Changeset d3a804f5 for src/libcfa/concurrency

Timestamp:

Feb 14, 2017, 1:19:28 PM (9 years ago)

Author:

Aaron Moss <a3moss@…>

Branches:

ADT, aaron-thesis, arm-eh, ast-experimental, cleanup-dtors, deferred_resn, demangler, enum, forall-pointer-decay, jacob/cs343-translation, jenkins-sandbox, master, new-ast, new-ast-unique-expr, new-env, no_list, persistent-indexer, pthread-emulation, qualifiedEnum, resolv-new, with_gc

Children:

97d246d

Parents:

f923b5f (diff), eafb094 (diff)
Note: this is a merge changeset, the changes displayed below correspond to the merge itself.
Use the (diff) links above to see all the changes relative to each parent.

Message:

Merge branch 'master' of plg.uwaterloo.ca:software/cfa/cfa-cc

Location:

src/libcfa/concurrency

Files:

• coroutines (modified) (4 diffs)
• coroutines.c (modified) (1 diff)
• invoke.c (modified) (1 diff)
• invoke.h (modified) (4 diffs)
• kernel (modified) (4 diffs)
• kernel.c (modified) (21 diffs)
• kernel_private.h (added)
• threads.c (modified) (6 diffs)

src/libcfa/concurrency/coroutines

@@ -36 +36 @@
 void ?{}(coStack_t * this);
 void ?{}(coroutine * this);
+void ?{}(coroutine * this, const char * name);
 void ^?{}(coStack_t * this);
 void ^?{}(coroutine * this);
@@ -76 +77 @@
                 "Possible cause is a suspend executed in a member called by a coroutine user rather than by the coroutine main.",
                 src->name, src );
-        assertf( src->last->notHalted,
+        assertf( src->last->state != Halted,
                 "Attempt by coroutine \"%.256s\" (%p) to suspend back to terminated coroutine \"%.256s\" (%p).\n"
                 "Possible cause is terminated coroutine's main routine has already returned.",
@@ -97 +98 @@
       // not resuming self ?
         if ( src != dst ) {
-                assertf( dst->notHalted ,
+                assertf( dst->state != Halted ,
                         "Attempt by coroutine %.256s (%p) to resume terminated coroutine %.256s (%p).\n"
                         "Possible cause is terminated coroutine's main routine has already returned.",
@@ -115 +116 @@
       // not resuming self ?
         if ( src != dst ) {
-                assertf( dst->notHalted ,
+                assertf( dst->state != Halted ,
                         "Attempt by coroutine %.256s (%p) to resume terminated coroutine %.256s (%p).\n"
                         "Possible cause is terminated coroutine's main routine has already returned.",

src/libcfa/concurrency/coroutines.c

@@ -61 +61 @@
 
 void ?{}(coroutine* this) {
-        this->name = "Anonymous Coroutine";
+        this{ "Anonymous Coroutine" };
+}
+
+void ?{}(coroutine* this, const char * name) {
+        this->name = name;
         this->errno_ = 0;
         this->state = Start;
-      this->notHalted = true;
         this->starter = NULL;
         this->last = NULL;

src/libcfa/concurrency/invoke.c

@@ -48 +48 @@
       main( this );
 
-      cor->state = Halt;
-      cor->notHalted = false;
+      cor->state = Halted;
 
       //Final suspend, should never return

src/libcfa/concurrency/invoke.h

@@ -30 +30 @@
       #define SCHEDULER_CAPACITY 10
 
+      struct spinlock {
+            volatile int lock;
+      };
+
       struct simple_thread_list {
             struct thread * head;
             struct thread ** tail;
+      };
+
+      struct signal_once {
+            volatile bool condition;
+            struct spinlock lock;
+            struct simple_thread_list blocked;
       };
 
@@ -40 +50 @@
             void append( struct simple_thread_list *, struct thread * );
             struct thread * pop_head( struct simple_thread_list * );
+
+            void ?{}(spinlock * this);
+            void ^?{}(spinlock * this);
+
+            void ?{}(signal_once * this);
+            void ^?{}(signal_once * this);
       }
       #endif
@@ -53 +69 @@
       };
 
-      enum coroutine_state { Start, Inactive, Active, Halt, Primed };
+      enum coroutine_state { Halted, Start, Inactive, Active, Primed };
 
       struct coroutine {
@@ -60 +76 @@
             int errno_;                         // copy of global UNIX variable errno
             enum coroutine_state state; // current execution status for coroutine
-            bool notHalted;                     // indicate if execuation state is not halted
-
             struct coroutine *starter;  // first coroutine to resume this one
             struct coroutine *last;             // last coroutine to resume this one
       };
 
-      struct simple_lock {
-        struct simple_thread_list blocked;
-      };
-
       struct thread {
-            struct coroutine c;
-            struct simple_lock lock;
-            struct thread * next;
+            struct coroutine c;           // coroutine body used to store context
+            struct signal_once terminated;// indicate if execuation state is not halted
+            struct thread * next;         // instrusive link field for threads
       };
 

src/libcfa/concurrency/kernel

@@ -9 +9 @@
 //
 // Author           : Thierry Delisle
-// Created On       : Tue Jan 17 12:27:26 2016
+// Created On       : Tue Jan 17 12:27:26 2017
 // Last Modified By : Thierry Delisle
 // Last Modified On : --
@@ -27 +27 @@
 
 //-----------------------------------------------------------------------------
+// Locks
+void lock( spinlock * );
+void unlock( spinlock * );
+
+void wait( signal_once * );
+void signal( signal_once * );
+
+//-----------------------------------------------------------------------------
 // Cluster
 struct cluster {
         simple_thread_list ready_queue;
-        pthread_spinlock_t lock;
+        spinlock lock;
 };
 
@@ -38 +46 @@
 //-----------------------------------------------------------------------------
 // Processor
+enum FinishOpCode { No_Action, Release, Schedule, Release_Schedule };
+struct FinishAction {
+        FinishOpCode action_code;
+        thread * thrd;
+        spinlock * lock;
+};
+static inline void ?{}(FinishAction * this) { 
+        this->action_code = No_Action;
+        this->thrd = NULL;
+        this->lock = NULL;
+}
+static inline void ^?{}(FinishAction * this) {}
+
 struct processor {
-        struct processorCtx_t * ctx;
+        struct processorCtx_t * runner;
         cluster * cltr;
         coroutine * current_coroutine;
         thread * current_thread;
         pthread_t kernel_thread;
-        simple_lock lock;
-        volatile bool terminated;
+        
+        signal_once terminated;
+        volatile bool is_terminated;
+
+        struct FinishAction finish;
 };
 
@@ -51 +75 @@
 void ?{}(processor * this, cluster * cltr);
 void ^?{}(processor * this);
-
-
-//-----------------------------------------------------------------------------
-// Locks
-
-void ?{}(simple_lock * this);
-void ^?{}(simple_lock * this);
-
-void lock( simple_lock * );
-void unlock( simple_lock * );
-
-struct pthread_spinlock_guard {
-        pthread_spinlock_t * lock;
-};
-
-static inline void ?{}( pthread_spinlock_guard * this, pthread_spinlock_t * lock ) {
-        this->lock = lock;
-        pthread_spin_lock( this->lock );
-}
-
-static inline void ^?{}( pthread_spinlock_guard * this ) {
-        pthread_spin_unlock( this->lock );
-}
-
-// //Simple spinlock implementation from 
-// //http://stackoverflow.com/questions/1383363/is-my-spin-lock-implementation-correct-and-optimal
-// //Not optimal but correct
-// #define VOL 
-
-// struct simple_spinlock {
-//      VOL int lock;
-// };
-
-// extern VOL int __sync_lock_test_and_set( VOL int *, VOL int);
-// extern void __sync_synchronize();
-
-// static inline void lock( simple_spinlock * this ) {
-//     while (__sync_lock_test_and_set(&this->lock, 1)) {
-//         // Do nothing. This GCC builtin instruction
-//         // ensures memory barrier.
-//     }
-// }
-
-// static inline void unlock( simple_spinlock * this ) {
-//     __sync_synchronize(); // Memory barrier.
-//     this->lock = 0;
-// }
 
 #endif //KERNEL_H

src/libcfa/concurrency/kernel.c

@@ -9 +9 @@
 //
 // Author           : Thierry Delisle
-// Created On       : Tue Jan 17 12:27:26 2016
+// Created On       : Tue Jan 17 12:27:26 2017
 // Last Modified By : Thierry Delisle
 // Last Modified On : --
@@ -20 +20 @@
 
 //Header
-#include "kernel"
+#include "kernel_private.h"
 
 //C Includes
 #include <stddef.h>
 extern "C" {
+#include <fenv.h>
 #include <sys/resource.h>
 }
@@ -30 +31 @@
 //CFA Includes
 #include "libhdr.h"
-#include "threads"
 
 //Private includes
@@ -38 +38 @@
 //-----------------------------------------------------------------------------
 // Kernel storage
-struct processorCtx_t {
-        processor * proc;
-        coroutine c;
-};
-
-DECL_COROUTINE(processorCtx_t);
-
 #define KERNEL_STORAGE(T,X) static char X##_storage[sizeof(T)]
 
@@ -92 +85 @@
 
         rlimit r;
-        int ret = getrlimit( RLIMIT_STACK, &r);
+        getrlimit( RLIMIT_STACK, &r);
         this->size = r.rlim_cur;
 
@@ -114 +107 @@
         this->name = "Main Thread";
         this->errno_ = 0;
-        this->state = Inactive;
-        this->notHalted = true;
+        this->state = Start;
 }
 
@@ -127 +119 @@
         (&this->c){};
         this->proc = proc;
-        proc->ctx = this;
+        proc->runner = this;
 }
 
@@ -133 +125 @@
         (&this->c){ info };
         this->proc = proc;
-        proc->ctx = this;
-}
-
-void start(processor * this);
+        proc->runner = this;
+}
 
 void ?{}(processor * this) {
@@ -146 +136 @@
         this->current_coroutine = NULL;
         this->current_thread = NULL;
-        (&this->lock){};
-        this->terminated = false;
+        (&this->terminated){};
+        this->is_terminated = false;
 
         start( this );
 }
 
-void ?{}(processor * this, cluster * cltr, processorCtx_t * ctx) {
+void ?{}(processor * this, cluster * cltr, processorCtx_t * runner) {
         this->cltr = cltr;
         this->current_coroutine = NULL;
         this->current_thread = NULL;
-        (&this->lock){};
-        this->terminated = false;
-
-        this->ctx = ctx;
-        LIB_DEBUG_PRINTF("Kernel : constructing processor context %p\n", ctx);
-        ctx{ this };
+        (&this->terminated){};
+        this->is_terminated = false;
+
+        this->runner = runner;
+        LIB_DEBUG_PRINTF("Kernel : constructing processor context %p\n", runner);
+        runner{ this };
 }
 
 void ^?{}(processor * this) {
-        if( ! this->terminated ) {
+        if( ! this->is_terminated ) {
                 LIB_DEBUG_PRINTF("Kernel : core %p signaling termination\n", this);
-                this->terminated = true;
-                lock( &this->lock );
+                this->is_terminated = true;
+                wait( &this->terminated );
         }
 }
@@ -174 +164 @@
 void ?{}(cluster * this) {
         ( &this->ready_queue ){};
-        pthread_spin_init( &this->lock, PTHREAD_PROCESS_PRIVATE );
+        ( &this->lock ){};
 }
 
 void ^?{}(cluster * this) {
-        pthread_spin_destroy( &this->lock );
-}
-
-//-----------------------------------------------------------------------------
-// Processor running routines
-void main(processorCtx_t * ctx);
-thread * nextThread(cluster * this);
-void scheduleInternal(processor * this, thread * dst);
-void spin(processor * this, unsigned int * spin_count);
-
-void main(processorCtx_t * ctx) {
-        processor * this = ctx->proc;
+        
+}
+
+//=============================================================================================
+// Kernel Scheduling logic
+//=============================================================================================
+//Main of the processor contexts
+void main(processorCtx_t * runner) {
+        processor * this = runner->proc;
         LIB_DEBUG_PRINTF("Kernel : core %p starting\n", this);
 
         thread * readyThread = NULL;
-        for( unsigned int spin_count = 0; ! this->terminated; spin_count++ ) {
-                
+        for( unsigned int spin_count = 0; ! this->is_terminated; spin_count++ ) 
+        {
                 readyThread = nextThread( this->cltr );
 
-                if(readyThread) {
-                        scheduleInternal(this, readyThread);
+                if(readyThread) 
+                {
+                        runThread(this, readyThread);
+
+                        //Some actions need to be taken from the kernel
+                        finishRunning(this);
+
                         spin_count = 0;
-                } else {
+                } 
+                else 
+                {
                         spin(this, &spin_count);
                 }               
@@ -206 +200 @@
 
         LIB_DEBUG_PRINTF("Kernel : core %p unlocking thread\n", this);
-        unlock( &this->lock );
+        signal( &this->terminated );
         LIB_DEBUG_PRINTF("Kernel : core %p terminated\n", this);
 }
 
-//Declarations for scheduleInternal
-extern void ThreadCtxSwitch(coroutine * src, coroutine * dst);
-
-// scheduleInternal runs a thread by context switching 
+// runThread runs a thread by context switching 
 // from the processor coroutine to the target thread 
-void scheduleInternal(processor * this, thread * dst) {
-        // coroutine * proc_ctx = get_coroutine(this->ctx);
-        // coroutine * thrd_ctx = get_coroutine(dst);
-
-        // //Update global state
-        // this->current_thread = dst;
-
-        // // Context Switch to the thread
-        // ThreadCtxSwitch(proc_ctx, thrd_ctx);
-        // // when ThreadCtxSwitch returns we are back in the processor coroutine
-
-        coroutine * proc_ctx = get_coroutine(this->ctx);
-        coroutine * thrd_ctx = get_coroutine(dst);
-      thrd_ctx->last = proc_ctx;
- 
-      // context switch to specified coroutine
-      // Which is now the current_coroutine
-      LIB_DEBUG_PRINTF("Kernel : switching to ctx %p (from %p, current %p)\n", thrd_ctx, proc_ctx, this->current_coroutine);
-      this->current_thread = dst;
-      this->current_coroutine = thrd_ctx;
-      CtxSwitch( proc_ctx->stack.context, thrd_ctx->stack.context );
-      this->current_coroutine = proc_ctx;
-      LIB_DEBUG_PRINTF("Kernel : returned from ctx %p (to %p, current %p)\n", thrd_ctx, proc_ctx, this->current_coroutine);
- 
-      // when CtxSwitch returns we are back in the processor coroutine
+void runThread(processor * this, thread * dst) {
+        coroutine * proc_cor = get_coroutine(this->runner);
+        coroutine * thrd_cor = get_coroutine(dst);
+        
+        //Reset the terminating actions here
+        this->finish.action_code = No_Action;
+
+        //Update global state
+        this->current_thread = dst;
+
+        // Context Switch to the thread
+        ThreadCtxSwitch(proc_cor, thrd_cor);
+        // when ThreadCtxSwitch returns we are back in the processor coroutine
+}
+
+// Once a thread has finished running, some of 
+// its final actions must be executed from the kernel
+void finishRunning(processor * this) {
+        if( this->finish.action_code == Release ) {
+                unlock( this->finish.lock );
+        }
+        else if( this->finish.action_code == Schedule ) {
+                ScheduleThread( this->finish.thrd );
+        }
+        else if( this->finish.action_code == Release_Schedule ) {
+                unlock( this->finish.lock );            
+                ScheduleThread( this->finish.thrd );
+        }
+        else {
+                assert(this->finish.action_code == No_Action);
+        }
 }
 
@@ -262 +259 @@
         processorCtx_t proc_cor_storage = { proc, &info };
 
+        LIB_DEBUG_PRINTF("Coroutine : created stack %p\n", proc_cor_storage.c.stack.base);
+
         //Set global state
-        proc->current_coroutine = &proc->ctx->c;
+        proc->current_coroutine = &proc->runner->c;
         proc->current_thread = NULL;
 
         //We now have a proper context from which to schedule threads
-        LIB_DEBUG_PRINTF("Kernel : core %p created (%p)\n", proc, proc->ctx);
+        LIB_DEBUG_PRINTF("Kernel : core %p created (%p, %p)\n", proc, proc->runner, &ctx);
 
         // SKULLDUGGERY: Since the coroutine doesn't have its own stack, we can't 
@@ -275 +274 @@
         proc_cor_storage.c.state = Active;
       main( &proc_cor_storage );
-      proc_cor_storage.c.state = Halt;
-      proc_cor_storage.c.notHalted = false;
+      proc_cor_storage.c.state = Halted;
 
         // Main routine of the core returned, the core is now fully terminated
-        LIB_DEBUG_PRINTF("Kernel : core %p main ended (%p)\n", proc, proc->ctx);        
+        LIB_DEBUG_PRINTF("Kernel : core %p main ended (%p)\n", proc, proc->runner);     
 
         return NULL;
@@ -287 +285 @@
         LIB_DEBUG_PRINTF("Kernel : Starting core %p\n", this);
         
-        pthread_attr_t attributes;
-        pthread_attr_init( &attributes );
-
-        pthread_create( &this->kernel_thread, &attributes, CtxInvokeProcessor, (void*)this );
-
-        pthread_attr_destroy( &attributes );
+        // pthread_attr_t attributes;
+        // pthread_attr_init( &attributes );
+
+        pthread_create( &this->kernel_thread, NULL, CtxInvokeProcessor, (void*)this );
+
+        // pthread_attr_destroy( &attributes );
 
         LIB_DEBUG_PRINTF("Kernel : core %p started\n", this);   
@@ -299 +297 @@
 //-----------------------------------------------------------------------------
 // Scheduler routines
-void thread_schedule( thread * thrd ) {
+void ScheduleThread( thread * thrd ) {
         assertf( thrd->next == NULL, "Expected null got %p", thrd->next );
         
-        pthread_spinlock_guard guard = { &systemProcessor->cltr->lock };
+        lock( &systemProcessor->cltr->lock );
         append( &systemProcessor->cltr->ready_queue, thrd );
+        unlock( &systemProcessor->cltr->lock );
 }
 
 thread * nextThread(cluster * this) {
-        pthread_spinlock_guard guard = { &this->lock };
-        return pop_head( &this->ready_queue );
+        lock( &this->lock );
+        thread * head = pop_head( &this->ready_queue );
+        unlock( &this->lock );
+        return head;
+}
+
+void ScheduleInternal() {
+        suspend();
+}
+
+void ScheduleInternal( spinlock * lock ) {
+        get_this_processor()->finish.action_code = Release;
+        get_this_processor()->finish.lock = lock;
+        suspend();
+}
+
+void ScheduleInternal( thread * thrd ) {
+        get_this_processor()->finish.action_code = Schedule;
+        get_this_processor()->finish.thrd = thrd;
+        suspend();
+}
+
+void ScheduleInternal( spinlock * lock, thread * thrd ) {
+        get_this_processor()->finish.action_code = Release_Schedule;
+        get_this_processor()->finish.lock = lock;
+        get_this_processor()->finish.thrd = thrd;
+        suspend();
 }
 
@@ -314 +338 @@
 // Kernel boot procedures
 void kernel_startup(void) {
-
+        LIB_DEBUG_PRINTF("Kernel : Starting\n");        
+
+        // Start by initializing the main thread
         // SKULLDUGGERY: the mainThread steals the process main thread 
         // which will then be scheduled by the systemProcessor normally
-        LIB_DEBUG_PRINTF("Kernel : Starting\n");        
-
+        mainThread = (thread *)&mainThread_storage;
         current_stack_info_t info;
-
-        // LIB_DEBUG_PRINTF("Kernel : core    base : %p \n", info.base );
-        // LIB_DEBUG_PRINTF("Kernel : core storage : %p \n", info.storage );
-        // LIB_DEBUG_PRINTF("Kernel : core    size : %x \n", info.size );
-        // LIB_DEBUG_PRINTF("Kernel : core   limit : %p \n", info.limit );
-        // LIB_DEBUG_PRINTF("Kernel : core context : %p \n", info.context );
-        // LIB_DEBUG_PRINTF("Kernel : core     top : %p \n", info.top );
-
-        // Start by initializing the main thread
-        mainThread = (thread *)&mainThread_storage;
         mainThread{ &info };
 
@@ -343 +358 @@
         // Add the main thread to the ready queue 
         // once resume is called on systemProcessor->ctx the mainThread needs to be scheduled like any normal thread
-        thread_schedule(mainThread);
+        ScheduleThread(mainThread);
 
         //initialize the global state variables
@@ -353 +368 @@
         // context. Hence, the main thread does not begin through CtxInvokeThread, like all other threads. The trick here is that
         // mainThread is on the ready queue when this call is made. 
-        resume(systemProcessor->ctx);
+        resume(systemProcessor->runner);
 
 
 
         // THE SYSTEM IS NOW COMPLETELY RUNNING
-
-
-
         LIB_DEBUG_PRINTF("Kernel : Started\n--------------------------------------------------\n\n");
 }
@@ -370 +382 @@
         // When its coroutine terminates, it return control to the mainThread
         // which is currently here
-        systemProcessor->terminated = true;
+        systemProcessor->is_terminated = true;
         suspend();
 
@@ -377 +389 @@
         // Destroy the system processor and its context in reverse order of construction
         // These were manually constructed so we need manually destroy them
-        ^(systemProcessor->ctx){};
+        ^(systemProcessor->runner){};
         ^(systemProcessor){};
 
@@ -389 +401 @@
 //-----------------------------------------------------------------------------
 // Locks
-void ?{}( simple_lock * this ) {
-        ( &this->blocked ){};
-}
-
-void ^?{}( simple_lock * this ) {
-
-}
-
-void lock( simple_lock * this ) {
+void ?{}( spinlock * this ) {
+        this->lock = 0;
+}
+void ^?{}( spinlock * this ) {
+
+}
+
+void lock( spinlock * this ) {
+        for ( unsigned int i = 1;; i += 1 ) {
+                if ( this->lock == 0 && __sync_lock_test_and_set_4( &this->lock, 1 ) == 0 ) break;
+        }
+}
+
+void unlock( spinlock * this ) {
+        __sync_lock_release_4( &this->lock );
+}
+
+void ?{}( signal_once * this ) {
+        this->condition = false;
+}
+void ^?{}( signal_once * this ) {
+
+}
+
+void wait( signal_once * this ) {
+        lock( &this->lock );
+        if( !this->condition ) {
+                append( &this->blocked, this_thread() );
+                ScheduleInternal( &this->lock );
+                lock( &this->lock );
+        }
+        unlock( &this->lock );
+}
+
+void signal( signal_once * this ) {
+        lock( &this->lock );
         {
-                pthread_spinlock_guard guard = { &systemCluster->lock };        //HUGE TEMP HACK which only works if we have a single cluster and is stupid
-                append( &this->blocked, this_thread() );
-        }
-        suspend();
-}
-
-void unlock( simple_lock * this ) {
-        thread * it;
-        while( it = pop_head( &this->blocked) ) {
-                thread_schedule( it );
-        }
+                this->condition = true;
+
+                thread * it;
+                while( it = pop_head( &this->blocked) ) {
+                        ScheduleThread( it );
+                }
+        }
+        unlock( &this->lock );
 }
 

src/libcfa/concurrency/threads.c

@@ -17 +17 @@
 #include "threads"
 
-#include "kernel"
+#include "kernel_private.h"
 #include "libhdr.h"
 
@@ -24 +24 @@
 
 extern "C" {
+        #include <fenv.h>
         #include <stddef.h>
 }
@@ -43 +44 @@
         (&this->c){};
         this->c.name = "Anonymous Coroutine";
-        (&this->lock){};
+        (&this->terminated){};
         this->next = NULL;
 }
@@ -71 +72 @@
 //-----------------------------------------------------------------------------
 // Starting and stopping threads
-extern "C" {
-      forall(dtype T | is_thread(T))
-      void CtxInvokeThread(T * this);
-}
-
-extern void thread_schedule( thread * );
-
 forall( dtype T | is_thread(T) )
 void start( T* this ) {
@@ -91 +85 @@
         CtxSwitch( thrd_c->last->stack.context, thrd_c->stack.context );
 
-        LIB_DEBUG_PRINTF("Thread started : %p (t %p, c %p)\n", this, thrd_c, thrd_h);
-
-        thread_schedule(thrd_h);
+        ScheduleThread(thrd_h);
 }
 
 forall( dtype T | is_thread(T) )
 void stop( T* this ) {
-        thread*  thrd = get_thread(this);
-        if( thrd->c.notHalted ) {
-                lock( &thrd->lock );
-        }
+        wait( & get_thread(this)->terminated ); 
 }
 
 void yield( void ) {
-        thread_schedule( this_thread() );
-        suspend();
+        ScheduleInternal( get_this_processor()->current_thread );
 }
 
 void ThreadCtxSwitch(coroutine* src, coroutine* dst) {
+        // set state of current coroutine to inactive
+        src->state = Inactive;
+        dst->state = Active;
+
+        //update the last resumer
         dst->last = src;
 
-        // set state of current coroutine to inactive
-        src->state = Inactive;
-
-        // set new coroutine that task is executing
+        // set new coroutine that the processor is executing
+        // and context switch to it
         get_this_processor()->current_coroutine = dst;  
-
-        // context switch to specified coroutine
         CtxSwitch( src->stack.context, dst->stack.context );
-        // when CtxSwitch returns we are back in the src coroutine
+        get_this_processor()->current_coroutine = src;  
 
         // set state of new coroutine to active
+        dst->state = Inactive;
         src->state = Active;
 }
@@ -130 +120 @@
 extern "C" {
         void __thread_signal_termination( thread * this ) {
-                this->c.state = Halt;
-                this->c.notHalted = false;
-                unlock( &this->lock );
+                this->c.state = Halted;
+                LIB_DEBUG_PRINTF("Thread end : %p\n", this);
+                signal( &this->terminated );    
         }
 }