[FFmpeg-devel] looking for comparison of intrinsics vs hand written asm

[Reimar Döffinger]

On Fri, Oct 30, 2009 at 04:34:04PM +0100, Attila Kinali wrote:
> I had discussion with some prof about code optimization
> and it came to a point where we disagreed on how big
> the difference between intrinsics and hand written asm
> is. I was quite sure that i've seen somewhere such a comparison
> done by someone with a clue (ie both intrinsics and asm had
> a good quality), but cannot find anything anymore.

Well, you could just look at the ASM code in FFmpeg that does
not use MMX/SSE/... like CABAC decoding, there the speedup with
intrinsics is obviously 0, I think we got about 20%.
For special cases, the same should also apply for code using
intrinsics, though those cases may be rarer.
The problem I see with intrinsics is that I doubt they keep up
when new SSE* version and instructions are added, binutils usually
support them before the CPU is released, and support is even in more
stable distros by the time the CPUs become common, how is the situation
here for intrinsics?