static Ticks ExitUserTime = 0;
static Ticks ExitElapsedTime = 0;
-static ullong GC_tot_alloc = 0;
-static ullong GC_tot_copied = 0;
+static StgWord64 GC_tot_alloc = 0;
+static StgWord64 GC_tot_copied = 0;
-static ullong GC_par_max_copied = 0;
-static ullong GC_par_avg_copied = 0;
+static StgWord64 GC_par_max_copied = 0;
+static StgWord64 GC_par_avg_copied = 0;
static Ticks GC_start_time = 0, GC_tot_time = 0; /* User GC Time */
static Ticks GCe_start_time = 0, GCe_tot_time = 0; /* Elapsed GC time */
PROF_VAL(RPe_tot_time + HCe_tot_time) - InitElapsedStamp;
if (MutElapsedTime < 0) { MutElapsedTime = 0; } /* sometimes -0.00 */
- MutUserTime = user - GC_tot_time - PROF_VAL(RP_tot_time + HC_tot_time) - InitUserTime;
+ MutUserTime = user - GC_tot_time -
+ PROF_VAL(RP_tot_time + HC_tot_time) - InitUserTime;
if (MutUserTime < 0) { MutUserTime = 0; }
#if USE_PAPI
}
}
-#if defined(PROFILING) || defined(DEBUG)
- GC_start_time = getProcessCPUTime(); // needed in mut_user_time_during_GC()
-#endif
-
- if (RtsFlags.GcFlags.giveStats != NO_GC_STATS) {
-#if !defined(PROFILING) && !defined(DEBUG)
- GC_start_time = getProcessCPUTime();
-#endif
- GCe_start_time = getProcessElapsedTime();
+ if (RtsFlags.GcFlags.giveStats != NO_GC_STATS
+ || RtsFlags.ProfFlags.doHeapProfile)
+ // heap profiling needs GC_tot_time
+ {
+ getProcessTimes(&GC_start_time, &GCe_start_time);
if (RtsFlags.GcFlags.giveStats) {
GC_start_faults = getPageFaults();
}
stat_endGC (lnat alloc, lnat live, lnat copied, lnat gen,
lnat max_copied, lnat avg_copied, lnat slop)
{
- if (RtsFlags.GcFlags.giveStats != NO_GC_STATS) {
+ if (RtsFlags.GcFlags.giveStats != NO_GC_STATS ||
+ RtsFlags.ProfFlags.doHeapProfile)
+ // heap profiling needs GC_tot_time
+ {
Ticks time, etime, gc_time, gc_etime;
getProcessTimes(&time, &etime);
GC_coll_times[gen] += gc_time;
GC_coll_etimes[gen] += gc_etime;
- GC_tot_copied += (ullong) copied;
- GC_tot_alloc += (ullong) alloc;
- GC_par_max_copied += (ullong) max_copied;
- GC_par_avg_copied += (ullong) avg_copied;
+ GC_tot_copied += (StgWord64) copied;
+ GC_tot_alloc += (StgWord64) alloc;
+ GC_par_max_copied += (StgWord64) max_copied;
+ GC_par_avg_copied += (StgWord64) avg_copied;
GC_tot_time += gc_time;
GCe_tot_time += gc_etime;
/* Report the value of a counter */
#define REPORT(counter) \
{ \
- ullong_format_string(counter,temp,rtsTrue/*commas*/); \
+ showStgWord64(counter,temp,rtsTrue/*commas*/); \
statsPrintf(" (" #counter ") : %s\n",temp); \
}
}
if (RtsFlags.GcFlags.giveStats >= SUMMARY_GC_STATS) {
- ullong_format_string(GC_tot_alloc*sizeof(W_),
+ showStgWord64(GC_tot_alloc*sizeof(W_),
temp, rtsTrue/*commas*/);
statsPrintf("%16s bytes allocated in the heap\n", temp);
- ullong_format_string(GC_tot_copied*sizeof(W_),
+ showStgWord64(GC_tot_copied*sizeof(W_),
temp, rtsTrue/*commas*/);
statsPrintf("%16s bytes copied during GC\n", temp);
if ( ResidencySamples > 0 ) {
- ullong_format_string(MaxResidency*sizeof(W_),
+ showStgWord64(MaxResidency*sizeof(W_),
temp, rtsTrue/*commas*/);
statsPrintf("%16s bytes maximum residency (%ld sample(s))\n",
temp, ResidencySamples);
}
- ullong_format_string(MaxSlop*sizeof(W_), temp, rtsTrue/*commas*/);
+ showStgWord64(MaxSlop*sizeof(W_), temp, rtsTrue/*commas*/);
statsPrintf("%16s bytes maximum slop\n", temp);
statsPrintf("%16ld MB total memory in use (%ld MB lost due to fragmentation)\n\n",
- mblocks_allocated * MBLOCK_SIZE_W / (1024 * 1024 / sizeof(W_)),
- (mblocks_allocated * MBLOCK_SIZE_W - hw_alloc_blocks * BLOCK_SIZE_W) / (1024 * 1024 / sizeof(W_)));
+ peak_mblocks_allocated * MBLOCK_SIZE_W / (1024 * 1024 / sizeof(W_)),
+ (peak_mblocks_allocated * BLOCKS_PER_MBLOCK * BLOCK_SIZE_W - hw_alloc_blocks * BLOCK_SIZE_W) / (1024 * 1024 / sizeof(W_)));
/* Print garbage collections in each gen */
for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
i++, task = task->all_link) {
statsPrintf(" Task %2d %-8s : %6.2fs (%6.2fs) %6.2fs (%6.2fs)\n",
i,
- (task->tso == NULL) ? "(worker)" : "(bound)",
+ (task->worker) ? "(worker)" : "(bound)",
TICK_TO_DBL(task->mut_time),
TICK_TO_DBL(task->mut_etime),
TICK_TO_DBL(task->gc_time),
{
nat i;
lnat sparks_created = 0;
+ lnat sparks_dud = 0;
lnat sparks_converted = 0;
- lnat sparks_pruned = 0;
+ lnat sparks_gcd = 0;
+ lnat sparks_fizzled = 0;
for (i = 0; i < n_capabilities; i++) {
sparks_created += capabilities[i].sparks_created;
+ sparks_dud += capabilities[i].sparks_dud;
sparks_converted += capabilities[i].sparks_converted;
- sparks_pruned += capabilities[i].sparks_pruned;
+ sparks_gcd += capabilities[i].sparks_gcd;
+ sparks_fizzled += capabilities[i].sparks_fizzled;
}
- statsPrintf(" SPARKS: %ld (%ld converted, %ld pruned)\n\n",
- sparks_created, sparks_converted, sparks_pruned);
+ statsPrintf(" SPARKS: %ld (%ld converted, %ld dud, %ld GC'd, %ld fizzled)\n\n",
+ sparks_created + sparks_dud, sparks_converted, sparks_dud, sparks_gcd, sparks_fizzled);
}
#endif
TICK_TO_DBL(GCe_tot_time)*100/TICK_TO_DBL(etime));
if (time - GC_tot_time - PROF_VAL(RP_tot_time + HC_tot_time) == 0)
- ullong_format_string(0, temp, rtsTrue/*commas*/);
+ showStgWord64(0, temp, rtsTrue/*commas*/);
else
- ullong_format_string(
- (ullong)((GC_tot_alloc*sizeof(W_))/
+ showStgWord64(
+ (StgWord64)((GC_tot_alloc*sizeof(W_))/
TICK_TO_DBL(time - GC_tot_time -
PROF_VAL(RP_tot_time + HC_tot_time))),
temp, rtsTrue/*commas*/);
#endif
#if defined(THREADED_RTS) && defined(PROF_SPIN)
{
- nat g, s;
+ nat g;
statsPrintf("gc_alloc_block_sync: %"FMT_Word64"\n", gc_alloc_block_sync.spin);
statsPrintf("whitehole_spin: %"FMT_Word64"\n", whitehole_spin);
for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
- for (s = 0; s < generations[g].n_steps; s++) {
- statsPrintf("gen[%d].steps[%d].sync_large_objects: %"FMT_Word64"\n", g, s, generations[g].steps[s].sync_large_objects.spin);
- }
+ statsPrintf("gen[%d].sync: %"FMT_Word64"\n", g, generations[g].sync.spin);
}
}
#endif
fmt2 = "%d GCs, %ld/%ld avg/max bytes residency (%ld samples), %luM in use, %.2f INIT (%.2f elapsed), %.2f MUT (%.2f elapsed), %.2f GC (%.2f elapsed) :ghc>>\n";
}
/* print the long long separately to avoid bugginess on mingwin (2001-07-02, mingw-0.5) */
- statsPrintf(fmt1, GC_tot_alloc*(ullong)sizeof(W_));
+ statsPrintf(fmt1, GC_tot_alloc*(StgWord64)sizeof(W_));
statsPrintf(fmt2,
total_collections,
ResidencySamples == 0 ? 0 :
AvgResidency*sizeof(W_)/ResidencySamples,
MaxResidency*sizeof(W_),
ResidencySamples,
- (unsigned long)(mblocks_allocated * MBLOCK_SIZE / (1024L * 1024L)),
+ (unsigned long)(peak_mblocks_allocated * MBLOCK_SIZE / (1024L * 1024L)),
TICK_TO_DBL(InitUserTime), TICK_TO_DBL(InitElapsedTime),
TICK_TO_DBL(MutUserTime), TICK_TO_DBL(MutElapsedTime),
TICK_TO_DBL(GC_tot_time), TICK_TO_DBL(GCe_tot_time));
void
statDescribeGens(void)
{
- nat g, s, mut, lge;
- lnat live, slop;
+ nat g, mut, lge, i;
+ lnat gen_slop;
lnat tot_live, tot_slop;
+ lnat gen_live, gen_blocks;
bdescr *bd;
- step *step;
-
+ generation *gen;
+
debugBelch(
-"-----------------------------------------------------------------\n"
-" Gen Max Mut-list Step Blocks Large Live Slop\n"
-" Blocks Bytes Objects \n"
-"-----------------------------------------------------------------\n");
+"----------------------------------------------------------\n"
+" Gen Max Mut-list Blocks Large Live Slop\n"
+" Blocks Bytes Objects \n"
+"----------------------------------------------------------\n");
tot_live = 0;
tot_slop = 0;
+
for (g = 0; g < RtsFlags.GcFlags.generations; g++) {
- mut = 0;
- for (bd = generations[g].mut_list; bd != NULL; bd = bd->link) {
- mut += (bd->free - bd->start) * sizeof(W_);
+ gen = &generations[g];
+
+ for (bd = gen->large_objects, lge = 0; bd; bd = bd->link) {
+ lge++;
}
- debugBelch("%5d %7d %9d", g, generations[g].max_blocks, mut);
+ gen_live = genLiveWords(gen);
+ gen_blocks = genLiveBlocks(gen);
- for (s = 0; s < generations[g].n_steps; s++) {
- step = &generations[g].steps[s];
- for (bd = step->large_objects, lge = 0; bd; bd = bd->link) {
- lge++;
- }
- live = step->n_words + countOccupied(step->large_objects);
- if (s != 0) {
- debugBelch("%23s","");
+ mut = 0;
+ for (i = 0; i < n_capabilities; i++) {
+ mut += countOccupied(capabilities[i].mut_lists[g]);
+ gen_live += gcThreadLiveWords(i,g);
+ gen_blocks += gcThreadLiveBlocks(i,g);
}
- slop = (step->n_blocks + step->n_large_blocks) * BLOCK_SIZE_W - live;
- debugBelch("%6d %8d %8d %8ld %8ld\n", s, step->n_blocks, lge,
- live*sizeof(W_), slop*sizeof(W_));
- tot_live += live;
- tot_slop += slop;
- }
+
+ debugBelch("%5d %7d %9d", g, gen->max_blocks, mut);
+
+ gen_slop = gen_blocks * BLOCK_SIZE_W - gen_live;
+
+ debugBelch("%8ld %8d %8ld %8ld\n", gen_blocks, lge,
+ gen_live*sizeof(W_), gen_slop*sizeof(W_));
+ tot_live += gen_live;
+ tot_slop += gen_slop;
}
- debugBelch("-----------------------------------------------------------------\n");
- debugBelch("%48s%8ld %8ld\n","",tot_live*sizeof(W_),tot_slop*sizeof(W_));
- debugBelch("-----------------------------------------------------------------\n");
+ debugBelch("----------------------------------------------------------\n");
+ debugBelch("%41s%8ld %8ld\n","",tot_live*sizeof(W_),tot_slop*sizeof(W_));
+ debugBelch("----------------------------------------------------------\n");
debugBelch("\n");
}
-------------------------------------------------------------------------- */
extern HsInt64 getAllocations( void )
-{ return (HsInt64)total_allocated * sizeof(W_); }
+{ return (HsInt64)GC_tot_alloc * sizeof(W_); }
/* -----------------------------------------------------------------------------
Dumping stuff in the stats file, or via the debug message interface