This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: x86_64 varargs setup jump table
On Tue, Jul 20, 2010 at 4:05 PM, Richard Henderson <rth@redhat.com> wrote:
> On 07/20/2010 09:32 AM, Richard Henderson wrote:
>> On 07/19/2010 02:13 PM, Richard Henderson wrote:
>>> This bootstraps; regression test starting now.
>>>
>>> Obviously there's some patterns in i386.md that should be removed
>>> along with this, were this patch to go in.
>>
>> A slightly different patch that passes regression testing.
>> This also vanishes the patterns that should go.
>
> It was pointed out to me on IRC that x86_64 bootstrap is still broken.
> So unless there are objections to this patch, I'll commit it tomorrow
> morning (GMT+7). ?We can debate the performance impact afterward.
>
Here are results on Intel Core i7:
Old: Gcc 4.6.0 revision 162269
New: Gcc 4.6.0 revision 162269 + this patch.
New/Old
-O2:
164.gzip 0.595829%
175.vpr 0.123508%
176.gcc 0.580271%
181.mcf -0.724263%
186.crafty 0.222717%
197.parser 0.100806%
252.eon -0.159744%
253.perlbmk -6.45709%
254.gap 0.417202%
255.vortex 0.127226%
256.bzip2 -0.167926%
300.twolf 0%
SPECint_base2000 -0.461082%
168.wupwise -0.0897666%
171.swim -0.210859%
172.mgrid 0.304298%
173.applu 1.39241%
177.mesa -0.0538938%
178.galgel 0.0946074%
179.art -0.244662%
183.equake -1.24064%
187.facerec 0.50797%
188.ammp 0.532092%
189.lucas -0.149785%
191.fma3d 0.841751%
200.sixtrack -0.0752445%
301.apsi -0.209018%
SPECfp_base2000 0.0985068%
400.perlbench 3.35821%
401.bzip2 1.11732%
403.gcc -0.413223%
429.mcf 0.425532%
445.gobmk 0.41841%
456.hmmer 0%
458.sjeng 0%
462.libquantum -1.81818%
464.h264ref 0%
471.omnetpp 0%
473.astar 0%
483.xalancbmk 0.3861%
SPECint(R)_base2006 0.41841%
410.bwaves -0.287356%
416.gamess 0%
433.milc 1.08696%
434.zeusmp 0%
435.gromacs -0.490196%
436.cactusADM -4.67836%
437.leslie3d -0.502513%
444.namd 0%
447.dealII 0%
450.soplex -1.14504%
453.povray -0.358423%
454.calculix 0%
459.GemsFDTD -1.32159%
465.tonto 0.44843%
470.lbm -10%
481.wrf 0%
482.sphinx3 0.621118%
SPECfp(R)_base2006 -0.865801%
-O3:
164.gzip -0.0487567%
175.vpr 0.204499%
176.gcc -0.389509%
181.mcf -0.721092%
186.crafty 0%
197.parser 0.328638%
252.eon -0.021796%
253.perlbmk 0.107817%
254.gap 0.607806%
255.vortex 0.176278%
256.bzip2 0%
300.twolf 0%
SPECint_base2000 0.018909%
168.wupwise 0.404722%
171.swim -0.289321%
172.mgrid 0.662495%
173.applu 0%
177.mesa 0.0272109%
178.galgel 0.0126374%
179.art 0.155699%
183.equake 3.73021%
187.facerec 0.27248%
188.ammp 0.064%
189.lucas 0.432331%
191.fma3d 0.658436%
200.sixtrack -0.142653%
301.apsi -0.536572%
SPECfp_base2000 0.385806%
400.perlbench 2.22222%
401.bzip2 6.89655%
403.gcc -0.809717%
429.mcf 0%
445.gobmk 0%
456.hmmer 0%
458.sjeng 0%
462.libquantum -2.86396%
464.h264ref 0%
471.omnetpp 0%
473.astar 0%
483.xalancbmk -0.352113%
SPECint(R)_base2006 0.403226%
410.bwaves -7.14286%
416.gamess 0%
433.milc -1.02564%
434.zeusmp -0.865801%
435.gromacs 0%
436.cactusADM 0%
437.leslie3d -1.66667%
444.namd 0%
447.dealII -0.595238%
450.soplex -0.754717%
453.povray 0.373134%
454.calculix 0.480769%
459.GemsFDTD 0.444444%
465.tonto 0.44843%
470.lbm -0.438596%
481.wrf 0%
482.sphinx3 -1.08696%
SPECfp(R)_base2006 -0.78125%
Results look OK. Can't tell why big changes in 410.bwaves,
400.perlbench and 401.bzip2.
--
H.J.