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.
(07-31-2010, 01:54 AM)erdo309 Wrote: [ -> ]Just a question what is the differenz between SSE3, SSE4.1, SSE4.2, SSSE3
to regular ? sorry im new to dolphin.

In my builds, Regular means it was compiled with some optimizations, but not with SSE. The ones with SSE also have those optimizations but also have the SSE instruction sets.

The higher the number, the better it will run, as long as your CPU supports it.


(07-31-2010, 05:20 AM)Terranigma Freak Wrote: [ -> ]Speaking of SSE, I want to know which version is best for me. CPU-Z says I have something called SSE4A, what is that? I see SSE4.1 and 4.2, but not SSE4A.

SSE4A is a set of instructions specific to AMD processors only, unlike SSE3, SSSE3, SSE4.1, and SSE4.2 which can work on both Intel and AMD processors.


SSE instructions help the CPU to handle large amounts of data quicker.
Do you know why the SSE4.2 build works for me? My processor shouldn't support it : http://valid.canardpc.com/show_oc.php?id=1317840.
When I launch the exe only the 4.1 is showed, so why do we need a 4.1 build if a 4.2 works? Huh
[color=#FF0000]REQUEST:[/color] Can you compile future SVNs with SSE4a? I know SSE4a is AMD only, BUT alot of people are using the new AMD CPUs and SSE 4.1 and 4.2 are INTEL only aswell.

(07-31-2010, 07:05 AM)Lectrode Wrote: [ -> ]SSE4A is a set of instructions specific to AMD processors only, unlike SSE3, SSSE3, SSE4.1, and SSE4.2 which can work on both Intel and AMD processors.


SSE instructions help the CPU to handle large amounts of data quicker.

SSE4.1 and SSE4.2 are ONLY for Intel CPUs...
(07-31-2010, 11:12 AM)bm01 Wrote: [ -> ]Do you know why the SSE4.2 build works for me? My processor shouldn't support it : http://valid.canardpc.com/show_oc.php?id=1317840.
When I launch the exe only the 4.1 is showed, so why do we need a 4.1 build if a 4.2 works? Huh

Did you try both the 4.1 and 4.2? I have no idea but I don't think 4.2 should work on a CPU that only supports 4.1.
Like I said, I have no idea. I don't have access to a computer that can handle either, so I can't do any testing.


(07-31-2010, 12:01 PM)ICAC123 Wrote: [ -> ]SSE4.1 and SSE4.2 are ONLY for Intel CPUs...



Quote:SSE4 is an instruction set used in the AMD K10 (K8L) and Intel Core microarchitecture.

Quote:AMD currently only supports 4 instructions from the SSE4 instruction set, but have also added two new SSE instructions that is named SSE4a.

This is based completely on speculation, but if it runs on the computer, the computer probably supports it. I know when I was compiling specifically for intel, it did not run on AMD. I'm assuming that the opposite is true if I compile SSE4a which is specific to AMD processors. (That is, that SSE4a won't work on Intel processors).
The SSE4.2 builds are probably working fine because no SSE4.2 instructions are being used, once an SSE4.2 instruction tries to execute on hardware that doesn't support that specific instruction the emulator will crash. The only reason SSE4.1 is working and nets a speed increase is because nodchip actually hard coded SSSE3 and SSE4.1 optimizations into the emulator, so I'm sure if you tried to run the SSE4.1 builds on hardware that only supports SSE, SSE2, and SSE3 the emulator will likely crash once these instructions are executed.

You'll find that the same is also true for Pcsx2 and Gsdx.
(07-31-2010, 12:42 PM)Lectrode Wrote: [ -> ]This is based completely on speculation, but if it runs on the computer, the computer probably supports it. I know when I was compiling specifically for intel, it did not run on AMD. I'm assuming that the opposite is true if I compile SSE4a which is specific to AMD processors. (That is, that SSE4a won't work on Intel processors).

Yeah, it would be nice to see an SSE4a for AMD users if it's possible. Targeting specific CPUs might make it run smoother for different people with different brand CPUs.
Lots of people making request lately but not thanks. Thanks, bro keep it up.
@Xtreme2Damax: Thx for clearing that up.

(08-01-2010, 07:03 AM)Terranigma Freak Wrote: [ -> ]Yeah, it would be nice to see an SSE4a for AMD users if it's possible. Targeting specific CPUs might make it run smoother for different people with different brand CPUs.

From what I can tell, building SSE4a just is not worth the time and energy. It's got even less speed-enhancing instructions than SSE4.2, which only has 7 compared to SSE4.1 which has 47. SSE4a has 6 total instructions, 4 of which you already get in the rest of SSE4 (4.1 & 4.2)



(08-01-2010, 07:33 AM)obscured Wrote: [ -> ]Lots of people making request lately but not thanks. Thanks, bro keep it up.

No problem. Happy Gaming Smile
Well it's OK if you dont want to make a SSE4a Version you are already providing INTEL specific versions. Hope someone else from the community can do it: the demand for a SSE4a version is certainly there and I think alot of ppl will download it.

BTW Lectrode, when will you compile the next build, svn 6023 is already out.
compiling 6023 as we speak

EDIT: R6023 is up and ready for download.

Also, I started roughly about 10:30pm (my time) and finished 7 min ago (1:40am). So in case anyone is wondering, it takes about 3 hours to compile all the different types I'm doing now. I REALLY don't want to add any more. Not only that, but SSE4a requires a completely different compiler.


I hope you're right for your sake about someone else doing it, although I did hear somewhere that ICC has the best optimizations out there, even for AMD. I don't know if it's true or not but I'll let someone else figure that out. For now, I'm going to bed.