Implement a new heap-tuning option: -H

-H alone causes the RTS to use a larger nursery, but without exceeding
the amount of memory that the application is already using.  It trades
off GC time against locality: the default setting is to use a
fixed-size 512k nursery, but this is sometimes worse than using a very
large nursery despite the worse locality.

Not all programs get faster, but some programs that use large heaps do
much better with -H.  e.g. this helps a lot with #3061 (binary-trees),
though not as much as specifying -H<large>.  Typically using -H<large>
is better than plain -H, because the runtime doesn't know ahead of
time how much memory you want to use.

Should -H be on by default?  I'm not sure, it makes some programs go
slower, but others go faster.

diff --git a/includes/rts/Flags.h b/includes/rts/Flags.h
index e78bed2..b861461 100644 (file)
--- a/includes/rts/Flags.h
+++ b/includes/rts/Flags.h
@@ -34,6 +34,7 @@ struct GC_FLAGS {
     nat     minAllocAreaSize;   /* in *blocks* */
     nat     minOldGenSize;      /* in *blocks* */
     nat     heapSizeSuggestion; /* in *blocks* */
+    rtsBool heapSizeSuggestionAuto;
     double  oldGenFactor;
     double  pcFreeHeap;
 

diff --git a/rts/RtsFlags.c b/rts/RtsFlags.c
index a6dbaf0..a646e57 100644 (file)
--- a/rts/RtsFlags.c
+++ b/rts/RtsFlags.c
@@ -70,6 +70,7 @@ void initRtsFlagsDefaults(void)
     RtsFlags.GcFlags.minOldGenSize      = (1024 * 1024)       / BLOCK_SIZE;
     RtsFlags.GcFlags.maxHeapSize       = 0;    /* off by default */
     RtsFlags.GcFlags.heapSizeSuggestion        = 0;    /* none */
+    RtsFlags.GcFlags.heapSizeSuggestionAuto = rtsFalse;
     RtsFlags.GcFlags.pcFreeHeap                = 3;    /* 3% */
     RtsFlags.GcFlags.oldGenFactor       = 2;
     RtsFlags.GcFlags.generations        = 2;
@@ -690,9 +691,13 @@ error = rtsTrue;
                break;
 
              case 'H':
-               RtsFlags.GcFlags.heapSizeSuggestion =
-                    (nat)(decodeSize(rts_argv[arg], 2, BLOCK_SIZE, HS_WORD_MAX) / BLOCK_SIZE);
-               break;
+                  if (rts_argv[arg][2] == '\0') {
+                      RtsFlags.GcFlags.heapSizeSuggestionAuto = rtsTrue;
+                  } else {
+                      RtsFlags.GcFlags.heapSizeSuggestion =
+                          (nat)(decodeSize(rts_argv[arg], 2, BLOCK_SIZE, HS_WORD_MAX) / BLOCK_SIZE);
+                  }
+                  break;
 
 #ifdef RTS_GTK_FRONTPANEL
              case 'f':

diff --git a/rts/sm/GC.c b/rts/sm/GC.c
index 5097483..0593bd7 100644 (file)
--- a/rts/sm/GC.c
+++ b/rts/sm/GC.c
@@ -1548,6 +1548,10 @@ resize_generations (void)
        size = stg_max(live * RtsFlags.GcFlags.oldGenFactor,
                       RtsFlags.GcFlags.minOldGenSize);
        
+        if (RtsFlags.GcFlags.heapSizeSuggestionAuto) {
+            RtsFlags.GcFlags.heapSizeSuggestion = size;
+        }
+
        // minimum size for generation zero
        min_alloc = stg_max((RtsFlags.GcFlags.pcFreeHeap * max) / 200,
                            RtsFlags.GcFlags.minAllocAreaSize);