Another new code generator status report

[Edward Z. Yang]

Oh yeah, sorry ^_^

Columns:
1. Percent size change of binary
2. Size of the old binary
3. Delta between old binary and new binary

Where "old" is "compiled with old codegen" and vice versa.
Thus,

>    0.2699   7024   1896 ./stranal/should_run/strun001.o

means that strun001.o was 7kb when compiled with the old code
gen, and gained 2kb with the new codegen, making for a 27% increase
in size.

Edward

Excerpts from Johan Tibell's message of Sat May 14 11:36:23 -0400 2011:
> Hi,
> 
> Do you mind clarifying what the different numbers mean?
> 
> Thanks!
> 
> On Sat, May 14, 2011 at 4:39 PM, Edward Z. Yang <ezyang at mit.edu> wrote:
> > I spent some time last evening and today fixing some more low
> > hanging fruit (still waiting for a big enough chunk of free time to
> > tackle the spill and stack layout rewrites), and here are the latest
> > statistics:
> >
> >    0.2558   5160   1320 ./deSugar/should_compile/ds018.o
> >    0.2565   7204   1848 ./gadt/red-black.o
> >    0.2640   1576    416 ./typecheck/should_compile/tc008.o
> >    0.2649   2416    640 ./typecheck/should_compile/tc236.o
> >    0.2683 101532  27244 ./dph/sumnats/SumNatsVect.o
> >    0.2699   7024   1896 ./stranal/should_run/strun001.o
> >    0.2746   1792    492 ./deSugar/should_compile/ds004.o
> >    0.2835   1552    440 ./deSugar/should_compile/ds058.o
> >    0.2854  12248   3496 ./ffi/should_compile/cc004.o
> >    0.2871   3204    920 ./typecheck/should_compile/T3342.o
> >    0.2880  35024  10088 ./typecheck/should_compile/tc095.o
> >    0.2884   6020   1736 ./deSugar/should_compile/ds036.o
> >    0.3341   3460   1156 ./haddock/should_compile_flag_haddock/haddockA033.o
> >    0.3451   2712    936 ./simplCore/should_compile/simpl016.o
> >    Average percent change: 0.045714553
> >
> > For comparison, here is what we had last April 18th:
> >
> >    0.2830   7024   1988 ./stranal/should_run/strun001.o
> >    0.2833 110516  31308 ./indexed-types/should_compile/T3787.o
> >    0.2835   1552    440 ./deSugar/should_compile/ds058.o
> >    0.2871   3204    920 ./typecheck/should_compile/T3342.o
> >    0.2884   6020   1736 ./deSugar/should_compile/ds036.o
> >    0.2924   1792    524 ./deSugar/should_compile/ds004.o
> >    0.2962  12248   3628 ./ffi/should_compile/cc004.o
> >    0.3024  35024  10592 ./typecheck/should_compile/tc095.o
> >    0.3134  94508  29616 ./dph/sumnats/SumNatsVect.o
> >    0.3341   3460   1156 ./haddock/should_compile_flag_haddock/haddockA033.o
> >    0.3451   2712    936 ./simplCore/should_compile/simpl016.o
> >    0.3920   7744   3036 ./gadt/red-black.o
> >    0.4525 208856  94504 ./deSugar/should_run/dsrun023.o
> >    0.5459   2220   1212 ./deSugar/should_compile/ds005.o
> >    1.1125 700616 779448 ./perf/compiler/T3294.o
> >    Average percent change: 0.04710892
> >
> > The average percent change didn't change much, but I managed to kill
> > off some of the big outliers.
> >
> > Cheers,
> > Edward
> >
> > _______________________________________________
> > Cvs-ghc mailing list
> > Cvs-ghc at haskell.org
> > http://www.haskell.org/mailman/listinfo/cvs-ghc
> >