flac dev - Mar 2014 - About "attempt to fix differences between x86 FPU and SSE calculations"

More specifically, about this patch:
http://git.xiph.org/?p=flac.git;a=commitdiff;h=70b078cfd5f9d4b0692c33f018cac3c652b14f90

I downloaded the latest code from git (flac-70b078c), disabled
all SSE optimizations in the code and compiled it (GCC 4.8.2).
This patch doesn't change FLAC output.

Either gcc is too smart and optimizes this new code back to the old,
or this fix is MSVS-specific. Or both.

Check with -mfpmath=387 to be sure that x87 FPU code is used and not some
SSE optim made by GCC


On Fri, Mar 21, 2014 at 6:28 PM, lvqcl <lvqcl.mail at gmail.com> wrote:
> More specifically, about this patch:
>
http://git.xiph.org/?p=flac.git;a=commitdiff;h=70b078cfd5f9d4b0692c33f018cac3c652b14f90
>
> I downloaded the latest code from git (flac-70b078c), disabled
> all SSE optimizations in the code and compiled it (GCC 4.8.2).
> This patch doesn't change FLAC output.
>
> Either gcc is too smart and optimizes this new code back to the old,
> or this fix is MSVS-specific. Or both.
> _______________________________________________
> flac-dev mailing list
> flac-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/flac-dev
>


-- 
Olivier TRISTAN
uvi.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.xiph.org/pipermail/flac-dev/attachments/20140321/99085e96/attachment.htm

Olivier Tristan <o.tristan at uvi.net> ?????(?) ? ????? ?????? Fri, 21 Mar
2014 22:41:00 +0400:
> Check with -mfpmath=387 to be sure that x87 FPU code is used and not some
> SSE optim made by GCC
I added "XIPH_ADD_CFLAGS([-mfpmath=387])" into configure.ac
Still the result is different from SSE version.

---------------

MSVS adds two instructions to the generated code after the patch:

	fld	DWORD PTR [eax]
	inc	ecx
	fmul	ST(0), ST(1)
	add	eax, 4
	fstp	DWORD PTR tv2337[esp+20]    <- this: (copy from FP stack to tmp)
	fld	DWORD PTR tv2337[esp+20]    <- and this (copy from tmp to FP stack)
	fadd	DWORD PTR [ebx+ecx*4-4]
	fstp	DWORD PTR [ebx+ecx*4-4]


However GCC doesn't do this:

	lea	ecx, [eax+2]
	fld	DWORD PTR [edx+ecx*4]
	fmul	st, st(1)
	fadd	DWORD PTR [ebx+ecx*4]
	fstp	DWORD PTR [ebx+ecx*4]

Also MSVS doesn't add these instructions if Floating Point Model is set to
Fast (/fp:fast).