Hi, I compiled the lasest libvorbis
and used the library to encode a WAV file ( 44100Hz Stereo, 16bit LPCM ) to a vorbis file ( quality 0.2 ).
To my surpise, when I tried the SSE/SSE2 vesion,
the encoding speed is slow:
1.20 standard: 22.10x
1.22 standard: 24.55x
1.22 SSE: 23.92x
1.22 SEE2: 19.15x
It is nice to find that version 1.22 is a bit faster than the previous,
but the SEE/SEE2 optimization is puzzling.
I use Core2 Duo processor and also wonder what will happen when the SSE2 one is run on a computer which does not suport SSE2.
Full Version: libvorbis SSE slow the encoding
QUOTE (quadamage @ Jul 3 2009, 21:18) 
Hi, I compiled the lasest libvorbis
and used the library to encode a WAV file ( 44100Hz Stereo, 16bit LPCM ) to a vorbis file ( quality 0.2 ).
To my surpise, when I tried the SSE/SSE2 vesion,
the encoding speed is slow:
1.20 standard: 22.10x
1.22 standard: 24.55x
1.22 SSE: 23.92x
1.22 SEE2: 19.15x
It is nice to find that version 1.22 is a bit faster than the previous,
but the SEE/SEE2 optimization is puzzling.
I use Core2 Duo processor and also wonder what will happen when the SSE2 one is run on a computer which does not suport SSE2.
and used the library to encode a WAV file ( 44100Hz Stereo, 16bit LPCM ) to a vorbis file ( quality 0.2 ).
To my surpise, when I tried the SSE/SSE2 vesion,
the encoding speed is slow:
1.20 standard: 22.10x
1.22 standard: 24.55x
1.22 SSE: 23.92x
1.22 SEE2: 19.15x
It is nice to find that version 1.22 is a bit faster than the previous,
but the SEE/SEE2 optimization is puzzling.
I use Core2 Duo processor and also wonder what will happen when the SSE2 one is run on a computer which does not suport SSE2.
What flags did you use? what version of the compiler?
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.