widen the loads and stores to 128-bit.
this was added, but not enabled in:
493a857 Add some sse2 code for intra prediction.
Change-Id: I277d7db608a7db7d75cc0bde86f48fa66ad487e4
specialize qw/vp9_dc_left_predictor_16x16/;
add_proto qw/void vp9_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-specialize qw/vp9_dc_128_predictor_16x16/;
+specialize qw/vp9_dc_128_predictor_16x16/, "$sse2_x86inc";
add_proto qw/void vp9_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
specialize qw/vp9_d207_predictor_32x32/, "$ssse3_x86inc";
REP_RET
-INIT_MMX sse
+INIT_XMM sse2
cglobal dc_128_predictor_16x16, 4, 5, 3, dst, stride, above, left, goffset
GET_GOT goffsetq
DEFINE_ARGS dst, stride, stride3, lines4
lea stride3q, [strideq*3]
mov lines4d, 4
- movq m0, [GLOBAL(dc_128)]
+ mova m0, [GLOBAL(dc_128)]
.loop:
mova [dstq ], m0
mova [dstq+strideq ], m0