Video encoders and SSE4

[ElAguila]

Which video encoders support the SSE4 instruction set? I currently have tmpgenc plus and the main encoder 1.5 and I know the tmpgenc version I have doesn't but I am not sure about the main encoder.

[LoRd_MuldeR]

I think x264 can already use SSE4, if available. At least it detects the X264_CPU_SSE4 flag ;)

However the question is not, which encoder does "support" SSE4, the question is which encoder actually benefits from SSE4 instructions - that is: better quality and/or higher speed...

[Dark Shikari]

The only one I know that uses it is DivX, and only as a useless exhaustive search mode that's far slower than the regular mode anyways.

I don't know any MPEG-2 encoders that support it; I doubt there's many plans to since I still have yet to find a single useful instruction in SSE4 for video encoding.

[Sagekilla]

DS, is the SSE4 ESA slower than the ESA ("SEA") we have on x264?

[Dark Shikari]

DS, is the SSE4 ESA slower than the ESA ("SEA") we have on x264?Yes, by a very large factor. In tests, I think the SSE4 ESA was only about 30% faster than naive exhaustive at best, while SEA is 7 times faster.

Intel's much larger difference (~2.6x) vs naive exhaustive is probably because their naive exhaustive search didn't compensate for cacheline splits.

[zambelli]

I can give you some info about the VC-1 Encoder SDK (http://www.microsoft.com/downloads/details.aspx?familyid=5031c859-e8da-41bc-98e3-c13add5378b0&displaylang=en&tm)...

The commercial Enterprise version of the VC-1 SDK supports SSE4 only in interlaced encoding mode with about a 15% perf advantage. All other versions and modes use SSE3 and SSSE3 optimizations.

[Sharktooth]

the problem is those optimizations could be potentially less then or equally effective as mmx/sse1/sse2...