On Jun 09, 2005 Simon Marlow wrote:
On 09 June 2005 11:47, Serge D. Mechveliani wrote:
A difference of 2.5Mb is really very small. If you could show that the space usage is linear in one case and flat in another, then we have a problem.
But a ratio 2600 K / 200 K for n = 55000 is not so small.
I never tested this, but what if in this mergeSort-like example the space sizes are space.6.4(n) = 13*space.5.02(n)
(which has the above ratio for n = 55000) ?
If you can demonstrate that GHC 6.4 uses 13 times more space than 5.02 on your example, then that's a performance bug - but all you showed so far is that 6.4 uses 2.5Mb more space. i.e. you need to do a few more measurements for different values of n.
I hope, this will be impossible to demonstrate. This measurement report for 55000 was stored several years ago. Now, I indeed, intend to compare 6.4 to 5.02 on several examples, downloaded ghc-5.02.3 and tried to make it with ghc-6.4 on recent Debian Linux. But it falied. It looks difficult to port 5.02.x to recent Debian Linux. ---------------------------------------------------- ---------------------------------------------------- ... ...
primop-primop-info.hs-incl ../../glafp-utils/mkdependC/mkdependC -f ...
/usr/bin/ghc -M -optdep-f -optdep.depend ...
-recomp -Rghc-timing -H16M -O utils/Bag.lhs utils/BitSet.lhs
...
/StixPrim.lhs main/Config.hs rename/ParseIface.hs parser/Parser.hs main/ParsePkgConf.hs
utils/FastTypes.lhs:41: warning: extra tokens at end of #else directive
utils/FastTypes.lhs:62: warning: extra tokens at end of #endif directive
...
nativeGen/MachMisc.lhs:72: unterminated argument list invoking macro "IF_ARCH_alpha"
<