Changes in libcfa/src/concurrency/ready_queue.cfa [0fb3ee5:a2a4566]

File:

• libcfa/src/concurrency/ready_queue.cfa (modified) (8 diffs)

libcfa/src/concurrency/ready_queue.cfa

@@ -303 +303 @@
                         lanes.help[idx].dst = 0;
                         lanes.help[idx].tri = 0;
+                }
+
+                caches = alloc( cpu_info.llc_count );
+                for( idx; (size_t)cpu_info.llc_count ) {
+                        (caches[idx]){};
                 }
         #else
@@ -399 +404 @@
                 /* 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);
+                __atomic_store_n(&lanes.caches[this / READYQ_SHARD_FACTOR].id, this_cache, __ATOMIC_RELAXED);
 
                 const unsigned long long ctsc = rdtscl();
@@ -505 +506 @@
         }
 
+        static inline int pop_getcpu(processor * proc, __ready_queue_caches_t * caches) {
+                const int prv = proc->rdq.cpu;
+                const int cpu = __kernel_getcpu();
+                if( prv != proc->rdq.cpu ) {
+                        unsigned pidx = cpu_info.llc_map[prv].cache;
+                        /* paranoid */ verify(pidx < cpu_info.llc_count);
+
+                        unsigned nidx = cpu_info.llc_map[cpu].cache;
+                        /* paranoid */ verify(pidx < cpu_info.llc_count);
+
+                        depart(caches[pidx]);
+                        arrive(caches[nidx]);
+
+                        __STATS( /* cpu migs++ */ )
+                }
+                return proc->rdq.cpu = cpu;
+        }
+
         // Pop from the ready queue from a given cluster
         __attribute__((hot)) thread$ * pop_fast(struct cluster * cltr) with (cltr->ready_queue) {
@@ -511 +530 @@
 
                 processor * const proc = kernelTLS().this_processor;
-                const int cpu = __kernel_getcpu();
+                const int cpu = pop_getcpu( proc, caches );
+                // const int cpu = __kernel_getcpu();
                 /* paranoid */ verify(cpu >= 0);
                 /* paranoid */ verify(cpu < cpu_info.hthrd_count);
@@ -528 +548 @@
                         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);
+                                unsigned long long tsc = moving_average(ctsc - ts(lanes.data[start + i]), lanes.tscs[start + i].ma);
                                 if(tsc > max) max = tsc;
                         }
@@ -549 +569 @@
                         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);
+                                unsigned long long tsc = moving_average(ctsc - ts(lanes.data[start + i]), lanes.tscs[start + i].ma);
                                 if(tsc > max) max = tsc;
                         }
@@ -557 +577 @@
                                 proc->rdq.target = MAX;
                                 lanes.help[target / READYQ_SHARD_FACTOR].tri++;
-                                if(moving_average(ctsc, lanes.tscs[target].tv, lanes.tscs[target].ma) > cutoff) {
+                                if(moving_average(ctsc - lanes.tscs[target].tv, lanes.tscs[target].ma) > cutoff) {
+                                        __STATS( __tls_stats()->ready.pop.helped[target]++; )
                                         thread$ * t = try_pop(cltr, target __STATS(, __tls_stats()->ready.pop.help));
                                         proc->rdq.last = target;
@@ -566 +587 @@
 
                         unsigned last = proc->rdq.last;
-                        if(last != MAX && moving_average(ctsc, lanes.tscs[last].tv, lanes.tscs[last].ma) > cutoff) {
+                        if(last != MAX && moving_average(ctsc - lanes.tscs[last].tv, lanes.tscs[last].ma) > cutoff) {
+                                __STATS( __tls_stats()->ready.pop.helped[last]++; )
                                 thread$ * t = try_pop(cltr, last __STATS(, __tls_stats()->ready.pop.help));
                                 if(t) return t;