GitLab

Try running any program with -G1 and a large -H option -- for example, cichelli from the nofib suite:

$ ./cichelli  +RTS -s -H32M -G1
[output snipped]
41,211,752 bytes allocated in the heap
      19,329,984 bytes copied during GC
       1,760,916 bytes maximum residency (34 sample(s))
          36,724 bytes maximum slop
               7 MB total memory in use (0 MB lost due to fragmentation)

Here, the suggestion of 32M seems to have been ignored, as the heap is at most 7M during execution. Whereas without -G1:

./cichelli  +RTS -s -H32M
[output snipped]
41,212,300 bytes allocated in the heap
       1,121,304 bytes copied during GC
          49,484 bytes maximum residency (1 sample(s))
          36,824 bytes maximum slop
              31 MB total memory in use (0 MB lost due to fragmentation)

Here, the size suggestion is used and a 32M heap is allocated.

Shouldn't the GC use the heap size suggestion even when there is only one generation?

Trac metadata

Trac field	Value
Version	6.10.1
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Runtime System
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture