                         for (auto asst : param->assertions) {
                                 ty->assertions.emplace_back(new ast::VariableExpr({}, asst));
 …
+        }
         virtual void visit( const EnumInstType * old ) override final {
                 ast::EnumInstType * ty;
+        virtual void visit( const EnumInstType * old ) override final { // Here is visiting the EnumInst Decl not the usage.
+                ast::EnumInstType * ty;
                 if ( old->baseEnum ) {
                         ty = new ast::EnumInstType{
+                        ty = new ast::EnumInstType{ // Probably here: missing the specification of the base
                                 GET_ACCEPT_1( baseEnum, EnumDecl ),
                                 cv( old ),

src/AST/Decl.cpp

-              rba897d21
+              r2e9b59b
+        }
         for (auto & tp : this->type_params) {
                 ftype->forall.emplace_back(new TypeInstType(tp->name, tp));
+                ftype->forall.emplace_back(new TypeInstType(tp));
                 for (auto & ap: tp->assertions) {
                         ftype->assertions.emplace_back(new VariableExpr(loc, ap));
 …
         auto it = enumValues.find( enumerator->name );
         if ( it != enumValues.end() ) {
+                value = it->second;
+                // Handle typed enum by casting the value in (C++) compiler
+                // if ( base ) { // A typed enum
+                //      if ( const BasicType * bt = dynamic_cast<const BasicType *>(base) ) {
+                //              switch( bt->kind ) {
+                //                      case BasicType::Kind::Bool:     value = (bool) it->second; break;
+                //                      case BasicType::Kind::Char: value = (char) it->second; break;
+                //                      case BasicType::Kind::SignedChar: value = (signed char) it->second; break;
+                //                      case BasicType::Kind::UnsignedChar: value = (unsigned char) it->second; break;
+                //                      case BasicType::Kind::ShortSignedInt: value = (short signed int) it->second; break;
+                //                      case BasicType::Kind::SignedInt: value = (signed int) it->second; break;
+                //                      case BasicType::Kind::UnsignedInt: value = (unsigned int) it->second; break;
+                //                      case BasicType::Kind::LongSignedInt: value = (long signed int) it->second; break;
+                //                      case BasicType::Kind::LongUnsignedInt: value = (long unsigned int) it->second; break;
+                //                      case BasicType::Kind::LongLongSignedInt: value = (long long signed int) it->second; break;
+                //                      case BasicType::Kind::LongLongUnsignedInt: value = (long long unsigned int) it->second; break;
+                //                      // TODO: value should be able to handle long long unsigned int
+                //                      default:
+                //                      value = it->second;
+                //              }
+                //      }
+                // } else {
+                        value = it->second;
+                //}
                 return true;
+        }

src/AST/Decl.hpp

-              rba897d21
+              r2e9b59b
 class EnumDecl final : public AggregateDecl {
 public:
+        ptr<Type> base;
         EnumDecl( const CodeLocation& loc, const std::string& name,
+                std::vector<ptr<Attribute>>&& attrs = {}, Linkage::Spec linkage = Linkage::Cforall )
+        : AggregateDecl( loc, name, std::move(attrs), linkage ), enumValues() {}
+                std::vector<ptr<Attribute>>&& attrs = {}, Linkage::Spec linkage = Linkage::Cforall, Type * base = nullptr,
+                 std::unordered_map< std::string, long long > enumValues = std::unordered_map< std::string, long long >() )
+        : AggregateDecl( loc, name, std::move(attrs), linkage ), base(base), enumValues(enumValues) {}
         /// gets the integer value for this enumerator, returning true iff value found
+        // Maybe it is not used in producing the enum value
         bool valueOf( const Decl * enumerator, long long& value ) const;
 …
         const char * typeString() const override { return aggrString( Enum ); }
+        bool isTyped() {return base && base.get();}
 private:

src/AST/Expr.hpp

-              rba897d21
+              r2e9b59b
 };
+class DimensionExpr final : public Expr {
+public:
+        std::string name;
+        DimensionExpr( const CodeLocation & loc, std::string name )
+        : Expr( loc ), name( name ) {}
+        const Expr * accept( Visitor & v ) const override { return v.visit( this ); }
+private:
+        DimensionExpr * clone() const override { return new DimensionExpr{ *this }; }
+        MUTATE_FRIEND
+};
 /// A GCC "asm constraint operand" used in an asm statement, e.g. `[output] "=f" (result)`.
 /// https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Machine-Constraints.html#Machine-Constraints

src/AST/Fwd.hpp

-              rba897d21
+              r2e9b59b
 class ForStmt;
 class SwitchStmt;
 class CaseStmt;
+class CaseClause;
 class BranchStmt;
 class ReturnStmt;
 class ThrowStmt;
 class TryStmt;
 class CatchStmt;
 class FinallyStmt;
+class CatchClause;
+class FinallyClause;
 class SuspendStmt;
 class WaitForStmt;
 …
 class CommaExpr;
 class TypeExpr;
+class DimensionExpr;
 class AsmExpr;
 class ImplicitCopyCtorExpr;
 …
 class TranslationUnit;
+// TODO: Get from the TranslationUnit:
+extern ptr<Type> sizeType;
+extern const FunctionDecl * dereferenceOperator;
+extern const StructDecl   * dtorStruct;
+extern const FunctionDecl * dtorStructDestroy;
+class TranslationGlobal;
+}

src/AST/GenericSubstitution.cpp

rba897d21	r2e9b59b
45	45	visit_children = false;
46	46	const AggregateDecl * aggr = ty->aggr();
47		sub = TypeSubstitution~~{ aggr->params.begin(), aggr->params.end(), ty->params.begin() }~~;
	47	sub = TypeSubstitution( aggr->params, ty->params );
48	48	}
49	49

src/AST/Node.cpp

-              rba897d21
+              r2e9b59b
 // Author           : Thierry Delisle
 // Created On       : Thu May 16 14:16:00 2019
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Tue Feb  1 09:09:39 2022
 // Update Count     : 3
+// Last Modified By : Andrew Beach
+// Last Modified On : Fri Mar 25 10:30:00 2022
+// Update Count     : 4
 //
 …
 #include <csignal>  // MEMORY DEBUG -- for raise
 #include <iostream>
+#include <utility>
 #include "Attribute.hpp"
 …
 void ast::ptr_base<node_t, ref_t>::_check() const {
         // if(node) assert(node->was_ever_strong == false || node->strong_count > 0);
+}
+template< typename node_t, enum ast::Node::ref_type ref_t >
+void ast::ptr_base<node_t, ref_t>::swap( ptr_base & other ) noexcept {
+        std::swap( this->node, other.node );
+        _trap( this->node );
+        _trap( other.node );
+}
 …
 template class ast::ptr_base< ast::SwitchStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::SwitchStmt, ast::Node::ref_type::strong >;
 template class ast::ptr_base< ast::CaseStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::CaseStmt, ast::Node::ref_type::strong >;
+template class ast::ptr_base< ast::CaseClause, ast::Node::ref_type::weak >;
+template class ast::ptr_base< ast::CaseClause, ast::Node::ref_type::strong >;
 template class ast::ptr_base< ast::BranchStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::BranchStmt, ast::Node::ref_type::strong >;
 …
 template class ast::ptr_base< ast::TryStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::TryStmt, ast::Node::ref_type::strong >;
 template class ast::ptr_base< ast::CatchStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::CatchStmt, ast::Node::ref_type::strong >;
 template class ast::ptr_base< ast::FinallyStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::FinallyStmt, ast::Node::ref_type::strong >;
+template class ast::ptr_base< ast::CatchClause, ast::Node::ref_type::weak >;
+template class ast::ptr_base< ast::CatchClause, ast::Node::ref_type::strong >;
+template class ast::ptr_base< ast::FinallyClause, ast::Node::ref_type::weak >;
+template class ast::ptr_base< ast::FinallyClause, ast::Node::ref_type::strong >;
 template class ast::ptr_base< ast::WaitForStmt, ast::Node::ref_type::weak >;
 template class ast::ptr_base< ast::WaitForStmt, ast::Node::ref_type::strong >;

src/AST/Node.hpp

-              rba897d21
+              r2e9b59b
 // Created On       : Wed May 8 10:27:04 2019
 // Last Modified By : Andrew Beach
 // Last Modified On : Fri Jun 5 9:47:00 2020
 // Update Count     : 6
+// Last Modified On : Fri Mar 25 10:33:00 2022
+// Update Count     : 7
 //
 …
 /// Mutate a node field (only clones if not equal to existing value)
 template<typename node_t, typename parent_t, typename field_t, typename assn_t>
 const node_t * mutate_field( const node_t * node, field_t parent_t::* field, assn_t && val ) {
+template<typename node_t, typename super_t, typename field_t, typename assn_t>
+const node_t * mutate_field( const node_t * node, field_t super_t::* field, assn_t && val ) {
         // skip mutate if equivalent
         if ( node->*field == val ) return node;
 …
 /// Mutate a single index of a node field (only clones if not equal to existing value)
 template<typename node_t, typename parent_t, typename coll_t, typename ind_t, typename field_t>
+template<typename node_t, typename super_t, typename coll_t, typename ind_t, typename field_t>
 const node_t * mutate_field_index(
         const node_t * node, coll_t parent_t::* field, ind_t i, field_t && val
+        const node_t * node, coll_t super_t::* field, ind_t i, field_t && val
 ) {
         // skip mutate if equivalent
 …
 /// Mutate an entire indexed collection by cloning to accepted value
 template<typename node_t, typename parent_t, typename coll_t>
 const node_t * mutate_each( const node_t * node, coll_t parent_t::* field, Visitor & v ) {
+template<typename node_t, typename super_t, typename coll_t>
+const node_t * mutate_each( const node_t * node, coll_t super_t::* field, Visitor & v ) {
         for ( unsigned i = 0; i < (node->*field).size(); ++i ) {
                 node = mutate_field_index( node, field, i, (node->*field)[i]->accept( v ) );
 …
+        }
+        /// Swaps the nodes contained within two pointers.
+        void swap( ptr_base & other ) noexcept;
         const node_t * get() const { _check(); return  node; }
         const node_t * operator->() const { _check(); return  node; }
 …
 template< typename node_t >
 using readonly = ptr_base< node_t, Node::ref_type::weak >;
+/// Non-member swap that an participate in overload resolution.
+template< typename node_t, enum Node::ref_type ref_t >
+void swap( ptr_base< node_t, ref_t > & l, ptr_base< node_t, ref_t > & r ) {
+        l.swap( r );
+}
+}

src/AST/Pass.hpp

-              rba897d21
+              r2e9b59b
         const ast::Stmt *             visit( const ast::ForStmt              * ) override final;
         const ast::Stmt *             visit( const ast::SwitchStmt           * ) override final;
         const ast::Stmt *             visit( const ast::CaseStmt             * ) override final;
+        const ast::CaseClause *       visit( const ast::CaseClause           * ) override final;
         const ast::Stmt *             visit( const ast::BranchStmt           * ) override final;
         const ast::Stmt *             visit( const ast::ReturnStmt           * ) override final;
         const ast::Stmt *             visit( const ast::ThrowStmt            * ) override final;
         const ast::Stmt *             visit( const ast::TryStmt              * ) override final;
         const ast::Stmt *             visit( const ast::CatchStmt            * ) override final;
         const ast::Stmt *             visit( const ast::FinallyStmt          * ) override final;
+        const ast::CatchClause *      visit( const ast::CatchClause          * ) override final;
+        const ast::FinallyClause *    visit( const ast::FinallyClause        * ) override final;
         const ast::Stmt *             visit( const ast::SuspendStmt          * ) override final;
         const ast::Stmt *             visit( const ast::WaitForStmt          * ) override final;
 …
         const ast::Expr *             visit( const ast::CommaExpr            * ) override final;
         const ast::Expr *             visit( const ast::TypeExpr             * ) override final;
+        const ast::Expr *             visit( const ast::DimensionExpr        * ) override final;
         const ast::Expr *             visit( const ast::AsmExpr              * ) override final;
         const ast::Expr *             visit( const ast::ImplicitCopyCtorExpr * ) override final;

src/AST/Pass.impl.hpp

-              rba897d21
+              r2e9b59b
                         // Take all the elements that are different in 'values'
                         // and swap them into 'container'
                         if( values[i] != nullptr ) std::swap(container[i], values[i]);
+                        if( values[i] != nullptr ) swap(container[i], values[i]);
+                }
 …
         template< typename core_t >
         template<typename node_t, typename parent_t, typename child_t>
+        template<typename node_t, typename super_t, typename field_t>
         void ast::Pass< core_t >::maybe_accept(
                 const node_t * & parent,
                 child_t parent_t::*child
+                field_t super_t::*field
         ) {
                 static_assert( std::is_base_of<parent_t, node_t>::value, "Error deducing member object" );
                 if(__pass::skip(parent->*child)) return;
                 const auto & old_val = __pass::get(parent->*child, 0);
+                static_assert( std::is_base_of<super_t, node_t>::value, "Error deducing member object" );
+                if(__pass::skip(parent->*field)) return;
+                const auto & old_val = __pass::get(parent->*field, 0);
                 static_assert( !std::is_same<const ast::Node * &, decltype(old_val)>::value, "ERROR");
 …
                 if( new_val.differs ) {
                         auto new_parent = __pass::mutate<core_t>(parent);
                         new_val.apply(new_parent, child);
+                        new_val.apply(new_parent, field);
                         parent = new_parent;
+                }
 …
         template< typename core_t >
         template<typename node_t, typename parent_t, typename child_t>
+        template<typename node_t, typename super_t, typename field_t>
         void ast::Pass< core_t >::maybe_accept_as_compound(
                 const node_t * & parent,
                 child_t parent_t::*child
+                field_t super_t::*child
         ) {
                 static_assert( std::is_base_of<parent_t, node_t>::value, "Error deducing member object" );
+                static_assert( std::is_base_of<super_t, node_t>::value, "Error deducing member object" );
                 if(__pass::skip(parent->*child)) return;
 …
                         __pass::symtab::addId( core, 0, func );
                         if ( __visit_children() ) {
+                                // parameter declarations
+                                maybe_accept( node, &FunctionDecl::type_params );
+                                maybe_accept( node, &FunctionDecl::assertions );
                                 maybe_accept( node, &FunctionDecl::params );
                                 maybe_accept( node, &FunctionDecl::returns );
+                                // type params and assertions
+                                maybe_accept( node, &FunctionDecl::type_params );
+                                maybe_accept( node, &FunctionDecl::assertions );
+                                maybe_accept( node, &FunctionDecl::type );
                                 // First remember that we are now within a function.
                                 ValueGuard< bool > oldInFunction( inFunction );
 …
         if ( __visit_children() ) {
                 maybe_accept( node, &SwitchStmt::cond  );
                 maybe_accept( node, &SwitchStmt::stmts );
+                maybe_accept( node, &SwitchStmt::cases );
+        }
 …
 //--------------------------------------------------------------------------
 // CaseStmt
 template< typename core_t >
 const ast::Stmt * ast::Pass< core_t >::visit( const ast::CaseStmt * node ) {
         VISIT_START( node );
         if ( __visit_children() ) {
                 maybe_accept( node, &CaseStmt::cond  );
                 maybe_accept( node, &CaseStmt::stmts );
+        }
         VISIT_END( Stmt, node );
+// CaseClause
+template< typename core_t >
+const ast::CaseClause * ast::Pass< core_t >::visit( const ast::CaseClause * node ) {
+        VISIT_START( node );
+        if ( __visit_children() ) {
+                maybe_accept( node, &CaseClause::cond  );
+                maybe_accept( node, &CaseClause::stmts );
+        }
+        VISIT_END( CaseClause, node );
+}
 …
 //--------------------------------------------------------------------------
 // CatchStmt
 template< typename core_t >
 const ast::Stmt * ast::Pass< core_t >::visit( const ast::CatchStmt * node ) {
+// CatchClause
+template< typename core_t >
+const ast::CatchClause * ast::Pass< core_t >::visit( const ast::CatchClause * node ) {
         VISIT_START( node );
 …
                 // catch statements introduce a level of scope (for the caught exception)
                 guard_symtab guard { *this };
                 maybe_accept( node, &CatchStmt::decl );
                 maybe_accept( node, &CatchStmt::cond );
                 maybe_accept_as_compound( node, &CatchStmt::body );
+        }
         VISIT_END( Stmt, node );
+}
 //--------------------------------------------------------------------------
 // FinallyStmt
 template< typename core_t >
 const ast::Stmt * ast::Pass< core_t >::visit( const ast::FinallyStmt * node ) {
         VISIT_START( node );
         if ( __visit_children() ) {
                 maybe_accept( node, &FinallyStmt::body );
+        }
         VISIT_END( Stmt, node );
+                maybe_accept( node, &CatchClause::decl );
+                maybe_accept( node, &CatchClause::cond );
+                maybe_accept_as_compound( node, &CatchClause::body );
+        }
+        VISIT_END( CatchClause, node );
+}
+//--------------------------------------------------------------------------
+// FinallyClause
+template< typename core_t >
+const ast::FinallyClause * ast::Pass< core_t >::visit( const ast::FinallyClause * node ) {
+        VISIT_START( node );
+        if ( __visit_children() ) {
+                maybe_accept( node, &FinallyClause::body );
+        }
+        VISIT_END( FinallyClause, node );
+}
 …
                         auto n = __pass::mutate<core_t>(node);
                         for(size_t i = 0; i < new_clauses.size(); i++) {
                                 if(new_clauses.at(i).target.func != nullptr) std::swap(n->clauses.at(i).target.func, new_clauses.at(i).target.func);
+                                if(new_clauses.at(i).target.func != nullptr) swap(n->clauses.at(i).target.func, new_clauses.at(i).target.func);
                                 for(size_t j = 0; j < new_clauses.at(i).target.args.size(); j++) {
                                         if(new_clauses.at(i).target.args.at(j) != nullptr) std::swap(n->clauses.at(i).target.args.at(j), new_clauses.at(i).target.args.at(j));
+                                        if(new_clauses.at(i).target.args.at(j) != nullptr) swap(n->clauses.at(i).target.args.at(j), new_clauses.at(i).target.args.at(j));
+                                }
                                 if(new_clauses.at(i).stmt != nullptr) std::swap(n->clauses.at(i).stmt, new_clauses.at(i).stmt);
                                 if(new_clauses.at(i).cond != nullptr) std::swap(n->clauses.at(i).cond, new_clauses.at(i).cond);
+                                if(new_clauses.at(i).stmt != nullptr) swap(n->clauses.at(i).stmt, new_clauses.at(i).stmt);
+                                if(new_clauses.at(i).cond != nullptr) swap(n->clauses.at(i).cond, new_clauses.at(i).cond);
+                        }
                         node = n;
 …
+                }
                 maybe_accept( node, &TypeExpr::type );
+        }
+        VISIT_END( Expr, node );
+}
+//--------------------------------------------------------------------------
+// DimensionExpr
+template< typename core_t >
+const ast::Expr * ast::Pass< core_t >::visit( const ast::DimensionExpr * node ) {
+        VISIT_START( node );
+        if ( __visit_children() ) {
+                guard_symtab guard { *this };
+                maybe_accept( node, &DimensionExpr::result );
+        }
 …
         if ( __visit_children() ) {
                 // xxx - should PointerType visit/mutate dimension?
+                maybe_accept( node, &PointerType::dimension );
                 maybe_accept( node, &PointerType::base );
+        }
 …
         if ( __visit_children() ) {
+                {
+                        bool mutated = false;
+                        std::unordered_map< ast::TypeInstType::TypeEnvKey, ast::ptr< ast::Type > > new_map;
+                        for ( const auto & p : node->typeEnv ) {
+                                guard_symtab guard { *this };
+                                auto new_node = p.second->accept( *this );
+                                if (new_node != p.second) mutated = true;
+                                new_map.insert({ p.first, new_node });
+                        }
+                        if (mutated) {
+                                auto new_node = __pass::mutate<core_t>( node );
+                                new_node->typeEnv.swap( new_map );
+                                node = new_node;
+                        }
+                bool mutated = false;
+                std::unordered_map< ast::TypeInstType::TypeEnvKey, ast::ptr< ast::Type > > new_map;
+                for ( const auto & p : node->typeEnv ) {
+                        guard_symtab guard { *this };
+                        auto new_node = p.second->accept( *this );
+                        if (new_node != p.second) mutated = true;
+                        new_map.insert({ p.first, new_node });
+                }
+                if (mutated) {
+                        auto new_node = __pass::mutate<core_t>( node );
+                        new_node->typeEnv.swap( new_map );
+                        node = new_node;
+                }
+        }

src/AST/Pass.proto.hpp

-              rba897d21
+              r2e9b59b
 struct PureVisitor;
+template<typename node_t>
+node_t * deepCopy( const node_t * localRoot );
 namespace __pass {
 …
                 static inline auto addStructFwd( core_t & core, int, const ast::StructDecl * decl ) -> decltype( core.symtab.addStruct( decl ), void() ) {
                         ast::StructDecl * fwd = new ast::StructDecl( decl->location, decl->name );
+                        fwd->params = decl->params;
+                        for ( const auto & param : decl->params ) {
+                                fwd->params.push_back( deepCopy( param.get() ) );
+                        }
                         core.symtab.addStruct( fwd );
+                }
 …
                 template<typename core_t>
                 static inline auto addUnionFwd( core_t & core, int, const ast::UnionDecl * decl ) -> decltype( core.symtab.addUnion( decl ), void() ) {
+                        UnionDecl * fwd = new UnionDecl( decl->location, decl->name );
+                        fwd->params = decl->params;
+                        ast::UnionDecl * fwd = new ast::UnionDecl( decl->location, decl->name );
+                        for ( const auto & param : decl->params ) {
+                                fwd->params.push_back( deepCopy( param.get() ) );
+                        }
                         core.symtab.addUnion( fwd );
+                }

src/AST/Print.cpp

-              rba897d21
+              r2e9b59b
+                }
+                auto ptrToEnum = dynamic_cast<const ast::EnumDecl *>(node);
+                if ( ! short_mode && ptrToEnum && ptrToEnum->base ) {
+                        os << endl << indent << ".. with (enum) base" << endl;
+                        ++indent;
+                        ptrToEnum->base->accept( *this );
+                        --indent;
+                }
                 os << endl;
+        }
 …
                 ++indent;
                 for ( const ast::Stmt * stmt : node->stmts ) {
+                for ( const ast::CaseClause * stmt : node->cases ) {
                         stmt->accept( *this );
+                }
 …
+        }
         virtual const ast::Stmt * visit( const ast::CaseStmt * node ) override final {
+        virtual const ast::CaseClause * visit( const ast::CaseClause * node ) override final {
                 if ( node->isDefault() ) {
                         os << indent << "Default ";
 …
                 os << indent-1 << "... and handlers:" << endl;
                 for ( const ast::CatchStmt * stmt : node->handlers ) {
+                for ( const ast::CatchClause * stmt : node->handlers ) {
                         os << indent;
                         stmt->accept( *this );
 …
+        }
         virtual const ast::Stmt * visit( const ast::CatchStmt * node ) override final {
+        virtual const ast::CatchClause * visit( const ast::CatchClause * node ) override final {
                 os << "Catch ";
                 switch ( node->kind ) {
 …
+        }
         virtual const ast::Stmt * visit( const ast::FinallyStmt * node ) override final {
+        virtual const ast::FinallyClause * visit( const ast::FinallyClause * node ) override final {
                 os << "Finally Statement" << endl;
                 os << indent << "... with block:" << endl;
 …
         virtual const ast::Expr * visit( const ast::TypeExpr * node ) override final {
                 safe_print( node->type );
+                postprint( node );
+                return node;
+        }
+        virtual const ast::Expr * visit( const ast::DimensionExpr * node ) override final {
+                os << "Type-Sys Value: " << node->name;
                 postprint( node );

src/AST/Stmt.hpp

-              rba897d21
+              r2e9b59b
 // Author           : Aaron B. Moss
 // Created On       : Wed May  8 13:00:00 2019
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Wed Feb  2 20:06:41 2022
 // Update Count     : 34
+// Last Modified By : Andrew Beach
+// Last Modified On : Mon Mar 28  9:50:00 2022
+// Update Count     : 35
 //
 …
   private:
         Stmt * clone() const override = 0;
+        MUTATE_FRIEND
+};
+// Base statement component node (only serves to group them).
+class StmtClause : public ParseNode {
+  public:
+        // This is for non-statements that still belong with the statements,
+        // but are not statements, usually some sort of clause. Often these can
+        // (and should) be folded into the approprate parent node, but if they
+        // cannot be, they are sub-types of this type, for organization.
+    StmtClause( const CodeLocation & loc )
+                : ParseNode(loc) {}
+  private:
+        StmtClause * clone() const override = 0;
         MUTATE_FRIEND
 };
 …
   public:
         ptr<Expr> cond;
+        std::vector<ptr<CaseClause>> cases;
+        SwitchStmt( const CodeLocation & loc, const Expr * cond,
+                                const std::vector<ptr<CaseClause>> && cases,
+                                const std::vector<Label> && labels = {} )
+                : Stmt(loc, std::move(labels)), cond(cond), cases(std::move(cases)) {}
+        const Stmt * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        SwitchStmt * clone() const override { return new SwitchStmt{ *this }; }
+        MUTATE_FRIEND
+};
+// Case label: case ...: or default:
+class CaseClause final : public StmtClause {
+  public:
+        // Null for the default label.
+        ptr<Expr> cond;
         std::vector<ptr<Stmt>> stmts;
+        SwitchStmt( const CodeLocation & loc, const Expr * cond, const std::vector<ptr<Stmt>> && stmts,
+                                const std::vector<Label> && labels = {} )
+                : Stmt(loc, std::move(labels)), cond(cond), stmts(std::move(stmts)) {}
+        const Stmt * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        SwitchStmt * clone() const override { return new SwitchStmt{ *this }; }
+        MUTATE_FRIEND
+};
+// Case label: case ...: or default:
+class CaseStmt final : public Stmt {
+  public:
+        // Null for the default label.
+        ptr<Expr> cond;
+        std::vector<ptr<Stmt>> stmts;
+        CaseStmt( const CodeLocation & loc, const Expr * cond, const std::vector<ptr<Stmt>> && stmts,
+                          const std::vector<Label> && labels = {} )
+                : Stmt(loc, std::move(labels)), cond(cond), stmts(std::move(stmts)) {}
+        CaseClause( const CodeLocation & loc, const Expr * cond, const std::vector<ptr<Stmt>> && stmts )
+                : StmtClause(loc), cond(cond), stmts(std::move(stmts)) {}
         bool isDefault() const { return !cond; }
         const Stmt * accept( Visitor & v ) const override { return v.visit( this ); }
   private:
         CaseStmt * clone() const override { return new CaseStmt{ *this }; }
+        const CaseClause * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        CaseClause * clone() const override { return new CaseClause{ *this }; }
         MUTATE_FRIEND
 };
 …
   public:
         ptr<CompoundStmt> body;
         std::vector<ptr<CatchStmt>> handlers;
         ptr<FinallyStmt> finally;
+        std::vector<ptr<CatchClause>> handlers;
+        ptr<FinallyClause> finally;
         TryStmt( const CodeLocation & loc, const CompoundStmt * body,
                          const std::vector<ptr<CatchStmt>> && handlers, const FinallyStmt * finally,
+                         const std::vector<ptr<CatchClause>> && handlers, const FinallyClause * finally,
                          const std::vector<Label> && labels = {} )
                 : Stmt(loc, std::move(labels)), body(body), handlers(std::move(handlers)), finally(finally) {}
 …
 // Catch clause of try statement
 class CatchStmt final : public Stmt {
+class CatchClause final : public StmtClause {
   public:
         ptr<Decl> decl;
 …
         ExceptionKind kind;
         CatchStmt( const CodeLocation & loc, ExceptionKind kind, const Decl * decl, const Expr * cond,
                            const Stmt * body, const std::vector<Label> && labels = {} )
                 : Stmt(loc, std::move(labels)), decl(decl), cond(cond), body(body), kind(kind) {}
         const Stmt * accept( Visitor & v ) const override { return v.visit( this ); }
   private:
         CatchStmt * clone() const override { return new CatchStmt{ *this }; }
+        CatchClause( const CodeLocation & loc, ExceptionKind kind, const Decl * decl, const Expr * cond,
+                           const Stmt * body )
+                : StmtClause(loc), decl(decl), cond(cond), body(body), kind(kind) {}
+        const CatchClause * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        CatchClause * clone() const override { return new CatchClause{ *this }; }
         MUTATE_FRIEND
 };
 // Finally clause of try statement
 class FinallyStmt final : public Stmt {
+class FinallyClause final : public StmtClause {
   public:
         ptr<CompoundStmt> body;
+        FinallyStmt( const CodeLocation & loc, const CompoundStmt * body,
+                                 std::vector<Label> && labels = {} )
+                : Stmt(loc, std::move(labels)), body(body) {}
+        const Stmt * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        FinallyStmt * clone() const override { return new FinallyStmt{ *this }; }
+        FinallyClause( const CodeLocation & loc, const CompoundStmt * body )
+                : StmtClause(loc), body(body) {}
+        const FinallyClause * accept( Visitor & v ) const override { return v.visit( this ); }
+  private:
+        FinallyClause * clone() const override { return new FinallyClause{ *this }; }
         MUTATE_FRIEND
 };

src/AST/TranslationUnit.hpp

-              rba897d21
+              r2e9b59b
 // Created On       : Tue Jun 11 15:30:00 2019
 // Last Modified By : Andrew Beach
 // Last Modified On : Tue Jun 11 15:42:00 2019
 // Update Count     : 0
+// Last Modified On : Tue Mar 11 11:19:00 2022
+// Update Count     : 1
 //
 …
 namespace ast {
+class TranslationGlobal {
+public:
+        std::map< UniqueId, Decl * > idMap;
+        ptr<Type> sizeType;
+        const FunctionDecl * dereference;
+        const StructDecl * dtorStruct;
+        const FunctionDecl * dtorDestroy;
+};
 class TranslationUnit {
 public:
         std::list< ptr< Decl > > decls;
+        struct Global {
+                std::map< UniqueId, Decl * > idMap;
+                ptr<Type> sizeType;
+                const FunctionDecl * dereference;
+                const StructDecl * dtorStruct;
+                const FunctionDecl * dtorDestroy;
+        } global;
+        TranslationGlobal global;
 };

src/AST/Type.cpp

-              rba897d21
+              r2e9b59b
 // --- TypeInstType
+TypeInstType::TypeInstType( const TypeDecl * b,
+        CV::Qualifiers q, std::vector<ptr<Attribute>> && as )
+: BaseInstType( b->name, q, move(as) ), base( b ), kind( b->kind ) {}
 void TypeInstType::set_base( const TypeDecl * b ) {
         base = b;

src/AST/Type.hpp

-              rba897d21
+              r2e9b59b
                 std::vector<ptr<Attribute>> && as = {} )
         : BaseInstType( n, q, std::move(as) ), base( b ), kind( b->kind ) {}
+        TypeInstType( const TypeDecl * b,
+                CV::Qualifiers q = {}, std::vector<ptr<Attribute>> && as = {} );
         TypeInstType( const std::string& n, TypeDecl::Kind k, CV::Qualifiers q = {},
                 std::vector<ptr<Attribute>> && as = {} )

src/AST/TypeSubstitution.hpp

-              rba897d21
+              r2e9b59b
   public:
         TypeSubstitution();
+        template< typename FormalContainer, typename ActualContainer >
+        TypeSubstitution( FormalContainer formals, ActualContainer actuals );
         template< typename FormalIterator, typename ActualIterator >
         TypeSubstitution( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin );
 …
         bool empty() const;
+        template< typename FormalContainer, typename ActualContainer >
+        void addAll( FormalContainer formals, ActualContainer actuals );
         template< typename FormalIterator, typename ActualIterator >
         void add( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin );
+        void addAll( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin );
         /// create a new TypeSubstitution using bindings from env containing all of the type variables in expr
 …
 };
+template< typename FormalContainer, typename ActualContainer >
+TypeSubstitution::TypeSubstitution( FormalContainer formals, ActualContainer actuals ) {
+        assert( formals.size() == actuals.size() );
+        addAll( formals.begin(), formals.end(), actuals.begin() );
+}
+template< typename FormalIterator, typename ActualIterator >
+TypeSubstitution::TypeSubstitution( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin ) {
+        addAll( formalBegin, formalEnd, actualBegin );
+}
+template< typename FormalContainer, typename ActualContainer >
+void TypeSubstitution::addAll( FormalContainer formals, ActualContainer actuals ) {
+        assert( formals.size() == actuals.size() );
+        addAll( formals.begin(), formals.end(), actuals.begin() );
+}
 // this is the only place where type parameters outside a function formal may be substituted.
 template< typename FormalIterator, typename ActualIterator >
 void TypeSubstitution::add( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin ) {
+void TypeSubstitution::addAll( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin ) {
         // FormalIterator points to a TypeDecl
         // ActualIterator points to a Type
 …
                         } // if
                 } else {
+                        // Is this an error?
                 } // if
         } // for
+}
-template< typename FormalIterator, typename ActualIterator >
-TypeSubstitution::TypeSubstitution( FormalIterator formalBegin, FormalIterator formalEnd, ActualIterator actualBegin ) {
-        add( formalBegin, formalEnd, actualBegin );
+}
 } // namespace ast

src/AST/Util.cpp

-              rba897d21
+              r2e9b59b
 // Created On       : Wed Jan 19  9:46:00 2022
 // Last Modified By : Andrew Beach
 // Last Modified On : Fri Feb 18  9:42:00 2022
 // Update Count     : 0
+// Last Modified On : Fri Mar 11 18:07:00 2022
+// Update Count     : 1
 //
 #include "Util.hpp"
-#include "Decl.hpp"
 #include "Node.hpp"
+#include "ParseNode.hpp"
 #include "Pass.hpp"
 #include "TranslationUnit.hpp"
-#include "Common/ScopedMap.h"
 #include <vector>
 …
 };
+/// Check that every note that can has a set CodeLocation.
+struct SetCodeLocationsCore {
+        void previsit( const ParseNode * node ) {
+                assert( node->location.isSet() );
+        }
+};
 struct InvariantCore {
         // To save on the number of visits: this is a kind of composed core.
         // None of the passes should make changes so ordering doesn't matter.
         NoStrongCyclesCore no_strong_cycles;
+        SetCodeLocationsCore set_code_locations;
         void previsit( const Node * node ) {
                 no_strong_cycles.previsit( node );
+        }
+        void previsit( const ParseNode * node ) {
+                no_strong_cycles.previsit( node );
+                set_code_locations.previsit( node );
+        }

src/AST/Visitor.hpp

-              rba897d21
+              r2e9b59b
     virtual const ast::Stmt *             visit( const ast::ForStmt              * ) = 0;
     virtual const ast::Stmt *             visit( const ast::SwitchStmt           * ) = 0;
     virtual const ast::Stmt *             visit( const ast::CaseStmt             * ) = 0;
+    virtual const ast::CaseClause *       visit( const ast::CaseClause           * ) = 0;
     virtual const ast::Stmt *             visit( const ast::BranchStmt           * ) = 0;
     virtual const ast::Stmt *             visit( const ast::ReturnStmt           * ) = 0;
     virtual const ast::Stmt *             visit( const ast::ThrowStmt            * ) = 0;
     virtual const ast::Stmt *             visit( const ast::TryStmt              * ) = 0;
     virtual const ast::Stmt *             visit( const ast::CatchStmt            * ) = 0;
     virtual const ast::Stmt *             visit( const ast::FinallyStmt          * ) = 0;
+    virtual const ast::CatchClause *      visit( const ast::CatchClause          * ) = 0;
+    virtual const ast::FinallyClause *    visit( const ast::FinallyClause        * ) = 0;
     virtual const ast::Stmt *             visit( const ast::SuspendStmt          * ) = 0;
     virtual const ast::Stmt *             visit( const ast::WaitForStmt          * ) = 0;
 …
     virtual const ast::Expr *             visit( const ast::CommaExpr            * ) = 0;
     virtual const ast::Expr *             visit( const ast::TypeExpr             * ) = 0;
+    virtual const ast::Expr *             visit( const ast::DimensionExpr        * ) = 0;
     virtual const ast::Expr *             visit( const ast::AsmExpr              * ) = 0;
     virtual const ast::Expr *             visit( const ast::ImplicitCopyCtorExpr * ) = 0;

src/CodeGen/CodeGenerator.cc

-              rba897d21
+              r2e9b59b
         void CodeGenerator::postvisit( EnumDecl * enumDecl ) {
                 extension( enumDecl );
-                output << "enum ";
-                genAttributes( enumDecl->get_attributes() );
-                output << enumDecl->get_name();
                 std::list< Declaration* > &memb = enumDecl->get_members();
+                if ( ! memb.empty() ) {
+                        output << " {" << endl;
+                        ++indent;
+                if (enumDecl->base && ! memb.empty()) {
+                        unsigned long long last_val = -1;
                         for ( std::list< Declaration* >::iterator i = memb.begin(); i != memb.end();  i++) {
                                 ObjectDecl * obj = dynamic_cast< ObjectDecl* >( *i );
                                 assert( obj );
+                                output << indent << mangleName( obj );
+                                if ( obj->get_init() ) {
+                                        output << " = ";
+                                        obj->get_init()->accept( *visitor );
+                                } // if
+                                output << "," << endl;
+                                output << "static const ";
+                                output << genType(enumDecl->base, "", options) << " ";
+                                output << mangleName( obj ) << " ";
+                                output << " = ";
+                                output << "(" << genType(enumDecl->base, "", options) << ")";
+                                if ( (BasicType *)(enumDecl->base) && ((BasicType *)(enumDecl->base))->isWholeNumber() ) {
+                                        if ( obj->get_init() ) {
+                                                obj->get_init()->accept( *visitor );
+                                                last_val = ((ConstantExpr *)(((SingleInit *)(obj->init))->value))->constant.get_ival();
+                                        } else {
+                                                output << ++last_val;
+                                        } // if
+                                } else {
+                                        if ( obj->get_init() ) {
+                                                obj->get_init()->accept( *visitor );
+                                        } else {
+                                                // Should not reach here!
+                                        }
+                                }
+                                output << ";" << endl;
                         } // for
+                } else {
+                        output << "enum ";
+                        genAttributes( enumDecl->get_attributes() );
+                        output << enumDecl->get_name();
+                        if ( ! memb.empty() ) {
+                                output << " {" << endl;
+                                ++indent;
+                                for ( std::list< Declaration* >::iterator i = memb.begin(); i != memb.end();  i++) {
+                                        ObjectDecl * obj = dynamic_cast< ObjectDecl* >( *i );
+                                        assert( obj );
+                                        output << indent << mangleName( obj );
+                                        if ( obj->get_init() ) {
+                                                output << " = ";
+                                                obj->get_init()->accept( *visitor );
+                                        } // if
+                                        output << "," << endl;
+                                } // for
                         --indent;
                         output << indent << "}";
+                        } // if
                 } // if
+        }
 …
                                 des->accept( *visitor );
                         } else {
                                 // otherwise, it has to be a ConstantExpr or CastExpr, initializing array eleemnt
+                                // otherwise, it has to be a ConstantExpr or CastExpr, initializing array element
                                 output << "[";
                                 des->accept( *visitor );
 …
                         output << opInfo->symbol;
                 } else {
+                        // if (dynamic_cast<EnumInstType *>(variableExpr->get_var()->get_type())
+                        // && dynamic_cast<EnumInstType *>(variableExpr->get_var()->get_type())->baseEnum->base) {
+                        //      output << '(' <<genType(dynamic_cast<EnumInstType *>(variableExpr->get_var()->get_type())->baseEnum->base, "", options) << ')';
+                        // }
                         output << mangleName( variableExpr->get_var() );
                 } // if

src/CodeGen/FixMain.cc

-              rba897d21
+              r2e9b59b
+}
 ObjectDecl * charStarObj() {
+ObjectDecl * makeArgvObj() {
         return new ObjectDecl(
                 "", Type::StorageClasses(), LinkageSpec::Cforall, 0,
 …
         main_type->get_returnVals().push_back( signedIntObj() );
         main_type->get_parameters().push_back( signedIntObj() );
         main_type->get_parameters().push_back( charStarObj() );
+        main_type->get_parameters().push_back( makeArgvObj() );
         return create_mangled_main_function_name( main_type );
+}

src/CodeGen/GenType.cc

-              rba897d21
+              r2e9b59b
         void GenType::postvisit( EnumInstType * enumInst ) {
+                typeString = enumInst->name + " " + typeString;
+                if ( options.genC ) typeString = "enum " + typeString;
+                if ( enumInst->baseEnum->base ) {
+                        typeString = genType(enumInst->baseEnum->base, "", options) + typeString;
+                } else {
+                        typeString = enumInst->name + " " + typeString;
+                        if ( options.genC ) {
+                                typeString = "enum " + typeString;
+                        }
+                }
                 handleQualifiers( enumInst );
+        }

src/Common/CodeLocationTools.cpp

-              rba897d21
+              r2e9b59b
 // Author           : Andrew Beach
 // Created On       : Fri Dec  4 15:42:00 2020
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Tue Feb  1 09:14:39 2022
 // Update Count     : 3
+// Last Modified By : Andrew Beach
+// Last Modified On : Mon Mar 14 15:14:00 2022
+// Update Count     : 4
 //
 …
     macro(ForStmt, Stmt) \
     macro(SwitchStmt, Stmt) \
     macro(CaseStmt, Stmt) \
+    macro(CaseClause, CaseClause) \
     macro(BranchStmt, Stmt) \
     macro(ReturnStmt, Stmt) \
     macro(ThrowStmt, Stmt) \
     macro(TryStmt, Stmt) \
     macro(CatchStmt, Stmt) \
     macro(FinallyStmt, Stmt) \
+    macro(CatchClause, CatchClause) \
+    macro(FinallyClause, FinallyClause) \
     macro(SuspendStmt, Stmt) \
     macro(WaitForStmt, Stmt) \
 …
     macro(CommaExpr, Expr) \
     macro(TypeExpr, Expr) \
+    macro(DimensionExpr, Expr) \
     macro(AsmExpr, Expr) \
     macro(ImplicitCopyCtorExpr, Expr) \
 …
 };
+class LocalFillCore : public ast::WithGuards {
+        CodeLocation const * parent;
+public:
+        LocalFillCore( CodeLocation const & location ) : parent( &location ) {
+                assert( location.isSet() );
+        }
+        template<typename node_t>
+        auto previsit( node_t const * node )
+                        -> typename std::enable_if<has_code_location<node_t>::value, node_t const *>::type {
+                if ( node->location.isSet() ) {
+                        GuardValue( parent ) = &node->location;
+                        return node;
+                } else {
+                        node_t * mut = ast::mutate( node );
+                        mut->location = *parent;
+                        return mut;
+                }
+        }
+};
 } // namespace
 …
         ast::Pass<FillCore>::run( unit );
+}
+ast::Node const * localFillCodeLocations(
+                CodeLocation const & location , ast::Node const * node ) {
+        ast::Pass<LocalFillCore> visitor( location );
+        return node->accept( visitor );
+}

src/Common/CodeLocationTools.hpp

-              rba897d21
+              r2e9b59b
 // Created On       : Fri Dec  4 15:35:00 2020
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Dec  9  9:53:00 2020
 // Update Count     : 1
+// Last Modified On : Mon Mar 14 15:14:00 2022
+// Update Count     : 2
 //
 #pragma once
+struct CodeLocation;
 namespace ast {
+        class Node;
         class TranslationUnit;
+}
 …
 // Assign a nearby code-location to any unset code locations in the forest.
 void forceFillCodeLocations( ast::TranslationUnit & unit );
+// Fill in code-locations with a parent code location,
+// using the provided CodeLocation as the base.
+ast::Node const *
+        localFillCodeLocations( CodeLocation const &, ast::Node const * );

src/Common/Eval.cc

rba897d21	r2e9b59b
112	112	}
113	113
114		void postvisit( const ast::VariableExpr * expr ) {
	114	void postvisit( const ast::VariableExpr * expr ) { // No hit
115	115	if ( const ast::EnumInstType * inst = dynamic_cast<const ast::EnumInstType *>(expr->result.get()) ) {
116	116	if ( const ast::EnumDecl * decl = inst->base ) {

src/Common/Examine.cc

-              rba897d21
+              r2e9b59b
 // file "LICENCE" distributed with Cforall.
 //
 // Examine.h --
+// Examine.cc -- Helpers for examining AST code.
 //
 // Author           : Andrew Beach
 // Created On       : Wed Sept 2 14:02 2020
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Sep  8 12:15 2020
 // Update Count     : 0
+// Last Modified On : Fri Dec 10 10:27 2021
+// Update Count     : 1
 //
 #include "Common/Examine.h"
+#include "AST/Type.hpp"
 #include "CodeGen/OperatorTable.h"
+#include "InitTweak/InitTweak.h"
 DeclarationWithType * isMainFor( FunctionDecl * func, AggregateDecl::Aggregate kind ) {
 …
 namespace {
+// getTypeofThis but does some extra checks used in this module.
+const ast::Type * getTypeofThisSolo( const ast::FunctionDecl * func ) {
+        if ( 1 != func->params.size() ) {
+                return nullptr;
+        }
+        auto ref = func->type->params.front().as<ast::ReferenceType>();
+        return (ref) ? ref->base : nullptr;
+}
+}
+const ast::DeclWithType * isMainFor(
+                const ast::FunctionDecl * func, ast::AggregateDecl::Aggregate kind ) {
+        if ( "main" != func->name ) return nullptr;
+        if ( 1 != func->params.size() ) return nullptr;
+        auto param = func->params.front();
+        auto type = dynamic_cast<const ast::ReferenceType *>( param->get_type() );
+        if ( !type ) return nullptr;
+        auto obj = type->base.as<ast::StructInstType>();
+        if ( !obj ) return nullptr;
+        if ( kind != obj->base->kind ) return nullptr;
+        return param;
+}
+namespace {
         Type * getDestructorParam( FunctionDecl * func ) {
                 if ( !CodeGen::isDestructor( func->name ) ) return nullptr;
 …
                 return nullptr;
+        }
+const ast::Type * getDestructorParam( const ast::FunctionDecl * func ) {
+        if ( !CodeGen::isDestructor( func->name ) ) return nullptr;
+        //return InitTweak::getParamThis( func )->type;
+        return getTypeofThisSolo( func );
+}
+}
 …
         return false;
+}
+bool isDestructorFor(
+                const ast::FunctionDecl * func, const ast::StructDecl * type_decl ) {
+        if ( const ast::Type * type = getDestructorParam( func ) ) {
+                auto stype = dynamic_cast<const ast::StructInstType *>( type );
+                return stype && stype->base.get() == type_decl;
+        }
+        return false;
+}

src/Common/Examine.h

-              rba897d21
+              r2e9b59b
 // file "LICENCE" distributed with Cforall.
 //
 // Examine.h --
+// Examine.h -- Helpers for examining AST code.
 //
 // Author           : Andrew Beach
 // Created On       : Wed Sept 2 13:57 2020
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Sep  8 12:08 2020
 // Update Count     : 0
+// Last Modified On : Fri Dec 10 10:28 2021
+// Update Count     : 1
 //
+#include "AST/Decl.hpp"
 #include "SynTree/Declaration.h"
 /// Check if this is a main function for a type of an aggregate kind.
 DeclarationWithType * isMainFor( FunctionDecl * func, AggregateDecl::Aggregate kind );
+const ast::DeclWithType * isMainFor(
+        const ast::FunctionDecl * func, ast::AggregateDecl::Aggregate kind );
 // Returns a pointer to the parameter if true, nullptr otherwise.
 /// Check if this function is a destructor for the given structure.
 bool isDestructorFor( FunctionDecl * func, StructDecl * type_decl );
+bool isDestructorFor(
+        const ast::FunctionDecl * func, const ast::StructDecl * type );

src/Common/PassVisitor.impl.h

rba897d21	r2e9b59b
754	754
755	755	// unlike structs, traits, and unions, enums inject their members into the global scope
	756	// if ( node->base ) maybeAccept_impl( node->base, *this ); // Need this? Maybe not?
756	757	maybeAccept_impl( node->parameters, *this );
757	758	maybeAccept_impl( node->members , *this );

src/Concurrency/Keywords.cc

-              rba897d21
+              r2e9b59b
                                         //new TypeofType( noQualifiers, args.front()->clone() )
                                         new TypeofType( noQualifiers, new UntypedExpr(
                                                         new NameExpr( "__get_type" ),
+                                                        new NameExpr( "__get_mutexstmt_lock_type" ),
                                                         { args.front()->clone() }
+                                                )
 …
                                 map_range < std::list<Initializer*> > ( args, [](Expression * var ){
                                         return new SingleInit( new UntypedExpr(
                                                         new NameExpr( "__get_ptr" ),
+                                                        new NameExpr( "__get_mutexstmt_lock_ptr" ),
                                                         { var }
                                         ) );
 …
                 TypeExpr * lock_type_expr = new TypeExpr(
                         new TypeofType( noQualifiers, new UntypedExpr(
                                 new NameExpr( "__get_type" ),
+                                new NameExpr( "__get_mutexstmt_lock_type" ),
                                 { args.front()->clone() }
+                                )

src/Concurrency/KeywordsNew.cpp

-              rba897d21
+              r2e9b59b
 // Created On       : Tue Nov 16  9:53:00 2021
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Dec  1 11:24:00 2021
 // Update Count     : 1
+// Last Modified On : Fri Mar 11 10:40:00 2022
+// Update Count     : 2
 //
+#include <iostream>
 #include "Concurrency/Keywords.h"
 …
 #include "AST/Copy.hpp"
 #include "AST/Decl.hpp"
+#include "AST/Expr.hpp"
 #include "AST/Pass.hpp"
 #include "AST/Stmt.hpp"
+#include "AST/DeclReplacer.hpp"
 #include "AST/TranslationUnit.hpp"
 #include "CodeGen/OperatorTable.h"
+#include "Common/Examine.h"
 #include "Common/utility.h"
+#include "Common/UniqueName.h"
+#include "ControlStruct/LabelGeneratorNew.hpp"
 #include "InitTweak/InitTweak.h"
+#include "Virtual/Tables.h"
 namespace Concurrency {
 …
 namespace {
+inline static bool isThread( const ast::DeclWithType * decl ) {
+// --------------------------------------------------------------------------
+// Loose Helper Functions:
+/// Detect threads constructed with the keyword thread.
+bool isThread( const ast::DeclWithType * decl ) {
         auto baseType = decl->get_type()->stripDeclarator();
         auto instType = dynamic_cast<const ast::StructInstType *>( baseType );
 …
+}
+/// Get the virtual type id if given a type name.
+std::string typeIdType( std::string const & exception_name ) {
+        return exception_name.empty() ? std::string()
+                : Virtual::typeIdType( exception_name );
+}
+/// Get the vtable type name if given a type name.
+std::string vtableTypeName( std::string const & exception_name ) {
+        return exception_name.empty() ? std::string()
+                : Virtual::vtableTypeName( exception_name );
+}
+static ast::Type * mutate_under_references( ast::ptr<ast::Type>& type ) {
+        ast::Type * mutType = type.get_and_mutate();
+        for ( ast::ReferenceType * mutRef
+                ; (mutRef = dynamic_cast<ast::ReferenceType *>( mutType ))
+                ; mutType = mutRef->base.get_and_mutate() );
+        return mutType;
+}
+// Describe that it adds the generic parameters and the uses of the generic
+// parameters on the function and first "this" argument.
+ast::FunctionDecl * fixupGenerics(
+                const ast::FunctionDecl * func, const ast::StructDecl * decl ) {
+        const CodeLocation & location = decl->location;
+        // We have to update both the declaration
+        auto mutFunc = ast::mutate( func );
+        auto mutType = mutFunc->type.get_and_mutate();
+        if ( decl->params.empty() ) {
+                return mutFunc;
+        }
+        assert( 0 != mutFunc->params.size() );
+        assert( 0 != mutType->params.size() );
+        // Add the "forall" clause information.
+        for ( const ast::ptr<ast::TypeDecl> & typeParam : decl->params ) {
+                auto typeDecl = ast::deepCopy( typeParam );
+                mutFunc->type_params.push_back( typeDecl );
+                mutType->forall.push_back( new ast::TypeInstType( typeDecl ) );
+                for ( auto & assertion : typeDecl->assertions ) {
+                        mutFunc->assertions.push_back( assertion );
+                        mutType->assertions.emplace_back(
+                                new ast::VariableExpr( location, assertion ) );
+                }
+                typeDecl->assertions.clear();
+        }
+        // Even chain_mutate is not powerful enough for this:
+        ast::ptr<ast::Type>& paramType = strict_dynamic_cast<ast::ObjectDecl *>(
+                mutFunc->params[0].get_and_mutate() )->type;
+        auto paramTypeInst = strict_dynamic_cast<ast::StructInstType *>(
+                mutate_under_references( paramType ) );
+        auto typeParamInst = strict_dynamic_cast<ast::StructInstType *>(
+                mutate_under_references( mutType->params[0] ) );
+        for ( const ast::ptr<ast::TypeDecl> & typeDecl : mutFunc->type_params ) {
+                paramTypeInst->params.push_back(
+                        new ast::TypeExpr( location, new ast::TypeInstType( typeDecl ) ) );
+                typeParamInst->params.push_back(
+                        new ast::TypeExpr( location, new ast::TypeInstType( typeDecl ) ) );
+        }
+        return mutFunc;
+}
 // --------------------------------------------------------------------------
+struct MutexKeyword final {
+struct ConcurrentSueKeyword : public ast::WithDeclsToAdd<> {
+        ConcurrentSueKeyword(
+                std::string&& type_name, std::string&& field_name,
+                std::string&& getter_name, std::string&& context_error,
+                std::string&& exception_name,
+                bool needs_main, ast::AggregateDecl::Aggregate cast_target
+        ) :
+                type_name( type_name ), field_name( field_name ),
+                getter_name( getter_name ), context_error( context_error ),
+                exception_name( exception_name ),
+                typeid_name( typeIdType( exception_name ) ),
+                vtable_name( vtableTypeName( exception_name ) ),
+                needs_main( needs_main ), cast_target( cast_target )
+        {}
+        virtual ~ConcurrentSueKeyword() {}
+        const ast::Decl * postvisit( const ast::StructDecl * decl );
+        const ast::DeclWithType * postvisit( const ast::FunctionDecl * decl );
+        const ast::Expr * postvisit( const ast::KeywordCastExpr * expr );
+        struct StructAndField {
+                const ast::StructDecl * decl;
+                const ast::ObjectDecl * field;
+        };
+        const ast::StructDecl * handleStruct( const ast::StructDecl * );
+        void handleMain( const ast::FunctionDecl *, const ast::StructInstType * );
+        void addTypeId( const ast::StructDecl * );
+        void addVtableForward( const ast::StructDecl * );
+        const ast::FunctionDecl * forwardDeclare( const ast::StructDecl * );
+        StructAndField addField( const ast::StructDecl * );
+        void addGetRoutines( const ast::ObjectDecl *, const ast::FunctionDecl * );
+        void addLockUnlockRoutines( const ast::StructDecl * );
+private:
+        const std::string type_name;
+        const std::string field_name;
+        const std::string getter_name;
+        const std::string context_error;
+        const std::string exception_name;
+        const std::string typeid_name;
+        const std::string vtable_name;
+        const bool needs_main;
+        const ast::AggregateDecl::Aggregate cast_target;
+        const ast::StructDecl   * type_decl = nullptr;
+        const ast::FunctionDecl * dtor_decl = nullptr;
+        const ast::StructDecl * except_decl = nullptr;
+        const ast::StructDecl * typeid_decl = nullptr;
+        const ast::StructDecl * vtable_decl = nullptr;
+};
+// Handles thread type declarations:
+//
+// thread Mythread {                         struct MyThread {
+//  int data;                                  int data;
+//  a_struct_t more_data;                      a_struct_t more_data;
+//                                =>             thread$ __thrd_d;
+// };                                        };
+//                                           static inline thread$ * get_thread( MyThread * this ) { return &this->__thrd_d; }
+//
+struct ThreadKeyword final : public ConcurrentSueKeyword {
+        ThreadKeyword() : ConcurrentSueKeyword(
+                "thread$",
+                "__thrd",
+                "get_thread",
+                "thread keyword requires threads to be in scope, add #include <thread.hfa>\n",
+                "ThreadCancelled",
+                true,
+                ast::AggregateDecl::Thread )
+        {}
+        virtual ~ThreadKeyword() {}
+};
+// Handles coroutine type declarations:
+//
+// coroutine MyCoroutine {                   struct MyCoroutine {
+//  int data;                                  int data;
+//  a_struct_t more_data;                      a_struct_t more_data;
+//                                =>             coroutine$ __cor_d;
+// };                                        };
+//                                           static inline coroutine$ * get_coroutine( MyCoroutine * this ) { return &this->__cor_d; }
+//
+struct CoroutineKeyword final : public ConcurrentSueKeyword {
+        CoroutineKeyword() : ConcurrentSueKeyword(
+                "coroutine$",
+                "__cor",
+                "get_coroutine",
+                "coroutine keyword requires coroutines to be in scope, add #include <coroutine.hfa>\n",
+                "CoroutineCancelled",
+                true,
+                ast::AggregateDecl::Coroutine )
+        {}
+        virtual ~CoroutineKeyword() {}
+};
+// Handles monitor type declarations:
+//
+// monitor MyMonitor {                       struct MyMonitor {
+//  int data;                                  int data;
+//  a_struct_t more_data;                      a_struct_t more_data;
+//                                =>             monitor$ __mon_d;
+// };                                        };
+//                                           static inline monitor$ * get_coroutine( MyMonitor * this ) {
+//                                               return &this->__cor_d;
+//                                           }
+//                                           void lock(MyMonitor & this) {
+//                                               lock(get_monitor(this));
+//                                           }
+//                                           void unlock(MyMonitor & this) {
+//                                               unlock(get_monitor(this));
+//                                           }
+//
+struct MonitorKeyword final : public ConcurrentSueKeyword {
+        MonitorKeyword() : ConcurrentSueKeyword(
+                "monitor$",
+                "__mon",
+                "get_monitor",
+                "monitor keyword requires monitors to be in scope, add #include <monitor.hfa>\n",
+                "",
+                false,
+                ast::AggregateDecl::Monitor )
+        {}
+        virtual ~MonitorKeyword() {}
+};
+// Handles generator type declarations:
+//
+// generator MyGenerator {                   struct MyGenerator {
+//  int data;                                  int data;
+//  a_struct_t more_data;                      a_struct_t more_data;
+//                                =>             int __generator_state;
+// };                                        };
+//
+struct GeneratorKeyword final : public ConcurrentSueKeyword {
+        GeneratorKeyword() : ConcurrentSueKeyword(
+                "generator$",
+                "__generator_state",
+                "get_generator",
+                "Unable to find builtin type generator$\n",
+                "",
+                true,
+                ast::AggregateDecl::Generator )
+        {}
+        virtual ~GeneratorKeyword() {}
+};
+const ast::Decl * ConcurrentSueKeyword::postvisit(
+                const ast::StructDecl * decl ) {
+        if ( !decl->body ) {
+                return decl;
+        } else if ( cast_target == decl->kind ) {
+                return handleStruct( decl );
+        } else if ( type_name == decl->name ) {
+                assert( !type_decl );
+                type_decl = decl;
+        } else if ( exception_name == decl->name ) {
+                assert( !except_decl );
+                except_decl = decl;
+        } else if ( typeid_name == decl->name ) {
+                assert( !typeid_decl );
+                typeid_decl = decl;
+        } else if ( vtable_name == decl->name ) {
+                assert( !vtable_decl );
+                vtable_decl = decl;
+        }
+        return decl;
+}
+// Try to get the full definition, but raise an error on conflicts.
+const ast::FunctionDecl * getDefinition(
+                const ast::FunctionDecl * old_decl,
+                const ast::FunctionDecl * new_decl ) {
+        if ( !new_decl->stmts ) {
+                return old_decl;
+        } else if ( !old_decl->stmts ) {
+                return new_decl;
+        } else {
+                assert( !old_decl->stmts || !new_decl->stmts );
+                return nullptr;
+        }
+}
+const ast::DeclWithType * ConcurrentSueKeyword::postvisit(
+                const ast::FunctionDecl * decl ) {
+        if ( type_decl && isDestructorFor( decl, type_decl ) ) {
+                // Check for forward declarations, try to get the full definition.
+                dtor_decl = (dtor_decl) ? getDefinition( dtor_decl, decl ) : decl;
+        } else if ( !vtable_name.empty() && decl->has_body() ) {
+                if (const ast::DeclWithType * param = isMainFor( decl, cast_target )) {
+                        if ( !vtable_decl ) {
+                                SemanticError( decl, context_error );
+                        }
+                        // Should be safe because of isMainFor.
+                        const ast::StructInstType * struct_type =
+                                static_cast<const ast::StructInstType *>(
+                                        static_cast<const ast::ReferenceType *>(
+                                                param->get_type() )->base.get() );
+                        handleMain( decl, struct_type );
+                }
+        }
+        return decl;
+}
+const ast::Expr * ConcurrentSueKeyword::postvisit(
+                const ast::KeywordCastExpr * expr ) {
+        if ( cast_target == expr->target ) {
+                // Convert `(thread &)ex` to `(thread$ &)*get_thread(ex)`, etc.
+                if ( !type_decl || !dtor_decl ) {
+                        SemanticError( expr, context_error );
+                }
+                assert( nullptr == expr->result );
+                auto cast = ast::mutate( expr );
+                cast->result = new ast::ReferenceType( new ast::StructInstType( type_decl ) );
+                cast->concrete_target.field  = field_name;
+                cast->concrete_target.getter = getter_name;
+                return cast;
+        }
+        return expr;
+}
+const ast::StructDecl * ConcurrentSueKeyword::handleStruct(
+                const ast::StructDecl * decl ) {
+        assert( decl->body );
+        if ( !type_decl || !dtor_decl ) {
+                SemanticError( decl, context_error );
+        }
+        if ( !exception_name.empty() ) {
+                if( !typeid_decl || !vtable_decl ) {
+                        SemanticError( decl, context_error );
+                }
+                addTypeId( decl );
+                addVtableForward( decl );
+        }
+        const ast::FunctionDecl * func = forwardDeclare( decl );
+        StructAndField addFieldRet = addField( decl );
+        decl = addFieldRet.decl;
+        const ast::ObjectDecl * field = addFieldRet.field;
+        addGetRoutines( field, func );
+        // Add routines to monitors for use by mutex stmt.
+        if ( ast::AggregateDecl::Monitor == cast_target ) {
+                addLockUnlockRoutines( decl );
+        }
+        return decl;
+}
+void ConcurrentSueKeyword::handleMain(
+                const ast::FunctionDecl * decl, const ast::StructInstType * type ) {
+        assert( vtable_decl );
+        assert( except_decl );
+        const CodeLocation & location = decl->location;
+        std::vector<ast::ptr<ast::Expr>> poly_args = {
+                new ast::TypeExpr( location, type ),
+        };
+        ast::ObjectDecl * vtable_object = Virtual::makeVtableInstance(
+                location,
+                "_default_vtable_object_declaration",
+                new ast::StructInstType( vtable_decl, copy( poly_args ) ),
+                type,
+                nullptr
+        );
+        declsToAddAfter.push_back( vtable_object );
+        declsToAddAfter.push_back(
+                new ast::ObjectDecl(
+                        location,
+                        Virtual::concurrentDefaultVTableName(),
+                        new ast::ReferenceType( vtable_object->type, ast::CV::Const ),
+                        new ast::SingleInit( location,
+                                new ast::VariableExpr( location, vtable_object ) ),
+                        ast::Storage::Classes(),
+                        ast::Linkage::Cforall
+                )
+        );
+        declsToAddAfter.push_back( Virtual::makeGetExceptionFunction(
+                location,
+                vtable_object,
+                new ast::StructInstType( except_decl, copy( poly_args ) )
+        ) );
+}
+void ConcurrentSueKeyword::addTypeId( const ast::StructDecl * decl ) {
+        assert( typeid_decl );
+        const CodeLocation & location = decl->location;
+        ast::StructInstType * typeid_type =
+                new ast::StructInstType( typeid_decl, ast::CV::Const );
+        typeid_type->params.push_back(
+                new ast::TypeExpr( location, new ast::StructInstType( decl ) ) );
+        declsToAddBefore.push_back(
+                Virtual::makeTypeIdInstance( location, typeid_type ) );
+        // If the typeid_type is going to be kept, the other reference will have
+        // been made by now, but we also get to avoid extra mutates.
+        ast::ptr<ast::StructInstType> typeid_cleanup = typeid_type;
+}
+void ConcurrentSueKeyword::addVtableForward( const ast::StructDecl * decl ) {
+        assert( vtable_decl );
+        const CodeLocation& location = decl->location;
+        std::vector<ast::ptr<ast::Expr>> poly_args = {
+                new ast::TypeExpr( location, new ast::StructInstType( decl ) ),
+        };
+        declsToAddBefore.push_back( Virtual::makeGetExceptionForward(
+                location,
+                new ast::StructInstType( vtable_decl, copy( poly_args ) ),
+                new ast::StructInstType( except_decl, copy( poly_args ) )
+        ) );
+        ast::ObjectDecl * vtable_object = Virtual::makeVtableForward(
+                location,
+                "_default_vtable_object_declaration",
+                new ast::StructInstType( vtable_decl, std::move( poly_args ) )
+        );
+        declsToAddBefore.push_back( vtable_object );
+        declsToAddBefore.push_back(
+                new ast::ObjectDecl(
+                        location,
+                        Virtual::concurrentDefaultVTableName(),
+                        new ast::ReferenceType( vtable_object->type, ast::CV::Const ),
+                        nullptr,
+                        ast::Storage::Extern,
+                        ast::Linkage::Cforall
+                )
+        );
+}
+const ast::FunctionDecl * ConcurrentSueKeyword::forwardDeclare(
+                const ast::StructDecl * decl ) {
+        const CodeLocation & location = decl->location;
+        ast::StructDecl * forward = ast::deepCopy( decl );
+        {
+                // If removing members makes ref-count go to zero, do not free.
+                ast::ptr<ast::StructDecl> forward_ptr = forward;
+                forward->body = false;
+                forward->members.clear();
+                forward_ptr.release();
+        }
+        ast::ObjectDecl * this_decl = new ast::ObjectDecl(
+                location,
+                "this",
+                new ast::ReferenceType( new ast::StructInstType( decl ) ),
+                nullptr,
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall
+        );
+        ast::ObjectDecl * ret_decl = new ast::ObjectDecl(
+                location,
+                "ret",
+                new ast::PointerType( new ast::StructInstType( type_decl ) ),
+                nullptr,
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall
+        );
+        ast::FunctionDecl * get_decl = new ast::FunctionDecl(
+                location,
+                getter_name,
+                {}, // forall
+                { this_decl }, // params
+                { ret_decl }, // returns
+                nullptr, // stmts
+                ast::Storage::Static,
+                ast::Linkage::Cforall,
+                { new ast::Attribute( "const" ) },
+                ast::Function::Inline
+        );
+        get_decl = fixupGenerics( get_decl, decl );
+        ast::FunctionDecl * main_decl = nullptr;
+        if ( needs_main ) {
+                // `this_decl` is copied here because the original was used above.
+                main_decl = new ast::FunctionDecl(
+                        location,
+                        "main",
+                        {},
+                        { ast::deepCopy( this_decl ) },
+                        {},
+                        nullptr,
+                        ast::Storage::Classes(),
+                        ast::Linkage::Cforall
+                );
+                main_decl = fixupGenerics( main_decl, decl );
+        }
+        declsToAddBefore.push_back( forward );
+        if ( needs_main ) declsToAddBefore.push_back( main_decl );
+        declsToAddBefore.push_back( get_decl );
+        return get_decl;
+}
+ConcurrentSueKeyword::StructAndField ConcurrentSueKeyword::addField(
+                const ast::StructDecl * decl ) {
+        const CodeLocation & location = decl->location;
+        ast::ObjectDecl * field = new ast::ObjectDecl(
+                location,
+                field_name,
+                new ast::StructInstType( type_decl ),
+                nullptr,
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall
+        );
+        auto mutDecl = ast::mutate( decl );
+        mutDecl->members.push_back( field );
+        return {mutDecl, field};
+}
+void ConcurrentSueKeyword::addGetRoutines(
+                const ast::ObjectDecl * field, const ast::FunctionDecl * forward ) {
+        // Say it is generated at the "same" places as the forward declaration.
+        const CodeLocation & location = forward->location;
+        const ast::DeclWithType * param = forward->params.front();
+        ast::Stmt * stmt = new ast::ReturnStmt( location,
+                new ast::AddressExpr( location,
+                        new ast::MemberExpr( location,
+                                field,
+                                new ast::CastExpr( location,
+                                        new ast::VariableExpr( location, param ),
+                                        ast::deepCopy( param->get_type()->stripReferences() ),
+                                        ast::ExplicitCast
+                                )
+                        )
+                )
+        );
+        ast::FunctionDecl * decl = ast::deepCopy( forward );
+        decl->stmts = new ast::CompoundStmt( location, { stmt } );
+        declsToAddAfter.push_back( decl );
+}
+void ConcurrentSueKeyword::addLockUnlockRoutines(
+                const ast::StructDecl * decl ) {
+        // This should only be used on monitors.
+        assert( ast::AggregateDecl::Monitor == cast_target );
+        const CodeLocation & location = decl->location;
+        // The parameter for both routines.
+        ast::ObjectDecl * this_decl = new ast::ObjectDecl(
+                location,
+                "this",
+                new ast::ReferenceType( new ast::StructInstType( decl ) ),
+                nullptr,
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall
+        );
+        ast::FunctionDecl * lock_decl = new ast::FunctionDecl(
+                location,
+                "lock",
+                { /* forall */ },
+                {
+                        // Copy the declaration of this.
+                        ast::deepCopy( this_decl ),
+                },
+                { /* returns */ },
+                nullptr,
+                ast::Storage::Static,
+                ast::Linkage::Cforall,
+                { /* attributes */ },
+                ast::Function::Inline
+        );
+        lock_decl = fixupGenerics( lock_decl, decl );
+        lock_decl->stmts = new ast::CompoundStmt( location, {
+                new ast::ExprStmt( location,
+                        new ast::UntypedExpr( location,
+                                new ast::NameExpr( location, "lock" ),
+                                {
+                                        new ast::UntypedExpr( location,
+                                                new ast::NameExpr( location, "get_monitor" ),
+                                                { new ast::VariableExpr( location,
+                                                        InitTweak::getParamThis( lock_decl ) ) }
+                                        )
+                                }
+                        )
+                )
+        } );
+        ast::FunctionDecl * unlock_decl = new ast::FunctionDecl(
+                location,
+                "unlock",
+                { /* forall */ },
+                {
+                        // Last use, consume the declaration of this.
+                        this_decl,
+                },
+                { /* returns */ },
+                nullptr,
+                ast::Storage::Static,
+                ast::Linkage::Cforall,
+                { /* attributes */ },
+                ast::Function::Inline
+        );
+        unlock_decl = fixupGenerics( unlock_decl, decl );
+        unlock_decl->stmts = new ast::CompoundStmt( location, {
+                new ast::ExprStmt( location,
+                        new ast::UntypedExpr( location,
+                                new ast::NameExpr( location, "unlock" ),
+                                {
+                                        new ast::UntypedExpr( location,
+                                                new ast::NameExpr( location, "get_monitor" ),
+                                                { new ast::VariableExpr( location,
+                                                        InitTweak::getParamThis( unlock_decl ) ) }
+                                        )
+                                }
+                        )
+                )
+        } );
+        declsToAddAfter.push_back( lock_decl );
+        declsToAddAfter.push_back( unlock_decl );
+}
+// --------------------------------------------------------------------------
+struct SuspendKeyword final :
+                public ast::WithStmtsToAdd<>, public ast::WithGuards {
+        SuspendKeyword() = default;
+        virtual ~SuspendKeyword() = default;
+        void previsit( const ast::FunctionDecl * );
+        const ast::DeclWithType * postvisit( const ast::FunctionDecl * );
+        const ast::Stmt * postvisit( const ast::SuspendStmt * );
+private:
+        bool is_real_suspend( const ast::FunctionDecl * );
+        const ast::Stmt * make_generator_suspend( const ast::SuspendStmt * );
+        const ast::Stmt * make_coroutine_suspend( const ast::SuspendStmt * );
+        struct LabelPair {
+                ast::Label obj;
+                int idx;
+        };
+        LabelPair make_label(const ast::Stmt * stmt ) {
+                labels.push_back( ControlStruct::newLabel( "generator", stmt ) );
+                return { labels.back(), int(labels.size()) };
+        }
+        const ast::DeclWithType * in_generator = nullptr;
+        const ast::FunctionDecl * decl_suspend = nullptr;
+        std::vector<ast::Label> labels;
+};
+void SuspendKeyword::previsit( const ast::FunctionDecl * decl ) {
+        GuardValue( in_generator ); in_generator = nullptr;
+        // If it is the real suspend, grab it if we don't have one already.
+        if ( is_real_suspend( decl ) ) {
+                decl_suspend = decl_suspend ? decl_suspend : decl;
+                return;
+        }
+        // Otherwise check if this is a generator main and, if so, handle it.
+        auto param = isMainFor( decl, ast::AggregateDecl::Generator );
+        if ( !param ) return;
+        if ( 0 != decl->returns.size() ) {
+                SemanticError( decl->location, "Generator main must return void" );
+        }
+        in_generator = param;
+        GuardValue( labels ); labels.clear();
+}
+const ast::DeclWithType * SuspendKeyword::postvisit(
+                const ast::FunctionDecl * decl ) {
+        // Only modify a full definition of a generator with states.
+        if ( !decl->stmts || !in_generator || labels.empty() ) return decl;
+        const CodeLocation & location = decl->location;
+        // Create a new function body:
+        // static void * __generator_labels[] = {&&s0, &&s1, ...};
+        // void * __generator_label = __generator_labels[GEN.__generator_state];
+        // goto * __generator_label;
+        // s0: ;
+        // OLD_BODY
+        // This is the null statement inserted right before the body.
+        ast::NullStmt * noop = new ast::NullStmt( location );
+        noop->labels.push_back( ControlStruct::newLabel( "generator", noop ) );
+        const ast::Label & first_label = noop->labels.back();
+        // Add each label to the init, starting with the first label.
+        std::vector<ast::ptr<ast::Init>> inits = {
+                new ast::SingleInit( location,
+                        new ast::LabelAddressExpr( location, copy( first_label ) ) ) };
+        // Then go through all the stored labels, and clear the store.
+        for ( auto && label : labels ) {
+                inits.push_back( new ast::SingleInit( label.location,
+                        new ast::LabelAddressExpr( label.location, std::move( label )
+                        ) ) );
+        }
+        labels.clear();
+        // Then construct the initializer itself.
+        auto init = new ast::ListInit( location, std::move( inits ) );
+        ast::ObjectDecl * generatorLabels = new ast::ObjectDecl(
+                location,
+                "__generator_labels",
+                new ast::ArrayType(
+                        new ast::PointerType( new ast::VoidType() ),
+                        nullptr,
+                        ast::FixedLen,
+                        ast::DynamicDim
+                ),
+                init,
+                ast::Storage::Classes(),
+                ast::Linkage::AutoGen
+        );
+        ast::ObjectDecl * generatorLabel = new ast::ObjectDecl(
+                location,
+                "__generator_label",
+                new ast::PointerType( new ast::VoidType() ),
+                new ast::SingleInit( location,
+                        new ast::UntypedExpr( location,
+                                new ast::NameExpr( location, "?[?]" ),
+                                {
+                                        // TODO: Could be a variable expr.
+                                        new ast::NameExpr( location, "__generator_labels" ),
+                                        new ast::UntypedMemberExpr( location,
+                                                new ast::NameExpr( location, "__generator_state" ),
+                                                new ast::VariableExpr( location, in_generator )
+                                        )
+                                }
+                        )
+                ),
+                ast::Storage::Classes(),
+                ast::Linkage::AutoGen
+        );
+        ast::BranchStmt * theGoTo = new ast::BranchStmt(
+                location, new ast::VariableExpr( location, generatorLabel )
+        );
+        // The noop goes here in order.
+        ast::CompoundStmt * body = new ast::CompoundStmt( location, {
+                { new ast::DeclStmt( location, generatorLabels ) },
+                { new ast::DeclStmt( location, generatorLabel ) },
+                { theGoTo },
+                { noop },
+                { decl->stmts },
+        } );
+        auto mutDecl = ast::mutate( decl );
+        mutDecl->stmts = body;
+        return mutDecl;
+}
+const ast::Stmt * SuspendKeyword::postvisit( const ast::SuspendStmt * stmt ) {
+        switch ( stmt->type ) {
+        case ast::SuspendStmt::None:
+                // Use the context to determain the implicit target.
+                if ( in_generator ) {
+                        return make_generator_suspend( stmt );
+                } else {
+                        return make_coroutine_suspend( stmt );
+                }
+        case ast::SuspendStmt::Coroutine:
+                return make_coroutine_suspend( stmt );
+        case ast::SuspendStmt::Generator:
+                // Generator suspends must be directly in a generator.
+                if ( !in_generator ) SemanticError( stmt->location, "'suspend generator' must be used inside main of generator type." );
+                return make_generator_suspend( stmt );
+        }
+        assert( false );
+        return stmt;
+}
+/// Find the real/official suspend declaration.
+bool SuspendKeyword::is_real_suspend( const ast::FunctionDecl * decl ) {
+        return ( !decl->linkage.is_mangled
+                && 0 == decl->params.size()
+                && 0 == decl->returns.size()
+                && "__cfactx_suspend" == decl->name );
+}
+const ast::Stmt * SuspendKeyword::make_generator_suspend(
+                const ast::SuspendStmt * stmt ) {
+        assert( in_generator );
+        // Target code is:
+        //   GEN.__generator_state = X;
+        //   THEN
+        //   return;
+        //   __gen_X:;
+        const CodeLocation & location = stmt->location;
+        LabelPair label = make_label( stmt );
+        // This is the context saving statement.
+        stmtsToAddBefore.push_back( new ast::ExprStmt( location,
+                new ast::UntypedExpr( location,
+                        new ast::NameExpr( location, "?=?" ),
+                        {
+                                new ast::UntypedMemberExpr( location,
+                                        new ast::NameExpr( location, "__generator_state" ),
+                                        new ast::VariableExpr( location, in_generator )
+                                ),
+                                ast::ConstantExpr::from_int( location, label.idx ),
+                        }
+                )
+        ) );
+        // The THEN component is conditional (return is not).
+        if ( stmt->then ) {
+                stmtsToAddBefore.push_back( stmt->then.get() );
+        }
+        stmtsToAddBefore.push_back( new ast::ReturnStmt( location, nullptr ) );
+        // The null statement replaces the old suspend statement.
+        return new ast::NullStmt( location, { label.obj } );
+}
+const ast::Stmt * SuspendKeyword::make_coroutine_suspend(
+                const ast::SuspendStmt * stmt ) {
+        // The only thing we need from the old statement is the location.
+        const CodeLocation & location = stmt->location;
+        if ( !decl_suspend ) {
+                SemanticError( location, "suspend keyword applied to coroutines requires coroutines to be in scope, add #include <coroutine.hfa>\n" );
+        }
+        if ( stmt->then ) {
+                SemanticError( location, "Compound statement following coroutines is not implemented." );
+        }
+        return new ast::ExprStmt( location,
+                new ast::UntypedExpr( location,
+                        ast::VariableExpr::functionPointer( location, decl_suspend ) )
+        );
+}
+// --------------------------------------------------------------------------
+struct MutexKeyword final : public ast::WithDeclsToAdd<> {
         const ast::FunctionDecl * postvisit( const ast::FunctionDecl * decl );
         void postvisit( const ast::StructDecl * decl );
 …
         ast::CompoundStmt * addStatements( const ast::CompoundStmt * body, const std::vector<ast::ptr<ast::Expr>> & args );
         ast::CompoundStmt * addThreadDtorStatements( const ast::FunctionDecl* func, const ast::CompoundStmt * body, const std::vector<const ast::DeclWithType *> & args );
+        ast::ExprStmt * genVirtLockUnlockExpr( const std::string & fnName, ast::ptr<ast::Expr> expr, const CodeLocation & location, ast::Expr * param);
+        ast::IfStmt * genTypeDiscrimLockUnlock( const std::string & fnName, const std::vector<ast::ptr<ast::Expr>> & args, const CodeLocation & location, ast::UntypedExpr * thisParam );
 private:
         const ast::StructDecl * monitor_decl = nullptr;
 …
         static ast::ptr<ast::Type> generic_func;
+        UniqueName mutex_func_namer = UniqueName("__lock_unlock_curr");
 };
 …
 const ast::Stmt * MutexKeyword::postvisit( const ast::MutexStmt * stmt ) {
+        if ( !lock_guard_decl ) {
+                SemanticError( stmt->location, "mutex stmt requires a header, add #include <mutex_stmt.hfa>\n" );
+        }
         ast::CompoundStmt * body =
                         new ast::CompoundStmt( stmt->location, { stmt->stmt } );
         addStatements( body, stmt->mutexObjs );
         return body;
+        return addStatements( body, stmt->mutexObjs );;
+}
 …
+                                {
                                         new ast::SingleInit( location,
                                                 new ast::AddressExpr(
+                                                new ast::AddressExpr( location,
                                                         new ast::VariableExpr( location, monitor ) ) ),
                                         new ast::SingleInit( location,
 …
+}
+// generates a cast to the void ptr to the appropriate lock type and dereferences it before calling lock or unlock on it
+// used to undo the type erasure done by storing all the lock pointers as void
+ast::ExprStmt * MutexKeyword::genVirtLockUnlockExpr( const std::string & fnName, ast::ptr<ast::Expr> expr, const CodeLocation & location, ast::Expr * param ) {
+        return new ast::ExprStmt( location,
+                new ast::UntypedExpr( location,
+                        new ast::NameExpr( location, fnName ), {
+                                ast::UntypedExpr::createDeref(
+                                        location,
+                                        new ast::CastExpr( location,
+                                                param,
+                                                new ast::PointerType( new ast::TypeofType( new ast::UntypedExpr(
+                                                        expr->location,
+                                                        new ast::NameExpr( expr->location, "__get_mutexstmt_lock_type" ),
+                                                        { expr }
+                                                ) ) ),
+                                                ast::GeneratedFlag::ExplicitCast
+                                        )
+                                )
+                        }
+                )
+        );
+}
+ast::IfStmt * MutexKeyword::genTypeDiscrimLockUnlock( const std::string & fnName, const std::vector<ast::ptr<ast::Expr>> & args, const CodeLocation & location, ast::UntypedExpr * thisParam ) {
+        ast::IfStmt * outerLockIf = nullptr;
+        ast::IfStmt * lastLockIf = nullptr;
+        //adds an if/elif clause for each lock to assign type from void ptr based on ptr address
+        for ( long unsigned int i = 0; i < args.size(); i++ ) {
+                ast::UntypedExpr * ifCond = new ast::UntypedExpr( location,
+                        new ast::NameExpr( location, "?==?" ), {
+                                ast::deepCopy( thisParam ),
+                                new ast::CastExpr( location, new ast::AddressExpr( location, args.at(i) ), new ast::PointerType( new ast::VoidType() ))
+                        }
+                );
+                ast::IfStmt * currLockIf = new ast::IfStmt(
+                        location,
+                        ifCond,
+                        genVirtLockUnlockExpr( fnName, args.at(i), location, ast::deepCopy( thisParam ) )
+                );
+                if ( i == 0 ) {
+                        outerLockIf = currLockIf;
+                } else {
+                        // add ifstmt to else of previous stmt
+                        lastLockIf->else_ = currLockIf;
+                }
+                lastLockIf = currLockIf;
+        }
+        return outerLockIf;
+}
 ast::CompoundStmt * MutexKeyword::addStatements(
                 const ast::CompoundStmt * body,
                 const std::vector<ast::ptr<ast::Expr>> & args ) {
-        ast::CompoundStmt * mutBody = ast::mutate( body );
         // Code is generated near the beginning of the compound statement.
+        const CodeLocation & location = mutBody->location;
+        const CodeLocation & location = body->location;
+                // final body to return
+        ast::CompoundStmt * newBody = new ast::CompoundStmt( location );
+        // std::string lockFnName = mutex_func_namer.newName();
+        // std::string unlockFnName = mutex_func_namer.newName();
         // Make pointer to the monitors.
 …
                 new ast::ArrayType(
                         new ast::PointerType(
+                                new ast::TypeofType(
+                                        new ast::UntypedExpr(
+                                                location,
+                                                new ast::NameExpr( location, "__get_type" ),
+                                                { args.front() }
+                                        )
+                                )
+                                new ast::VoidType()
                         ),
                         ast::ConstantExpr::from_ulong( location, args.size() ),
 …
                                                 new ast::UntypedExpr(
                                                         expr->location,
                                                         new ast::NameExpr( expr->location, "__get_ptr" ),
+                                                        new ast::NameExpr( expr->location, "__get_mutexstmt_lock_ptr" ),
                                                         { expr }
+                                                )
 …
         ast::StructInstType * lock_guard_struct =
                         new ast::StructInstType( lock_guard_decl );
+        ast::TypeExpr * lock_type_expr = new ast::TypeExpr(
+                location,
+                new ast::TypeofType(
+                        new ast::UntypedExpr(
+                                location,
+                                new ast::NameExpr( location, "__get_type" ),
+                                { args.front() }
+                        )
+                )
+        );
+        lock_guard_struct->params.push_back( lock_type_expr );
+        // In reverse order:
+        // use try stmts to lock and finally to unlock
+        ast::TryStmt * outerTry = nullptr;
+        ast::TryStmt * currentTry;
+        ast::CompoundStmt * lastBody = nullptr;
+        // adds a nested try stmt for each lock we are locking
+        for ( long unsigned int i = 0; i < args.size(); i++ ) {
+                ast::UntypedExpr * innerAccess = new ast::UntypedExpr(
+                        location,
+                        new ast::NameExpr( location,"?[?]" ), {
+                                new ast::NameExpr( location, "__monitors" ),
+                                ast::ConstantExpr::from_int( location, i )
+                        }
+                );
+                // make the try body
+                ast::CompoundStmt * currTryBody = new ast::CompoundStmt( location );
+                ast::IfStmt * lockCall = genTypeDiscrimLockUnlock( "lock", args, location, innerAccess );
+                currTryBody->push_back( lockCall );
+                // make the finally stmt
+                ast::CompoundStmt * currFinallyBody = new ast::CompoundStmt( location );
+                ast::IfStmt * unlockCall = genTypeDiscrimLockUnlock( "unlock", args, location, innerAccess );
+                currFinallyBody->push_back( unlockCall );
+                // construct the current try
+                currentTry = new ast::TryStmt(
+                        location,
+                        currTryBody,
+                        {},
+                        new ast::FinallyClause( location, currFinallyBody )
+                );
+                if ( i == 0 ) outerTry = currentTry;
+                else {
+                        // pushback try into the body of the outer try
+                        lastBody->push_back( currentTry );
+                }
+                lastBody = currTryBody;
+        }
+        // push body into innermost try body
+        if ( lastBody != nullptr ) {
+                lastBody->push_back( body );
+                newBody->push_front( outerTry );
+        }
         // monitor_guard_t __guard = { __monitors, # };
         mutBody->push_front(
+        newBody->push_front(
                 new ast::DeclStmt(
                         location,
 …
         // monitor$ * __monitors[] = { get_monitor(a), get_monitor(b) };
+        mutBody->push_front( new ast::DeclStmt( location, monitors ) );
+        return mutBody;
+        newBody->push_front( new ast::DeclStmt( location, monitors ) );
+        // // The parameter for both __lock_curr/__unlock_curr routines.
+        // ast::ObjectDecl * this_decl = new ast::ObjectDecl(
+        //      location,
+        //      "this",
+        //      new ast::PointerType( new ast::VoidType() ),
+        //      nullptr,
+        //      {},
+        //      ast::Linkage::Cforall
+        // );
+        // ast::FunctionDecl * lock_decl = new ast::FunctionDecl(
+        //      location,
+        //      lockFnName,
+        //      { /* forall */ },
+        //      {
+        //              // Copy the declaration of this.
+        //              this_decl,
+        //      },
+        //      { /* returns */ },
+        //      nullptr,
+        //      0,
+        //      ast::Linkage::Cforall,
+        //      { /* attributes */ },
+        //      ast::Function::Inline
+        // );
+        // ast::FunctionDecl * unlock_decl = new ast::FunctionDecl(
+        //      location,
+        //      unlockFnName,
+        //      { /* forall */ },
+        //      {
+        //              // Copy the declaration of this.
+        //              ast::deepCopy( this_decl ),
+        //      },
+        //      { /* returns */ },
+        //      nullptr,
+        //      0,
+        //      ast::Linkage::Cforall,
+        //      { /* attributes */ },
+        //      ast::Function::Inline
+        // );
+        // ast::IfStmt * outerLockIf = nullptr;
+        // ast::IfStmt * outerUnlockIf = nullptr;
+        // ast::IfStmt * lastLockIf = nullptr;
+        // ast::IfStmt * lastUnlockIf = nullptr;
+        // //adds an if/elif clause for each lock to assign type from void ptr based on ptr address
+        // for ( long unsigned int i = 0; i < args.size(); i++ ) {
+        //      ast::VariableExpr * thisParam = new ast::VariableExpr( location, InitTweak::getParamThis( lock_decl ) );
+        //      ast::UntypedExpr * ifCond = new ast::UntypedExpr( location,
+        //              new ast::NameExpr( location, "?==?" ), {
+        //                      thisParam,
+        //                      new ast::CastExpr( location, new ast::AddressExpr( location, args.at(i) ), new ast::PointerType( new ast::VoidType() ))
+        //              }
+        //      );
+        //      ast::IfStmt * currLockIf = new ast::IfStmt(
+        //              location,
+        //              ast::deepCopy( ifCond ),
+        //              genVirtLockUnlockExpr( "lock", args.at(i), location, ast::deepCopy( thisParam ) )
+        //      );
+        //      ast::IfStmt * currUnlockIf = new ast::IfStmt(
+        //              location,
+        //              ifCond,
+        //              genVirtLockUnlockExpr( "unlock", args.at(i), location, ast::deepCopy( thisParam ) )
+        //      );
+        //      if ( i == 0 ) {
+        //              outerLockIf = currLockIf;
+        //              outerUnlockIf = currUnlockIf;
+        //      } else {
+        //              // add ifstmt to else of previous stmt
+        //              lastLockIf->else_ = currLockIf;
+        //              lastUnlockIf->else_ = currUnlockIf;
+        //      }
+        //      lastLockIf = currLockIf;
+        //      lastUnlockIf = currUnlockIf;
+        // }
+        // // add pointer typing if/elifs to body of routines
+        // lock_decl->stmts = new ast::CompoundStmt( location, { outerLockIf } );
+        // unlock_decl->stmts = new ast::CompoundStmt( location, { outerUnlockIf } );
+        // // add routines to scope
+        // declsToAddBefore.push_back( lock_decl );
+        // declsToAddBefore.push_back( unlock_decl );
+        // newBody->push_front(new ast::DeclStmt( location, lock_decl ));
+        // newBody->push_front(new ast::DeclStmt( location, unlock_decl ));
+        return newBody;
+}
 …
 // --------------------------------------------------------------------------
+// Interface Functions:
 void implementKeywords( ast::TranslationUnit & translationUnit ) {
+        (void)translationUnit;
+        assertf(false, "Apply Keywords not implemented." );
+        ast::Pass<ThreadKeyword>::run( translationUnit );
+        ast::Pass<CoroutineKeyword>::run( translationUnit );
+        ast::Pass<MonitorKeyword>::run( translationUnit );
+        ast::Pass<GeneratorKeyword>::run( translationUnit );
+        ast::Pass<SuspendKeyword>::run( translationUnit );
+}

src/ControlStruct/ExceptTranslateNew.cpp

-              rba897d21
+              r2e9b59b
 // Author           : Andrew Beach
 // Created On       : Mon Nov  8 11:53:00 2021
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Mon Jan 31 18:49:58 2022
 // Update Count     : 1
+// Last Modified By : Andrew Beach
+// Last Modified On : Fri Mar 11 17:51:00 2022
+// Update Count     : 2
 //
 …
 namespace {
+        typedef std::list<ast::CatchStmt*> CatchList;
+        void split( CatchList& allHandlers, CatchList& terHandlers,
+                                CatchList& resHandlers ) {
+                while ( !allHandlers.empty() ) {
+                        ast::CatchStmt * stmt = allHandlers.front();
+                        allHandlers.pop_front();
+                        if (stmt->kind == ast::ExceptionKind::Terminate) {
+                                terHandlers.push_back(stmt);
+                        } else {
+                                resHandlers.push_back(stmt);
+                        }
+                }
+        }
+        typedef std::list<ast::CatchClause*> CatchList;
         void appendDeclStmt( ast::CompoundStmt * block, ast::DeclWithType * item ) {
 …
         {}
         void previsit( const ast::CatchStmt * stmt );
+        void previsit( const ast::CatchClause * stmt );
         const ast::Stmt * postvisit( const ast::ThrowStmt * stmt );
 };
 …
+}
 void TranslateThrowsCore::previsit( const ast::CatchStmt * stmt ) {
+void TranslateThrowsCore::previsit( const ast::CatchClause * stmt ) {
         // Validate the statement's form.
         const ast::ObjectDecl * decl = stmt->decl.as<ast::ObjectDecl>();
 …
         ast::FunctionDecl * create_terminate_catch( CatchList &handlers );
         ast::CompoundStmt * create_single_matcher(
                 const ast::DeclWithType * except_obj, ast::CatchStmt * modded_handler );
+                const ast::DeclWithType * except_obj, ast::CatchClause * modded_handler );
         ast::FunctionDecl * create_terminate_match( CatchList &handlers );
         ast::CompoundStmt * create_terminate_caller( CodeLocation loc, ast::FunctionDecl * try_wrapper,
 …
         ast::Stmt * create_resume_rethrow( const ast::ThrowStmt * throwStmt );
+        // Types used in translation, make sure to use clone.
+        // Types used in translation, first group are internal.
+        ast::ObjectDecl * make_index_object( CodeLocation const & ) const;
+        ast::ObjectDecl * make_exception_object( CodeLocation const & ) const;
+        ast::ObjectDecl * make_bool_object( CodeLocation const & ) const;
+        ast::ObjectDecl * make_voidptr_object( CodeLocation const & ) const;
+        ast::ObjectDecl * make_unused_index_object( CodeLocation const & ) const;
         // void (*function)();
         ast::FunctionDecl * try_func_t;
+        ast::FunctionDecl * make_try_function( CodeLocation const & ) const;
         // void (*function)(int, exception);
         ast::FunctionDecl * catch_func_t;
+        ast::FunctionDecl * make_catch_function( CodeLocation const & ) const;
         // int (*function)(exception);
         ast::FunctionDecl * match_func_t;
+        ast::FunctionDecl * make_match_function( CodeLocation const & ) const;
         // bool (*function)(exception);
         ast::FunctionDecl * handle_func_t;
+        ast::FunctionDecl * make_handle_function( CodeLocation const & ) const;
         // void (*function)(__attribute__((unused)) void *);
+        ast::FunctionDecl * finally_func_t;
+        ast::StructInstType * create_except_type() {
+                assert( except_decl );
+                return new ast::StructInstType( except_decl );
+        }
+        void init_func_types();
+        ast::FunctionDecl * make_finally_function( CodeLocation const & ) const;
 public:
 …
 };
+void TryMutatorCore::init_func_types() {
+ast::ObjectDecl * TryMutatorCore::make_index_object(
+                CodeLocation const & location ) const {
+        return new ast::ObjectDecl(
+                location,
+                "__handler_index",
+                new ast::BasicType(ast::BasicType::SignedInt),
+                nullptr, //init
+                ast::Storage::Classes{},
+                ast::Linkage::Cforall
+                );
+}
+ast::ObjectDecl * TryMutatorCore::make_exception_object(
+                CodeLocation const & location ) const {
         assert( except_decl );
+        ast::ObjectDecl index_obj(
+                {},
+                "__handler_index",
+                new ast::BasicType(ast::BasicType::SignedInt)
+                );
+        ast::ObjectDecl exception_obj(
+                {},
+        return new ast::ObjectDecl(
+                location,
                 "__exception_inst",
                 new ast::PointerType(
                         new ast::StructInstType( except_decl )
                         ),
+                NULL
+                );
+        ast::ObjectDecl bool_obj(
+                {},
+                nullptr, //init
+                ast::Storage::Classes{},
+                ast::Linkage::Cforall
+                );
+}
+ast::ObjectDecl * TryMutatorCore::make_bool_object(
+                CodeLocation const & location ) const {
+        return new ast::ObjectDecl(
+                location,
                 "__ret_bool",
                 new ast::BasicType( ast::BasicType::Bool ),
 …
                 std::vector<ast::ptr<ast::Attribute>>{ new ast::Attribute( "unused" ) }
                 );
+        ast::ObjectDecl voidptr_obj(
+                {},
+}
+ast::ObjectDecl * TryMutatorCore::make_voidptr_object(
+                CodeLocation const & location ) const {
+        return new ast::ObjectDecl(
+                location,
                 "__hook",
                 new ast::PointerType(
 …
                 std::vector<ast::ptr<ast::Attribute>>{ new ast::Attribute( "unused" ) }
                 );
+        ast::ObjectDecl unused_index_obj(
+                {},
+}
+ast::ObjectDecl * TryMutatorCore::make_unused_index_object(
+                CodeLocation const & location ) const {
+        return new ast::ObjectDecl(
+                location,
                 "__handler_index",
                 new ast::BasicType(ast::BasicType::SignedInt),
 …
                 std::vector<ast::ptr<ast::Attribute>>{ new ast::Attribute( "unused" ) }
         );
+        //unused_index_obj->attributes.push_back( new Attribute( "unused" ) );
+        try_func_t = new ast::FunctionDecl(
+                {},
+}
+ast::FunctionDecl * TryMutatorCore::make_try_function(
+                CodeLocation const & location ) const {
+        return new ast::FunctionDecl(
+                location,
                 "try",
                 {}, //forall
 …
                 ast::Linkage::Cforall
         );
+        catch_func_t = new ast::FunctionDecl(
+                {},
+}
+ast::FunctionDecl * TryMutatorCore::make_catch_function(
+                CodeLocation const & location ) const {
+        return new ast::FunctionDecl(
+                location,
                 "catch",
                 {}, //forall
                 {ast::deepCopy(&index_obj), ast::deepCopy(&exception_obj)},//param
+                { make_index_object( location ), make_exception_object( location ) },
                 {}, //return void
                 nullptr,
 …
                 ast::Linkage::Cforall
         );
+        match_func_t = new ast::FunctionDecl(
+                {},
+}
+ast::FunctionDecl * TryMutatorCore::make_match_function(
+                CodeLocation const & location ) const {
+        return new ast::FunctionDecl(
+                location,
                 "match",
                 {}, //forall
                 {ast::deepCopy(&exception_obj)},
                 {ast::deepCopy(&unused_index_obj)},
+                { make_exception_object( location ) },
+                { make_unused_index_object( location ) },
                 nullptr,
                 ast::Storage::Classes{},
                 ast::Linkage::Cforall
         );
+        handle_func_t = new ast::FunctionDecl(
+                {},
+}
+ast::FunctionDecl * TryMutatorCore::make_handle_function(
+                CodeLocation const & location ) const {
+        return new ast::FunctionDecl(
+                location,
                 "handle",
                 {}, //forall
                 {ast::deepCopy(&exception_obj)},
                 {ast::deepCopy(&bool_obj)},
+                { make_exception_object( location ) },
+                { make_bool_object( location ) },
                 nullptr,
                 ast::Storage::Classes{},
                 ast::Linkage::Cforall
         );
+        finally_func_t = new ast::FunctionDecl(
+                {},
+}
+ast::FunctionDecl * TryMutatorCore::make_finally_function(
+                CodeLocation const & location ) const {
+        return new ast::FunctionDecl(
+                location,
                 "finally",
                 {}, //forall
                 {ast::deepCopy(&voidptr_obj)},
+                { make_voidptr_object( location ) },
                 {}, //return void
                 nullptr,
 …
                 ast::Linkage::Cforall
         );
-        //catch_func_t.get_parameters().push_back( index_obj.clone() );
-        //catch_func_t.get_parameters().push_back( exception_obj.clone() );
-        //match_func_t.get_returnVals().push_back( unused_index_obj );
-        //match_func_t.get_parameters().push_back( exception_obj.clone() );
-        //handle_func_t.get_returnVals().push_back( bool_obj.clone() );
-        //handle_func_t.get_parameters().push_back( exception_obj.clone() );
-        //finally_func_t.get_parameters().push_back( voidptr_obj.clone() );
+}
 // TryStmt Mutation Helpers
-/*
-ast::CompoundStmt * TryMutatorCore::take_try_block( ast::TryStmt *tryStmt ) {
-        ast::CompoundStmt * block = tryStmt->body;
-        tryStmt->body = nullptr;
-        return block;
+}
-*/
 ast::FunctionDecl * TryMutatorCore::create_try_wrapper(
                 const ast::CompoundStmt *body ) {
         ast::FunctionDecl * ret = ast::deepCopy(try_func_t);
+        ast::FunctionDecl * ret = make_try_function( body->location );
         ret->stmts = body;
         return ret;
 …
 ast::FunctionDecl * TryMutatorCore::create_terminate_catch(
                 CatchList &handlers ) {
         std::vector<ast::ptr<ast::Stmt>> handler_wrappers;
+        std::vector<ast::ptr<ast::CaseClause>> handler_wrappers;
         assert (!handlers.empty());
         const CodeLocation loc = handlers.front()->location;
         ast::FunctionDecl * func_t = ast::deepCopy(catch_func_t);
+        ast::FunctionDecl * func_t = make_catch_function( loc );
         const ast::DeclWithType * index_obj = func_t->params.front();
         const ast::DeclWithType * except_obj = func_t->params.back();
 …
         for ( ; it != handlers.end() ; ++it ) {
                 ++index;
                 ast::CatchStmt * handler = *it;
+                ast::CatchClause * handler = *it;
                 const CodeLocation loc = handler->location;
 …
                 // handler->body = nullptr;
                 handler_wrappers.push_back( new ast::CaseStmt(loc,
+                handler_wrappers.push_back( new ast::CaseClause(loc,
                         ast::ConstantExpr::from_int(loc, index) ,
                         { block, new ast::ReturnStmt( loc, nullptr ) }
 …
         // TODO: Some sort of meaningful error on default perhaps?
+        /*
+        std::list<Statement*> stmt_handlers;
+        while ( !handler_wrappers.empty() ) {
+                stmt_handlers.push_back( handler_wrappers.front() );
+                handler_wrappers.pop_front();
+        }
+        */
+        ast::SwitchStmt * handler_lookup = new ast::SwitchStmt(loc,
+        ast::SwitchStmt * handler_lookup = new ast::SwitchStmt( loc,
                 new ast::VariableExpr( loc, index_obj ),
                 std::move(handler_wrappers)
                 );
+        ast::CompoundStmt * body = new ast::CompoundStmt(loc,
+                {handler_lookup});
+        ast::CompoundStmt * body = new ast::CompoundStmt( loc, {handler_lookup} );
         func_t->stmts = body;
 …
 // except_obj is referenced, modded_handler will be freed.
 ast::CompoundStmt * TryMutatorCore::create_single_matcher(
                 const ast::DeclWithType * except_obj, ast::CatchStmt * modded_handler ) {
+                const ast::DeclWithType * except_obj, ast::CatchClause * modded_handler ) {
         // {
         //     `modded_handler.decl`
 …
         // Check for type match.
         ast::VirtualCastExpr * vcex = new ast::VirtualCastExpr(loc,
+        ast::VirtualCastExpr * vcex = new ast::VirtualCastExpr(loc,
                 new ast::VariableExpr(loc, except_obj ),
                 local_except->get_type()
 …
+        }
         // Construct the match condition.
         block->push_back( new ast::IfStmt(loc,
+        block->push_back( new ast::IfStmt(loc,
                 cond, modded_handler->body, nullptr ) );
-        // xxx - how does this work in new ast
-        //modded_handler->set_decl( nullptr );
-        //modded_handler->set_cond( nullptr );
-        //modded_handler->set_body( nullptr );
-        //delete modded_handler;
         return block;
+}
 …
         ast::CompoundStmt * body = new ast::CompoundStmt(loc);
         ast::FunctionDecl * func_t = ast::deepCopy(match_func_t);
+        ast::FunctionDecl * func_t = make_match_function( loc );
         const ast::DeclWithType * except_obj = func_t->params.back();
 …
         for ( it = handlers.begin() ; it != handlers.end() ; ++it ) {
                 ++index;
                 ast::CatchStmt * handler = *it;
+                ast::CatchClause * handler = *it;
                 // Body should have been taken by create_terminate_catch.
 …
+        }
         body->push_back( new ast::ReturnStmt(loc,
+        body->push_back( new ast::ReturnStmt(loc,
                 ast::ConstantExpr::from_int( loc, 0 ) ));
 …
         ast::CompoundStmt * body = new ast::CompoundStmt(loc);
         ast::FunctionDecl * func_t = ast::deepCopy(handle_func_t);
+        ast::FunctionDecl * func_t = make_handle_function( loc );
         const ast::DeclWithType * except_obj = func_t->params.back();
         CatchList::iterator it;
         for ( it = handlers.begin() ; it != handlers.end() ; ++it ) {
                 ast::CatchStmt * handler = *it;
+                ast::CatchClause * handler = *it;
                 const CodeLocation loc = handler->location;
                 // Modifiy body.
                 ast::CompoundStmt * handling_code;
                 if (handler->body.as<ast::CompoundStmt>()) {
                         handling_code =
                         strict_dynamic_cast<ast::CompoundStmt*>( handler->body.get_and_mutate() );
+                        handling_code = strict_dynamic_cast<ast::CompoundStmt*>(
+                                handler->body.get_and_mutate() );
                 } else {
                         handling_code = new ast::CompoundStmt(loc);
 …
                 ast::TryStmt * tryStmt ) {
         // void finally() { `finally->block` }
         const ast::FinallyStmt * finally = tryStmt->finally;
+        const ast::FinallyClause * finally = tryStmt->finally;
         const ast::CompoundStmt * body = finally->body;
         ast::FunctionDecl * func_t = ast::deepCopy(finally_func_t);
+        ast::FunctionDecl * func_t = make_finally_function( tryStmt->location );
         func_t->stmts = body;
-        // finally->set_block( nullptr );
-        // delete finally;
         tryStmt->finally = nullptr;
         return func_t;
 …
         const CodeLocation loc = finally_wrapper->location;
-        // Make Cleanup Attribute.
-        /*
-        std::list< ast::Attribute * > attributes;
+        {
-                std::list<  > attr_params;
-                attr_params.push_back( nameOf( finally_wrapper ) );
-                attributes.push_back( new Attribute( "cleanup", attr_params ) );
+        }
-        */
         return new ast::ObjectDecl(
                 loc,
 …
         // return false;
         const CodeLocation loc = throwStmt->location;
         ast::Stmt * result = new ast::ReturnStmt(loc,
+        ast::Stmt * result = new ast::ReturnStmt(loc,
                 ast::ConstantExpr::from_bool( loc, false )
                 );
         result->labels = throwStmt->labels;
-        // delete throwStmt; done by postvisit
         return result;
+}
 …
                 assert( nullptr == except_decl );
                 except_decl = structDecl;
-                init_func_types();
         } else if ( structDecl->name == "__cfaehm_try_resume_node" ) {
                 assert( nullptr == node_decl );
 …
+                }
+        }
-        // split( mutStmt->handlers,
-        //              termination_handlers, resumption_handlers );
         if ( resumption_handlers.size() ) {

src/ControlStruct/LabelGeneratorNew.cpp

-              rba897d21
+              r2e9b59b
 // file "LICENCE" distributed with Cforall.
 //
 // LabelGenerator.cc --
+// LabelGeneratorNew.cpp --
 //
 // Author           : Peter A. Buhr
 // Created On       : Mon May 18 07:44:20 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Wed Feb  2 09:11:17 2022
 // Update Count     : 72
+// Last Modified By : Andrew Beach
+// Last Modified On : Mon Mar 28 10:03:00 2022
+// Update Count     : 73
 //
 …
 namespace ControlStruct {
+Label newLabel( const string & suffix, const Stmt * stmt ) {
+enum { size = 128 };
+static int newLabelPre( char buf[size], const string & suffix ) {
         static int current = 0;
-        assertf( stmt, "CFA internal error: parameter statement cannot be null pointer" );
-        enum { size = 128 };
-        char buf[size];                                                                         // space to build label
         int len = snprintf( buf, size, "__L%d__%s", current++, suffix.c_str() );
         assertf( len < size, "CFA Internal error: buffer overflow creating label" );
+        return len;
+}
+static Label newLabelPost( char buf[size], const CodeLocation & location ) {
+        Label ret_label( location, buf );
+        ret_label.attributes.push_back( new Attribute( "unused" ) );
+        return ret_label;
+}
+Label newLabel( const string & suffix, const Stmt * stmt ) {
+        // Buffer for string manipulation.
+        char buf[size];
+        assertf( stmt, "CFA internal error: parameter statement cannot be null pointer" );
+        int len = newLabelPre( buf, suffix );
         // What does this do?
 …
         } // if
+        Label ret_label( stmt->location, buf );
+        ret_label.attributes.push_back( new Attribute( "unused" ) );
+        return ret_label;
+        return newLabelPost( buf, stmt->location );
+}
+Label newLabel( const string & suffix, const CodeLocation & location ) {
+        // Buffer for string manipulation.
+        char buf[size];
+        newLabelPre( buf, suffix );
+        return newLabelPost( buf, location );
+}

src/ControlStruct/LabelGeneratorNew.hpp

-              rba897d21
+              r2e9b59b
 // Author           : Rodolfo G. Esteves
 // Created On       : Mon May 18 07:44:20 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Mon Jan 31 18:03:09 2022
 // Update Count     : 27
+// Last Modified By : Andrew Beach
+// Last Modified On : Fir Mar 25 15:40:00 2022
+// Update Count     : 28
 //
 …
 #include <string>                                                                               // for string
 class Statement;
+class CodeLocation;
 namespace ast {
+        class Label;
         class Stmt;
-        class Label;
 } // namespace ast
 namespace ControlStruct {
         ast::Label newLabel( const std::string &, const ast::Stmt * );
+        ast::Label newLabel( const std::string &, const CodeLocation & );
 } // namespace ControlStruct

src/ControlStruct/MultiLevelExit.cpp

-              rba897d21
+              r2e9b59b
 // Author           : Andrew Beach
 // Created On       : Mon Nov  1 13:48:00 2021
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Wed Feb  2 23:07:54 2022
 // Update Count     : 33
+// Last Modified By : Andrew Beach
+// Last Modified On : Mon Mar 28  9:42:00 2022
+// Update Count     : 34
 //
 …
         enum Kind {
                 ForStmtK, WhileDoStmtK, CompoundStmtK, IfStmtK, CaseStmtK, SwitchStmtK, TryStmtK
+                ForStmtK, WhileDoStmtK, CompoundStmtK, IfStmtK, CaseClauseK, SwitchStmtK, TryStmtK
         } kind;
 …
         Entry( const IfStmt *stmt, Label breakExit ) :
                 stmt( stmt ), firstTarget( breakExit ), secondTarget(), kind( IfStmtK ) {}
         Entry( const CaseStmt *stmt, Label fallExit ) :
                 stmt( stmt ), firstTarget( fallExit ), secondTarget(), kind( CaseStmtK ) {}
+        Entry( const CaseClause *, const CompoundStmt *stmt, Label fallExit ) :
+                stmt( stmt ), firstTarget( fallExit ), secondTarget(), kind( CaseClauseK ) {}
         Entry( const SwitchStmt *stmt, Label breakExit, Label fallDefaultExit ) :
                 stmt( stmt ), firstTarget( breakExit ), secondTarget( fallDefaultExit ), kind( SwitchStmtK ) {}
 …
         bool isContTarget() const { return kind <= WhileDoStmtK; }
         bool isBreakTarget() const { return kind != CaseStmtK; }
         bool isFallTarget() const { return kind == CaseStmtK; }
+        bool isBreakTarget() const { return kind != CaseClauseK; }
+        bool isFallTarget() const { return kind == CaseClauseK; }
         bool isFallDefaultTarget() const { return kind == SwitchStmtK; }
         // These routines set a target as being "used" by a BranchStmt
         Label useContExit() { assert( kind <= WhileDoStmtK ); return useTarget(secondTarget); }
         Label useBreakExit() { assert( kind != CaseStmtK ); return useTarget(firstTarget); }
         Label useFallExit() { assert( kind == CaseStmtK );  return useTarget(firstTarget); }
+        Label useBreakExit() { assert( kind != CaseClauseK ); return useTarget(firstTarget); }
+        Label useFallExit() { assert( kind == CaseClauseK );  return useTarget(firstTarget); }
         Label useFallDefaultExit() { assert( kind == SwitchStmtK ); return useTarget(secondTarget); }
         // These routines check if a specific label for a statement is used by a BranchStmt
         bool isContUsed() const { assert( kind <= WhileDoStmtK ); return secondTarget.used; }
         bool isBreakUsed() const { assert( kind != CaseStmtK ); return firstTarget.used; }
         bool isFallUsed() const { assert( kind == CaseStmtK ); return firstTarget.used; }
+        bool isBreakUsed() const { assert( kind != CaseClauseK ); return firstTarget.used; }
+        bool isFallUsed() const { assert( kind == CaseClauseK ); return firstTarget.used; }
         bool isFallDefaultUsed() const { assert( kind == SwitchStmtK ); return secondTarget.used; }
         void seenDefault() { fallDefaultValid = false; }
 …
         void previsit( const ForStmt * );
         const ForStmt * postvisit( const ForStmt * );
         const CaseStmt * previsit( const CaseStmt * );
+        const CaseClause * previsit( const CaseClause * );
         void previsit( const IfStmt * );
         const IfStmt * postvisit( const IfStmt * );
 …
         void previsit( const TryStmt * );
         void postvisit( const TryStmt * );
         void previsit( const FinallyStmt * );
+        void previsit( const FinallyClause * );
         const Stmt * mutateLoop( const Stmt * body, Entry& );
 …
                   auto switchStmt = strict_dynamic_cast< const SwitchStmt * >( targetEntry->stmt );
                   bool foundDefault = false;
+                  for ( auto subStmt : switchStmt->stmts ) {
+                          const CaseStmt * caseStmt = subStmt.strict_as<CaseStmt>();
+                  for ( auto caseStmt : switchStmt->cases ) {
                           if ( caseStmt->isDefault() ) {
                                   foundDefault = true;
 …
+}
 const CaseStmt * MultiLevelExitCore::previsit( const CaseStmt * stmt ) {
+const CaseClause * MultiLevelExitCore::previsit( const CaseClause * stmt ) {
         visit_children = false;
 …
         // The cond may not exist, but if it does update it now.
         visitor->maybe_accept( stmt, &CaseStmt::cond );
+        visitor->maybe_accept( stmt, &CaseClause::cond );
         // Just save the mutated node for simplicity.
         CaseStmt * mutStmt = mutate( stmt );
         Label fallLabel = newLabel( "fallThrough", stmt );
+        CaseClause * mutStmt = mutate( stmt );
+        Label fallLabel = newLabel( "fallThrough", stmt->location );
         if ( ! mutStmt->stmts.empty() ) {
+                // These should already be in a block.
+                auto first = mutStmt->stmts.front().get_and_mutate();
+                auto block = strict_dynamic_cast<CompoundStmt *>( first );
                 // Ensure that the stack isn't corrupted by exceptions in fixBlock.
                 auto guard = makeFuncGuard(
                         [&](){ enclosing_control_structures.emplace_back( mutStmt, fallLabel ); },
+                        [&](){ enclosing_control_structures.emplace_back( mutStmt, block, fallLabel ); },
                         [this](){ enclosing_control_structures.pop_back(); }
                         );
-                // These should already be in a block.
-                auto block = mutate( mutStmt->stmts.front().strict_as<CompoundStmt>() );
                 block->kids = fixBlock( block->kids, true );
 …
                 Entry & entry = enclosing_control_structures.back();
                 if ( entry.isFallUsed() ) {
                         mutStmt->stmts.push_back( labelledNullStmt( mutStmt->location, entry.useFallExit() ) );
+                        mutStmt->stmts.push_back( labelledNullStmt( block->location, entry.useFallExit() ) );
+                }
+        }
 …
+}
+bool isDefaultCase( const ptr<Stmt> & stmt ) {
+        const CaseStmt * caseStmt = stmt.strict_as<CaseStmt>();
+        return caseStmt->isDefault();
+static bool isDefaultCase( const ptr<CaseClause> & caseClause ) {
+        return caseClause->isDefault();
+}
 void MultiLevelExitCore::previsit( const SwitchStmt * stmt ) {
         Label label = newLabel( "switchBreak", stmt );
         auto it = find_if( stmt->stmts.rbegin(), stmt->stmts.rend(), isDefaultCase );
         const CaseStmt * defaultCase = it != stmt->stmts.rend() ? (it)->strict_as<CaseStmt>() : nullptr;
         Label defaultLabel = defaultCase ? newLabel( "fallThroughDefault", defaultCase ) : Label( stmt->location, "" );
+        auto it = find_if( stmt->cases.rbegin(), stmt->cases.rend(), isDefaultCase );
+        const CaseClause * defaultCase = it != stmt->cases.rend() ? (*it) : nullptr;
+        Label defaultLabel = defaultCase ? newLabel( "fallThroughDefault", defaultCase->location ) : Label( stmt->location, "" );
         enclosing_control_structures.emplace_back( stmt, label, defaultLabel );
         GuardAction( [this]() { enclosing_control_structures.pop_back(); } );
 …
         // Collect valid labels for fallthrough. It starts with all labels at this level, then remove as each is seen during
         // traversal.
+        for ( const Stmt * stmt : stmt->stmts ) {
+                auto * caseStmt = strict_dynamic_cast< const CaseStmt * >( stmt );
+        for ( const CaseClause * caseStmt : stmt->cases ) {
                 if ( caseStmt->stmts.empty() ) continue;
                 auto block = caseStmt->stmts.front().strict_as<CompoundStmt>();
 …
                 // exit label and break to the last case, create a default case if no cases.
                 SwitchStmt * mutStmt = mutate( stmt );
                 if ( mutStmt->stmts.empty() ) {
                         mutStmt->stmts.push_back( new CaseStmt( mutStmt->location, nullptr, {} ) );
+                }
                 auto caseStmt = mutStmt->stmts.back().strict_as<CaseStmt>();
+                if ( mutStmt->cases.empty() ) {
+                        mutStmt->cases.push_back( new CaseClause( mutStmt->location, nullptr, {} ) );
+                }
+                auto caseStmt = mutStmt->cases.back().get();
                 auto mutCase = mutate( caseStmt );
                 mutStmt->stmts.back() = mutCase;
+                mutStmt->cases.back() = mutCase;
                 Label label( mutCase->location, "breakLabel" );
 …
+}
 void MultiLevelExitCore::previsit( const FinallyStmt * ) {
+void MultiLevelExitCore::previsit( const FinallyClause * ) {
         GuardAction([this, old = move( enclosing_control_structures)](){ enclosing_control_structures = move(old); });
         enclosing_control_structures = vector<Entry>();

src/InitTweak/FixGlobalInit.cc

-              rba897d21
+              r2e9b59b
                 accept_all(translationUnit, fixer);
+                // Say these magic declarations come at the end of the file.
+                CodeLocation const & location = translationUnit.decls.back()->location;
                 if ( !fixer.core.initStmts.empty() ) {
                         std::vector<ast::ptr<ast::Expr>> ctorParams;
+                        if (inLibrary) ctorParams.emplace_back(ast::ConstantExpr::from_int({}, 200));
+                        auto initFunction = new ast::FunctionDecl({}, "__global_init__", {}, {}, {}, new ast::CompoundStmt({}, std::move(fixer.core.initStmts)),
+                                ast::Storage::Static, ast::Linkage::C, {new ast::Attribute("constructor", std::move(ctorParams))});
+                        if (inLibrary) ctorParams.emplace_back(ast::ConstantExpr::from_int(location, 200));
+                        auto initFunction = new ast::FunctionDecl(location,
+                                "__global_init__", {}, {}, {},
+                                new ast::CompoundStmt(location, std::move(fixer.core.initStmts)),
+                                ast::Storage::Static, ast::Linkage::C,
+                                {new ast::Attribute("constructor", std::move(ctorParams))});
                         translationUnit.decls.emplace_back( initFunction );
 …
                 if ( !fixer.core.destroyStmts.empty() ) {
                         std::vector<ast::ptr<ast::Expr>> dtorParams;
+                        if (inLibrary) dtorParams.emplace_back(ast::ConstantExpr::from_int({}, 200));
+                        auto destroyFunction = new ast::FunctionDecl({}, "__global_destroy__", {}, {}, {}, new ast::CompoundStmt({}, std::move(fixer.core.destroyStmts)),
+                                ast::Storage::Static, ast::Linkage::C, {new ast::Attribute("destructor", std::move(dtorParams))});
+                        if (inLibrary) dtorParams.emplace_back(ast::ConstantExpr::from_int(location, 200));
+                        auto destroyFunction = new ast::FunctionDecl( location,
+                                "__global_destroy__", {}, {}, {},
+                                new ast::CompoundStmt(location, std::move(fixer.core.destroyStmts)),
+                                ast::Storage::Static, ast::Linkage::C,
+                                {new ast::Attribute("destructor", std::move(dtorParams))});
                         translationUnit.decls.emplace_back(destroyFunction);

src/InitTweak/FixInitNew.cpp

-              rba897d21
+              r2e9b59b
 #include "CodeGen/GenType.h"           // for genPrettyType
 #include "CodeGen/OperatorTable.h"
+#include "Common/CodeLocationTools.hpp"
 #include "Common/PassVisitor.h"        // for PassVisitor, WithStmtsToAdd
 #include "Common/SemanticError.h"      // for SemanticError
 …
         /// generate/resolve copy construction expressions for each, and generate/resolve destructors for both
         /// arguments and return value temporaries
         struct ResolveCopyCtors final : public ast::WithGuards, public ast::WithStmtsToAdd<>, public ast::WithSymbolTable, public ast::WithShortCircuiting, public ast::WithVisitorRef<ResolveCopyCtors> {
+        struct ResolveCopyCtors final : public ast::WithGuards, public ast::WithStmtsToAdd<>, public ast::WithSymbolTable, public ast::WithShortCircuiting, public ast::WithVisitorRef<ResolveCopyCtors>, public ast::WithConstTranslationUnit {
                 const ast::Expr * postvisit( const ast::ImplicitCopyCtorExpr * impCpCtorExpr );
                 const ast::StmtExpr * previsit( const ast::StmtExpr * stmtExpr );
 …
         /// for any member that is missing a corresponding ctor/dtor call.
         /// error if a member is used before constructed
         struct GenStructMemberCalls final : public ast::WithGuards, public ast::WithShortCircuiting, public ast::WithSymbolTable, public ast::WithVisitorRef<GenStructMemberCalls> {
+        struct GenStructMemberCalls final : public ast::WithGuards, public ast::WithShortCircuiting, public ast::WithSymbolTable, public ast::WithVisitorRef<GenStructMemberCalls>, public ast::WithConstTranslationUnit {
                 void previsit( const ast::FunctionDecl * funcDecl );
                 const ast::DeclWithType * postvisit( const ast::FunctionDecl * funcDecl );
 …
         /// expands ConstructorExpr nodes into comma expressions, using a temporary for the first argument
         struct FixCtorExprs final : public ast::WithDeclsToAdd<>, public ast::WithSymbolTable, public ast::WithShortCircuiting {
+        struct FixCtorExprs final : public ast::WithDeclsToAdd<>, public ast::WithSymbolTable, public ast::WithShortCircuiting, public ast::WithConstTranslationUnit {
                 const ast::Expr * postvisit( const ast::ConstructorExpr * ctorExpr );
         };
 …
                 // (VariableExpr and already resolved expression)
                 CP_CTOR_PRINT( std::cerr << "ResolvingCtorDtor " << untyped << std::endl; )
                 ast::ptr<ast::Expr> resolved = ResolvExpr::findVoidExpression(untyped, symtab);
+                ast::ptr<ast::Expr> resolved = ResolvExpr::findVoidExpression(untyped, { symtab, transUnit().global } );
                 assert( resolved );
                 if ( resolved->env ) {
 …
                 ast::ptr<ast::Expr> guard = mutArg;
                 ast::ptr<ast::ObjectDecl> tmp = new ast::ObjectDecl({}, "__tmp", mutResult, nullptr );
+                ast::ptr<ast::ObjectDecl> tmp = new ast::ObjectDecl(loc, "__tmp", mutResult, nullptr );
                 // create and resolve copy constructor
 …
         ast::Expr * ResolveCopyCtors::destructRet( const ast::ObjectDecl * ret, const ast::Expr * arg ) {
+                auto global = transUnit().global;
                 // TODO: refactor code for generating cleanup attribute, since it's common and reused in ~3-4 places
                 // check for existing cleanup attribute before adding another(?)
                 // need to add __Destructor for _tmp_cp variables as well
                 assertf( ast::dtorStruct, "Destructor generation requires __Destructor definition." );
                 assertf( ast::dtorStruct->members.size() == 2, "__Destructor definition does not have expected fields." );
                 assertf( ast::dtorStructDestroy, "Destructor generation requires __destroy_Destructor." );
+                assertf( global.dtorStruct, "Destructor generation requires __Destructor definition." );
+                assertf( global.dtorStruct->members.size() == 2, "__Destructor definition does not have expected fields." );
+                assertf( global.dtorDestroy, "Destructor generation requires __destroy_Destructor." );
                 const CodeLocation loc = ret->location;
 …
                 auto dtorFunc = getDtorFunc( ret, new ast::ExprStmt(loc, dtor ), stmtsToAddBefore );
                 auto dtorStructType = new ast::StructInstType(ast::dtorStruct);
+                auto dtorStructType = new ast::StructInstType( global.dtorStruct );
                 // what does this do???
 …
                 static UniqueName namer( "_ret_dtor" );
                 auto retDtor = new ast::ObjectDecl(loc, namer.newName(), dtorStructType, new ast::ListInit(loc, { new ast::SingleInit(loc, ast::ConstantExpr::null(loc) ), new ast::SingleInit(loc, new ast::CastExpr( new ast::VariableExpr(loc, dtorFunc ), dtorType ) ) } ) );
                 retDtor->attributes.push_back( new ast::Attribute( "cleanup", { new ast::VariableExpr(loc, ast::dtorStructDestroy ) } ) );
+                retDtor->attributes.push_back( new ast::Attribute( "cleanup", { new ast::VariableExpr(loc, global.dtorDestroy ) } ) );
                 stmtsToAddBefore.push_back( new ast::DeclStmt(loc, retDtor ) );
                 if ( arg ) {
                         auto member = new ast::MemberExpr(loc, ast::dtorStruct->members.front().strict_as<ast::DeclWithType>(), new ast::VariableExpr(loc, retDtor ) );
+                        auto member = new ast::MemberExpr(loc, global.dtorStruct->members.front().strict_as<ast::DeclWithType>(), new ast::VariableExpr(loc, retDtor ) );
                         auto object = new ast::CastExpr( new ast::AddressExpr( new ast::VariableExpr(loc, ret ) ), new ast::PointerType(new ast::VoidType() ) );
                         ast::Expr * assign = createBitwiseAssignment( member, object );
 …
         // to prevent warnings ('_unq0' may be used uninitialized in this function),
         // insert an appropriate zero initializer for UniqueExpr temporaries.
         ast::Init * makeInit( const ast::Type * t ) {
+        ast::Init * makeInit( const ast::Type * t, CodeLocation const & loc ) {
                 if ( auto inst = dynamic_cast< const ast::StructInstType * >( t ) ) {
                         // initizer for empty struct must be empty
+                        if ( inst->base->members.empty() ) return new ast::ListInit({}, {});
+                        if ( inst->base->members.empty() ) {
+                                return new ast::ListInit( loc, {} );
+                        }
                 } else if ( auto inst = dynamic_cast< const ast::UnionInstType * >( t ) ) {
                         // initizer for empty union must be empty
+                        if ( inst->base->members.empty() ) return new ast::ListInit({}, {});
+                }
+                return new ast::ListInit( {}, { new ast::SingleInit( {}, ast::ConstantExpr::from_int({}, 0) ) } );
+                        if ( inst->base->members.empty() ) {
+                                return new ast::ListInit( loc, {} );
+                        }
+                }
+                return new ast::ListInit( loc, {
+                        new ast::SingleInit( loc, ast::ConstantExpr::from_int( loc, 0 ) )
+                } );
+        }
 …
                         } else {
                                 // expr isn't a call expr, so create a new temporary variable to use to hold the value of the unique expression
                                 mutExpr->object = new ast::ObjectDecl( mutExpr->location, toString("_unq", mutExpr->id), mutExpr->result, makeInit( mutExpr->result ) );
+                                mutExpr->object = new ast::ObjectDecl( mutExpr->location, toString("_unq", mutExpr->id), mutExpr->result, makeInit( mutExpr->result, mutExpr->location ) );
                                 mutExpr->var = new ast::VariableExpr( mutExpr->location, mutExpr->object );
+                        }
 …
                         auto guard = makeFuncGuard( [this]() { symtab.enterScope(); }, [this]() { symtab.leaveScope(); } );
                         symtab.addFunction( function );
+                        auto global = transUnit().global;
                         // need to iterate through members in reverse in order for
 …
                                                         static UniqueName memberDtorNamer = { "__memberDtor" };
                                                         assertf( ast::dtorStruct, "builtin __Destructor not found." );
                                                         assertf( ast::dtorStructDestroy, "builtin __destroy_Destructor not found." );
+                                                        assertf( global.dtorStruct, "builtin __Destructor not found." );
+                                                        assertf( global.dtorDestroy, "builtin __destroy_Destructor not found." );
                                                         ast::Expr * thisExpr = new ast::CastExpr( new ast::AddressExpr( new ast::VariableExpr(loc, thisParam ) ), new ast::PointerType( new ast::VoidType(), ast::CV::Qualifiers() ) );
 …
                                                         auto dtorType = new ast::PointerType( dtorFtype );
                                                         auto destructor = new ast::ObjectDecl(loc, memberDtorNamer.newName(), new ast::StructInstType( ast::dtorStruct ), new ast::ListInit(loc, { new ast::SingleInit(loc, thisExpr ), new ast::SingleInit(loc, new ast::CastExpr( dtorExpr, dtorType ) ) } ) );
                                                         destructor->attributes.push_back( new ast::Attribute( "cleanup", { new ast::VariableExpr({}, ast::dtorStructDestroy ) } ) );
+                                                        auto destructor = new ast::ObjectDecl(loc, memberDtorNamer.newName(), new ast::StructInstType( global.dtorStruct ), new ast::ListInit(loc, { new ast::SingleInit(loc, thisExpr ), new ast::SingleInit(loc, new ast::CastExpr( dtorExpr, dtorType ) ) } ) );
+                                                        destructor->attributes.push_back( new ast::Attribute( "cleanup", { new ast::VariableExpr( loc, global.dtorDestroy ) } ) );
                                                         mutStmts->push_front( new ast::DeclStmt(loc, destructor ) );
                                                         mutStmts->kids.splice( mutStmts->kids.begin(), stmtsToAdd );
 …
         const ast::Expr * GenStructMemberCalls::postvisit( const ast::UntypedExpr * untypedExpr ) {
-                // Expression * newExpr = untypedExpr;
                 // xxx - functions returning ast::ptr seems wrong...
+                auto res = ResolvExpr::findVoidExpression( untypedExpr, symtab );
+                return res.release();
+                // return newExpr;
+                auto res = ResolvExpr::findVoidExpression( untypedExpr, { symtab, transUnit().global } );
+                // Fix CodeLocation (at least until resolver is fixed).
+                auto fix = localFillCodeLocations( untypedExpr->location, res.release() );
+                return strict_dynamic_cast<const ast::Expr *>( fix );
+        }
 …
                 // resolve assignment and dispose of new env
                 auto resolved = ResolvExpr::findVoidExpression( assign, symtab );
+                auto resolved = ResolvExpr::findVoidExpression( assign, { symtab, transUnit().global } );
                 auto mut = resolved.get_and_mutate();
                 assertf(resolved.get() == mut, "newly resolved expression must be unique");

src/InitTweak/GenInit.cc

-              rba897d21
+              r2e9b59b
                                         retVal->location, "?{}", retVal, stmt->expr );
                                 assertf( ctorStmt,
                                         "ReturnFixer: genCtorDtor returned nllptr: %s / %s",
+                                        "ReturnFixer: genCtorDtor returned nullptr: %s / %s",
                                         toString( retVal ).c_str(),
                                         toString( stmt->expr ).c_str() );
                                         stmtsToAddBefore.push_back( ctorStmt );
+                                stmtsToAddBefore.push_back( ctorStmt );
                                 // Return the retVal object.
 …
         void genInit( ast::TranslationUnit & transUnit ) {
                 ast::Pass<HoistArrayDimension_NoResolve_New>::run( transUnit );
+                ast::Pass<ReturnFixer_New>::run( transUnit );
+        }
+        void fixReturnStatements( ast::TranslationUnit & transUnit ) {
                 ast::Pass<ReturnFixer_New>::run( transUnit );
+        }

src/InitTweak/GenInit.h

-              rba897d21
+              r2e9b59b
 // Created On       : Mon May 18 07:44:20 2015
 // Last Modified By : Andrew Beach
 // Last Modified On : Fri Oct 22 16:08:00 2021
 // Update Count     : 6
+// Last Modified On : Fri Mar 18 14:22:00 2022
+// Update Count     : 7
 //
 …
         /// Converts return statements into copy constructor calls on the hidden return variable
         void fixReturnStatements( std::list< Declaration * > & translationUnit );
+        void fixReturnStatements( ast::TranslationUnit & translationUnit );
         /// generates a single ctor/dtor statement using objDecl as the 'this' parameter and arg as the optional argument

src/InitTweak/InitTweak.cc

-              rba897d21
+              r2e9b59b
                                 loc, targetLabel.newName(), { new ast::Attribute{ "unused" } } };
                         std::vector< ast::ptr< ast::Stmt > > branches;
+                        std::vector< ast::ptr< ast::CaseClause > > branches;
                         for ( const ast::Init * init : *listInit ) {
                                 auto condition = ast::ConstantExpr::from_ulong( loc, cond );
 …
                                 stmts.emplace_back(
                                         new ast::BranchStmt{ loc, ast::BranchStmt::Break, switchLabel } );
                                 branches.emplace_back( new ast::CaseStmt{ loc, condition, std::move( stmts ) } );
+                                branches.emplace_back( new ast::CaseClause{ loc, condition, std::move( stmts ) } );
+                        }
                         out.emplace_back( new ast::SwitchStmt{ loc, index, std::move( branches ) } );

src/Parser/DeclarationNode.cc

-              rba897d21
+              r2e9b59b
         delete variable.initializer;
         delete type;
+//      delete type;
         delete bitfieldWidth;
 …
 } // DeclarationNode::newAggregate
 DeclarationNode * DeclarationNode::newEnum( const string * name, DeclarationNode * constants, bool body ) {
+DeclarationNode * DeclarationNode::newEnum( const string * name, DeclarationNode * constants, bool body, bool typed) {
         DeclarationNode * newnode = new DeclarationNode;
         newnode->type = new TypeData( TypeData::Enum );
 …
 } // DeclarationNode::newEnum
 DeclarationNode * DeclarationNode::newName( const string * name ) {
         DeclarationNode * newnode = new DeclarationNode;
 …
 } // DeclarationNode::newName
 DeclarationNode * DeclarationNode::newEnumConstant( const string * name, ExpressionNode * constant ) {
+DeclarationNode * DeclarationNode::newEnumConstant( const string * name, ExpressionNode * constant ) { // Marker
         DeclarationNode * newnode = newName( name );
         newnode->enumeratorValue.reset( constant );
 …
+}
+DeclarationNode * DeclarationNode::addEnumBase( DeclarationNode * o ) {
+        if ( o && o -> type)  {
+                type->base= o->type;
+        }
+        delete o;
+        return this;
+}
 DeclarationNode * DeclarationNode::addTypedef() {
         TypeData * newtype = new TypeData( TypeData::Symbolic );

src/Parser/ParseNode.h

-              rba897d21
+              r2e9b59b
         static DeclarationNode * newFunction( const std::string * name, DeclarationNode * ret, DeclarationNode * param, StatementNode * body );
         static DeclarationNode * newAggregate( AggregateDecl::Aggregate kind, const std::string * name, ExpressionNode * actuals, DeclarationNode * fields, bool body );
         static DeclarationNode * newEnum( const std::string * name, DeclarationNode * constants, bool body );
+        static DeclarationNode * newEnum( const std::string * name, DeclarationNode * constants, bool body, bool typed );
         static DeclarationNode * newEnumConstant( const std::string * name, ExpressionNode * constant );
         static DeclarationNode * newName( const std::string * );
 …
         DeclarationNode * addType( DeclarationNode * );
         DeclarationNode * addTypedef();
+        DeclarationNode * addEnumBase( DeclarationNode * );
         DeclarationNode * addAssertions( DeclarationNode * );
         DeclarationNode * addName( std::string * );

src/Parser/StatementNode.cc

rba897d21	r2e9b59b
366	366	} // maybe_build_compound
367	367
	368	// Question
368	369	Statement * build_asm( bool voltile, Expression * instruction, ExpressionNode * output, ExpressionNode * input, ExpressionNode * clobber, LabelNode * gotolabels ) {
369	370	list< Expression * > out, in;

src/Parser/TypeData.cc

-              rba897d21
+              r2e9b59b
 EnumDecl * buildEnum( const TypeData * td, std::list< Attribute * > attributes, LinkageSpec::Spec linkage ) {
         assert( td->kind == TypeData::Enum );
+        EnumDecl * ret = new EnumDecl( *td->enumeration.name, attributes, linkage );
+        Type * baseType = td->base ? typebuild(td->base) : nullptr;
+        EnumDecl * ret = new EnumDecl( *td->enumeration.name, attributes, linkage, baseType );
         buildList( td->enumeration.constants, ret->get_members() );
         list< Declaration * >::iterator members = ret->get_members().begin();
         for ( const DeclarationNode * cur = td->enumeration. constants; cur != nullptr; cur = dynamic_cast< DeclarationNode * >( cur->get_next() ), ++members ) {
+        for ( const DeclarationNode * cur = td->enumeration.constants; cur != nullptr; cur = dynamic_cast< DeclarationNode * >( cur->get_next() ), ++members ) {
                 if ( cur->has_enumeratorValue() ) {
                         ObjectDecl * member = dynamic_cast< ObjectDecl * >(* members);
                         member->set_init( new SingleInit( maybeMoveBuild< Expression >( cur->consume_enumeratorValue() ) ) );
+                } else {
+                        if ( baseType && (!dynamic_cast<BasicType *>(baseType) || !dynamic_cast<BasicType *>(baseType)->isWholeNumber())) {
+                                SemanticError( td->location, "A non whole number enum value decl must be explicitly initialized." );
+                        }
                 } // if
         } // for
         ret->set_body( td->enumeration.body );
+        ret->set_body( td->enumeration.body ); // Boolean; if it has body
         return ret;
 } // buildEnum

src/Parser/TypeData.h

rba897d21	r2e9b59b
132	132	Initializer * init = nullptr, std::list< class Attribute * > attributes = std::list< class Attribute * >() );
133	133	FunctionType * buildFunction( const TypeData * );
	134	Declaration * addEnumBase( Declaration , const TypeData );
134	135	void buildKRFunction( const TypeData::Function_t & function );
135	136

src/Parser/parser.yy

-              rba897d21
+              r2e9b59b
 // Created On       : Sat Sep  1 20:22:55 2001
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Fri Feb 25 17:54:56 2022
 // Update Count     : 5262
+// Last Modified On : Mon Mar 14 16:35:29 2022
+// Update Count     : 5276
 //
 …
                         // Historic, transitional: Disallow commas in subscripts.
                         // Switching to this behaviour may help check if a C compatibilty case uses comma-exprs in subscripts.
-                // { SemanticError( yylloc, "New array subscript is currently unimplemented." ); $$ = nullptr; }
                         // Current: Commas in subscripts make tuples.
                 { $$ = new ExpressionNode( build_binary_val( OperKinds::Index, $1, new ExpressionNode( build_tuple( (ExpressionNode *)($3->set_last( $5 ) ) )) ) ); }
 …
                 // equivalent to the old x[i,j].
                 { $$ = new ExpressionNode( build_binary_val( OperKinds::Index, $1, $3 ) ); }
+        | constant '[' assignment_expression ']'                        // 3[a], 'a'[a], 3.5[a]
+                { $$ = new ExpressionNode( build_binary_val( OperKinds::Index, $1, $3 ) ); }
+        | string_literal '[' assignment_expression ']'          // "abc"[3], 3["abc"]
+                { $$ = new ExpressionNode( build_binary_val( OperKinds::Index, new ExpressionNode( $1 ), $3 ) ); }
         | postfix_expression '{' argument_expression_list_opt '}' // CFA, constructor call
+                {
 …
+        ;
 enum_type:                                                                                              // enum
+enum_type: // static DeclarationNode * newEnum( const std::string * name, DeclarationNode * constants, bool body, bool typed );                                                                                         // enum
         ENUM attribute_list_opt '{' enumerator_list comma_opt '}'
                 { $$ = DeclarationNode::newEnum( nullptr, $4, true )->addQualifiers( $2 ); }
+                { $$ = DeclarationNode::newEnum( nullptr, $4, true, false )->addQualifiers( $2 ); }
         | ENUM attribute_list_opt identifier
                 { typedefTable.makeTypedef( *$3 ); }
           '{' enumerator_list comma_opt '}'
                 { $$ = DeclarationNode::newEnum( $3, $6, true )->addQualifiers( $2 ); }
+                { $$ = DeclarationNode::newEnum( $3, $6, true, false )->addQualifiers( $2 ); }
         | ENUM attribute_list_opt typedef_name                          // unqualified type name
           '{' enumerator_list comma_opt '}'
                 { $$ = DeclarationNode::newEnum( $3->name, $5, true )->addQualifiers( $2 ); }
+                { $$ = DeclarationNode::newEnum( $3->name, $5, true, false )->addQualifiers( $2 ); }
         | ENUM '(' cfa_abstract_parameter_declaration ')' attribute_list_opt '{' enumerator_list comma_opt '}'
+                {
+                        if ( $3->storageClasses.val != 0 || $3->type->qualifiers.val != 0 ) { SemanticError( yylloc, "storage-class and CV qualifiers are not meaningful for enumeration constants, which are const." ); }
+                        SemanticError( yylloc, "Typed enumeration is currently unimplemented." ); $$ = nullptr;
+                }
+        | ENUM '(' cfa_abstract_parameter_declaration ')' attribute_list_opt identifier attribute_list_opt
+                        if ( $3->storageClasses.val != 0 || $3->type->qualifiers.val != 0 )
+                        { SemanticError( yylloc, "storage-class and CV qualifiers are not meaningful for enumeration constants, which are const." ); }
+                        // SemanticError( yylloc, "Typed enumeration is currently unimplemented." ); $$ = nullptr;
+                        $$ = DeclarationNode::newEnum( nullptr, $7, true, true ) ->addQualifiers( $5 )  -> addEnumBase( $3 );
+                        // $$ = DeclarationNode::newEnum( nullptr, $7, true, true ) ->addQualifiers( $5 );
+                }
+        | ENUM '(' cfa_abstract_parameter_declaration ')' attribute_list_opt identifier attribute_list_opt // Question: why attributes/qualifier after identifier
+                {
                         if ( $3->storageClasses.val != 0 || $3->type->qualifiers.val != 0 ) { SemanticError( yylloc, "storage-class and CV qualifiers are not meaningful for enumeration constants, which are const." ); }
 …
           '{' enumerator_list comma_opt '}'
+                {
+                        SemanticError( yylloc, "Typed enumeration is currently unimplemented." ); $$ = nullptr;
+                        $$ = DeclarationNode::newEnum( $6, $10, true, true ) -> addQualifiers( $5 ) -> addQualifiers( $7 ) -> addEnumBase( $3 );
+                        // $$ = DeclarationNode::newEnum( $6, $10, true, true ) -> addQualifiers( $5 ) -> addQualifiers( $7 );
+                }
         | ENUM '(' cfa_abstract_parameter_declaration ')' attribute_list_opt typedef_name attribute_list_opt '{' enumerator_list comma_opt '}'
 …
                         if ( $3->storageClasses.val != 0 || $3->type->qualifiers.val != 0 ) { SemanticError( yylloc, "storage-class and CV qualifiers are not meaningful for enumeration constants, which are const." ); }
                         typedefTable.makeTypedef( *$6->name );
+                        SemanticError( yylloc, "Typed enumeration is currently unimplemented." ); $$ = nullptr;
+                        $$ = DeclarationNode::newEnum( $6->name, $9, true, true ) -> addQualifiers( $5 ) -> addQualifiers( $7 ) -> addEnumBase( $3 );
+                        // $$ = DeclarationNode::newEnum( $6->name, $9, true, true ) -> addQualifiers( $5 ) -> addQualifiers( $7 );
+                }
         | enum_type_nobody
 …
 enum_type_nobody:                                                                               // enum - {...}
         ENUM attribute_list_opt identifier
                 { typedefTable.makeTypedef( *$3 ); $$ = DeclarationNode::newEnum( $3, 0, false )->addQualifiers( $2 ); }
+                { typedefTable.makeTypedef( *$3 ); $$ = DeclarationNode::newEnum( $3, 0, false, false )->addQualifiers( $2 ); }
         | ENUM attribute_list_opt type_name                                     // qualified type name
                 { typedefTable.makeTypedef( *$3->type->symbolic.name ); $$ = DeclarationNode::newEnum( $3->type->symbolic.name, 0, false )->addQualifiers( $2 ); }
+                { typedefTable.makeTypedef( *$3->type->symbolic.name ); $$ = DeclarationNode::newEnum( $3->type->symbolic.name, 0, false, false )->addQualifiers( $2 ); }
+        ;

src/ResolvExpr/CandidateFinder.cpp

-              rba897d21
+              r2e9b59b
 // Created On       : Wed Jun 5 14:30:00 2019
 // Last Modified By : Andrew Beach
 // Last Modified On : Tue Oct  1 14:55:00 2019
 // Update Count     : 2
+// Last Modified On : Wed Mar 16 11:58:00 2022
+// Update Count     : 3
 //
 …
         /// Actually visits expressions to find their candidate interpretations
         class Finder final : public ast::WithShortCircuiting {
+                const ResolveContext & context;
                 const ast::SymbolTable & symtab;
         public:
 …
                 Finder( CandidateFinder & f )
                 : symtab( f.localSyms ), selfFinder( f ), candidates( f.candidates ), tenv( f.env ),
                   targetType( f.targetType ) {}
+                : context( f.context ), symtab( context.symtab ), selfFinder( f ),
+                  candidates( f.candidates ), tenv( f.env ), targetType( f.targetType ) {}
                 void previsit( const ast::Node * ) { visit_children = false; }
 …
                         Tuples::handleTupleAssignment( selfFinder, untypedExpr, argCandidates );
                         CandidateFinder funcFinder{ symtab, tenv };
+                        CandidateFinder funcFinder( context, tenv );
                         if (auto nameExpr = untypedExpr->func.as<ast::NameExpr>()) {
                                 auto kind = ast::SymbolTable::getSpecialFunctionKind(nameExpr->name);
 …
                         // find function operators
                         ast::ptr< ast::Expr > opExpr = new ast::NameExpr{ untypedExpr->location, "?()" };
                         CandidateFinder opFinder{ symtab, tenv };
+                        CandidateFinder opFinder( context, tenv );
                         // okay if there aren't any function operations
                         opFinder.find( opExpr, ResolvMode::withoutFailFast() );
 …
                 void postvisit( const ast::AddressExpr * addressExpr ) {
                         CandidateFinder finder{ symtab, tenv };
+                        CandidateFinder finder( context, tenv );
                         finder.find( addressExpr->arg );
 …
                         ast::ptr< ast::Type > toType = castExpr->result;
                         assert( toType );
                         toType = resolveTypeof( toType, symtab );
+                        toType = resolveTypeof( toType, context );
                         // toType = SymTab::validateType( castExpr->location, toType, symtab );
                         toType = adjustExprType( toType, tenv, symtab );
                         CandidateFinder finder{ symtab, tenv, toType };
+                        CandidateFinder finder( context, tenv, toType );
                         finder.find( castExpr->arg, ResolvMode::withAdjustment() );
 …
                 void postvisit( const ast::VirtualCastExpr * castExpr ) {
                         assertf( castExpr->result, "Implicit virtual cast targets not yet supported." );
                         CandidateFinder finder{ symtab, tenv };
+                        CandidateFinder finder( context, tenv );
                         // don't prune here, all alternatives guaranteed to have same type
                         finder.find( castExpr->arg, ResolvMode::withoutPrune() );
 …
                         auto target = inst->base.get();
                         CandidateFinder finder{ symtab, tenv };
+                        CandidateFinder finder( context, tenv );
                         auto pick_alternatives = [target, this](CandidateList & found, bool expect_ref) {
 …
                 void postvisit( const ast::UntypedMemberExpr * memberExpr ) {
                         CandidateFinder aggFinder{ symtab, tenv };
+                        CandidateFinder aggFinder( context, tenv );
                         aggFinder.find( memberExpr->aggregate, ResolvMode::withAdjustment() );
                         for ( CandidateRef & agg : aggFinder.candidates ) {
 …
                                 addCandidate(
                                         new ast::SizeofExpr{
                                                 sizeofExpr->location, resolveTypeof( sizeofExpr->type, symtab ) },
+                                                sizeofExpr->location, resolveTypeof( sizeofExpr->type, context ) },
                                         tenv );
                         } else {
                                 // find all candidates for the argument to sizeof
                                 CandidateFinder finder{ symtab, tenv };
+                                CandidateFinder finder( context, tenv );
                                 finder.find( sizeofExpr->expr );
                                 // find the lowest-cost candidate, otherwise ambiguous
 …
                                 addCandidate(
                                         new ast::AlignofExpr{
                                                 alignofExpr->location, resolveTypeof( alignofExpr->type, symtab ) },
+                                                alignofExpr->location, resolveTypeof( alignofExpr->type, context ) },
                                         tenv );
                         } else {
                                 // find all candidates for the argument to alignof
                                 CandidateFinder finder{ symtab, tenv };
+                                CandidateFinder finder( context, tenv );
                                 finder.find( alignofExpr->expr );
                                 // find the lowest-cost candidate, otherwise ambiguous
 …
                 void postvisit( const ast::LogicalExpr * logicalExpr ) {
                         CandidateFinder finder1{ symtab, tenv };
+                        CandidateFinder finder1( context, tenv );
                         finder1.find( logicalExpr->arg1, ResolvMode::withAdjustment() );
                         if ( finder1.candidates.empty() ) return;
                         CandidateFinder finder2{ symtab, tenv };
+                        CandidateFinder finder2( context, tenv );
                         finder2.find( logicalExpr->arg2, ResolvMode::withAdjustment() );
                         if ( finder2.candidates.empty() ) return;
 …
                 void postvisit( const ast::ConditionalExpr * conditionalExpr ) {
                         // candidates for condition
                         CandidateFinder finder1{ symtab, tenv };
+                        CandidateFinder finder1( context, tenv );
                         finder1.find( conditionalExpr->arg1, ResolvMode::withAdjustment() );
                         if ( finder1.candidates.empty() ) return;
                         // candidates for true result
                         CandidateFinder finder2{ symtab, tenv };
+                        CandidateFinder finder2( context, tenv );
                         finder2.find( conditionalExpr->arg2, ResolvMode::withAdjustment() );
                         if ( finder2.candidates.empty() ) return;
                         // candidates for false result
                         CandidateFinder finder3{ symtab, tenv };
+                        CandidateFinder finder3( context, tenv );
                         finder3.find( conditionalExpr->arg3, ResolvMode::withAdjustment() );
                         if ( finder3.candidates.empty() ) return;
 …
                 void postvisit( const ast::CommaExpr * commaExpr ) {
                         ast::TypeEnvironment env{ tenv };
                         ast::ptr< ast::Expr > arg1 = resolveInVoidContext( commaExpr->arg1, symtab, env );
                         CandidateFinder finder2{ symtab, env };
+                        ast::ptr< ast::Expr > arg1 = resolveInVoidContext( commaExpr->arg1, context, env );
+                        CandidateFinder finder2( context, env );
                         finder2.find( commaExpr->arg2, ResolvMode::withAdjustment() );
 …
                 void postvisit( const ast::ConstructorExpr * ctorExpr ) {
                         CandidateFinder finder{ symtab, tenv };
+                        CandidateFinder finder( context, tenv );
                         finder.find( ctorExpr->callExpr, ResolvMode::withoutPrune() );
                         for ( CandidateRef & r : finder.candidates ) {
 …
                 void postvisit( const ast::RangeExpr * rangeExpr ) {
                         // resolve low and high, accept candidates where low and high types unify
                         CandidateFinder finder1{ symtab, tenv };
+                        CandidateFinder finder1( context, tenv );
                         finder1.find( rangeExpr->low, ResolvMode::withAdjustment() );
                         if ( finder1.candidates.empty() ) return;
                         CandidateFinder finder2{ symtab, tenv };
+                        CandidateFinder finder2( context, tenv );
                         finder2.find( rangeExpr->high, ResolvMode::withAdjustment() );
                         if ( finder2.candidates.empty() ) return;
 …
                 void postvisit( const ast::UniqueExpr * unqExpr ) {
                         CandidateFinder finder{ symtab, tenv };
+                        CandidateFinder finder( context, tenv );
                         finder.find( unqExpr->expr, ResolvMode::withAdjustment() );
                         for ( CandidateRef & r : finder.candidates ) {
 …
                 void postvisit( const ast::StmtExpr * stmtExpr ) {
                         addCandidate( resolveStmtExpr( stmtExpr, symtab ), tenv );
+                        addCandidate( resolveStmtExpr( stmtExpr, context ), tenv );
+                }
 …
                         for ( const ast::InitAlternative & initAlt : initExpr->initAlts ) {
                                 // calculate target type
                                 const ast::Type * toType = resolveTypeof( initAlt.type, symtab );
+                                const ast::Type * toType = resolveTypeof( initAlt.type, context );
                                 // toType = SymTab::validateType( initExpr->location, toType, symtab );
                                 toType = adjustExprType( toType, tenv, symtab );
 …
                                 // types are not bound to the initialization type, since return type variables are
                                 // only open for the duration of resolving the UntypedExpr.
                                 CandidateFinder finder{ symtab, tenv, toType };
+                                CandidateFinder finder( context, tenv, toType );
                                 finder.find( initExpr->expr, ResolvMode::withAdjustment() );
                                 for ( CandidateRef & cand : finder.candidates ) {
 …
+                }
                 else {
                         satisfyAssertions(candidate, localSyms, satisfied, errors);
+                        satisfyAssertions(candidate, context.symtab, satisfied, errors);
                         needRecomputeKey = true;
+                }
 …
                         r->expr = ast::mutate_field(
                                 r->expr.get(), &ast::Expr::result,
                                 adjustExprType( r->expr->result, r->env, localSyms ) );
+                                adjustExprType( r->expr->result, r->env, context.symtab ) );
+                }
+        }
 …
         for ( const auto & x : xs ) {
                 out.emplace_back( localSyms, env );
+                out.emplace_back( context, env );
                 out.back().find( x, ResolvMode::withAdjustment() );

src/ResolvExpr/CandidateFinder.hpp

-              rba897d21
+              r2e9b59b
 // Created On       : Wed Jun 5 14:30:00 2019
 // Last Modified By : Andrew Beach
 // Last Modified On : Tue Oct  1  9:51:00 2019
 // Update Count     : 2
+// Last Modified On : Wed Mar 16 15:22:00 2022
+// Update Count     : 3
 //
 …
 namespace ResolvExpr {
+struct ResolveContext;
 /// Data to perform expression resolution
 struct CandidateFinder {
         CandidateList candidates;          ///< List of candidate resolutions
         const ast::SymbolTable & localSyms;   ///< Symbol table to lookup candidates
+        const ResolveContext & context;  ///< Information about where the canditates are being found.
         const ast::TypeEnvironment & env;  ///< Substitutions performed in this resolution
         ast::ptr< ast::Type > targetType;  ///< Target type for resolution
 …
         CandidateFinder(
                 const ast::SymbolTable & syms, const ast::TypeEnvironment & env,
+                const ResolveContext & context, const ast::TypeEnvironment & env,
                 const ast::Type * tt = nullptr )
         : candidates(), localSyms( syms ), env( env ), targetType( tt ) {}
+        : candidates(), context( context ), env( env ), targetType( tt ) {}
         /// Fill candidates with feasible resolutions for `expr`

src/ResolvExpr/CandidatePrinter.cpp

-              rba897d21
+              r2e9b59b
 // Created On       : Tue Nov  9  9:54:00 2021
 // Last Modified By : Andrew Beach
 // Last Modified On : Tue Nov  9 15:47:00 2021
 // Update Count     : 0
+// Last Modified On : Wed Mar 16 13:56:00 2022
+// Update Count     : 1
 //
 …
 #include "AST/TranslationUnit.hpp"
 #include "ResolvExpr/CandidateFinder.hpp"
+#include "ResolvExpr/Resolver.h"
 #include <iostream>
 …
 namespace {
+class CandidatePrintCore : public ast::WithSymbolTable {
+class CandidatePrintCore : public ast::WithSymbolTable,
+                public ast::WithConstTranslationUnit {
         std::ostream & os;
 public:
 …
         void postvisit( const ast::ExprStmt * stmt ) {
                 ast::TypeEnvironment env;
                 CandidateFinder finder( symtab, env );
+                CandidateFinder finder( { symtab, transUnit().global }, env );
                 finder.find( stmt->expr, ResolvMode::withAdjustment() );
                 int count = 1;

src/ResolvExpr/ConversionCost.cc

-              rba897d21
+              r2e9b59b
                 } else if ( dynamic_cast< const EnumInstType * >( dest ) ) {
                         // xxx - not positive this is correct, but appears to allow casting int => enum
+                        cost = Cost::unsafe;
+                        // TODO
+                        EnumDecl * decl = dynamic_cast< const EnumInstType * >( dest )->baseEnum;
+                        if ( decl->base ) {
+                                cost = Cost::infinity;
+                        } else {
+                                cost = Cost::unsafe;
+                        } // if
                 } // if
                 // no cases for zero_t/one_t because it should not be possible to convert int, etc. to zero_t/one_t.
 …
         } else if ( dynamic_cast< const ast::EnumInstType * >( dst ) ) {
                 // xxx - not positive this is correct, but appears to allow casting int => enum
+                cost = Cost::unsafe;
+                const ast::EnumDecl * decl = (dynamic_cast< const ast::EnumInstType * >( dst ))->base.get();
+                if ( decl->base ) {
+                        cost = Cost::infinity;
+                } else {
+                        cost = Cost::unsafe;
+                } // if
+        }
+}

src/ResolvExpr/RenameVars.h

-              rba897d21
+              r2e9b59b
         };
         const ast::Type * renameTyVars( const ast::Type *, RenameMode mode = GEN_USAGE, bool reset = true );
         /// resets internal state of renamer to avoid overflow
         void resetTyVarRenaming();
 } // namespace ResolvExpr

src/ResolvExpr/ResolveTypeof.cc

-              rba897d21
+              r2e9b59b
 // Author           : Richard C. Bilson
 // Created On       : Sun May 17 12:12:20 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Tue May 19 16:49:04 2015
 // Update Count     : 3
+// Last Modified By : Andrew Beach
+// Last Modified On : Wed Mar 16 16:09:00 2022
+// Update Count     : 4
 //
 …
 #include "AST/Node.hpp"
 #include "AST/Pass.hpp"
+#include "AST/TranslationUnit.hpp"
 #include "AST/Type.hpp"
 #include "AST/TypeEnvironment.hpp"
 …
 namespace {
         struct ResolveTypeof_new : public ast::WithShortCircuiting {
+                const ast::SymbolTable & localSymtab;
+                ResolveTypeof_new( const ast::SymbolTable & syms ) : localSymtab( syms ) {}
+                const ResolveContext & context;
+                ResolveTypeof_new( const ResolveContext & context ) :
+                        context( context ) {}
                 void previsit( const ast::TypeofType * ) { visit_children = false; }
 …
                                 ast::TypeEnvironment dummy;
                                 ast::ptr< ast::Expr > newExpr =
                                         resolveInVoidContext( typeofType->expr, localSymtab, dummy );
+                                        resolveInVoidContext( typeofType->expr, context, dummy );
                                 assert( newExpr->result && ! newExpr->result->isVoid() );
                                 newType = newExpr->result;
 …
 } // anonymous namespace
 const ast::Type * resolveTypeof( const ast::Type * type , const ast::SymbolTable & symtab ) {
         ast::Pass< ResolveTypeof_new > mutator{ symtab };
+const ast::Type * resolveTypeof( const ast::Type * type , const ResolveContext & context ) {
+        ast::Pass< ResolveTypeof_new > mutator( context );
         return type->accept( mutator );
+}
 struct FixArrayDimension {
+        // should not require a mutable symbol table - prevent pass template instantiation
+        const ast::SymbolTable & _symtab;
+        FixArrayDimension(const ast::SymbolTable & symtab): _symtab(symtab) {}
+        const ResolveContext & context;
+        FixArrayDimension(const ResolveContext & context) : context( context ) {}
         const ast::ArrayType * previsit (const ast::ArrayType * arrayType) {
                 if (!arrayType->dimension) return arrayType;
                 auto mutType = mutate(arrayType);
+                ast::ptr<ast::Type> sizetype = ast::sizeType ? ast::sizeType : new ast::BasicType(ast::BasicType::LongUnsignedInt);
+                mutType->dimension = findSingleExpression(arrayType->dimension, sizetype, _symtab);
+                auto globalSizeType = context.global.sizeType;
+                ast::ptr<ast::Type> sizetype = globalSizeType ? globalSizeType : new ast::BasicType(ast::BasicType::LongUnsignedInt);
+                mutType->dimension = findSingleExpression(arrayType->dimension, sizetype, context );
                 if (InitTweak::isConstExpr(mutType->dimension)) {
 …
 };
 const ast::Type * fixArrayType( const ast::Type * type, const ast::SymbolTable & symtab) {
         ast::Pass<FixArrayDimension> visitor {symtab};
+const ast::Type * fixArrayType( const ast::Type * type, const ResolveContext & context ) {
+        ast::Pass<FixArrayDimension> visitor(context);
         return type->accept(visitor);
+}
+const ast::ObjectDecl * fixObjectType( const ast::ObjectDecl * decl , const ast::SymbolTable & symtab ) {
+        if (!decl->isTypeFixed) {
+                auto mutDecl = mutate(decl);
+                auto resolvedType = resolveTypeof(decl->type, symtab);
+                resolvedType = fixArrayType(resolvedType, symtab);
+const ast::ObjectDecl * fixObjectType( const ast::ObjectDecl * decl , const ResolveContext & context ) {
+        if (decl->isTypeFixed) {
+                return decl;
+        }
+        auto mutDecl = mutate(decl);
+        {
+                auto resolvedType = resolveTypeof(decl->type, context);
+                resolvedType = fixArrayType(resolvedType, context);
                 mutDecl->type = resolvedType;
+                // check variable length if object is an array.
+                // xxx - should this be part of fixObjectType?
+                /*
+                if (auto arrayType = dynamic_cast<const ast::ArrayType *>(resolvedType)) {
+                        auto dimExpr = findSingleExpression(arrayType->dimension, ast::sizeType, symtab);
+                        if (auto varexpr = arrayType->dimension.as<ast::VariableExpr>()) {// hoisted previously
+                                if (InitTweak::isConstExpr(varexpr->var.strict_as<ast::ObjectDecl>()->init)) {
+                                        auto mutType = mutate(arrayType);
+                                        mutType->isVarLen = ast::LengthFlag::VariableLen;
+                                        mutDecl->type = mutType;
+                                }
+                        }
+                }
+                */
+                if (!mutDecl->name.empty())
+                        mutDecl->mangleName = Mangle::mangle(mutDecl); // do not mangle unnamed variables
+                mutDecl->type = renameTyVars(mutDecl->type, RenameMode::GEN_EXPR_ID);
+                mutDecl->isTypeFixed = true;
+                return mutDecl;
+        }
+        return decl;
+        }
+        // Do not mangle unnamed variables.
+        if (!mutDecl->name.empty()) {
+                mutDecl->mangleName = Mangle::mangle(mutDecl);
+        }
+        mutDecl->type = renameTyVars(mutDecl->type, RenameMode::GEN_EXPR_ID);
+        mutDecl->isTypeFixed = true;
+        return mutDecl;
+}

src/ResolvExpr/ResolveTypeof.h

-              rba897d21
+              r2e9b59b
 // file "LICENCE" distributed with Cforall.
 //
 // ResolveTypeof.h --
+// ResolveTypeof.h --
 //
 // Author           : Richard C. Bilson
 // Created On       : Sun May 17 12:14:53 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Sat Jul 22 09:38:35 2017
 // Update Count     : 3
+// Last Modified By : Andrew Beach
+// Last Modified On : Wed Mar 16 11:33:00 2022
+// Update Count     : 4
 //
 …
 namespace ast {
         class Type;
-        class SymbolTable;
         class ObjectDecl;
+}
 namespace ResolvExpr {
+        struct ResolveContext;
         Type *resolveTypeof( Type*, const SymTab::Indexer &indexer );
         const ast::Type * resolveTypeof( const ast::Type *, const ast::SymbolTable & );
         const ast::ObjectDecl * fixObjectType( const ast::ObjectDecl * decl , const ast::SymbolTable & symtab );
+        const ast::Type * resolveTypeof( const ast::Type *, const ResolveContext & );
+        const ast::ObjectDecl * fixObjectType( const ast::ObjectDecl * decl , const ResolveContext & );
 } // namespace ResolvExpr

src/ResolvExpr/Resolver.cc

-              rba897d21
+              r2e9b59b
 // Author           : Aaron B. Moss
 // Created On       : Sun May 17 12:17:01 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Tue Feb  1 16:27:14 2022
 // Update Count     : 245
+// Last Modified By : Andrew Beach
+// Last Modified On : Fri Mar 18 10:41:00 2022
+// Update Count     : 247
 //
 …
                 /// Calls the CandidateFinder and finds the single best candidate
                 CandidateRef findUnfinishedKindExpression(
                         const ast::Expr * untyped, const ast::SymbolTable & symtab, const std::string & kind,
+                        const ast::Expr * untyped, const ResolveContext & context, const std::string & kind,
                         std::function<bool(const Candidate &)> pred = anyCandidate, ResolvMode mode = {}
                 ) {
 …
                         ++recursion_level;
                         ast::TypeEnvironment env;
                         CandidateFinder finder{ symtab, env };
+                        CandidateFinder finder( context, env );
                         finder.find( untyped, recursion_level == 1 ? mode.atTopLevel() : mode );
                         --recursion_level;
 …
         ast::ptr< ast::Expr > resolveInVoidContext(
+                const ast::Expr * expr, const ast::SymbolTable & symtab, ast::TypeEnvironment & env
+                const ast::Expr * expr, const ResolveContext & context,
+                ast::TypeEnvironment & env
         ) {
                 assertf( expr, "expected a non-null expression" );
 …
                 ast::ptr< ast::CastExpr > untyped = new ast::CastExpr{ expr };
                 CandidateRef choice = findUnfinishedKindExpression(
                         untyped, symtab, "", anyCandidate, ResolvMode::withAdjustment() );
+                        untyped, context, "", anyCandidate, ResolvMode::withAdjustment() );
                 // a cast expression has either 0 or 1 interpretations (by language rules);
 …
                 /// context.
                 ast::ptr< ast::Expr > findVoidExpression(
                         const ast::Expr * untyped, const ast::SymbolTable & symtab
+                        const ast::Expr * untyped, const ResolveContext & context
                 ) {
                         ast::TypeEnvironment env;
                         ast::ptr< ast::Expr > newExpr = resolveInVoidContext( untyped, symtab, env );
+                        ast::ptr< ast::Expr > newExpr = resolveInVoidContext( untyped, context, env );
                         finishExpr( newExpr, env, untyped->env );
                         return newExpr;
 …
                 /// lowest cost, returning the resolved version
                 ast::ptr< ast::Expr > findKindExpression(
                         const ast::Expr * untyped, const ast::SymbolTable & symtab,
+                        const ast::Expr * untyped, const ResolveContext & context,
                         std::function<bool(const Candidate &)> pred = anyCandidate,
                         const std::string & kind = "", ResolvMode mode = {}
 …
                         if ( ! untyped ) return {};
                         CandidateRef choice =
                                 findUnfinishedKindExpression( untyped, symtab, kind, pred, mode );
+                                findUnfinishedKindExpression( untyped, context, kind, pred, mode );
                         ResolvExpr::finishExpr( choice->expr, choice->env, untyped->env );
                         return std::move( choice->expr );
 …
                 /// Resolve `untyped` to the single expression whose candidate is the best match
                 ast::ptr< ast::Expr > findSingleExpression(
                         const ast::Expr * untyped, const ast::SymbolTable & symtab
+                        const ast::Expr * untyped, const ResolveContext & context
                 ) {
                         Stats::ResolveTime::start( untyped );
                         auto res = findKindExpression( untyped, symtab );
+                        auto res = findKindExpression( untyped, context );
                         Stats::ResolveTime::stop();
                         return res;
 …
         ast::ptr< ast::Expr > findSingleExpression(
+                const ast::Expr * untyped, const ast::Type * type, const ast::SymbolTable & symtab
+                const ast::Expr * untyped, const ast::Type * type,
+                const ResolveContext & context
         ) {
                 assert( untyped && type );
                 ast::ptr< ast::Expr > castExpr = new ast::CastExpr{ untyped, type };
                 ast::ptr< ast::Expr > newExpr = findSingleExpression( castExpr, symtab );
                 removeExtraneousCast( newExpr, symtab );
+                ast::ptr< ast::Expr > newExpr = findSingleExpression( castExpr, context );
+                removeExtraneousCast( newExpr, context.symtab );
                 return newExpr;
+        }
 …
                 /// Resolve `untyped` as an integral expression, returning the resolved version
                 ast::ptr< ast::Expr > findIntegralExpression(
                         const ast::Expr * untyped, const ast::SymbolTable & symtab
+                        const ast::Expr * untyped, const ResolveContext & context
                 ) {
                         return findKindExpression( untyped, symtab, hasIntegralType, "condition" );
+                        return findKindExpression( untyped, context, hasIntegralType, "condition" );
+                }
 …
                 // for work previously in GenInit
                 static InitTweak::ManagedTypes_new managedTypes;
+                ResolveContext context;
                 bool inEnumDecl = false;
 …
         public:
                 static size_t traceId;
+                Resolver_new() = default;
+                Resolver_new( const ast::SymbolTable & syms ) { symtab = syms; }
+                Resolver_new( const ast::TranslationGlobal & global ) :
+                        context{ symtab, global } {}
+                Resolver_new( const ResolveContext & context ) :
+                        ast::WithSymbolTable{ context.symtab },
+                        context{ symtab, context.global } {}
                 const ast::FunctionDecl * previsit( const ast::FunctionDecl * );
 …
                 const ast::AsmStmt *         previsit( const ast::AsmStmt * );
                 const ast::IfStmt *          previsit( const ast::IfStmt * );
                 const ast::WhileDoStmt *       previsit( const ast::WhileDoStmt * );
+                const ast::WhileDoStmt *     previsit( const ast::WhileDoStmt * );
                 const ast::ForStmt *         previsit( const ast::ForStmt * );
                 const ast::SwitchStmt *      previsit( const ast::SwitchStmt * );
                 const ast::CaseStmt *        previsit( const ast::CaseStmt * );
+                const ast::CaseClause *      previsit( const ast::CaseClause * );
                 const ast::BranchStmt *      previsit( const ast::BranchStmt * );
                 const ast::ReturnStmt *      previsit( const ast::ReturnStmt * );
                 const ast::ThrowStmt *       previsit( const ast::ThrowStmt * );
                 const ast::CatchStmt *       previsit( const ast::CatchStmt * );
                 const ast::CatchStmt *       postvisit( const ast::CatchStmt * );
+                const ast::CatchClause *     previsit( const ast::CatchClause * );
+                const ast::CatchClause *     postvisit( const ast::CatchClause * );
                 const ast::WaitForStmt *     previsit( const ast::WaitForStmt * );
                 const ast::WithStmt *        previsit( const ast::WithStmt * );
 …
         void resolve( ast::TranslationUnit& translationUnit ) {
                 ast::Pass< Resolver_new >::run( translationUnit );
+                ast::Pass< Resolver_new >::run( translationUnit, translationUnit.global );
+        }
         ast::ptr< ast::Init > resolveCtorInit(
                 const ast::ConstructorInit * ctorInit, const ast::SymbolTable & symtab
+                const ast::ConstructorInit * ctorInit, const ResolveContext & context
         ) {
                 assert( ctorInit );
                 ast::Pass< Resolver_new > resolver{ symtab };
+                ast::Pass< Resolver_new > resolver( context );
                 return ctorInit->accept( resolver );
+        }
         const ast::Expr * resolveStmtExpr(
                 const ast::StmtExpr * stmtExpr, const ast::SymbolTable & symtab
+                const ast::StmtExpr * stmtExpr, const ResolveContext & context
         ) {
                 assert( stmtExpr );
                 ast::Pass< Resolver_new > resolver{ symtab };
+                ast::Pass< Resolver_new > resolver( context );
                 auto ret = mutate(stmtExpr->accept(resolver));
                 strict_dynamic_cast< ast::StmtExpr * >( ret )->computeResult();
 …
         namespace {
                 const ast::Attribute * handleAttribute(const CodeLocation & loc, const ast::Attribute * attr, const ast::SymbolTable & symtab) {
+                const ast::Attribute * handleAttribute(const CodeLocation & loc, const ast::Attribute * attr, const ResolveContext & context) {
                         std::string name = attr->normalizedName();
                         if (name == "constructor" || name == "destructor") {
                                 if (attr->params.size() == 1) {
                                         auto arg = attr->params.front();
                                         auto resolved = ResolvExpr::findSingleExpression( arg, new ast::BasicType( ast::BasicType::LongLongSignedInt ), symtab );
+                                        auto resolved = ResolvExpr::findSingleExpression( arg, new ast::BasicType( ast::BasicType::LongLongSignedInt ), context );
                                         auto result = eval(arg);
 …
                         for (auto & attr: mutDecl->attributes) {
                                 attr = handleAttribute(mutDecl->location, attr, symtab);
+                                attr = handleAttribute(mutDecl->location, attr, context );
+                        }
 …
                         for (auto & typeParam : mutDecl->type_params) {
                                 symtab.addType(typeParam);
                                 mutType->forall.emplace_back(new ast::TypeInstType(typeParam->name, typeParam));
+                                mutType->forall.emplace_back(new ast::TypeInstType(typeParam));
+                        }
                         for (auto & asst : mutDecl->assertions) {
                                 asst = fixObjectType(asst.strict_as<ast::ObjectDecl>(), symtab);
+                                asst = fixObjectType(asst.strict_as<ast::ObjectDecl>(), context);
                                 symtab.addId(asst);
                                 mutType->assertions.emplace_back(new ast::VariableExpr(functionDecl->location, asst));
 …
                         for (auto & param : mutDecl->params) {
                                 param = fixObjectType(param.strict_as<ast::ObjectDecl>(), symtab);
+                                param = fixObjectType(param.strict_as<ast::ObjectDecl>(), context);
                                 symtab.addId(param);
                                 paramTypes.emplace_back(param->get_type());
+                        }
                         for (auto & ret : mutDecl->returns) {
                                 ret = fixObjectType(ret.strict_as<ast::ObjectDecl>(), symtab);
+                                ret = fixObjectType(ret.strict_as<ast::ObjectDecl>(), context);
                                 returnTypes.emplace_back(ret->get_type());
+                        }
 …
                         // enumerator initializers should not use the enum type to initialize, since the
                         // enum type is still incomplete at this point. Use `int` instead.
+                        objectDecl = fixObjectType(objectDecl, symtab);
+                        currentObject = ast::CurrentObject{
+                                objectDecl->location, new ast::BasicType{ ast::BasicType::SignedInt } };
+                        if (dynamic_cast< const ast::EnumInstType * >( objectDecl->get_type() )->base->base) { // const ast::PointerType &
+                                // const ast::Type * enumBase =  (dynamic_cast< const ast::EnumInstType * >( objectDecl->get_type() )->base->base.get());
+                                // const ast::PointerType * enumBaseAsPtr = dynamic_cast<const ast::PointerType *>(enumBase);
+                                // if ( enumBaseAsPtr ) {
+                                //      const ast::Type * pointerBase = enumBaseAsPtr->base.get();
+                                //      if ( dynamic_cast<const ast::BasicType *>(pointerBase) ) {
+                                //              objectDecl = fixObjectType(objectDecl, context);
+                                //              if (dynamic_cast<const ast::BasicType *>(pointerBase)->kind == ast::BasicType::Char)
+                                //              currentObject = ast::CurrentObject{
+                                //                      objectDecl->location,  new ast::PointerType{
+                                //                              new ast::BasicType{ ast::BasicType::Char }
+                                //                      } };
+                                //      } else {
+                                //              objectDecl = fixObjectType(objectDecl, context);
+                                //              currentObject = ast::CurrentObject{objectDecl->location, new ast::BasicType{ ast::BasicType::SignedInt } };
+                                //      }
+                                // }
+                                objectDecl = fixObjectType( objectDecl, context );
+                                const ast::Type * enumBase =  (dynamic_cast< const ast::EnumInstType * >( objectDecl->get_type() )->base->base.get());
+                                currentObject = ast::CurrentObject{
+                                        objectDecl->location,
+                                        enumBase
+                                };
+                        } else {
+                                objectDecl = fixObjectType( objectDecl, context );
+                                currentObject = ast::CurrentObject{
+                                        objectDecl->location, new ast::BasicType{ ast::BasicType::SignedInt } };
+                        }
+                }
                 else {
                         if (!objectDecl->isTypeFixed) {
                                 auto newDecl = fixObjectType(objectDecl, symtab);
+                                auto newDecl = fixObjectType(objectDecl, context);
                                 auto mutDecl = mutate(newDecl);
 …
                         // nested type decls are hoisted already. no need to do anything
                         if (auto obj = member.as<ast::ObjectDecl>()) {
                                 member = fixObjectType(obj, symtab);
+                                member = fixObjectType(obj, context);
+                        }
+                }
 …
                 return ast::mutate_field(
                         assertDecl, &ast::StaticAssertDecl::cond,
                         findIntegralExpression( assertDecl->cond, symtab ) );
+                        findIntegralExpression( assertDecl->cond, context ) );
+        }
         template< typename PtrType >
         const PtrType * handlePtrType( const PtrType * type, const ast::SymbolTable & symtab ) {
+        const PtrType * handlePtrType( const PtrType * type, const ResolveContext & context ) {
                 if ( type->dimension ) {
                         ast::ptr< ast::Type > sizeType = ast::sizeType;
+                        ast::ptr< ast::Type > sizeType = context.global.sizeType;
                         ast::mutate_field(
                                 type, &PtrType::dimension,
                                 findSingleExpression( type->dimension, sizeType, symtab ) );
+                                findSingleExpression( type->dimension, sizeType, context ) );
+                }
                 return type;
 …
         const ast::ArrayType * Resolver_new::previsit( const ast::ArrayType * at ) {
                 return handlePtrType( at, symtab );
+                return handlePtrType( at, context );
+        }
         const ast::PointerType * Resolver_new::previsit( const ast::PointerType * pt ) {
                 return handlePtrType( pt, symtab );
+                return handlePtrType( pt, context );
+        }
 …
                 return ast::mutate_field(
                         exprStmt, &ast::ExprStmt::expr, findVoidExpression( exprStmt->expr, symtab ) );
+                        exprStmt, &ast::ExprStmt::expr, findVoidExpression( exprStmt->expr, context ) );
+        }
 …
                 asmExpr = ast::mutate_field(
                         asmExpr, &ast::AsmExpr::operand, findVoidExpression( asmExpr->operand, symtab ) );
+                        asmExpr, &ast::AsmExpr::operand, findVoidExpression( asmExpr->operand, context ) );
                 return asmExpr;
 …
         const ast::IfStmt * Resolver_new::previsit( const ast::IfStmt * ifStmt ) {
                 return ast::mutate_field(
                         ifStmt, &ast::IfStmt::cond, findIntegralExpression( ifStmt->cond, symtab ) );
+                        ifStmt, &ast::IfStmt::cond, findIntegralExpression( ifStmt->cond, context ) );
+        }
         const ast::WhileDoStmt * Resolver_new::previsit( const ast::WhileDoStmt * whileDoStmt ) {
                 return ast::mutate_field(
                         whileDoStmt, &ast::WhileDoStmt::cond, findIntegralExpression( whileDoStmt->cond, symtab ) );
+                        whileDoStmt, &ast::WhileDoStmt::cond, findIntegralExpression( whileDoStmt->cond, context ) );
+        }
 …
                 if ( forStmt->cond ) {
                         forStmt = ast::mutate_field(
                                 forStmt, &ast::ForStmt::cond, findIntegralExpression( forStmt->cond, symtab ) );
+                                forStmt, &ast::ForStmt::cond, findIntegralExpression( forStmt->cond, context ) );
+                }
                 if ( forStmt->inc ) {
                         forStmt = ast::mutate_field(
                                 forStmt, &ast::ForStmt::inc, findVoidExpression( forStmt->inc, symtab ) );
+                                forStmt, &ast::ForStmt::inc, findVoidExpression( forStmt->inc, context ) );
+                }
 …
                 switchStmt = ast::mutate_field(
                         switchStmt, &ast::SwitchStmt::cond,
                         findIntegralExpression( switchStmt->cond, symtab ) );
+                        findIntegralExpression( switchStmt->cond, context ) );
                 currentObject = ast::CurrentObject{ switchStmt->location, switchStmt->cond->result };
                 return switchStmt;
+        }
         const ast::CaseStmt * Resolver_new::previsit( const ast::CaseStmt * caseStmt ) {
+        const ast::CaseClause * Resolver_new::previsit( const ast::CaseClause * caseStmt ) {
                 if ( caseStmt->cond ) {
                         std::deque< ast::InitAlternative > initAlts = currentObject.getOptions();
 …
                         ast::ptr< ast::Expr > untyped =
                                 new ast::CastExpr{ caseStmt->location, caseStmt->cond, initAlts.front().type };
                         ast::ptr< ast::Expr > newExpr = findSingleExpression( untyped, symtab );
+                        ast::ptr< ast::Expr > newExpr = findSingleExpression( untyped, context );
                         // case condition cannot have a cast in C, so it must be removed here, regardless of
 …
+                        }
                         caseStmt = ast::mutate_field( caseStmt, &ast::CaseStmt::cond, newExpr );
+                        caseStmt = ast::mutate_field( caseStmt, &ast::CaseClause::cond, newExpr );
+                }
                 return caseStmt;
 …
                         branchStmt = ast::mutate_field(
                                 branchStmt, &ast::BranchStmt::computedTarget,
                                 findSingleExpression( branchStmt->computedTarget, target, symtab ) );
+                                findSingleExpression( branchStmt->computedTarget, target, context ) );
+                }
                 return branchStmt;
 …
                         returnStmt = ast::mutate_field(
                                 returnStmt, &ast::ReturnStmt::expr,
                                 findSingleExpression( returnStmt->expr, functionReturn, symtab ) );
+                                findSingleExpression( returnStmt->expr, functionReturn, context ) );
+                }
                 return returnStmt;
 …
                         throwStmt = ast::mutate_field(
                                 throwStmt, &ast::ThrowStmt::expr,
                                 findSingleExpression( throwStmt->expr, exceptType, symtab ) );
+                                findSingleExpression( throwStmt->expr, exceptType, context ) );
+                }
                 return throwStmt;
+        }
         const ast::CatchStmt * Resolver_new::previsit( const ast::CatchStmt * catchStmt ) {
+        const ast::CatchClause * Resolver_new::previsit( const ast::CatchClause * catchClause ) {
                 // Until we are very sure this invarent (ifs that move between passes have then)
                 // holds, check it. This allows a check for when to decode the mangling.
                 if ( auto ifStmt = catchStmt->body.as<ast::IfStmt>() ) {
+                if ( auto ifStmt = catchClause->body.as<ast::IfStmt>() ) {
                         assert( ifStmt->then );
+                }
                 // Encode the catchStmt so the condition can see the declaration.
                 if ( catchStmt->cond ) {
                         ast::CatchStmt * stmt = mutate( catchStmt );
                         stmt->body = new ast::IfStmt( stmt->location, stmt->cond, nullptr, stmt->body );
                         stmt->cond = nullptr;
                         return stmt;
+                }
                 return catchStmt;
+        }
         const ast::CatchStmt * Resolver_new::postvisit( const ast::CatchStmt * catchStmt ) {
+                if ( catchClause->cond ) {
+                        ast::CatchClause * clause = mutate( catchClause );
+                        clause->body = new ast::IfStmt( clause->location, clause->cond, nullptr, clause->body );
+                        clause->cond = nullptr;
+                        return clause;
+                }
+                return catchClause;
+        }
+        const ast::CatchClause * Resolver_new::postvisit( const ast::CatchClause * catchClause ) {
                 // Decode the catchStmt so everything is stored properly.
                 const ast::IfStmt * ifStmt = catchStmt->body.as<ast::IfStmt>();
+                const ast::IfStmt * ifStmt = catchClause->body.as<ast::IfStmt>();
                 if ( nullptr != ifStmt && nullptr == ifStmt->then ) {
                         assert( ifStmt->cond );
                         assert( ifStmt->else_ );
                         ast::CatchStmt * stmt = ast::mutate( catchStmt );
                         stmt->cond = ifStmt->cond;
                         stmt->body = ifStmt->else_;
+                        ast::CatchClause * clause = ast::mutate( catchClause );
+                        clause->cond = ifStmt->cond;
+                        clause->body = ifStmt->else_;
                         // ifStmt should be implicately deleted here.
                         return stmt;
+                }
                 return catchStmt;
+                        return clause;
+                }
+                return catchClause;
+        }
 …
                         ast::TypeEnvironment env;
                         CandidateFinder funcFinder{ symtab, env };
+                        CandidateFinder funcFinder( context, env );
                         // Find all candidates for a function in canonical form
 …
                                 );
                                 clause2.target.args.emplace_back( findSingleExpression( init, symtab ) );
+                                clause2.target.args.emplace_back( findSingleExpression( init, context ) );
+                        }
                         // Resolve the conditions as if it were an IfStmt, statements normally
                         clause2.cond = findSingleExpression( clause.cond, symtab );
+                        clause2.cond = findSingleExpression( clause.cond, context );
                         clause2.stmt = clause.stmt->accept( *visitor );
 …
                         ast::ptr< ast::Type > target =
                                 new ast::BasicType{ ast::BasicType::LongLongUnsignedInt };
                         timeout2.time = findSingleExpression( stmt->timeout.time, target, symtab );
                         timeout2.cond = findSingleExpression( stmt->timeout.cond, symtab );
+                        timeout2.time = findSingleExpression( stmt->timeout.time, target, context );
+                        timeout2.cond = findSingleExpression( stmt->timeout.cond, context );
                         timeout2.stmt = stmt->timeout.stmt->accept( *visitor );
 …
                         ast::WaitForStmt::OrElse orElse2;
                         orElse2.cond = findSingleExpression( stmt->orElse.cond, symtab );
+                        orElse2.cond = findSingleExpression( stmt->orElse.cond, context );
                         orElse2.stmt = stmt->orElse.stmt->accept( *visitor );
 …
                 for (auto & expr : exprs) {
                         // only struct- and union-typed expressions are viable candidates
                         expr = findKindExpression( expr, symtab, structOrUnion, "with expression" );
+                        expr = findKindExpression( expr, context, structOrUnion, "with expression" );
                         // if with expression might be impure, create a temporary so that it is evaluated once
 …
                 ast::ptr< ast::Expr > untyped = new ast::UntypedInitExpr{
                         singleInit->location, singleInit->value, currentObject.getOptions() };
                 ast::ptr<ast::Expr> newExpr = findSingleExpression( untyped, symtab );
+                ast::ptr<ast::Expr> newExpr = findSingleExpression( untyped, context );
                 const ast::InitExpr * initExpr = newExpr.strict_as< ast::InitExpr >();

src/ResolvExpr/Resolver.h

-              rba897d21
+              r2e9b59b
 // Author           : Richard C. Bilson
 // Created On       : Sun May 17 12:18:34 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Mon Feb 18 20:40:38 2019
 // Update Count     : 4
+// Last Modified By : Andrew Beach
+// Last Modified On : Wed Mar 16 11:32:00 2022
+// Update Count     : 5
 //
 …
 class Declaration;
 class Expression;
+class DeletedExpr;
 class StmtExpr;
+class Type;
 namespace SymTab {
         class Indexer;
 …
         class StmtExpr;
         class SymbolTable;
+        class TranslationGlobal;
         class TranslationUnit;
         class Type;
 …
         void resolveWithExprs( std::list< Declaration * > & translationUnit );
+        /// Helper Type: Passes around information between various sub-calls.
+        struct ResolveContext {
+                const ast::SymbolTable & symtab;
+                const ast::TranslationGlobal & global;
+        };
         /// Checks types and binds syntactic constructs to typed representations
         void resolve( ast::TranslationUnit& translationUnit );
 …
         /// context.
         ast::ptr< ast::Expr > resolveInVoidContext(
                 const ast::Expr * expr, const ast::SymbolTable & symtab, ast::TypeEnvironment & env );
+                const ast::Expr * expr, const ResolveContext &, ast::TypeEnvironment & env );
         /// Resolve `untyped` to the single expression whose candidate is the best match for the
         /// given type.
         ast::ptr< ast::Expr > findSingleExpression(
                 const ast::Expr * untyped, const ast::Type * type, const ast::SymbolTable & symtab );
+                const ast::Expr * untyped, const ast::Type * type, const ResolveContext & );
         ast::ptr< ast::Expr > findVoidExpression(
                 const ast::Expr * untyped, const ast::SymbolTable & symtab);
+                const ast::Expr * untyped, const ResolveContext & );
         /// Resolves a constructor init expression
         ast::ptr< ast::Init > resolveCtorInit(
                 const ast::ConstructorInit * ctorInit, const ast::SymbolTable & symtab );
+                const ast::ConstructorInit * ctorInit, const ResolveContext & context );
         /// Resolves a statement expression
         const ast::Expr * resolveStmtExpr(
                 const ast::StmtExpr * stmtExpr, const ast::SymbolTable & symtab );
+                const ast::StmtExpr * stmtExpr, const ResolveContext & context );
 } // namespace ResolvExpr

src/ResolvExpr/Unify.cc

rba897d21	r2e9b59b
943	943	// check that the other type is compatible and named the same
944	944	auto otherInst = dynamic_cast< const XInstType * >( other );
945		~~this->result = otherInst && inst->name == otherInst->name~~;
	945	if (otherInst && inst->name == otherInst->name) this->result = otherInst;
946	946	return otherInst;
947	947	}

src/SymTab/Validate.cc

-              rba897d21
+              r2e9b59b
                                 TranslateDimensionGenericParameters::translateDimensions( translationUnit );
                         });
+                        if (!useNewAST) {
                         Stats::Time::TimeBlock("Resolve Enum Initializers", [&]() {
                                 acceptAll( translationUnit, rei ); // must happen after translateDimensions because rei needs identifier lookup, which needs name mangling
                         });
+                        }
                         Stats::Time::TimeBlock("Check Function Returns", [&]() {
                                 ReturnChecker::checkFunctionReturns( translationUnit );
 …
                         });
+                }
+        }
+        static void decayForallPointers( std::list< Declaration * > & translationUnit ) {
+                PassVisitor<TraitExpander_old> te;
+                acceptAll( translationUnit, te );
+                PassVisitor<AssertionFixer_old> af;
+                acceptAll( translationUnit, af );
+                PassVisitor<CheckOperatorTypes_old> cot;
+                acceptAll( translationUnit, cot );
+                PassVisitor<FixUniqueIds_old> fui;
+                acceptAll( translationUnit, fui );
+        }
 …
+        }
-        void decayForallPointers( std::list< Declaration * > & translationUnit ) {
-                PassVisitor<TraitExpander_old> te;
-                acceptAll( translationUnit, te );
-                PassVisitor<AssertionFixer_old> af;
-                acceptAll( translationUnit, af );
-                PassVisitor<CheckOperatorTypes_old> cot;
-                acceptAll( translationUnit, cot );
-                PassVisitor<FixUniqueIds_old> fui;
-                acceptAll( translationUnit, fui );
+        }
-        void decayForallPointersA( std::list< Declaration * > & translationUnit ) {
-                PassVisitor<TraitExpander_old> te;
-                acceptAll( translationUnit, te );
+        }
-        void decayForallPointersB( std::list< Declaration * > & translationUnit ) {
-                PassVisitor<AssertionFixer_old> af;
-                acceptAll( translationUnit, af );
+        }
-        void decayForallPointersC( std::list< Declaration * > & translationUnit ) {
-                PassVisitor<CheckOperatorTypes_old> cot;
-                acceptAll( translationUnit, cot );
+        }
-        void decayForallPointersD( std::list< Declaration * > & translationUnit ) {
-                PassVisitor<FixUniqueIds_old> fui;
-                acceptAll( translationUnit, fui );
+        }
         void validate( std::list< Declaration * > &translationUnit, __attribute__((unused)) bool doDebug ) {
                 validate_A( translationUnit );
 …
                                         // need to resolve enumerator initializers early so that other passes that determine if an expression is constexpr have the appropriate information.
                                         SingleInit * init = strict_dynamic_cast<SingleInit *>( field->init );
+                                        ResolvExpr::findSingleExpression( init->value, new BasicType( Type::Qualifiers(), BasicType::SignedInt ), indexer );
+                                        if ( !enumDecl->base || dynamic_cast<BasicType *>(enumDecl->base))
+                                                ResolvExpr::findSingleExpression( init->value, new BasicType( Type::Qualifiers(), BasicType::SignedInt ), indexer );
+                                        else {
+                                                if (dynamic_cast<PointerType *>(enumDecl->base)) {
+                                                        auto typePtr = dynamic_cast<PointerType *>(enumDecl->base);
+                                                        ResolvExpr::findSingleExpression( init->value,
+                                                         new PointerType( Type::Qualifiers(), typePtr->base ), indexer );
+                                                } else {
+                                                        ResolvExpr::findSingleExpression( init->value, new BasicType( Type::Qualifiers(), BasicType::SignedInt ), indexer );
+                                                }
+                                        }
+                                }
+                        }
                 } // if
+        }
 …
                         declsToAddBefore.push_back( new UnionDecl( aggDecl->name, noAttributes, tyDecl->linkage ) );
                 } else if ( EnumInstType * enumDecl = dynamic_cast< EnumInstType * >( designatorType ) ) {
+                        declsToAddBefore.push_back( new EnumDecl( enumDecl->name, noAttributes, tyDecl->linkage ) );
+                        // declsToAddBefore.push_back( new EnumDecl( enumDecl->name, noAttributes, tyDecl->linkage, enumDecl->baseEnum->base ) );
+                        if (enumDecl->baseEnum) {
+                                declsToAddBefore.push_back( new EnumDecl( enumDecl->name, noAttributes, tyDecl->linkage, enumDecl->baseEnum->base ) );
+                        } else {
+                                declsToAddBefore.push_back( new EnumDecl( enumDecl->name, noAttributes, tyDecl->linkage ) );
+                        }
                 } // if
                 return tyDecl->clone();

src/SymTab/Validate.h

-              rba897d21
+              r2e9b59b
         void validate_E( std::list< Declaration * > &translationUnit );
         void validate_F( std::list< Declaration * > &translationUnit );
-        void decayForallPointers( std::list< Declaration * > & translationUnit );
-        void decayForallPointersA( std::list< Declaration * > & translationUnit );
-        void decayForallPointersB( std::list< Declaration * > & translationUnit );
-        void decayForallPointersC( std::list< Declaration * > & translationUnit );
-        void decayForallPointersD( std::list< Declaration * > & translationUnit );
         const ast::Type * validateType(

src/SynTree/AggregateDecl.cc

-              rba897d21
+              r2e9b59b
         } // if
         os << " with body " << has_body();
         if ( ! parameters.empty() ) {
                 os << endl << indent << "... with parameters" << endl;
 …
 const char * EnumDecl::typeString() const { return aggrString( Enum ); }
+void EnumDecl::print( std::ostream & os, Indenter indent ) const {
+        AggregateDecl::print(os, indent);
+        os << " with base? " << (base? "True" : "False") << std::endl;
+        if ( base ) {
+                os << "Base Type of Enum:" << std::endl;
+                base->print(os, indent);
+        }
+        os <<  std::endl << "End of EnumDecl::print" << std::endl;
+}
 const char * TraitDecl::typeString() const { return aggrString( Trait ); }

src/SynTree/BasicType.cc

-              rba897d21
+              r2e9b59b
+}
+bool BasicType::isWholeNumber() const {
+        return kind == Bool ||
+                kind ==Char ||
+                kind == SignedChar ||
+                kind == UnsignedChar ||
+                kind == ShortSignedInt ||
+                kind == ShortUnsignedInt ||
+                kind == SignedInt ||
+                kind == UnsignedInt ||
+                kind == LongSignedInt ||
+                kind == LongUnsignedInt ||
+                kind == LongLongSignedInt ||
+                kind ==LongLongUnsignedInt ||
+                kind == SignedInt128 ||
+                kind == UnsignedInt128;
+}
 bool BasicType::isInteger() const {
         return kind <= UnsignedInt128;

src/SynTree/Declaration.h

-              rba897d21
+              r2e9b59b
         virtual void print( std::ostream & os, Indenter indent = {} ) const override;
         virtual void printShort( std::ostream & os, Indenter indent = {} ) const override;
+        // TODO: Move to the right place
+        void checkAssignedValue() const;
 };
 …
         AggregateDecl * set_body( bool body ) { AggregateDecl::body = body; return this; }
         virtual void print( std::ostream & os, Indenter indent = {} ) const override final;
+        virtual void print( std::ostream & os, Indenter indent = {} ) const override;
         virtual void printShort( std::ostream & os, Indenter indent = {} ) const override;
   protected:
 …
         typedef AggregateDecl Parent;
   public:
+        EnumDecl( const std::string & name, const std::list< Attribute * > & attributes = std::list< class Attribute * >(), LinkageSpec::Spec linkage = LinkageSpec::Cforall ) : Parent( name, attributes, linkage ) {}
+        EnumDecl( const EnumDecl & other ) : Parent( other ) {}
+        EnumDecl( const std::string & name,
+         const std::list< Attribute * > & attributes = std::list< class Attribute * >(),
+          LinkageSpec::Spec linkage = LinkageSpec::Cforall,
+          Type * baseType = nullptr ) : Parent( name, attributes, linkage ) , base( baseType ){}
+        EnumDecl( const EnumDecl & other ) : Parent( other ), base( other.base ) {}
         bool valueOf( Declaration * enumerator, long long int & value );
 …
         virtual void accept( Visitor & v ) const override { v.visit( this ); }
         virtual Declaration * acceptMutator( Mutator & m )  override { return m.mutate( this ); }
+  private:
+        Type * base;
         std::unordered_map< std::string, long long int > enumValues;
+        virtual void print( std::ostream & os, Indenter indent = {} ) const override final;
+  private:
+        // std::unordered_map< std::string, long long int > enumValues;
         virtual const char * typeString() const override;
 };

src/SynTree/Type.h

rba897d21	r2e9b59b
268	268	virtual Type *acceptMutator( Mutator & m ) override { return m.mutate( this ); }
269	269	virtual void print( std::ostream & os, Indenter indent = {} ) const override;
270
	270	bool isWholeNumber() const;
271	271	bool isInteger() const;
272	272	};

src/SynTree/Visitor.h

-              rba897d21
+              r2e9b59b
         virtual void visit( UnionDecl * node ) { visit( const_cast<const UnionDecl *>(node) ); }
         virtual void visit( const UnionDecl * aggregateDecl ) = 0;
         virtual void visit( EnumDecl * node ) { visit( const_cast<const EnumDecl *>(node) ); }
+        virtual void visit( EnumDecl * node ) { visit( const_cast<const EnumDecl *>(node) ); } // Marker 1
         virtual void visit( const EnumDecl * aggregateDecl ) = 0;
         virtual void visit( TraitDecl * node ) { visit( const_cast<const TraitDecl *>(node) ); }
 …
         virtual void visit( UnionInstType * node ) { visit( const_cast<const UnionInstType *>(node) ); }
         virtual void visit( const UnionInstType * aggregateUseType ) = 0;
         virtual void visit( EnumInstType * node ) { visit( const_cast<const EnumInstType *>(node) ); }
+        virtual void visit( EnumInstType * node ) { visit( const_cast<const EnumInstType *>(node) ); } // Marker 2
         virtual void visit( const EnumInstType * aggregateUseType ) = 0;
         virtual void visit( TraitInstType * node ) { visit( const_cast<const TraitInstType *>(node) ); }

src/Tuples/TupleAssignment.cc

-              rba897d21
+              r2e9b59b
 // Author           : Rodolfo G. Esteves
 // Created On       : Mon May 18 07:44:20 2015
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Fri Dec 13 23:45:33 2019
 // Update Count     : 9
+// Last Modified By : Andrew Beach
+// Last Modified On : Wed Mar 16 14:06:00 2022
+// Update Count     : 10
 //
 …
                                         // resolve ctor/dtor for the new object
                                         ast::ptr< ast::Init > ctorInit = ResolvExpr::resolveCtorInit(
                                                         InitTweak::genCtorInit( location, ret ), spotter.crntFinder.localSyms );
+                                                        InitTweak::genCtorInit( location, ret ), spotter.crntFinder.context );
                                         // remove environments from subexpressions of stmtExpr
                                         ast::Pass< EnvRemover > rm{ env };
 …
                                         // resolve the cast expression so that rhsCand return type is bound by the cast
                                         // type as needed, and transfer the resulting environment
                                         ResolvExpr::CandidateFinder finder{ spotter.crntFinder.localSyms, env };
+                                        ResolvExpr::CandidateFinder finder( spotter.crntFinder.context, env );
                                         finder.find( rhsCand->expr, ResolvExpr::ResolvMode::withAdjustment() );
                                         assert( finder.candidates.size() == 1 );
 …
                                         // explode the LHS so that each field of a tuple-valued expr is assigned
                                         ResolvExpr::CandidateList lhs;
                                         explode( *lhsCand, crntFinder.localSyms, back_inserter(lhs), true );
+                                        explode( *lhsCand, crntFinder.context.symtab, back_inserter(lhs), true );
                                         for ( ResolvExpr::CandidateRef & cand : lhs ) {
                                                 // each LHS value must be a reference - some come in with a cast, if not
 …
                                                         if ( isTuple( rhsCand->expr ) ) {
                                                                 // multiple assignment
                                                                 explode( *rhsCand, crntFinder.localSyms, back_inserter(rhs), true );
+                                                                explode( *rhsCand, crntFinder.context.symtab, back_inserter(rhs), true );
                                                                 matcher.reset(
                                                                         new MultipleAssignMatcher{ *this, expr->location, lhs, rhs } );
 …
                                                         // multiple assignment
                                                         ResolvExpr::CandidateList rhs;
                                                         explode( rhsCand, crntFinder.localSyms, back_inserter(rhs), true );
+                                                        explode( rhsCand, crntFinder.context.symtab, back_inserter(rhs), true );
                                                         matcher.reset(
                                                                 new MultipleAssignMatcher{ *this, expr->location, lhs, rhs } );
 …
+                                )
                                 ResolvExpr::CandidateFinder finder{ crntFinder.localSyms, matcher->env };
+                                ResolvExpr::CandidateFinder finder( crntFinder.context, matcher->env );
                                 try {

src/Validate/Autogen.cpp

-              rba897d21
+              r2e9b59b
                 structInst.params.push_back( new ast::TypeExpr(
                         typeDecl->location,
                         new ast::TypeInstType( typeDecl->name, typeDecl )
+                        new ast::TypeInstType( typeDecl )
                 ) );
+        }
 …
                 unionInst.params.push_back( new ast::TypeExpr(
                         unionDecl->location,
                         new ast::TypeInstType( typeDecl->name, typeDecl )
+                        new ast::TypeInstType( typeDecl )
                 ) );
+        }

src/Validate/FindSpecialDeclsNew.cpp

-              rba897d21
+              r2e9b59b
 struct FindDeclsCore : public ast::WithShortCircuiting {
         ast::TranslationUnit::Global & global;
         FindDeclsCore( ast::TranslationUnit::Global & g ) : global( g ) {}
+        ast::TranslationGlobal & global;
+        FindDeclsCore( ast::TranslationGlobal & g ) : global( g ) {}
         void previsit( const ast::Decl * decl );
 …
         ast::Pass<FindDeclsCore>::run( translationUnit, translationUnit.global );
-        // TODO: When everything gets the globals from the translation unit,
-        // remove these.
-        ast::dereferenceOperator = translationUnit.global.dereference;
-        ast::dtorStruct = translationUnit.global.dtorStruct;
-        ast::dtorStructDestroy = translationUnit.global.dtorDestroy;
         // TODO: conditionally generate 'fake' declarations for missing features,
         // so that translation can proceed in the event that builtins, prelude,

src/Validate/ForallPointerDecay.cpp

-              rba897d21
+              r2e9b59b
         for ( auto & type_param : decl->type_params ) {
                 type->forall.emplace_back(
                         new ast::TypeInstType( type_param->name, type_param ) );
+                        new ast::TypeInstType( type_param ) );
+        }
         for ( auto & assertion : decl->assertions ) {
 …
                 AssertionList assertions;
                 // Substitute trait decl parameters for instance parameters.
+                ast::TypeSubstitution sub(
+                        inst->base->params.begin(),
+                        inst->base->params.end(),
+                        inst->params.begin()
+                );
+                ast::TypeSubstitution sub( inst->base->params, inst->params );
                 for ( const ast::ptr<ast::Decl> & decl : inst->base->members ) {
                         ast::ptr<ast::DeclWithType> copy =

src/Validate/module.mk

-              rba897d21
+              r2e9b59b
         Validate/ForallPointerDecay.cpp \
         Validate/ForallPointerDecay.hpp \
+        Validate/GenericParameter.cpp \
+        Validate/GenericParameter.hpp \
         Validate/HandleAttributes.cc \
         Validate/HandleAttributes.h \
 …
         Validate/LabelAddressFixer.cpp \
         Validate/LabelAddressFixer.hpp \
+        Validate/ReturnCheck.cpp \
+        Validate/ReturnCheck.hpp \
         Validate/FindSpecialDeclsNew.cpp \
         Validate/FindSpecialDecls.cc \

src/Virtual/Tables.cc

-              rba897d21
+              r2e9b59b
 // Created On       : Mon Aug 31 11:11:00 2020
 // Last Modified By : Andrew Beach
+// Last Modified On : Wed Apr 21 15:36:00 2021
+// Update Count     : 2
+//
+// Last Modified On : Fri Mar 11 10:40:00 2022
+// Update Count     : 3
+//
+#include "AST/Attribute.hpp"
+#include "AST/Copy.hpp"
+#include "AST/Decl.hpp"
+#include "AST/Expr.hpp"
+#include "AST/Init.hpp"
+#include "AST/Stmt.hpp"
+#include "AST/Type.hpp"
 #include <SynTree/Attribute.h>
 #include <SynTree/Declaration.h>
 …
+}
+static ast::ObjectDecl * makeVtableDeclaration(
+                CodeLocation const & location, std::string const & name,
+                ast::StructInstType const * type, ast::Init const * init ) {
+        ast::Storage::Classes storage;
+        if ( nullptr == init ) {
+                storage.is_extern = true;
+        }
+        return new ast::ObjectDecl(
+                location,
+                name,
+                type,
+                init,
+                storage,
+                ast::Linkage::Cforall
+        );
+}
 ObjectDecl * makeVtableForward( std::string const & name, StructInstType * type ) {
         assert( type );
         return makeVtableDeclaration( name, type, nullptr );
+}
+ast::ObjectDecl * makeVtableForward(
+                CodeLocation const & location, std::string const & name,
+                ast::StructInstType const * vtableType ) {
+        assert( vtableType );
+        return makeVtableDeclaration( location, name, vtableType, nullptr );
+}
 …
+}
+static std::vector<ast::ptr<ast::Init>> buildInits(
+                CodeLocation const & location,
+                //std::string const & name,
+                ast::StructInstType const * vtableType,
+                ast::Type const * objectType ) {
+        ast::StructDecl const * vtableStruct = vtableType->base;
+        std::vector<ast::ptr<ast::Init>> inits;
+        inits.reserve( vtableStruct->members.size() );
+        // This is designed to run before the resolver.
+        for ( auto field : vtableStruct->members ) {
+                if ( std::string( "parent" ) == field->name ) {
+                        // This will not work with polymorphic state.
+                        auto oField = field.strict_as<ast::ObjectDecl>();
+                        auto fieldType = oField->type.strict_as<ast::PointerType>();
+                        auto parentType = fieldType->base.strict_as<ast::StructInstType>();
+                        std::string const & parentInstance = instanceName( parentType->name );
+                        inits.push_back(
+                                        new ast::SingleInit( location, new ast::AddressExpr( new ast::NameExpr( location, parentInstance ) ) ) );
+                } else if ( std::string( "__cfavir_typeid" ) == field->name ) {
+                        std::string const & baseType = baseTypeName( vtableType->name );
+                        std::string const & typeId = typeIdName( baseType );
+                        inits.push_back( new ast::SingleInit( location, new ast::AddressExpr( new ast::NameExpr( location, typeId ) ) ) );
+                } else if ( std::string( "size" ) == field->name ) {
+                        inits.push_back( new ast::SingleInit( location, new ast::SizeofExpr( location, objectType )
+                        ) );
+                } else if ( std::string( "align" ) == field->name ) {
+                        inits.push_back( new ast::SingleInit( location,
+                                new ast::AlignofExpr( location, objectType )
+                        ) );
+                } else {
+                        inits.push_back( new ast::SingleInit( location,
+                                new ast::NameExpr( location, field->name )
+                        ) );
+                }
+                //ast::Expr * expr = buildInitExpr(...);
+                //inits.push_back( new ast::SingleInit( location, expr ) )
+        }
+        return inits;
+}
+ast::ObjectDecl * makeVtableInstance(
+                CodeLocation const & location,
+                std::string const & name,
+                ast::StructInstType const * vtableType,
+                ast::Type const * objectType,
+                ast::Init const * init ) {
+        assert( vtableType );
+        assert( objectType );
+        // Build the initialization.
+        if ( nullptr == init ) {
+                init = new ast::ListInit( location,
+                        buildInits( location, vtableType, objectType ) );
+        // The provided init should initialize everything except the parent
+        // pointer, the size-of and align-of fields. These should be inserted.
+        } else {
+                // Except this is not yet supported.
+                assert(false);
+        }
+        return makeVtableDeclaration( location, name, vtableType, init );
+}
 namespace {
         std::string const functionName = "get_exception_vtable";
 …
                 new ReferenceType( noQualifiers, vtableType ),
                 nullptr,
         { new Attribute("unused") }
+                { new Attribute("unused") }
         ) );
         type->parameters.push_back( new ObjectDecl(
 …
                 type,
                 nullptr
+        );
+}
+ast::FunctionDecl * makeGetExceptionForward(
+                CodeLocation const & location,
+                ast::Type const * vtableType,
+                ast::Type const * exceptType ) {
+        assert( vtableType );
+        assert( exceptType );
+        return new ast::FunctionDecl(
+                location,
+                functionName,
+                { /* forall */ },
+                { new ast::ObjectDecl(
+                        location,
+                        "__unused",
+                        new ast::PointerType( exceptType )
+                ) },
+                { new ast::ObjectDecl(
+                        location,
+                        "_retvalue",
+                        new ast::ReferenceType( vtableType )
+                ) },
+                nullptr,
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall,
+                { new ast::Attribute( "unused" ) }
         );
+}
 …
+}
+ast::FunctionDecl * makeGetExceptionFunction(
+                CodeLocation const & location,
+                ast::ObjectDecl const * vtableInstance, ast::Type const * exceptType ) {
+        assert( vtableInstance );
+        assert( exceptType );
+        ast::FunctionDecl * func = makeGetExceptionForward(
+                        location, ast::deepCopy( vtableInstance->type ), exceptType );
+        func->stmts = new ast::CompoundStmt( location, {
+                new ast::ReturnStmt( location, new ast::VariableExpr( location, vtableInstance ) )
+        } );
+        return func;
+}
 ObjectDecl * makeTypeIdInstance( StructInstType const * typeIdType ) {
         assert( typeIdType );
 …
+}
+}
+ast::ObjectDecl * makeTypeIdInstance(
+                CodeLocation const & location,
+                ast::StructInstType const * typeIdType ) {
+        assert( typeIdType );
+        ast::StructInstType * type = ast::mutate( typeIdType );
+        type->set_const( true );
+        std::string const & typeid_name = typeIdTypeToInstance( typeIdType->name );
+        return new ast::ObjectDecl(
+                location,
+                typeid_name,
+                type,
+                new ast::ListInit( location, {
+                        new ast::SingleInit( location,
+                                new ast::AddressExpr( location,
+                                        new ast::NameExpr( location, "__cfatid_exception_t" ) ) )
+                } ),
+                ast::Storage::Classes(),
+                ast::Linkage::Cforall,
+                nullptr,
+                { new ast::Attribute( "cfa_linkonce" ) }
+        );
+}
+}

src/Virtual/Tables.h

-              rba897d21
+              r2e9b59b
 // Created On       : Mon Aug 31 11:07:00 2020
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Apr 21 10:30:00 2021
 // Update Count     : 2
+// Last Modified On : Wec Dec  8 16:58:00 2021
+// Update Count     : 3
 //
 #include <list>  // for list
+#include <string>
+#include "AST/Fwd.hpp"
 class Declaration;
 class StructDecl;
 …
  * vtableType node is consumed.
  */
+ast::ObjectDecl * makeVtableForward(
+        CodeLocation const & location, std::string const & name,
+        ast::StructInstType const * vtableType );
 ObjectDecl * makeVtableInstance(
 …
  * vtableType and init (if provided) nodes are consumed.
  */
+ast::ObjectDecl * makeVtableInstance(
+        CodeLocation const & location,
+        std::string const & name,
+        ast::StructInstType const * vtableType,
+        ast::Type const * objectType,
+        ast::Init const * init = nullptr );
 // Some special code for how exceptions interact with virtual tables.
 …
  * linking the vtableType to the exceptType. Both nodes are consumed.
  */
+ast::FunctionDecl * makeGetExceptionForward(
+        CodeLocation const & location,
+        ast::Type const * vtableType,
+        ast::Type const * exceptType );
 FunctionDecl * makeGetExceptionFunction(
 …
  * exceptType node is consumed.
  */
+ast::FunctionDecl * makeGetExceptionFunction(
+        CodeLocation const & location,
+        ast::ObjectDecl const * vtableInstance, ast::Type const * exceptType );
 ObjectDecl * makeTypeIdInstance( StructInstType const * typeIdType );
 …
  * TODO: Should take the parent type. Currently locked to the exception_t.
  */
+ast::ObjectDecl * makeTypeIdInstance(
+        const CodeLocation & location, ast::StructInstType const * typeIdType );
+}

src/main.cc

-              rba897d21
+              r2e9b59b
 // Created On       : Fri May 15 23:12:02 2015
 // Last Modified By : Andrew Beach
 // Last Modified On : Wed Jan 26 14:09:00 2022
 // Update Count     : 670
+// Last Modified On : Wed Apr 13 11:11:00 2022
+// Update Count     : 672
 //
 …
 #include "AST/Convert.hpp"
-#include "AST/Print.hpp"
 #include "CompilationState.h"
 #include "../config.h"                      // for CFA_LIBDIR
 …
 #include "Tuples/Tuples.h"                  // for expandMemberTuples, expan...
 #include "Validate/Autogen.hpp"             // for autogenerateRoutines
+#include "Validate/GenericParameter.hpp"    // for fillGenericParameters, tr...
 #include "Validate/FindSpecialDecls.h"      // for findGlobalDecls
 #include "Validate/ForallPointerDecay.hpp"  // for decayForallPointers
 …
 #include "Validate/InitializerLength.hpp"   // for setLengthFromInitializer
 #include "Validate/LabelAddressFixer.hpp"   // for fixLabelAddresses
+#include "Validate/ReturnCheck.hpp"         // for checkReturnStatements
 #include "Virtual/ExpandCasts.h"            // for expandCasts
 …
                 PASS( "Validate-A", SymTab::validate_A( translationUnit ) );
                 PASS( "Validate-B", SymTab::validate_B( translationUnit ) );
-                PASS( "Validate-C", SymTab::validate_C( translationUnit ) );
                 CodeTools::fillLocations( translationUnit );
                 if( useNewAST ) {
-                        PASS( "Implement Concurrent Keywords", Concurrency::applyKeywords( translationUnit ) );
-                        //PASS( "Forall Pointer Decay - A", SymTab::decayForallPointersA( translationUnit ) );
-                        //PASS( "Forall Pointer Decay - B", SymTab::decayForallPointersB( translationUnit ) );
-                        //PASS( "Forall Pointer Decay - C", SymTab::decayForallPointersC( translationUnit ) );
-                        //PASS( "Forall Pointer Decay - D", SymTab::decayForallPointersD( translationUnit ) );
                         CodeTools::fillLocations( translationUnit );
 …
                         forceFillCodeLocations( transUnit );
+                        // Check as early as possible. Can't happen before
+                        // LinkReferenceToType, observed failing when attempted
+                        // before eliminateTypedef
+                        PASS( "Validate Generic Parameters", Validate::fillGenericParameters( transUnit ) );
+                        PASS( "Translate Dimensions", Validate::translateDimensionParameters( transUnit ) );
+                        PASS( "Check Function Returns", Validate::checkReturnStatements( transUnit ) );
+                        // Must happen before Autogen.
+                        PASS( "Fix Return Statements", InitTweak::fixReturnStatements( transUnit ) );
+                        PASS( "Implement Concurrent Keywords", Concurrency::implementKeywords( transUnit ) );
                         // Must be after implement concurrent keywords; because uniqueIds
 …
                         translationUnit = convert( move( transUnit ) );
                 } else {
+                        PASS( "Validate-C", SymTab::validate_C( translationUnit ) );
                         PASS( "Validate-D", SymTab::validate_D( translationUnit ) );
                         PASS( "Validate-E", SymTab::validate_E( translationUnit ) );
 …
                         PASS( "Translate Tries" , ControlStruct::translateTries( translationUnit ) );
+                }
                 PASS( "Gen Waitfor" , Concurrency::generateWaitFor( translationUnit ) );

tests/Makefile.am

-              rba897d21
+              r2e9b59b
 DEBUG_FLAGS=-debug -g -O0
 quick_test=avl_test operators numericConstants expression enum array typeof cast raii/dtor-early-exit raii/init_once attributes
+quick_test=avl_test operators numericConstants expression enum array typeof cast raii/dtor-early-exit raii/init_once attributes meta/dumpable
 concurrent=
 …
 PRETTY_PATH=mkdir -p $(dir $(abspath ${@})) && cd ${srcdir} &&
 .PHONY: list .validate
 .INTERMEDIATE: .validate .validate.cfa
+.PHONY: list .validate .test_makeflags
+.INTERMEDIATE: .validate .validate.cfa .test_makeflags
 EXTRA_PROGRAMS = avl_test linkonce .dummy_hack # build but do not install
 EXTRA_DIST = test.py \
 …
         @+${TEST_PY} --list ${concurrent}
+.test_makeflags:
+        @echo "${MAKEFLAGS}"
 .validate: .validate.cfa
         $(CFACOMPILE) .validate.cfa -fsyntax-only -Wall -Wextra -Werror

tests/PRNG.cfa

-              rba897d21
+              r2e9b59b
 // Created On       : Wed Dec 29 09:38:12 2021
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Sat Feb 12 12:23:57 2022
 // Update Count     : 342
+// Last Modified On : Sat Apr  9 15:21:14 2022
+// Update Count     : 344
 //
 …
 #include <mutex_stmt.hfa>
-// FIX ME: spurious characters appear in output
-Duration default_preemption() { return 0; }
 #ifdef TIME                                                                                             // use -O2 -nodebug
 #define STARTTIME start = timeHiRes()

tests/collections/.expect/vector-err-pass-perm-it-byval.txt

rba897d21	r2e9b59b
1		error: Unique best alternative includes deleted identifier in Generated Cast of:
	1	collections/vector-demo.cfa:95:1 error: Unique best alternative includes deleted identifier in Generated Cast of:
2	2	Application of
3	3	Deleted Expression

tests/concurrent/mutexstmt/.expect/locks.txt

-              rba897d21
+              r2e9b59b
 Start Test: multi lock deadlock/mutual exclusion
 End Test: multi lock deadlock/mutual exclusion
+Start Test: single scoped lock mutual exclusion
+End Test: single scoped lock mutual exclusion
+Start Test: multi scoped lock deadlock/mutual exclusion
+End Test: multi scoped lock deadlock/mutual exclusion
+Start Test: multi polymorphic lock deadlock/mutual exclusion
+End Test: multi polymorphic lock deadlock/mutual exclusion

tests/concurrent/mutexstmt/locks.cfa

-              rba897d21
+              r2e9b59b
 const unsigned int num_times = 10000;
+Duration default_preemption() { return 0; }
 single_acquisition_lock m1, m2, m3, m4, m5;
 …
+}
+void refTest( single_acquisition_lock & m ) {
+        mutex ( m ) {
+                assert(!insideFlag);
+                insideFlag = true;
+                assert(insideFlag);
+                insideFlag = false;
+        }
+}
 thread T_Multi {};
 void main( T_Multi & this ) {
         for (unsigned int i = 0; i < num_times; i++) {
+                refTest( m1 );
                 mutex ( m1 ) {
                         assert(!insideFlag);
 …
+}
+thread T_Mutex_Scoped {};
+single_acquisition_lock l1;
+linear_backoff_then_block_lock l2;
+owner_lock l3;
+void main( T_Mutex_Scoped & this ) {
+monitor monitor_t {};
+monitor_t l4;
+thread T_Multi_Poly {};
+void main( T_Multi_Poly & this ) {
         for (unsigned int i = 0; i < num_times; i++) {
+                {
+                        scoped_lock(single_acquisition_lock) s{m1};
+                        count++;
+                }
+                {
+                        scoped_lock(single_acquisition_lock) s{m1};
+                refTest( l1 );
+                mutex ( l1, l4 ) {
                         assert(!insideFlag);
                         insideFlag = true;
 …
                         insideFlag = false;
+                }
+        }
+}
+thread T_Multi_Scoped {};
+void main( T_Multi_Scoped & this ) {
+        for (unsigned int i = 0; i < num_times; i++) {
+                {
+                        scoped_lock(single_acquisition_lock) s{m1};
+                mutex ( l1, l2, l3 ) {
                         assert(!insideFlag);
                         insideFlag = true;
 …
                         insideFlag = false;
+                }
+                {
+                        scoped_lock(single_acquisition_lock) s1{m1};
+                        scoped_lock(single_acquisition_lock) s2{m2};
+                        scoped_lock(single_acquisition_lock) s3{m3};
+                        scoped_lock(single_acquisition_lock) s4{m4};
+                        scoped_lock(single_acquisition_lock) s5{m5};
+                mutex ( l3, l1, l4 ) {
                         assert(!insideFlag);
                         insideFlag = true;
 …
                         insideFlag = false;
+                }
+                {
+                        scoped_lock(single_acquisition_lock) s1{m1};
+                        scoped_lock(single_acquisition_lock) s3{m3};
+                        assert(!insideFlag);
+                        insideFlag = true;
+                        assert(insideFlag);
+                        insideFlag = false;
+                }
+                {
+                        scoped_lock(single_acquisition_lock) s1{m1};
+                        scoped_lock(single_acquisition_lock) s2{m2};
+                        scoped_lock(single_acquisition_lock) s4{m4};
+                        assert(!insideFlag);
+                        insideFlag = true;
+                        assert(insideFlag);
+                        insideFlag = false;
+                }
+                {
+                        scoped_lock(single_acquisition_lock) s1{m1};
+                        scoped_lock(single_acquisition_lock) s3{m3};
+                        scoped_lock(single_acquisition_lock) s4{m4};
+                        scoped_lock(single_acquisition_lock) s5{m5};
+                mutex ( l1, l2, l4 ) {
                         assert(!insideFlag);
                         insideFlag = true;
 …
 int num_tasks = 10;
 int main() {
         processor p[10];
+        processor p[num_tasks - 1];
         printf("Start Test: single lock mutual exclusion\n");
+        {
                 T_Mutex t[10];
+                T_Mutex t[num_tasks];
+        }
         assert(count == num_tasks * num_times);
 …
         printf("Start Test: multi lock deadlock/mutual exclusion\n");
+        {
                 T_Multi t[10];
+                T_Multi t[num_tasks];
+        }
         printf("End Test: multi lock deadlock/mutual exclusion\n");
+        count = 0;
+        printf("Start Test: single scoped lock mutual exclusion\n");
+        printf("Start Test: multi polymorphic lock deadlock/mutual exclusion\n");
+        {
                 T_Mutex_Scoped t[10];
+                T_Multi_Poly t[num_tasks];
+        }
+        assert(count == num_tasks * num_times);
+        printf("End Test: single scoped lock mutual exclusion\n");
+        printf("Start Test: multi scoped lock deadlock/mutual exclusion\n");
+        {
+                T_Multi_Scoped t[10];
+        }
+        printf("End Test: multi scoped lock deadlock/mutual exclusion\n");
+        printf("End Test: multi polymorphic lock deadlock/mutual exclusion\n");
+}

tests/designations.cfa

-              rba897d21
+              r2e9b59b
 // Created On       : Thu Jun 29 15:26:36 2017
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Thu Jul 27 11:46:35 2017
 // Update Count     : 3
+// Last Modified On : Mon Mar 28 22:41:55 2022
+// Update Count     : 15
 //
 …
 // is used for the designation syntax
 #ifdef __cforall
+#define DES :
+#define _ :
+#define AT @
 #else
+int printf(const char *, ...);
+#define DES =
+int printf( const char *, ...);
+#define _ =
+#define AT
 #endif
 const int indentAmt = 2;
 void indent(int level) {
         for (int i = 0; i < level; ++i) {
                 printf(" ");
+void indent( int level ) {
+        for ( int i = 0; i < level; ++i ) {
+                printf( " " );
+        }
+}
 …
         int * ptr;
 };
 void printA(struct A a, int level) {
         indent(level);
         printf("(A){ %d %d %p }\n", a.x, a.y, a.ptr);
+void printA( struct A a, int level ) {
+        indent( level );
+        printf( "(A){ %d %d %p }\n", a.x, a.y, a.ptr );
+}
 …
         struct A a0, a1;
 };
 void printB(struct B b, int level) {
         indent(level);
         printf("(B){\n");
         printA(b.a0, level+indentAmt);
         printA(b.a1, level+indentAmt);
         indent(level);
         printf("}\n");
+void printB( struct B b, int level ) {
+        indent( level );
+        printf( "(B){\n" );
+        printA( b.a0, level+indentAmt );
+        printA( b.a1, level+indentAmt );
+        indent( level );
+        printf( "}\n" );
+}
 …
         struct B b;
 };
 void printC(struct C c, int level) {
         indent(level);
         printf("(C){\n");
         indent(level+indentAmt);
         printf("(int[]{ %d %d %d }\n", c.arr[0], c.arr[1], c.arr[2]);
         printB(c.b, level+indentAmt);
         indent(level);
         printf("}\n");
+void printC( struct C c, int level ) {
+        indent( level );
+        printf( "(C){\n" );
+        indent( level+indentAmt );
+        printf( "(int[]{ %d %d %d }\n", c.arr[0], c.arr[1], c.arr[2]);
+        printB( c.b, level+indentAmt );
+        indent( level );
+        printf( "}\n" );
+}
 …
         };
 };
 void printD(struct D d, int level) {
         indent(level);
         printf("(D){ %d }\n", d.x);
+void printD( struct D d, int level ) {
+        indent( level);
+        printf( "(D ){ %d }\n", d.x );
+}
 …
     } m;
 };
 struct Fred s1 @= { .m.j : 3 };
 struct Fred s2 @= { .i : { [2] : 2 } };
+struct Fred s1 AT= { .m.j _ 3 };
+struct Fred s2 AT= { .i _ { [2] _ 2 } };
 int main() {
         // simple designation case - starting from beginning of structure, leaves ptr default-initialized (zero)
         struct A y0 = {
                 .x DES 2,
                 .y DES 3
+                .x _ 2,
+                .y _ 3
         };
 …
         // use designation to move to member y, leaving x default-initialized (zero)
         struct A y2 = {
                 .y DES 3,
+                .y _ 3,
         };
 …
 #endif
         printf("=====A=====\n");
         printA(y0, 0);
         printA(y1, 0);
         printA(y2, 0);
         printf("=====A=====\n\n");
+        printf( "=====A=====\n" );
+        printA( y0, 0 );
+        printA( y1, 0 );
+        printA( y2, 0 );
+        printf( "=====A=====\n\n" );
         // initialize only first element (z0.a.x), leaving everything else default-initialized (zero), no nested curly-braces
 …
                 { 3 }, // z1.a0
                 { 4 }, // z1.a1
                 .a0 DES { 5 }, // z1.a0
+                .a0 _ { 5 }, // z1.a0
                 { 6 }, // z1.a1
                 .a0.y DES 2, // z1.a0.y
+                .a0.y _ 2, // z1.a0.y
 , // z1.a0.ptr
         };
 …
         };
         printf("=====B=====\n");
         printB(z0, 0);
         printB(z1, 0);
         printB(z2, 0);
         printB(z3, 0);
         printB(z5, 0);
         printB(z6, 0);
         printf("=====B=====\n\n");
+        printf( "=====B=====\n" );
+        printB( z0, 0 );
+        printB( z1, 0 );
+        printB( z2, 0 );
+        printB( z3, 0 );
+        printB( z5, 0 );
+        printB( z6, 0 );
+        printf( "=====B=====\n\n" );
         // TODO: what about extra things in a nested init? are empty structs skipped??
 …
         };
         printf("=====C=====\n");
         printC(c1, 0);
         printf("=====C=====\n\n");
+        printf( "=====C=====\n" );
+        printC( c1, 0 );
+        printf( "=====C=====\n\n" );
 #if ERROR
 …
 #endif
         // array designation
         int i[2] = { [1] : 3 };
+        int i[2] = { [1] _ 3 };
         // allowed to have 'too many' initialized lists - essentially they are ignored.
         int i1 = { 3 };
 …
         // doesn't work yet.
         // designate unnamed object's members
         // struct D d = { .x DES 3 };
+        // struct D d = { .x _ 3 };
 #if ERROR
         struct D d1 = { .y DES 3 };
+        struct D d1 = { .y _ 3 };
 #endif
 …
         // move cursor to e4.b.a0.x and initialize until e3.b.a1.ptr inclusive
         union E e3 = {
                 .b.a0.x DES 2, 3, 0, 5, 6, 0
         };
         printf("=====E=====\n");
         printA(e0.a, 0);
         printA(e1.a, 0);
         printA(e2.a, 0);
         printB(e3.b, 0);
         printf("=====E=====\n\n");
+                .b.a0.x _ 2, 3, 0, 5, 6, 0
+        };
+        printf( "=====E=====\n" );
+        printA( e0.a, 0 );
+        printA( e1.a, 0 );
+        printA( e2.a, 0 );
+        printB( e3.b, 0 );
+        printf( "=====E=====\n\n" );
         // special case of initialization: char[] can be initialized with a string literal
         const char * str0 = "hello";
         char str1[] = "hello";
+        const char c1[] = "abc";
+        const char c2[] = { 'a', 'b', 'c' };
+        const char c3[][2] = { { 'a', 'b' }, { 'c', 'd'}, { 'c', 'd'} };
+        const char c2[] = "abc";
+        const char c3[] = { 'a', 'b', 'c' };
+        const char c4[][2] = { { 'a', 'b' }, { 'c', 'd'}, { 'c', 'd'} };
+        // more cases
+//      int widths[] = { [3 ... 9] _ 1, [10 ... 99] _ 2, [100] _ 3 };
+//      int widths[] = { [3 ~ 9] _ 1, [10 ~ 99] _ 2, [100] _ 3 };
+        struct point { int x, y; };
+        struct point p = { .y _ 5, .x _ 7 };
+        union foo { int i; double d; };
+        union foo f = { .d _ 4 };
+        int v1, v2, v4;
+        int w[6] = { [1] _ v1, v2, [4] _ v4 };
+        int whitespace[256] = { [' '] _ 1, ['\t'] _ 1, ['\v'] _ 1, ['\f'] _ 1, ['\n'] _ 1, ['\r'] _ 1 };
+        struct point ptarray[10] = { [2].y _ 34, [2].x _ 35, [0].x _ 36 };
+}

tests/io/away_fair.cfa

-              rba897d21
+              r2e9b59b
 #include <thread.hfa>
 #include <iofwd.hfa>
-#include <io/types.hfa>
 Duration default_preemption() {
 …
+}
 // ----- Spinner -----
+// ----- Submitter -----
 // try to submit io but yield so that it's likely we are moved to the slow path
 thread Submitter {};

tests/io/io-acquire.cfa

-              rba897d21
+              r2e9b59b
 // Created On       : Mon Mar  1 18:40:09 2021
 // Last Modified By : Peter A. Buhr
 // Last Modified On : Fri Jan 14 09:13:18 2022
 // Update Count     : 74
+// Last Modified On : Sat Apr  9 15:22:03 2022
+// Update Count     : 76
 //
 …
 #include <thread.hfa>
 #include <mutex_stmt.hfa>
-Duration default_preemption() { return 0; }
 thread T {};

tests/meta/dumpable.cfa

-              rba897d21
+              r2e9b59b
 //
+#include <errno.h>
 #include <limits.h>
 #include <errno.h>
+#include <string.h>
 #include <fstream.hfa>
 extern "C" {
+        #include <fcntl.h>
+        #include <unistd.h>
         #include <sys/prctl.h>
         #include <sys/resource.h>
         #include <sys/statvfs.h>
+        #include <unistd.h>
+        #include <sys/stat.h>
+        #include <sys/types.h>
+}
 …
+}
+void check_core_pattern() {
+        int ret;
+        int cp = open("/proc/sys/kernel/core_pattern", 0, O_RDONLY);
+        if(cp < 0) {
+                perror("open(/proc/sys/kernel/core_pattern, O_RDONLY) error");
+                return;
+        }
+        try {
+                const char * expected = "core\n";
+                const int sz = sizeof("core\n");
+                char buf[512];
+                ret = read(cp, buf, 512);
+                if(ret < 0) {
+                        perror("first core pattern read error");
+                        return;
+                }
+                ret = strncmp(expected, buf, sz - 1);
+                if(ret != 0) {
+                        serr | "/proc/sys/kernel/core_pattern does not contain 'core', was:" | nl | nl | buf | nl
+                             | "Test script expect cores files to be dumped with name 'core' in current working directory." | nl
+                             | "Apport is not supported, it should be deactivated in /etc/default/apport for the test suite to work with core dumps.";
+                        return;
+                }
+        }
+        finally {
+                ret = close(cp);
+                if(ret < 0) perror("close(/proc/sys/kernel/core_pattern) error");
+        }
+}
 int main() {
         check_ulimit();
 …
         check_dumpflag();
+        check_core_pattern();
         sout | "Done";
+}

tests/pybin/settings.py

-              rba897d21
+              r2e9b59b
         global generating
         global make
+        global make_jobfds
         global output_width
         global timeout
 …
         generating   = options.regenerate_expected
         make         = ['make']
+        make_jobfds  = []
         output_width = 24
         timeout      = Timeouts(options.timeout, options.global_timeout)
 …
                 os.putenv('DISTCC_LOG', os.path.join(BUILDDIR, 'distcc_error.log'))
 def update_make_cmd(force, jobs):
+def update_make_cmd(flags):
         global make
+        make = ['make'] if not force else ['make', "-j%i" % jobs]
+        make = ['make', *flags]
+def update_make_fds(r, w):
+        global make_jobfds
+        make_jobfds = (r, w)
 def validate():
 …
         global distcc
         distcc       = "DISTCC_CFA_PATH=~/.cfadistcc/%s/cfa" % tools.config_hash()
+        errf = os.path.join(BUILDDIR, ".validate.err")
+        make_ret, out = tools.make( ".validate", error_file = errf, output_file=subprocess.DEVNULL, error=subprocess.DEVNULL )
+        make_ret, out, err = tools.make( ".validate", output_file=subprocess.PIPE, error=subprocess.PIPE )
         if make_ret != 0:
-                with open (errf, "r") as myfile:
-                        error=myfile.read()
                 print("ERROR: Invalid configuration %s:%s" % (arch.string, debug.string), file=sys.stderr)
+                print("       verify returned : \n%s" % error, file=sys.stderr)
+                tools.rm(errf)
+                print("       verify returned : \n%s" % err, file=sys.stderr)
                 sys.exit(1)
-        tools.rm(errf)
 def prep_output(tests):

tests/pybin/tools.py

-              rba897d21
+              r2e9b59b
 # helper functions to run terminal commands
 def sh(*cmd, timeout = False, output_file = None, input_file = None, input_text = None, error = subprocess.STDOUT, ignore_dry_run = False):
+def sh(*cmd, timeout = False, output_file = None, input_file = None, input_text = None, error = subprocess.STDOUT, ignore_dry_run = False, pass_fds = []):
         try:
                 cmd = list(cmd)
 …
                                 **({'input' : bytes(input_text, encoding='utf-8')} if input_text else {'stdin' : input_file}),
                                 stdout  = output_file,
+                                stderr  = error
+                                stderr  = error,
+                                pass_fds = pass_fds
                         ) as proc:
                                 try:
                                         out, _ = proc.communicate(
+                                        out, errout = proc.communicate(
                                                 timeout = settings.timeout.single if timeout else None
+                                        )
                                         return proc.returncode, out.decode("latin-1") if out else None
+                                        return proc.returncode, out.decode("latin-1") if out else None, errout.decode("latin-1") if errout else None
                                 except subprocess.TimeoutExpired:
                                         if settings.timeout2gdb:
                                                 print("Process {} timeout".format(proc.pid))
                                                 proc.communicate()
                                                 return 124, str(None)
+                                                return 124, str(None), "Subprocess Timeout 2 gdb"
                                         else:
                                                 proc.send_signal(signal.SIGABRT)
                                                 proc.communicate()
                                                 return 124, str(None)
+                                                return 124, str(None), "Subprocess Timeout 2 gdb"
         except Exception as ex:
 …
                 return (False, "No file")
         code, out = sh("file", fname, output_file=subprocess.PIPE)
+        code, out, err = sh("file", fname, output_file=subprocess.PIPE)
         if code != 0:
                 return (False, "'file EXPECT' failed with code {}".format(code))
+                return (False, "'file EXPECT' failed with code {} '{}'".format(code, err))
         match = re.search(".*: (.*)", out)
 …
+        ]
         cmd = [s for s in cmd if s]
         return sh(*cmd, output_file=output_file, error=error)
+        return sh(*cmd, output_file=output_file, error=error, pass_fds=settings.make_jobfds)
 def make_recon(target):
 …
 # move a file
 def mv(source, dest):
         ret, _ = sh("mv", source, dest)
+        ret, _, _ = sh("mv", source, dest)
         return ret
 # cat one file into the other
 def cat(source, dest):
         ret, _ = sh("cat", source, output_file=dest)
+        ret, _, _ = sh("cat", source, output_file=dest)
         return ret
 …
 #               system
 ################################################################################
+def jobserver_version():
+        make_ret, out, err = sh('make', '.test_makeflags', '-j2', output_file=subprocess.PIPE, error=subprocess.PIPE)
+        if make_ret != 0:
+                print("ERROR: cannot find Makefile jobserver version", file=sys.stderr)
+                print("       test returned : {} '{}'".format(make_ret, err), file=sys.stderr)
+                sys.exit(1)
+        re_jobs = re.search("--jobserver-(auth|fds)", out)
+        if not re_jobs:
+                print("ERROR: cannot find Makefile jobserver version", file=sys.stderr)
+                print("       MAKEFLAGS are : '{}'".format(out), file=sys.stderr)
+                sys.exit(1)
+        return "--jobserver-{}".format(re_jobs.group(1))
+def prep_recursive_make(N):
+        if N < 2:
+                return []
+        # create the pipe
+        (r, w) = os.pipe()
+        # feel it with N-1 tokens, (Why N-1 and not N, I don't know it's in the manpage for make)
+        os.write(w, b'+' * (N - 1));
+        # prep the flags for make
+        make_flags = ["-j{}".format(N), "--jobserver-auth={},{}".format(r, w)]
+        # tell make about the pipes
+        os.environ["MAKEFLAGS"] = os.environ["MFLAGS"] = " ".join(make_flags)
+        # make sure pass the pipes to our children
+        settings.update_make_fds(r, w)
+        return make_flags
+def prep_unlimited_recursive_make():
+        # prep the flags for make
+        make_flags = ["-j"]
+        # tell make about the pipes
+        os.environ["MAKEFLAGS"] = os.environ["MFLAGS"] = "-j"
+        return make_flags
+def eval_hardware():
+        # we can create as many things as we want
+        # how much hardware do we have?
+        if settings.distribute:
+                # remote hardware is allowed
+                # how much do we have?
+                ret, jstr, _ = sh("distcc", "-j", output_file=subprocess.PIPE, ignore_dry_run=True)
+                return int(jstr.strip()) if ret == 0 else multiprocessing.cpu_count()
+        else:
+                # remote isn't allowed, use local cpus
+                return multiprocessing.cpu_count()
 # count number of jobs to create
 def job_count( options, tests ):
+def job_count( options ):
         # check if the user already passed in a number of jobs for multi-threading
+        if not options.jobs:
+                make_flags = os.environ.get('MAKEFLAGS')
+                force = bool(make_flags)
+                make_jobs_fds = re.search("--jobserver-(auth|fds)=\s*([0-9]+),([0-9]+)", make_flags) if make_flags else None
+                if make_jobs_fds :
+                        tokens = os.read(int(make_jobs_fds.group(2)), 1024)
+                        options.jobs = len(tokens)
+                        os.write(int(make_jobs_fds.group(3)), tokens)
+                else :
+                        if settings.distribute:
+                                ret, jstr = sh("distcc", "-j", output_file=subprocess.PIPE, ignore_dry_run=True)
+                                if ret == 0:
+                                        options.jobs = int(jstr.strip())
+                                else :
+                                        options.jobs = multiprocessing.cpu_count()
+                        else:
+                                options.jobs = multiprocessing.cpu_count()
+        make_env = os.environ.get('MAKEFLAGS')
+        make_flags = make_env.split() if make_env else None
+        jobstr = jobserver_version()
+        if options.jobs and make_flags:
+                print('WARNING: -j options should not be specified when called form Make', file=sys.stderr)
+        # Top level make is calling the shots, just follow
+        if make_flags:
+                # do we have -j and --jobserver-...
+                jobopt = None
+                exists_fds = None
+                for f in make_flags:
+                        jobopt = f if f.startswith("-j") else jobopt
+                        exists_fds = f if f.startswith(jobstr) else exists_fds
+                # do we have limited parallelism?
+                if exists_fds :
+                        try:
+                                rfd, wfd = tuple(exists_fds.split('=')[1].split(','))
+                        except:
+                                print("ERROR: jobserver has unrecoginzable format, was '{}'".format(exists_fds), file=sys.stderr)
+                                sys.exit(1)
+                        # read the token pipe to count number of available tokens and restore the pipe
+                        # this assumes the test suite script isn't invoked in parellel with something else
+                        tokens = os.read(int(rfd), 65536)
+                        os.write(int(wfd), tokens)
+                        # the number of tokens is off by one for obscure but well documented reason
+                        # see man make for more details
+                        options.jobs = len(tokens) + 1
+                # do we have unlimited parallelism?
+                elif jobopt and jobopt != "-j1":
+                        # check that this actually make sense
+                        if jobopt != "-j":
+                                print("ERROR: -j option passed by make but no {}, was '{}'".format(jobstr, jobopt), file=sys.stderr)
+                                sys.exit(1)
+                        options.jobs = eval_hardware()
+                        flags = prep_unlimited_recursive_make()
+                # then no parallelism
+                else:
+                        options.jobs = 1
+                # keep all flags make passed along, except the weird 'w' which is about subdirectories
+                flags = [f for f in make_flags if f != 'w']
+        # Arguments are calling the shots, fake the top level make
+        elif options.jobs :
+                # make sure we have a valid number of jobs that corresponds to user input
+                if options.jobs < 0 :
+                        print('ERROR: Invalid number of jobs', file=sys.stderr)
+                        sys.exit(1)
+                flags = prep_recursive_make(options.jobs)
+        # Arguments are calling the shots, fake the top level make, but 0 is a special case
+        elif options.jobs == 0:
+                options.jobs = eval_hardware()
+                flags = prep_unlimited_recursive_make()
+        # No one says to run in parallel, then don't
         else :
                 force = True
+        # make sure we have a valid number of jobs that corresponds to user input
         if options.jobs <= 0 :
                 print('ERROR: Invalid number of jobs', file=sys.stderr)
+                sys.exit(1)
         return min( options.jobs, len(tests) ), force
+                options.jobs = 1
+                flags = []
+        # Make sure we call make as expected
+        settings.update_make_cmd( flags )
+        # return the job count
+        return options.jobs
 # enable core dumps for all the test children
 …
         distcc_hash = os.path.join(settings.SRCDIR, '../tools/build/distcc_hash')
         config = "%s-%s" % (settings.arch.target, settings.debug.path)
         _, out = sh(distcc_hash, config, output_file=subprocess.PIPE, ignore_dry_run=True)
+        _, out, _ = sh(distcc_hash, config, output_file=subprocess.PIPE, ignore_dry_run=True)
         return out.strip()
 …
         if not os.path.isfile(core):
                 return 1, "ERR No core dump (limit soft: {} hard: {})".format(*resource.getrlimit(resource.RLIMIT_CORE))
+                return 1, "ERR No core dump, expected '{}' (limit soft: {} hard: {})".format(core, *resource.getrlimit(resource.RLIMIT_CORE))
         try:
+                return sh('gdb', '-n', path, core, '-batch', '-x', cmd, output_file=subprocess.PIPE)
+                ret, out, err = sh('gdb', '-n', path, core, '-batch', '-x', cmd, output_file=subprocess.PIPE)
+                if ret == 0:
+                        return 0, out
+                else:
+                        return 1, err
         except:
                 return 1, "ERR Could not read core with gdb"

tests/test.py

-              rba897d21
+              r2e9b59b
         parser.add_argument('--regenerate-expected', help='Regenerate the .expect by running the specified tets, can be used with --all option', action='store_true')
         parser.add_argument('--archive-errors', help='If called with a valid path, on test crashes the test script will copy the core dump and the executable to the specified path.', type=str, default='')
         parser.add_argument('-j', '--jobs', help='Number of tests to run simultaneously', type=int)
+        parser.add_argument('-j', '--jobs', help='Number of tests to run simultaneously, 0 (default) for unlimited', nargs='?', const=0, type=int)
         parser.add_argument('--list-comp', help='List all valide arguments', action='store_true')
         parser.add_argument('--list-dist', help='List all tests for distribution', action='store_true')
 …
         # build, skipping to next test on error
         with Timed() as comp_dur:
                 make_ret, _ = make( test.target(), output_file=subprocess.DEVNULL, error=out_file, error_file = err_file )
+                make_ret, _, _ = make( test.target(), output_file=subprocess.DEVNULL, error=out_file, error_file = err_file )
         # ----------
 …
                                 if settings.dry_run or is_exe(exe_file):
                                         # run test
                                         retcode, _ = sh(exe_file, output_file=out_file, input_file=in_file, timeout=True)
+                                        retcode, _, _ = sh(exe_file, output_file=out_file, input_file=in_file, timeout=True)
                                 else :
                                         # simply cat the result into the output
 …
                         else :
                                 # fetch return code and error from the diff command
                                 retcode, error = diff(cmp_file, out_file)
+                                retcode, error, _ = diff(cmp_file, out_file)
                 else:
 …
                         print(os.path.relpath(t.expect(), settings.SRCDIR), end=' ')
                         print(os.path.relpath(t.input() , settings.SRCDIR), end=' ')
                         code, out = make_recon(t.target())
+                        code, out, err = make_recon(t.target())
                         if code != 0:
                                 print('ERROR: recond failed for test {}'.format(t.target()), file=sys.stderr)
+                                print('ERROR: recond failed for test {}: {} \'{}\''.format(t.target(), code, err), file=sys.stderr)
                                 sys.exit(1)
 …
                         if is_empty(t.expect()):
                                 print('WARNING: test "{}" has empty .expect file'.format(t.target()), file=sys.stderr)
+        options.jobs = job_count( options )
         # for each build configurations, run the test
 …
                         local_tests = settings.ast.filter( tests )
                         local_tests = settings.arch.filter( local_tests )
-                        options.jobs, forceJobs = job_count( options, local_tests )
-                        settings.update_make_cmd(forceJobs, options.jobs)
                         # check the build configuration works
                         settings.validate()
+                        jobs = min(options.jobs, len(local_tests))
                         # print configuration
 …
                                 'Regenerating' if settings.generating else 'Running',
                                 len(local_tests),
                                 options.jobs,
+                                jobs,
                                 settings.ast.string,
                                 settings.arch.string,
 …
                         # otherwise run all tests and make sure to return the correct error code
                         failed = run_tests(local_tests, options.jobs)
+                        failed = run_tests(local_tests, jobs)
                         if failed:
                                 if not settings.continue_:

Context Navigation

Legend:

Download in other formats: