On Wed, 21 Sep 2005 13:57:57 +1000 Daniel Kasak <[EMAIL PROTECTED]>
babbled:
> Greetings.
>
> Having nothing better to do, I decided to test various versions of gcc,
> using evas as the benchmark.
>
> My system:
>
> 1Ghz G4 Powerbook
> 1GB RAM
> Radeon R250 ( FireGL 9000 )
>
> Gentoo Linux ( PPC )
> 2.6.13-gentoo kernel
> CFLAGS: "-O2 -mtune=7450 -mcpu=7450 -maltivec -mabi-altivec -pipe"
> NPTL
> xorg-x11-6.8.99.15
>
> The benchmark:
>
> Basically I ran evas_gl_x11_test 4 times, followed by running
> evas_software_x11_test 4 times.
> I restarted Enlightenment each time I began a new benchmark.
> Enlightenment had a static background and 3 gadgets ( battery, cpufreq,
> clock ). When I recompiled evas, I also recompiled eet, edb and imlib2
> first.
>
> The results ( figures are average fps / evas bench scores ):
>
> *Test Run
> * *gcc-3.4.4
> * *gcc-4.0.2-pre20050917
> * *gcc-4.1.0-beta20050902
> *
> GL 1
> 657.747 / 10.962
> 657.634 / 10.961
> 677.965 / 11.299
> GL 2
> 659.212 / 10.987
> 658.078 / 10.968
> 682.894 / 11.382
> GL 3
> 646.534 / 10.776
> 658.933 / 10.982
> 647.282 / 10.788
> GL 4
> 645.827 / 10.764
> 659.841 / 10.997
> 681.061 / 11.351
> Software 1
> 134.229 / 2.237
> 138.874 / 2.315
> 137.560 / 2.293
> Software 2
> 134.280 / 2.238
> 138.891 / 2.315
> 137.183 / 2.286
> Software 3
> 133.731 / 2.229
> 138.056 / 2.301
> 137.940 / 2.299
> Software 4
> 134.237 / 2.237
> 138.877 / 2.315
> 138.366 / 2.306
>
>
> So there.
>
> Feel free to post your own benchmarks :)
> If anyone's wondering - yes Enlightenment is just as stable when
> compiled with each version of gcc ( in the past I've compiled *almost*
> all the EFL with gcc-4.1.0 and run E without anything abnormal happening ).
>
> I'm interested to see of the x86 version of gcc is already showing
> improvements over 3.x. I know Apple put a lot of effort into gcc-4.x, so
> maybe the PPC version is ahead a little? I suppose I can always to the
> same on my work PC if no-one else will :)
if you are on x86 you want to stick to gcc 3.x - 4.x has shown a good 10% drop
in speed consistently in benchmarks comapred to 3.3 and 3.4 (same optimisation
options - highest level for my cpu - a p4): -O2 -march=pentium4
-fomit-frame-pointer that will give best results for p4's i have found but on
4.x its 10% slower than 3.x - with -O0 (no optimisation at all) gcc4's code
output is HALF the speed of gcc3.3 and 3.4.
on x86 - beyond some new c feature support, gcc 4.x is a downgrade and if you
care about performance, is not a good idea. i cannot vouch for its correctness
when generating code either, but i'd be inclined to trust the more stable 3.x
series for code correctness - but thats just a hunch/instinct.
you should see no real changes in the gl engine based on compiler as almost all
the work is done in the drivers and hardware. there is not much point
benchmarking it unless u are trying to benchmark gfx cards and their drivers.
you may find you benchmarks are quite low on ppc too:
p4 3.4ghz, nvidia 6600gt agp8x:
gcc version 4.0.2 20050821 (prerelease) (Debian 4.0.1-6)
GL:
# EVAS BENCH: 70.812
Software X11:
# EVAS BENCH: 5.642
--
------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler) [EMAIL PROTECTED]
裸好多 [EMAIL PROTECTED]
Tokyo, Japan (東京 日本)
-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
enlightenment-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/enlightenment-users
