2012/1/12 Christophe Gisquet <[email protected]>:
> Unfortunately, the SSSE3 version is not as
> effective as the previous values made it look, and I'm wondering if
> the few cycles shaved are worth the object size.

Checked the above: this occurred because I switched to another
sequence which didn't use the fastpath. The good thing is that it
justifies again a SSSE3 version.

Summary of changes since previous patch:
try to factorize a bit more
use movh/movhps
fix number of gprs used

Attachment: 0001-rv34-x86-SIMD-for-biweight.patch
Description: Binary data

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to