On 14/03/14 7:56 AM, Christophe Gisquet wrote: > Hi, > > 2014-03-04 3:25 GMT+01:00 James Almer <[email protected]>: >> -INIT_XMM sse2 >> +%macro SETZERO 1 >> +%if cpuflag(sse2) >> + pxor %1, %1 >> +%else >> + xorps %1, %1, %1 >> +%endif >> +%endmacro >> + >> +%macro SHUF 2 >> +%if cpuflag(sse2) >> + pshufd %1, %2, q0123 >> +%else >> + mova %1, %2 >> + shufps %1, %1, q0123 >> +%endif >> +%endmacro > > We already discussed this, and indeed it is worth having SSE2 > (integer) instructions instead of pure (float) SSE ones for the SSE2 > version as they are actually faster. OK from me then for the asm. > > Not sure if the C part still applies cleanly, but this should be minor.
It doesn't. I'll rebase and send the patchset again with some other changes later. _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
