This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH,RFC] Disallow reordering of x87 insns while scheduling
- From: "Vladimir N. Makarov" <vmakarov at redhat dot com>
- To: Roger Sayle <roger at eyesopen dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 19 Apr 2005 16:29:20 -0400
- Subject: Re: [PATCH,RFC] Disallow reordering of x87 insns while scheduling
- References: <Pine.LNX.4.44.0504182107570.22733-100000@www.eyesopen.com>
Roger Sayle wrote:
Hi Vlad,
On Mon, 18 Apr 2005, Vladimir N. Makarov wrote:
I think the most important platform is Pentium-M right now. So I
ran SPECFP95 on my 1.6Ghz Pentium-M notebook (sorry SPECFp2000 is too
much for the notebook).
A big degradation for wave5 (in size and rate) is surprise for me.
Text segment size (base is the first, peak is the second).
----------------CFP95-----------------
11.395% 111099 123759 146.wave5
Could you describe which compiler flags and options you used in
your SPEC benchmarking, and against which branch you tested?
Could you double check your numbers and see if they're reproduceable?
I suspect that the incomprehensible 10% code growth from simply
disabling scheduling is probably likely to be the result of some
silly mistake.
It looks like vawe5 as the last test was taken from another run (I
belived that vawe5 was rebuilt. I was my flight to San-Francicso, I
had close to discharged battery and run spec without full rebuilding).
Sorry, about this. Here is the result of the new run.
So it looks like it is worth to use your patch at least for Pentium-M.
Vlad
Base Base Base Peak Peak Peak
Benchmarks Ref Time Run Time Ratio Ref Time Run Time Ratio
------------ -------- -------- -------- -------- -------- --------
101.tomcatv 3700 37.0 100 3700 37.1 99.6*
101.tomcatv 3700 36.5 101 3700 37.2 99.4
101.tomcatv 3700 36.7 101 * 3700 36.6 101
102.swim 8600 54.8 157 * 8600 54.3 158
102.swim 8600 54.8 157 8600 54.2 159
102.swim 8600 54.6 157 8600 54.2 159 *
103.su2cor 1400 26.7 52.4* 1400 26.4 53.1
103.su2cor 1400 26.7 52.5 1400 26.5 52.8*
103.su2cor 1400 26.7 52.4 1400 26.8 52.3
104.hydro2d 2400 50.7 47.4* 2400 50.2 47.8
104.hydro2d 2400 50.9 47.2 2400 50.2 47.8*
104.hydro2d 2400 50.5 47.5 2400 50.3 47.7
107.mgrid 2500 66.0 37.9 2500 65.8 38.0
107.mgrid 2500 65.8 38.0 2500 65.9 37.9*
107.mgrid 2500 65.9 38.0* 2500 66.2 37.8
110.applu 2200 55.7 39.5 2200 55.2 39.8*
110.applu 2200 55.4 39.7* 2200 55.3 39.8
110.applu 2200 55.3 39.8 2200 55.2 39.8
125.turb3d 4100 74.4 55.1 4100 71.8 57.1
125.turb3d 4100 74.4 55.1* 4100 71.5 57.4
125.turb3d 4100 74.7 54.9 4100 71.5 57.3*
141.apsi 2100 31.9 65.8 2100 31.6 66.5
141.apsi 2100 32.0 65.7 2100 31.6 66.5*
141.apsi 2100 31.9 65.8* 2100 31.5 66.6
145.fpppp 9600 86.1 111 9600 74.4 129
145.fpppp 9600 86.5 111 9600 74.1 130 *
145.fpppp 9600 86.2 111 * 9600 74.1 130
146.wave5 3000 33.1 90.7 3000 33.8 88.6
146.wave5 3000 33.0 90.9 3000 33.7 89.0*
146.wave5 3000 33.0 90.8* 3000 33.7 89.1
========================================================================
101.tomcatv 3700 36.7 101 3700 37.1 99.6
102.swim 8600 54.8 157 8600 54.2 159
103.su2cor 1400 26.7 52.4 1400 26.5 52.8
104.hydro2d 2400 50.7 47.4 2400 50.2 47.8
107.mgrid 2500 65.9 38.0 2500 65.9 37.9
110.applu 2200 55.4 39.7 2200 55.2 39.8
125.turb3d 4100 74.4 55.1 4100 71.5 57.3
141.apsi 2100 31.9 65.8 2100 31.6 66.5
145.fpppp 9600 86.2 111 9600 74.1 130
146.wave5 3000 33.0 90.8 3000 33.7 89.0
SPECfp_base95 (Geom. Mean) 68.2
SPECfp95 (Geom. Mean) 69.5
----------------CFP95-----------------
-1.857% 120630 118390 141.apsi
-0.808% 5943 5895 101.tomcatv
-0.581% 8265 8217 102.swim
-2.275% 111099 108571 146.wave5
0.000% 13137 13137 107.mgrid
-0.534% 35948 35756 104.hydro2d
-3.317% 62703 60623 110.applu
-7.324% 62697 58105 145.fpppp
-1.330% 38482 37970 103.su2cor
-2.036% 30649 30025 125.turb3d
Average = -1.54328%
real 0m44.888s
user 0m44.307s
sys 0m0.532s
real 0m44.881s
user 0m44.294s
sys 0m0.530s