This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: RFA: patch fixing PR37948
- From: Vladimir Makarov <vmakarov at redhat dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: gcc-patches <gcc-patches at gcc dot gnu dot org>, Jeffrey Law <law at redhat dot com>, Kenneth Zadeck <Kenneth dot Zadeck at naturalbridge dot com>
- Date: Tue, 11 Nov 2008 16:45:27 -0500
- Subject: Re: RFA: patch fixing PR37948
- References: <49185F85.4050602@redhat.com> <6dc9ffc80811110723m582dc1d5ya7bca47bf3afb852@mail.gmail.com>
H.J. Lu wrote:
On Mon, Nov 10, 2008 at 8:21 AM, Vladimir Makarov <vmakarov@redhat.com> wrote:
The following patch solves PR37948. The analysis of the problem can be
found
I applied this patch and patch for PR37397:
http://gcc.gnu.org/ml/gcc-patches/2008-11/msg00373.html
on ira-merge branch. I got slower SPEC CPU 2K integer at -O2 on Intel64:
-O2/-O2 -fno-ira
164.gzip -13.6086%
175.vpr 1.98564%
176.gcc -2.571%
181.mcf -2.03085%
186.crafty -0.0267738%
197.parser 0.42261%
252.eon 1.66472%
253.perlbmk -0.419036%
254.gap 1.52114%
255.vortex -0.382458%
256.bzip2 0.620347%
300.twolf 1.4111%
SPECint_base2000 -1.04107%
Sorry, I can not confirm this. Huge degradation on gzip is especially
suspicious for me. Here are my benchmarks results on 2.66Ghz Core2 in
64-bit mode. For the first table, base is -O2 -fno-ira, peak is -O2 on
trunk with patch for PR37397 (Chaitin trick). For the second table,
peak is -O2 on trunk with the both patches.
Estimated Estimated
Base Base Base Peak Peak Peak
Benchmarks Ref Time Run Time Ratio Ref Time Run Time Ratio
------------ -------- -------- -------- -------- -------- --------
164.gzip 1400 101 1381* 1400 101 1385*
164.gzip 1400 101 1381 1400 101 1384
164.gzip 1400 102 1375 1400 101 1385
175.vpr 1400 86.1 1626 1400 84.7 1653*
175.vpr 1400 85.9 1630* 1400 84.6 1654
175.vpr 1400 85.5 1638 1400 84.8 1650
176.gcc 1100 63.4 1734 1100 60.9 1807
176.gcc 1100 63.6 1730 1100 60.8 1809*
176.gcc 1100 63.5 1731* 1100 60.8 1810
181.mcf 1800 107 1681 1800 107 1681
181.mcf 1800 107 1678* 1800 108 1661
181.mcf 1800 108 1672 1800 107 1675*
186.crafty 1000 38.7 2585 1000 38.3 2609*
186.crafty 1000 38.7 2585* 1000 38.4 2607
186.crafty 1000 39.0 2566 1000 38.3 2610
197.parser 1800 147 1227 1800 147 1228*
197.parser 1800 147 1221 1800 146 1232
197.parser 1800 147 1224* 1800 147 1228
252.eon 1300 50.7 2565 1300 53.4 2436*
252.eon 1300 50.5 2575 1300 53.4 2436
252.eon 1300 50.6 2570* 1300 53.6 2426
253.perlbmk 1800 72.3 2490 1800 68.6 2625*
253.perlbmk 1800 71.9 2505 1800 69.2 2602
253.perlbmk 1800 72.1 2497* 1800 68.5 2627
254.gap 1100 54.6 2015* 1100 54.1 2033
254.gap 1100 54.6 2015 1100 53.7 2050
254.gap 1100 55.3 1991 1100 53.7 2049*
255.vortex 1900 84.2 2257 1900 82.7 2296
255.vortex 1900 84.0 2262* 1900 82.8 2295*
255.vortex 1900 83.8 2266 1900 83.1 2285
256.bzip2 1500 80.5 1864* 1500 80.5 1864*
256.bzip2 1500 80.4 1865 1500 80.3 1867
256.bzip2 1500 80.6 1861 1500 80.7 1858
300.twolf 3000 116 2591 3000 116 2588*
300.twolf 3000 116 2585 3000 117 2568
300.twolf 3000 116 2588* 3000 116 2588
========================================================================
164.gzip 1400 101 1381* 1400 101 1385*
175.vpr 1400 85.9 1630* 1400 84.7 1653*
176.gcc 1100 63.5 1731* 1100 60.8 1809*
181.mcf 1800 107 1678* 1800 107 1675*
186.crafty 1000 38.7 2585* 1000 38.3 2609*
197.parser 1800 147 1224* 1800 147 1228*
252.eon 1300 50.6 2570* 1300 53.4 2436*
253.perlbmk 1800 72.1 2497* 1800 68.6 2625*
254.gap 1100 54.6 2015* 1100 53.7 2049*
255.vortex 1900 84.0 2262* 1900 82.8 2295*
256.bzip2 1500 80.5 1864* 1500 80.5 1864*
300.twolf 3000 116 2588* 3000 116 2588*
Est. SPECint_base2000 1944
Est. SPECint2000 1961
Estimated Estimated
Base Base Base Peak Peak Peak
Benchmarks Ref Time Run Time Ratio Ref Time Run Time Ratio
------------ -------- -------- -------- -------- -------- --------
164.gzip 1400 102 1372
164.gzip 1400 101 1386
164.gzip 1400 101 1382*
175.vpr 1400 84.9 1650
175.vpr 1400 84.9 1648*
175.vpr 1400 85.0 1648
176.gcc 1100 60.5 1819*
176.gcc 1100 60.5 1818
176.gcc 1100 60.3 1823
181.mcf 1800 108 1673
181.mcf 1800 108 1666
181.mcf 1800 108 1673*
186.crafty 1000 38.5 2596
186.crafty 1000 38.3 2610
186.crafty 1000 38.5 2597*
197.parser 1800 146 1234
197.parser 1800 146 1232*
197.parser 1800 147 1228
252.eon 1300 50.5 2572
252.eon 1300 50.4 2579
252.eon 1300 50.4 2577*
253.perlbmk 1800 70.4 2555
253.perlbmk 1800 69.0 2609
253.perlbmk 1800 69.3 2598*
254.gap 1100 53.7 2047*
254.gap 1100 53.7 2048
254.gap 1100 53.8 2046
255.vortex 1900 83.0 2288
255.vortex 1900 83.4 2277
255.vortex 1900 83.1 2285*
256.bzip2 1500 80.5 1863
256.bzip2 1500 80.7 1859*
256.bzip2 1500 81.1 1850
300.twolf 3000 117 2564
300.twolf 3000 117 2570*
300.twolf 3000 116 2580
========================================================================
164.gzip 1400 101 1382*
175.vpr 1400 84.9 1648*
176.gcc 1100 60.5 1819*
181.mcf 1800 108 1673*
186.crafty 1000 38.5 2597*
197.parser 1800 146 1232*
252.eon 1300 50.4 2577*
253.perlbmk 1800 69.3 2598*
254.gap 1100 53.7 2047*
255.vortex 1900 83.1 2285*
256.bzip2 1500 80.7 1859*
300.twolf 3000 117 2570*
Est. SPECint_base2000 0
Est. SPECint2000 1965