sse3 support
-
- Newbie
- Posts: 4
- Joined: Thu Mar 21, 2019 11:37 am
sse3 support
Hi,
Are there any known issues in wine with SSE3 instructions?
Replacing
pshufd $68,%xmm0,%xmm9
by
movddup %xmm0,%xmm9
seems to cause program to crash:
002b:err:seh:setup_exception stack overflow 2256 bytes in thread 002b eip 000000007bc9e279 esp 0000000000140d40 stack 0x140000-0x141000-0x340000
Running wine-4.0 under Fedora 29 on Intel® Core™2 Duo CPU E8600 @ 3.33GHz × 2 processor.
Thank you,
David
Are there any known issues in wine with SSE3 instructions?
Replacing
pshufd $68,%xmm0,%xmm9
by
movddup %xmm0,%xmm9
seems to cause program to crash:
002b:err:seh:setup_exception stack overflow 2256 bytes in thread 002b eip 000000007bc9e279 esp 0000000000140d40 stack 0x140000-0x141000-0x340000
Running wine-4.0 under Fedora 29 on Intel® Core™2 Duo CPU E8600 @ 3.33GHz × 2 processor.
Thank you,
David
- DarkShadow44
- Level 8
- Posts: 1207
- Joined: Tue Nov 22, 2016 5:39 pm
Re: sse3 support
I doubt this is an issue with SSE3. Why are you replacing instructions like that?
-
- Newbie
- Posts: 4
- Joined: Thu Mar 21, 2019 11:37 am
Re: sse3 support
Instruction replacement is an attempt to perform optimizations.
- DarkShadow44
- Level 8
- Posts: 1207
- Joined: Tue Nov 22, 2016 5:39 pm
Re: sse3 support
And how do you do that? Compiler options?
-
- Newbie
- Posts: 4
- Joined: Thu Mar 21, 2019 11:37 am
Re: sse3 support
Optimizations performed by the code optimizer/auto parallelizer I wrote; see www.dalsoft.com.
The code optimizer was developed on Linux and I am porting it to Windows. I am doing the port on Linux ( that is Wine is used ).
Now that it appears that Wine is not a problem, I will need to check the other elements involved:
1. development tools ( mingW )
2. optimizer itself ( although problem doesn't seem to appear in the Linux version ).
Being on the subject, how well the parallel code is handled by Wine? What is the performance penalty compared to ( the run ) on Windows?
The code optimizer was developed on Linux and I am porting it to Windows. I am doing the port on Linux ( that is Wine is used ).
Now that it appears that Wine is not a problem, I will need to check the other elements involved:
1. development tools ( mingW )
2. optimizer itself ( although problem doesn't seem to appear in the Linux version ).
Being on the subject, how well the parallel code is handled by Wine? What is the performance penalty compared to ( the run ) on Windows?
- DarkShadow44
- Level 8
- Posts: 1207
- Joined: Tue Nov 22, 2016 5:39 pm
Re: sse3 support
Depends. Would be different from application to application, why don't you just test?david.livshin wrote:Being on the subject, how well the parallel code is handled by Wine? What is the performance penalty compared to ( the run ) on Windows?
Does an optimized program work on windows?
-
- Newbie
- Posts: 4
- Joined: Thu Mar 21, 2019 11:37 am
Re: sse3 support
Test, how? I don't have an access to 64-bit Windows machine, that is why am using Wine.DarkShadow44 wrote: Depends. Would be different from application to application, why don't you just test?
Optimizer works under Wine meaning, I hope, it works on Windows. Optimizer generates x86 assembly file which may be used to create an executable for Windows. See www.dalsoft.com/documentation_manual.html for more.DarkShadow44 wrote:Does an optimized program work on windows?
- DarkShadow44
- Level 8
- Posts: 1207
- Joined: Tue Nov 22, 2016 5:39 pm
Re: sse3 support
Well, it doesn't seem to work perfectly under wine, that's why I wondered if it works on windows. If it doesn't work under wine, it might also not work under windows. To make sure, you'd need to test on windows.
Re: sse3 support
Hello DarkShadow, i am using a haswell cpu, so compile wine with -march=native will cause issue? What about -O2 option?DarkShadow44 wrote:I doubt this is an issue with SSE3. Why are you replacing instructions like that?
Safe and suggested optimize options?
- DarkShadow44
- Level 8
- Posts: 1207
- Joined: Tue Nov 22, 2016 5:39 pm
Re: sse3 support
Just compiling wine with different options does not cause issues. In the rare cases that it does, it's a bug that will be fixed.tim110011 wrote:Hello DarkShadow, i am using a haswell cpu, so compile wine with -march=native will cause issue? What about -O2 option?DarkShadow44 wrote:I doubt this is an issue with SSE3. Why are you replacing instructions like that?
Safe and suggested optimize options?