On Mon, 14 Jan 2013, Daniel Kang wrote: > +INIT_MMX mmx > +; get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size) > +cglobal get_pixels, 3,4 > + movsxdifnidn r2, r2d > + add r0, 128 > + mov r3, -128 > + pxor m7, m7 > +.loop: > + movu m0, [r1] > + movu m2, [r1+r2] > + movu m1, m0 > + movu m3, m2 > + punpcklbw m0, m7 > + punpckhbw m1, m7 > + punpcklbw m2, m7 > + punpckhbw m3, m7 > + movu [r0+r3+ 0], m0 > + movu [r0+r3+ 8], m1 > + movu [r0+r3+16], m2 > + movu [r0+r3+24], m3 > + lea r1, [r1+r2*2] > + add r3, 32 > + js .loop > + REP_RET
Is this really unaligned? --Loren Merritt _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
