On 06/04/2011 07:10 PM, Loren Merritt wrote: > On Sat, 4 Jun 2011, Justin Ruggles wrote: >> + ; create mask of all 1's if coef == 0 >> + pxor m1, m1 >> + pcmpeqd m1, m0 >> + ; calculate log2 of each coef >> + cvtdq2ps m0, m0 >> + psrld m0, 23 >> + pand m0, m2 >> + psubd m0, m3 >> + ; set zero coefs to -1 >> + por m0, m1 >> + ; m0 = 23 - m0 >> + mova m1, m0 >> + mova m0, m4 >> + psubd m0, m1 > > pslld m0, 1 > por m0, pd_1 > cvtdq2ps m0, m0 > psrld m0, 23 > mova m1, pd_151 > psubd m1, m0 > > (untested)
Wow, that does work. Athlon64 SSE2: my version - 19737 your version - 16257 Thanks! I'll include this in the next patch set. -Justin _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
