/favor isn't in the ICC docs so I doubt its doing anything other than silently ignoring it (all 3 builds are probably the same size). And a FYI to anyone using it.
Quote:Note that the recommendation for AMD Phenom and later AMD processors is to use the default of '/favor:blend' and not '/favor:AMD64'. The "AMD64" setting is really intended only for older Athlon, Opteron, and Sempron processors.
(03-30-2012, 12:10 AM)sixor Wrote: [ -> ]thanks lectrode, long time without using dolphin, glad you are still doing your work
working fine in my phenom2 965be, placebo effect or not i like your builds
your welcome
I sure hope it's not the placebo effect. If it is, then I and a good number of other
people could possibly be going insane
(03-30-2012, 08:42 AM)lamedude Wrote: [ -> ]/favor isn't in the ICC docs so I doubt its doing anything other than silently ignoring it (all 3 builds are probably the same size). And a FYI to anyone using it.
Quote:Note that the recommendation for AMD Phenom and later AMD processors is to use the default of '/favor:blend' and not '/favor:AMD64'. The "AMD64" setting is really intended only for older Athlon, Opteron, and Sempron processors.
Thanks for the info. These were not built with ICC but with VS, so the
option was not ignored.
The main point of this test is to see if the /favor option has a significant
speed increase over vanilla builds
I tested the T3 BETA builds and uploaded the benchmark results to
Mediafire. If you really want me to, I can paste each file into code tags, but I didn't think it would be necessary. There is a text file in each that describes the test conditions, but here are my specs anyway:
i5-2500k@3.4GHz
Radeon HD 6870
2x4Gb 2133MHz
I'll probably do some more testing later, perhaps with an overclock.
Hi~Lectrode
Do you know how to get r7719's source and re-compile it ?
I want to re-compile r7719(Texcache-Rewrite) myself
(03-30-2012, 08:49 AM)Lectrode Wrote: [ -> ] (03-30-2012, 08:42 AM)lamedude Wrote: [ -> ]/favor isn't in the ICC docs so I doubt its doing anything other than silently ignoring it (all 3 builds are probably the same size). And a FYI to anyone using it.
Thanks for the info. These were not built with ICC but with VS, so the
option was not ignored.
The main point of this test is to see if the /favor option has a significant
speed increase over vanilla builds
I did not know this
If I did I never would have asked you to try it
(03-30-2012, 12:39 PM)slmpika Wrote: [ -> ]I tested the T3 BETA builds and uploaded the benchmark results to Mediafire. If you really want me to, I can paste each file into code tags, but I didn't think it would be necessary. There is a text file in each that describes the test conditions, but here are my specs anyway:
i5-2500k@3.4GHz
Radeon HD 6870
2x4Gb 2133MHz
I'll probably do some more testing later, perhaps with an overclock.
Thanks for testing!
I need at least one more test to determine the usefulness of this option.
If someone with an AMD processor could
test these that'd be great!
(03-30-2012, 10:53 PM)ExtremeDude2 Wrote: [ -> ]I did not know this If I did I never would have asked you to try it
Well then do your homework next time
(03-30-2012, 02:02 PM)disraelie Wrote: [ -> ]Hi~Lectrode
Do you know how to get r7719's source and re-compile it ?
I want to re-compile r7719(Texcache-Rewrite) myself
Actually this is something that I've been trying to do as well.
I've tried the following with no success:
Code:
git reset 84fef03d0bd7
git checkout -b r7719 84fef03d0bd7
git checkout 84fef03d0bd7
I'm working on a solution.
thank your reply
seems it is difficult to re-compile the old revision(Texcache-Rewrite)
I'm sorry if you've mentioned this somewhere in this 136 pages thread, but how are you doing these optimizations anyways? I'd like to compile my own builds with these optimizations (except AVX) and a few hacks of my choice and requesting a custom build every few revisions seems like a hassle for you and me.
Lectrode's ICC builds perform about the same as Mamario's builds on AMD K10 CPUs (Athlon II / Phenom II / Llano APU). Zero improvement in those extremely slow CPU-heavy scenes (SMG Spin Dig Galaxy, TLS Lazulis City, MKWii mode select / options screens, etc.) where even a slight boost in FPS matters.
Looks like Mamario's MSVS x64 builds are still the best choice for those with AMD CPUs (same speed, but less issues).
The only thing that makes Lectrode's builds superior is the inclusion of
custom patches
There was a time when (properly patched) ICC builds ran *much faster* than MSVS/GCC builds on
old AMD (32-bit, single core, SSE1) CPUs. This is not the case with modern AMD 64-bit CPUs and latest versions of ICC.
(03-31-2012, 06:31 AM)Lectrode Wrote: [ -> ]I need at least one more test to determine the usefulness of this option.
If someone with an AMD processor could test these that'd be great!
Optimization Test 3: The third test build (TS3) seems a tiny bit (~0.5 fps) faster than T1 and T2 , about the same speed as the 3.0-555 ICC L102 x64 build.