Hi there,
John33 just did some ICL7 compiles for Pentium4 (SSE2 support) over in the ogg section with great success imho, time savings were around 25-35%
I don't know whether mppenc already uses this feature, but if not it would be nice to see the performance in mpc aswell. (encoder, decoder and replaygain tool)

Thx in advance

I think MPC already has the most critical parts optimized in SSE2.

Menno

I tried this, the performance increase was < 1%. And no, MPC is not SSE2 optimized in any way.

'this' will by no means be a sufficient pronoun. Please tell us what 'this' is. I doubt you have access to the mppenc source code. Are you talking about the oggenc performance on your machine?

I've been told that it's partially SSE optimized, but also that it's not currently SSE2 optimized. I would very much like it to be, however . . .

Some critical parts are written in Assembler.

The mppenc sourcecode, at least for the pre-SV8 encoder, has been available for a few weeks now.

http://www.personal.uni-jena.de/~pfk/MPP/s...src/sv8.tar.bz2

The source is available for usage, it's just not Open Source yet... Frank still has to decide on a license and hasn't had time lately AFAIK.

Like Dibrom says, the SV8 source has been available for a while now, and it is a simple task to disassemble mppenc.exe and see it uses no sse2 instructions to do its processing.

I performed a very quick and possibly inaccurate test with mppenc SV7.15/8. P4-optimized vs. non-optimized, and there was no difference in speed. I don't really know why this is. If it's true that parts of it are written in assembler, then there's absolutely nothing the compiler can do. This isn't saying that mppenc can't benefit from SSE2. What it's saying is that currently, Frank Klemm explicitly doesn't allow the compiler to do so. It's up to him if he wishes to implement any SSE2 instructions. I think he should, especially with Hammer right around the corner. Of course, if he could have left the code generic, ICL7 could SSE2 optimize for him, in about 5 seconds.