Port pred8x8l_down_right_sse2/ssse3 (H.264 intra prediction) from x264
(authors: Jason, Loren, Holger) to FFmpeg. Patch by Daniel Kang <daniel dot d dot kang at gmail com>, as part of Google's GCI 2010. Originally committed as revision 26145 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
parent
ec44dd5fc2
commit
602a4cb25a
@ -1605,6 +1605,123 @@ cglobal pred8x8l_down_right_mmxext, 4,5
|
||||
por mm0, mm1
|
||||
movq [r0+r3*1], mm0
|
||||
RET
|
||||
|
||||
%macro PRED8x8L_DOWN_RIGHT 1
|
||||
cglobal pred8x8l_down_right_%1, 4,5
|
||||
sub r0, r3
|
||||
lea r4, [r0+r3*2]
|
||||
movq mm0, [r0+r3*1-8]
|
||||
punpckhbw mm0, [r0+r3*0-8]
|
||||
movq mm1, [r4+r3*1-8]
|
||||
punpckhbw mm1, [r0+r3*2-8]
|
||||
mov r4, r0
|
||||
punpckhwd mm1, mm0
|
||||
lea r0, [r0+r3*4]
|
||||
movq mm2, [r0+r3*1-8]
|
||||
punpckhbw mm2, [r0+r3*0-8]
|
||||
lea r0, [r0+r3*2]
|
||||
movq mm3, [r0+r3*1-8]
|
||||
punpckhbw mm3, [r0+r3*0-8]
|
||||
punpckhwd mm3, mm2
|
||||
punpckhdq mm3, mm1
|
||||
lea r0, [r0+r3*2]
|
||||
movq mm0, [r0+r3*0-8]
|
||||
movq mm1, [r4]
|
||||
mov r0, r4
|
||||
movq mm4, mm3
|
||||
movq mm2, mm3
|
||||
PALIGNR mm4, mm0, 7, mm0
|
||||
PALIGNR mm1, mm2, 1, mm2
|
||||
test r1, r1
|
||||
jz .fix_lt_1
|
||||
jmp .do_left
|
||||
.fix_lt_1:
|
||||
movq mm5, mm3
|
||||
pxor mm5, mm4
|
||||
psrlq mm5, 56
|
||||
psllq mm5, 48
|
||||
pxor mm1, mm5
|
||||
jmp .do_left
|
||||
.fix_lt_2:
|
||||
movq mm5, mm3
|
||||
pxor mm5, mm2
|
||||
psllq mm5, 56
|
||||
psrlq mm5, 56
|
||||
pxor mm2, mm5
|
||||
test r2, r2
|
||||
jnz .do_top
|
||||
.fix_tr_1:
|
||||
movq mm5, mm3
|
||||
pxor mm5, mm1
|
||||
psrlq mm5, 56
|
||||
psllq mm5, 56
|
||||
pxor mm1, mm5
|
||||
jmp .do_top
|
||||
.do_left:
|
||||
movq mm0, mm4
|
||||
PRED4x4_LOWPASS mm2, mm1, mm4, mm3, mm5
|
||||
movq mm4, mm0
|
||||
movq mm7, mm2
|
||||
movq2dq xmm3, mm2
|
||||
PRED4x4_LOWPASS mm1, mm3, mm0, mm4, mm5
|
||||
psllq mm1, 56
|
||||
PALIGNR mm7, mm1, 7, mm3
|
||||
movq2dq xmm1, mm7
|
||||
movq mm0, [r0-8]
|
||||
movq mm3, [r0]
|
||||
movq mm1, [r0+8]
|
||||
movq mm2, mm3
|
||||
movq mm4, mm3
|
||||
PALIGNR mm2, mm0, 7, mm0
|
||||
PALIGNR mm1, mm4, 1, mm4
|
||||
test r1, r1
|
||||
jz .fix_lt_2
|
||||
test r2, r2
|
||||
jz .fix_tr_1
|
||||
.do_top:
|
||||
PRED4x4_LOWPASS mm4, mm2, mm1, mm3, mm5
|
||||
movq2dq xmm4, mm4
|
||||
lea r1, [r0+r3*2]
|
||||
movdqa xmm0, xmm3
|
||||
pslldq xmm4, 8
|
||||
por xmm3, xmm4
|
||||
lea r2, [r1+r3*2]
|
||||
pslldq xmm4, 1
|
||||
por xmm1, xmm4
|
||||
psrldq xmm0, 7
|
||||
pslldq xmm0, 15
|
||||
psrldq xmm0, 7
|
||||
por xmm1, xmm0
|
||||
lea r0, [r2+r3*2]
|
||||
movdqa xmm2, xmm3
|
||||
psrldq xmm2, 1
|
||||
INIT_XMM
|
||||
PRED4x4_LOWPASS xmm0, xmm1, xmm2, xmm3, xmm4
|
||||
movdqa xmm1, xmm0
|
||||
psrldq xmm1, 1
|
||||
movq [r0+r3*2], xmm0
|
||||
movq [r0+r3*1], xmm1
|
||||
psrldq xmm0, 2
|
||||
psrldq xmm1, 2
|
||||
movq [r2+r3*2], xmm0
|
||||
movq [r2+r3*1], xmm1
|
||||
psrldq xmm0, 2
|
||||
psrldq xmm1, 2
|
||||
movq [r1+r3*2], xmm0
|
||||
movq [r1+r3*1], xmm1
|
||||
psrldq xmm0, 2
|
||||
psrldq xmm1, 2
|
||||
movq [r4+r3*2], xmm0
|
||||
movq [r4+r3*1], xmm1
|
||||
RET
|
||||
%endmacro
|
||||
|
||||
INIT_MMX
|
||||
%define PALIGNR PALIGNR_MMX
|
||||
PRED8x8L_DOWN_RIGHT sse2
|
||||
INIT_MMX
|
||||
%define PALIGNR PALIGNR_SSSE3
|
||||
PRED8x8L_DOWN_RIGHT ssse3
|
||||
%endif
|
||||
|
||||
;-----------------------------------------------------------------------------
|
||||
|
@ -70,6 +70,8 @@ void ff_pred8x8l_vertical_ssse3 (uint8_t *src, int has_topleft, int has_topri
|
||||
void ff_pred8x8l_down_left_sse2 (uint8_t *src, int has_topleft, int has_topright, int stride);
|
||||
void ff_pred8x8l_down_left_ssse3 (uint8_t *src, int has_topleft, int has_topright, int stride);
|
||||
void ff_pred8x8l_down_right_mmxext (uint8_t *src, int has_topleft, int has_topright, int stride);
|
||||
void ff_pred8x8l_down_right_sse2 (uint8_t *src, int has_topleft, int has_topright, int stride);
|
||||
void ff_pred8x8l_down_right_ssse3 (uint8_t *src, int has_topleft, int has_topright, int stride);
|
||||
void ff_pred4x4_dc_mmxext (uint8_t *src, const uint8_t *topright, int stride);
|
||||
void ff_pred4x4_down_left_mmxext (uint8_t *src, const uint8_t *topright, int stride);
|
||||
void ff_pred4x4_tm_vp8_mmx (uint8_t *src, const uint8_t *topright, int stride);
|
||||
@ -149,6 +151,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
|
||||
h->pred16x16[DC_PRED8x8 ] = ff_pred16x16_dc_sse2;
|
||||
#if CONFIG_GPL
|
||||
h->pred8x8l [DIAG_DOWN_LEFT_PRED] = ff_pred8x8l_down_left_sse2;
|
||||
h->pred8x8l [DIAG_DOWN_RIGHT_PRED] = ff_pred8x8l_down_right_sse2;
|
||||
#endif
|
||||
if (codec_id == CODEC_ID_VP8) {
|
||||
h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_sse2;
|
||||
@ -175,6 +178,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
|
||||
h->pred8x8l [HOR_PRED ] = ff_pred8x8l_horizontal_ssse3;
|
||||
h->pred8x8l [VERT_PRED ] = ff_pred8x8l_vertical_ssse3;
|
||||
h->pred8x8l [DIAG_DOWN_LEFT_PRED] = ff_pred8x8l_down_left_ssse3;
|
||||
h->pred8x8l [DIAG_DOWN_RIGHT_PRED] = ff_pred8x8l_down_right_ssse3;
|
||||
#endif
|
||||
if (codec_id == CODEC_ID_VP8) {
|
||||
h->pred8x8 [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_ssse3;
|
||||
|
Loading…
Reference in New Issue
Block a user