X264 VFW SSE3 Pentium 4

[cakuhnen]

Dear Friends

I have make this only for test if yours have a Pentium 3.0 or above
you can try this build, but sorry only VFW not CLI, i'm working in CLI
to add more fast compression assembler CPU optimizations.
To install, first uninstall the current yours have installed then
extract the contents of rar arquive to a directory and right click in the file x264vfw.inf and select install, this will install the codec in your windows system folder, if XP ask you for not certification (WHQL) simple ignore this and select Yes.
I have added advanced optimizations and the compiler i prefer Intel C++ Compiler, but if yours don't have a computer that support SSE3 newer install this release, wait... i'm adding multiple processors optimizations and i will release in next days...
You need any Zip program like Winzip and the 7Zip to extract the arquive.

Thank's

cakuhnen

[Sirber]

Any link? :D

Will you add these code to the SVN once it's "stable" ?

Great work!!! :cool:

[Koti]

3.2 p4-e waiting patiently to be abused :D :D

Thx I look forward to trying :)

[Sirber]

Which source rev?

[Manao]

Sorry to be blunt, but where is the source code ? And why the zip / Zz thinggy ?

Oh, and since i don't have a PIVe, out of curiosity, what speed gain is to be expected ?

[lexor]

so yeah, you didn't say if you
1) just enabled an option in the compiler for SSE3 support... or
2) you actually wrote some asm code by hand that utilizes SSE3 instructions.

If (1) then it's useless and won't give jack to anyone (by which I mean <1% difference), if (2) then huge kudos to you for great work, but you really should have stated that, because there is a lot of wanna-be's all over internet doing (1) for different open source projects.

[Sirber]

DLL is 120KB bigger. Should have new code :)

[edit]

In fact, maybe not. Could be staticly linked things...

[BBugsBunny]

Did some tests with a small sample clip. (Dual Xeon 3.6 GHz, w2k, VirtualDub 1.6.5) Clip: 480x640 uncompressed avi
Used X264VFW_rev218_mmx.exe and the SSE3 version from this thread. Both say core 23. But I'm not sure if both are from the same build.
I did a 2pass encode with the default settings and fast first pass disabled.

1. pass:
MMX: 9,8 fps
SSE3: 10,3 fps

2. pass
MMX: 16,9 fps
SSE3: 18,4 fps

The resulting avi files I got are not identical. The SSE3 avi is 8 Bytes smaller.

Although the speed increase seems to be small, when encoding about 2 hours of video it should make a difference of about 15 Minutes in the first pass and that is quite a bit.

So hopefully the development of the SSE3 version will continue!

[Sharktooth]

BBugsBunny can you please compare the speed with this build?

http://www.webalice.it/f.corriga/x264/X264VFW_rev220_sse3.exe - VFW
http://www.webalice.it/f.corriga/x264/X264CLI_rev220_sse3.rar - CLI


EDIT: Updated links

[JoeBG]

@ sharktooth

Are they also usable for SSE2 users or exclusivly for SSE3?

[Sirber]

They've been compiled for P4 SSE3, might crash on your non-SSE3 machine.

[Sharktooth]

Originally posted by JoeBG
@ sharktooth

Are they also usable for SSE2 users or exclusivly for SSE3?
It may work... or may crash. It depends on gcc and how it choosen to optimize the code.

[BBugsBunny]

@ sharktooth:

I've now tested this version:
http://www.aziendeassociate.com/x264/X264VFW_rev220_sse3.exe
and it seems to give me the same results as this version: X264VFW_rev218_mmx.exe
1. pass: 9,8 fps
2. pass: 16,9 fps
The other SSE3 version seems to have some more optimizations or is it from another build?

[Sirber]

ICL8 is generaly faster than GCC. Since x264 is only optimized for MMX, ICL8 compile for p2 would give the same speed boost IMO.