On Mon, Jun 25, 2012 at 4:00 PM, Loren Merritt <[email protected]> wrote: > On Mon, 25 Jun 2012, Jason Garrett-Glaser wrote: > >>> + packssdw m0, m3 ; m0 = 0, 6, 12, 18, 3, 9, 15, 21 >>> + packssdw m1, m4 ; m1 = 1, 7, 13, 19, 4, 10, 16, 22 >>> + packssdw m2, m5 ; m2 = 2, 8, 14, 20, 5, 11, 17, 23 >>> + ; unpack words: >>> + movhlps m3, m0 ; m3 = 3, 9, 15, 21, x, x, x, x >>> + punpcklwd m0, m1 ; m0 = 0, 1, 6, 7, 12, 13, 18, 19 >>> + punpckhwd m1, m2 ; m1 = 4, 5, 10, 11, 16, 17, 22, 23 >>> + punpcklwd m2, m3 ; m2 = 2, 3, 8, 9, 14, 15, 20, 21 >>> + ; unpack dwords: >>> + punpckhdq m3, m0, m2 ; m3 = 12, 13, 14, 15, 18, 19, 20, 21 > > If you swap the inputs here > >>> + punpckldq m2, m1 ; m2 = 2, 3, 4, 5, 8, 9, 10, 11 >>> + ; shuffle qwords: >>> + movhlps m4, m2 ; m4 = 8, 9, 10, 11, x, x, x, x >>> + movlhps m4, m3 ; m4 = 8, 9, 10, 11, 12, 13, 14, 15 >>> (final) > > then you can get away with one less op here > >>> + movlhps m2, m0 ; m2 = 2, 3, 4, 5, 0, 1, 6, 7 >>> + movhlps m3, m1 ; m3 = 16, 17, 22, 23, 18, 19, 20, 21 > > at the cost of changing this to shufps or pblendw or vpperm. > Which should be a good tradeoff on nehalem+ and k10+, but bad on > conroe/penryn/k8. > >>> + ; shuffle dwords: >>> + pshufd m2, m2, q3102 ; m2 = 0, 1, 2, 3, 4, 5, 6, 7 >>> (final) >>> + pshufd m3, m3, q1320 ; m3 = 16, 17, 18, 19, 20, 21, 22, 23 >>> (final) >> >> Could some shuffps be used here instead? > > No, shufps has the constraint that the bottom half must come from one > source reg and the top half from the other. vpperm would work.
I meant in place of some of the previous movhlps/movlhps stuff as well, not merely the pshufd. Jason _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
