Dolphin, the GameCube and Wii emulator - Forums

Full Version: 3.0-776 Up! Lectrode's Dolphin Builds (SSE 3,S3,4.1,4.2,AVX ICC Optimized)+OpenMP
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
(06-20-2011, 12:42 PM)KnilDark Wrote: [ -> ]a friend with Intel processor worked well.
k thx. just making sure this issue was with AMD only.


Ok, built them.
First one has /arch:SSE3
Second has /arch:SSE2 (normal)

They're in the same archive:
http://www.megaupload.com/?d=2MS36SYC

Please let me know if either work. If they don't, then
I must not have installed something correctly when I
re-installed my compiler recently.
(06-20-2011, 01:00 PM)Lectrode Wrote: [ -> ]Ok, built them.
First one has /arch:SSE3
Second has /arch:SSE2 (normal)

Please let me know if either work. If they don't, then
I must not have installed something correctly when I
re-installed my compiler recently.

Ok, the First bluild doesn't works for me, but the Second build works fine.
I've noticed that are x86 versions.

Note: I have an x64 AMD processor with SSE3 support and Windows 7 x64.
(06-20-2011, 01:56 PM)KnilDark Wrote: [ -> ]Ok, the First bluild doesn't works for me, but the Second build works fine.
I've noticed that are x86 versions.
Note: I have an x64 AMD processor with SSE3 support and Windows 7 x64.
K. Rolled back revisions until I get a r7608 L102 compiled.

bleh

I have no idea why that doesn't work. My logic says it should. >.<
But, since it doesn't, I guess I won't be including that.

Also, even though you have x64 processor, x86 is basically the
"processor norm" meaning that x86 apps can run on both x86
and x64. However, x64 apps cannot run on x86.
(06-20-2011, 03:21 PM)Lectrode Wrote: [ -> ]K. Rolled back revisions until I get a r7608 L102 compiled.
I have no idea why that doesn't work. My logic says it should. >.<
But, since it doesn't, I guess I won't be including that.

Also, even though you have x64 processor, x86 is basically the
"processor norm" meaning that x86 apps can run on both x86
and x64. However, x64 apps cannot run on x86.

I know that Big Grin. Thanks anyway Smile
I hope i've helped. Waiting for the new version. Wink Thx again.

r7608 L102 is up and ready for Download!

It would be just my luck that the only optimization that I can't test
that I feel certain that I could just start including would fail on me. Dodgy

I still don't get why it didn't work. Oh well. On to other possible
optimizations.

...

<rant>
Ugh, I still don't get it. Making the baseline code path include support
for SSE3 (raising the minimum requirement to SSE3) should have no
reason to not run on AMD processors. From what I've read, the /arch
option was specifically made for non-Intel processors, while the
/Qx and /Qax options were made for Intel. Not only that, but by default
it uses the /arch option, but with SSE2 instead. Changing that to SSE3
should not have disabled it from running on AMD processors that
support SSE3
. Undecided

If anyone knows why it would behave like that, I would much appreciate
an explanation. I can't seem to find any reason why it would do that.
</rant>


hmmm.... unless one of the manual SSE3 optimizations conflicted with
the automated ones... or the ICC compiler incorrectly applied a SSE3
instruction... incompatible code... (as the wheels start whirring again).

Meh, tired. will think more about tomorrow. It's past 2am and I have early
day tmrw; I'm goin to bed Tongue
OK Lectrode, BL 103 WORKS FOR ME (AMD) here is how: You must patch the .exe with a ICC pather (like the one someone posted a link to somewhere in this thread).
Edit: OK I found the post:

(01-25-2011, 02:54 AM)lamedude Wrote: [ -> ]You can try using this to remove the Intel checks.
i don´t know if it helps you, but i always use your releases sse3 32bit, in windows732 in my amd 965be, cero problem

i don´t really know if it is using sse3 or not, but it works nice
(06-20-2011, 09:38 PM)ryancollins Wrote: [ -> ]OK Lectrode, BL 103 WORKS FOR ME (AMD) here is how: You must patch the .exe with a ICC pather (like the one someone posted a link to somewhere in this thread).
Edit: OK I found the post:
(01-25-2011, 02:54 AM)lamedude Wrote: [ -> ]You can try using this to remove the Intel checks.
Unfortunately, doing that to a L103 build is exactly the same as doing that to a L102 build.
You're not fixing the baseline code path, your making the
/QaxSSE3,SSSE3,SSE4.1,SSE4.2,and AVX paths runnable on an AMD processor.
Intel signed an agreement to no longer make apps run slower on other processors.
If they keep their word/contract (and even if they don't), then enabling the /Qax
paths may introduce critical errors when run on an AMD processor. If the app tries
to use something specific to an Intel processor, the app will crash.

The problem with the L103 build is that the baseline code path is corrupted
when compiled with SSE3 instead of SSE2. ICC is, yet again, not liking some of
the code. This means that if I'm going to be able to increase the baseline code
support to SSE3, I'm going to have to find the non-compatible source code.


(06-21-2011, 03:03 AM)sixor Wrote: [ -> ]i don´t know if it helps you, but i always use your releases sse3 32bit, in windows732 in my amd 965be, cero problem
i don´t really know if it is using sse3 or not, but it works nice
It is almost certainly not using the SSE3, the highest it is using is SSE2. HOWEVER,
The ICC optimizations still cause dolphin to run very noticeably faster even on AMD
processors. I would go so far as to say that the main cause of the speed increase
is the ICC optimizations. The SSE is probably just the icing on the cake Smile
Even so, the SSE Instruction sets still make a difference in speed and therefore still
a good thing to have.
I have not actually sat down and tested a non-SSE ICC build against its SSE counterpart
one right after the other. I did it once, but it was a very informal/non-data-collecting test.
I may do that now soon, or at least let you guys test them.

Meh, side tracked lol. Thx for comment sixor Smile
May I try out the previous L103 build? I didn't have the opportunity to download the x86 version (due to issues with Megaupload) and I have an Intel processor.

If there's anything that I can do to help, please give me some instructions on what 'tests' to run.

Sorry if I'm late to the scene.
(06-21-2011, 04:00 PM)Decimus10 Wrote: [ -> ]May I try out the previous L103 build? I didn't have the opportunity to download the x86 version (due to issues with Megaupload) and I have an Intel processor.

If there's anything that I can do to help, please give me some instructions on what 'tests' to run.

Sorry if I'm late to the scene.

But the L103 build has only extra optimizations for AMD processors, as I have understood, will not help you nothing compared with the L102 build.