Mikolaj / ghc

Mirror of ghc repository. DO NOT SUBMIT PULL REQUESTS HERE
http://www.haskell.org/ghc/
Other
1 stars 0 forks source link

Fix the GC.c calculations of par_tot_copied #2

Closed Mikolaj closed 12 years ago

Mikolaj commented 12 years ago

15:48 @mikolaj it can be fixed by changing the following lines: 15:48 @mikolaj rts/sm/GC.c- if (n_gc_threads == 1) { 15:48 @mikolaj rts/sm/GC.c- par_max_copied = copied; 15:48 @mikolaj rts/sm/GC.c: par_tot_copied = copied; 15:48 @mikolaj rts/sm/GC.c- } else { 15:48 @mikolaj rts/sm/GC.c: par_tot_copied = copied; 15:48 @mikolaj rts/sm/GC.c- } 15:48 @mikolaj to read: 15:48 @mikolaj rts/sm/GC.c: par_tot_copied = copied; 15:48 @mikolaj rts/sm/GC.c- if (n_gc_threads == 1) { 15:48 @mikolaj rts/sm/GC.c- par_max_copied = 0; 15:48 @mikolaj rts/sm/GC.c: par_tot_copied = 0: 15:48 @mikolaj rts/sm/GC.c: } 15:48 @mikolaj I checked if we can drop par_tot_copied altogether, but we can't, because it differs from the 'copie d' field of GCStatsGHC in that it does not contain the 'mutable list', for which we can't determine which core "copie d" it, as far as I understand, so it'd skew the Parallel GC work balance