On 26/05/2011 18:58, Ian Lynagh wrote:
Hi all,
Need to dash now, so haven't checked properly, but I think there's a
serious space regression in the HEAD. Be careful if doing a
multi-threaded validate, or you may find yourself swapping heavily!
The nightly build time jumped by ~30 mins or so last night.
nofib compile times are up by 20%, comparing HEAD before the generics
merge with last night's HEAD:
-1 s.d. ----- +8.7%
+1 s.d. ----- +31.9%
Average ----- +19.7%
My standard compile-performance test shows some increase when compiling
Cabal, but only back to where 7.0.3 was.
7.0.3:
13,377,509,376 bytes allocated in the heap
250 MB total memory in use (0 MB lost due to fragmentation)
Total time 24.46s ( 27.57s elapsed)
HEAD before the generics merge:
13,158,729,120 bytes allocated in the heap
225 MB total memory in use (0 MB lost due to fragmentation)
Total time 23.94s ( 27.07s elapsed)
HEAD now:
13,223,847,896 bytes allocated in the heap
250 MB total memory in use (0 MB lost due to fragmentation)
Total time 24.24s ( 27.40s elapsed)
So it seems that large --make compilations aren't affected much, but
short one-module compilations (as in nofib) are affected quite a lot.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
