(02-13-2011, 11:37 AM)Lectrode Wrote: [ -> ]I will try to have the latest revision up as soon as I can, may take longer since r7129 drops VS2008 support, and ICC 11.1.065 can't integrate with VS2010. So far, I have not been able to get ICC XE to compile correctly (All success, but doesn't run). Then again, I was having the same issue with ICC 11.1.065.
Can you compile r7128 wile you are trying to figure it out?
(02-13-2011, 11:53 AM)ryancollins Wrote: [ -> ]Can you compile r7128 wile you are trying to figure it out?
Unfortunately, the problem did not originate with r7129. I've never gotten ICC XE to work. I've only successfully compiled with 11.1.065. I'm guessing that there are multiple things keeping ICC XE from compiling correctly, whereas there may have been only 1 thing with ICC 11.1.065.
Basically, even if I get revisions before r7129 to compile with ICC 11.1.065, I'll probably still have a fix or few to find to make ICC XE work.
since the plug-ins have been merged into dolphin can they still be sse optimized ?
I've have always noticed speed increase with ICC compiled builds vs standard VS2008 and that is why I only use Lectrodes builds.
If you ever do finally do get ICC working properly with VS2010, make sure you see if you can compile an AVX build.
People say it wont make any difference, and maybe it wont, but the linpack tests Ive been running where i get 120 gflops(with AVX) vs 65 gflops(no avx instuctions) say that Dolphin may actually see a nice speed increase.
(02-13-2011, 12:11 PM)mfuhreck Wrote: [ -> ]I've have always noticed speed increase with ICC compiled builds vs standard VS2008 and that is why I only use Lectrodes builds.
If you ever do finally do get ICC working properly with VS2010, make sure you see if you can compile an AVX build.
People say it wont make any difference, and maybe it wont, but the linpack tests Ive been running where i get 120 gflops(with AVX) vs 65 gflops(no avx instuctions) say that Dolphin may actually see a nice speed increase.
I will keep that in mind. Thanks for the suggestion
Until dolphin start changing some of the source code from an sse type format to an avx type format you won't see any difference. I can promise you that now. I also strongly believe that the speedup offered by ICC is completely based on the more severe general runtime optimizations not sse code. I could be wrong. Lectrode do you use ICC for your sse2 builds or just visual studios?
(02-13-2011, 12:36 PM)NaturalViolence Wrote: [ -> ]Until dolphin start changing some of the source code from an sse type format to an avx type format you won't see any difference. I can promise you that now. I also strongly believe that the speedup offered by ICC is completely based on the more severe general runtime optimizations not sse code. I could be wrong. Lectrode do you use ICC for your sse2 builds or just visual studios?
Just visual studio for the Regular builds. They're probably not noticeably different from mamario's, but they're compiled with some of Visual Studio's optimizations in addition to the default settings.
I would be interested in testing a build compiled with ICC using the sse2 flag vs. sse3/sse4.1. I'm willing to bet their would be no difference in performance so long as it's compiled with ICC instead of vs.
i found that any kind of SSE 4.2 builds tend to crash on me a moment or another, where the normal build is ok. My Core i5 support SSE 4.2 though.
hello,
i have error with openal32.dll, windows xp 32 bits is good for your dolphin ?
thk