From 015821229f96bf7e677f2a711a58dbea3009f574 Mon Sep 17 00:00:00 2001 From: "Ronald S. Bultje" Date: Tue, 12 Mar 2013 07:28:12 -0700 Subject: [PATCH] vp3: Use full transpose for all IDCTs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This way, the special IDCT permutations are no longer needed. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Also remove the unused type == 0 cases from the plain C version of the idct. Signed-off-by: Martin Storsjö --- libavcodec/arm/vp3dsp_init_arm.c | 1 - libavcodec/bfin/vp3_bfin.c | 1 - libavcodec/ppc/vp3dsp_altivec.c | 1 - libavcodec/vp3.c | 13 ++-- libavcodec/vp3dsp.c | 92 +++++++++-------------- libavcodec/vp3dsp.h | 2 - libavcodec/vp5.c | 2 +- libavcodec/vp56.c | 9 ++- libavcodec/vp56.h | 2 +- libavcodec/vp6.c | 4 +- libavcodec/x86/vp3dsp.asm | 123 ++++++++++++++++++++----------- libavcodec/x86/vp3dsp_init.c | 2 - 12 files changed, 135 insertions(+), 117 deletions(-) diff --git a/libavcodec/arm/vp3dsp_init_arm.c b/libavcodec/arm/vp3dsp_init_arm.c index 5a7950c17c..dfd6078caa 100644 --- a/libavcodec/arm/vp3dsp_init_arm.c +++ b/libavcodec/arm/vp3dsp_init_arm.c @@ -41,6 +41,5 @@ av_cold void ff_vp3dsp_init_arm(VP3DSPContext *c, int flags) c->idct_dc_add = ff_vp3_idct_dc_add_neon; c->v_loop_filter = ff_vp3_v_loop_filter_neon; c->h_loop_filter = ff_vp3_h_loop_filter_neon; - c->idct_perm = FF_TRANSPOSE_IDCT_PERM; } } diff --git a/libavcodec/bfin/vp3_bfin.c b/libavcodec/bfin/vp3_bfin.c index 86c4b23b41..a8cdcb6a45 100644 --- a/libavcodec/bfin/vp3_bfin.c +++ b/libavcodec/bfin/vp3_bfin.c @@ -61,6 +61,5 @@ av_cold void ff_vp3dsp_init_bfin(VP3DSPContext *c, int flags) if (!(flags & CODEC_FLAG_BITEXACT)) { c->idct_add = bfin_vp3_idct_add; c->idct_put = bfin_vp3_idct_put; - c->idct_perm = FF_TRANSPOSE_IDCT_PERM; } } diff --git a/libavcodec/ppc/vp3dsp_altivec.c b/libavcodec/ppc/vp3dsp_altivec.c index e1e8988a53..c512047b82 100644 --- a/libavcodec/ppc/vp3dsp_altivec.c +++ b/libavcodec/ppc/vp3dsp_altivec.c @@ -184,7 +184,6 @@ av_cold void ff_vp3dsp_init_ppc(VP3DSPContext *c, int flags) if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) { c->idct_put = vp3_idct_put_altivec; c->idct_add = vp3_idct_add_altivec; - c->idct_perm = FF_TRANSPOSE_IDCT_PERM; } #endif } diff --git a/libavcodec/vp3.c b/libavcodec/vp3.c index 18b9cb89bf..7c26609b09 100644 --- a/libavcodec/vp3.c +++ b/libavcodec/vp3.c @@ -136,6 +136,7 @@ typedef struct Vp3DecodeContext { ThreadFrame current_frame; int keyframe; uint8_t idct_permutation[64]; + uint8_t idct_scantable[64]; DSPContext dsp; VideoDSPContext vdsp; VP3DSPContext vp3dsp; @@ -173,8 +174,6 @@ typedef struct Vp3DecodeContext { int8_t (*motion_val[2])[2]; - ScanTable scantable; - /* tables */ uint16_t coded_dc_scale_factor[64]; uint32_t coded_ac_scale_factor[64]; @@ -1351,7 +1350,7 @@ static inline int vp3_dequant(Vp3DecodeContext *s, Vp3Fragment *frag, int plane, int inter, int16_t block[64]) { int16_t *dequantizer = s->qmat[frag->qpi][inter][plane]; - uint8_t *perm = s->scantable.permutated; + uint8_t *perm = s->idct_scantable; int i = 0; do { @@ -1700,8 +1699,12 @@ static av_cold int vp3_decode_init(AVCodecContext *avctx) ff_videodsp_init(&s->vdsp, 8); ff_vp3dsp_init(&s->vp3dsp, avctx->flags); - ff_init_scantable_permutation(s->idct_permutation, s->vp3dsp.idct_perm); - ff_init_scantable(s->idct_permutation, &s->scantable, ff_zigzag_direct); + for (i = 0; i < 64; i++) { +#define T(x) (x >> 3) | ((x & 7) << 3) + s->idct_permutation[i] = T(i); + s->idct_scantable[i] = T(ff_zigzag_direct[i]); +#undef T + } /* initialize to an impossible value which will force a recalculation * in the first frame decode */ diff --git a/libavcodec/vp3dsp.c b/libavcodec/vp3dsp.c index d1a7db957d..94de0e5b96 100644 --- a/libavcodec/vp3dsp.c +++ b/libavcodec/vp3dsp.c @@ -54,11 +54,12 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int /* Inverse DCT on the rows now */ for (i = 0; i < 8; i++) { /* Check for non-zero values */ - if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) { - A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]); - B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]); - C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]); - D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]); + if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] | + ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) { + A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]); + B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]); + C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]); + D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]); Ad = M(xC4S4, (A - C)); Bd = M(xC4S4, (B - D)); @@ -66,11 +67,11 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int Cd = A + C; Dd = B + D; - E = M(xC4S4, (ip[0] + ip[4])); - F = M(xC4S4, (ip[0] - ip[4])); + E = M(xC4S4, (ip[0 * 8] + ip[4 * 8])); + F = M(xC4S4, (ip[0 * 8] - ip[4 * 8])); - G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]); - H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]); + G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]); + H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]); Ed = E - G; Gd = E + G; @@ -82,33 +83,33 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int Hd = Bd + H; /* Final sequence of operations over-write original inputs. */ - ip[0] = Gd + Cd ; - ip[7] = Gd - Cd ; + ip[0 * 8] = Gd + Cd ; + ip[7 * 8] = Gd - Cd ; - ip[1] = Add + Hd; - ip[2] = Add - Hd; + ip[1 * 8] = Add + Hd; + ip[2 * 8] = Add - Hd; - ip[3] = Ed + Dd ; - ip[4] = Ed - Dd ; + ip[3 * 8] = Ed + Dd ; + ip[4 * 8] = Ed - Dd ; - ip[5] = Fd + Bdd; - ip[6] = Fd - Bdd; + ip[5 * 8] = Fd + Bdd; + ip[6 * 8] = Fd - Bdd; } - ip += 8; /* next row */ + ip += 1; /* next row */ } ip = input; for ( i = 0; i < 8; i++) { /* Check for non-zero values (bitwise or faster than ||) */ - if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] | - ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) { + if ( ip[1] | ip[2] | ip[3] | + ip[4] | ip[5] | ip[6] | ip[7] ) { - A = M(xC1S7, ip[1*8]) + M(xC7S1, ip[7*8]); - B = M(xC7S1, ip[1*8]) - M(xC1S7, ip[7*8]); - C = M(xC3S5, ip[3*8]) + M(xC5S3, ip[5*8]); - D = M(xC3S5, ip[5*8]) - M(xC5S3, ip[3*8]); + A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]); + B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]); + C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]); + D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]); Ad = M(xC4S4, (A - C)); Bd = M(xC4S4, (B - D)); @@ -116,16 +117,16 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int Cd = A + C; Dd = B + D; - E = M(xC4S4, (ip[0*8] + ip[4*8])) + 8; - F = M(xC4S4, (ip[0*8] - ip[4*8])) + 8; + E = M(xC4S4, (ip[0] + ip[4])) + 8; + F = M(xC4S4, (ip[0] - ip[4])) + 8; if(type==1){ //HACK E += 16*128; F += 16*128; } - G = M(xC2S6, ip[2*8]) + M(xC6S2, ip[6*8]); - H = M(xC6S2, ip[2*8]) - M(xC2S6, ip[6*8]); + G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]); + H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]); Ed = E - G; Gd = E + G; @@ -137,19 +138,7 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int Hd = Bd + H; /* Final sequence of operations over-write original inputs. */ - if(type==0){ - ip[0*8] = (Gd + Cd ) >> 4; - ip[7*8] = (Gd - Cd ) >> 4; - - ip[1*8] = (Add + Hd ) >> 4; - ip[2*8] = (Add - Hd ) >> 4; - - ip[3*8] = (Ed + Dd ) >> 4; - ip[4*8] = (Ed - Dd ) >> 4; - - ip[5*8] = (Fd + Bdd ) >> 4; - ip[6*8] = (Fd - Bdd ) >> 4; - }else if(type==1){ + if (type == 1) { dst[0*stride] = av_clip_uint8((Gd + Cd ) >> 4); dst[7*stride] = av_clip_uint8((Gd - Cd ) >> 4); @@ -176,16 +165,7 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int } } else { - if(type==0){ - ip[0*8] = - ip[1*8] = - ip[2*8] = - ip[3*8] = - ip[4*8] = - ip[5*8] = - ip[6*8] = - ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20); - }else if(type==1){ + if (type == 1) { dst[0*stride]= dst[1*stride]= dst[2*stride]= @@ -193,10 +173,10 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int dst[4*stride]= dst[5*stride]= dst[6*stride]= - dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20)); + dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20)); }else{ - if(ip[0*8]){ - int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20); + if(ip[0]){ + int v= ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20); dst[0*stride] = av_clip_uint8(dst[0*stride] + v); dst[1*stride] = av_clip_uint8(dst[1*stride] + v); dst[2*stride] = av_clip_uint8(dst[2*stride] + v); @@ -209,7 +189,7 @@ static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int } } - ip++; /* next column */ + ip += 8; /* next column */ dst++; } } @@ -307,8 +287,6 @@ av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags) c->v_loop_filter = vp3_v_loop_filter_c; c->h_loop_filter = vp3_h_loop_filter_c; - c->idct_perm = FF_NO_IDCT_PERM; - if (ARCH_ARM) ff_vp3dsp_init_arm(c, flags); if (ARCH_BFIN) diff --git a/libavcodec/vp3dsp.h b/libavcodec/vp3dsp.h index 755271d37c..39c4408c0f 100644 --- a/libavcodec/vp3dsp.h +++ b/libavcodec/vp3dsp.h @@ -43,8 +43,6 @@ typedef struct VP3DSPContext { void (*idct_dc_add)(uint8_t *dest, int line_size, int16_t *block); void (*v_loop_filter)(uint8_t *src, int stride, int *bounding_values); void (*h_loop_filter)(uint8_t *src, int stride, int *bounding_values); - - int idct_perm; } VP3DSPContext; void ff_vp3dsp_init(VP3DSPContext *c, int flags); diff --git a/libavcodec/vp5.c b/libavcodec/vp5.c index 742262be05..1415428ee8 100644 --- a/libavcodec/vp5.c +++ b/libavcodec/vp5.c @@ -173,7 +173,7 @@ static void vp5_parse_coeff(VP56Context *s) { VP56RangeCoder *c = &s->c; VP56Model *model = s->modelp; - uint8_t *permute = s->scantable.permutated; + uint8_t *permute = s->idct_scantable; uint8_t *model1, *model2; int coeff, sign, coeff_idx; int b, i, cg, idx, ctx, ctx_last; diff --git a/libavcodec/vp56.c b/libavcodec/vp56.c index b93f75da7e..1a83f00f47 100644 --- a/libavcodec/vp56.c +++ b/libavcodec/vp56.c @@ -263,7 +263,7 @@ static VP56mb vp56_decode_mv(VP56Context *s, int row, int col) static void vp56_add_predictors_dc(VP56Context *s, VP56Frame ref_frame) { - int idx = s->scantable.permutated[0]; + int idx = s->idct_scantable[0]; int b; for (b=0; b<6; b++) { @@ -661,8 +661,11 @@ av_cold int ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha) ff_videodsp_init(&s->vdsp, 8); ff_vp3dsp_init(&s->vp3dsp, avctx->flags); ff_vp56dsp_init(&s->vp56dsp, avctx->codec->id); - ff_init_scantable_permutation(s->dsp.idct_permutation, s->vp3dsp.idct_perm); - ff_init_scantable(s->dsp.idct_permutation, &s->scantable,ff_zigzag_direct); + for (i = 0; i < 64; i++) { +#define T(x) (x >> 3) | ((x & 7) << 3) + s->idct_scantable[i] = T(ff_zigzag_direct[i]); +#undef T + } for (i = 0; i < FF_ARRAY_ELEMS(s->frames); i++) { s->frames[i] = av_frame_alloc(); diff --git a/libavcodec/vp56.h b/libavcodec/vp56.h index 80ede6a0aa..7fe6cf3eb4 100644 --- a/libavcodec/vp56.h +++ b/libavcodec/vp56.h @@ -100,7 +100,7 @@ struct vp56_context { VideoDSPContext vdsp; VP3DSPContext vp3dsp; VP56DSPContext vp56dsp; - ScanTable scantable; + uint8_t idct_scantable[64]; AVFrame *frames[4]; uint8_t *edge_emu_buffer_alloc; uint8_t *edge_emu_buffer; diff --git a/libavcodec/vp6.c b/libavcodec/vp6.c index 2e25a55f3f..54dc3781ce 100644 --- a/libavcodec/vp6.c +++ b/libavcodec/vp6.c @@ -368,7 +368,7 @@ static unsigned vp6_get_nb_null(VP56Context *s) static void vp6_parse_coeff_huffman(VP56Context *s) { VP56Model *model = s->modelp; - uint8_t *permute = s->scantable.permutated; + uint8_t *permute = s->idct_scantable; VLC *vlc_coeff; int coeff, sign, coeff_idx; int b, cg, idx; @@ -428,7 +428,7 @@ static void vp6_parse_coeff(VP56Context *s) { VP56RangeCoder *c = s->ccp; VP56Model *model = s->modelp; - uint8_t *permute = s->scantable.permutated; + uint8_t *permute = s->idct_scantable; uint8_t *model1, *model2, *model3; int coeff, sign, coeff_idx; int b, i, cg, idx, ctx; diff --git a/libavcodec/x86/vp3dsp.asm b/libavcodec/x86/vp3dsp.asm index 078e9db99a..fc8a047224 100644 --- a/libavcodec/x86/vp3dsp.asm +++ b/libavcodec/x86/vp3dsp.asm @@ -501,22 +501,22 @@ cglobal vp3_h_loop_filter, 3, 4 ; at this point, function has completed dequantization + dezigzag + ; partial transposition; now do the idct itself -%define I(x) [%1+16* x ] -%define J(x) [%1+16*(x-4)+8] +%define I(x) [%1+16*x] +%define J(x) [%1+16*x] RowIDCT Transpose -%define I(x) [%1+16* x +64] -%define J(x) [%1+16*(x-4)+72] +%define I(x) [%1+16*x+8] +%define J(x) [%1+16*x+8] RowIDCT Transpose -%define I(x) [%1+16*x] -%define J(x) [%1+16*x] +%define I(x) [%1+16* x] +%define J(x) [%1+16*(x-4)+8] ColumnIDCT -%define I(x) [%1+16*x+8] -%define J(x) [%1+16*x+8] +%define I(x) [%1+16* x +64] +%define J(x) [%1+16*(x-4)+72] ColumnIDCT %endif ; mmsize == 16/8 %endmacro @@ -534,10 +534,17 @@ cglobal vp3_idct_put, 3, 4, 9 mova m1, [r2+mmsize*2+%%i] mova m2, [r2+mmsize*4+%%i] mova m3, [r2+mmsize*6+%%i] +%if mmsize == 8 + packsswb m0, [r2+mmsize*8+%%i] + packsswb m1, [r2+mmsize*10+%%i] + packsswb m2, [r2+mmsize*12+%%i] + packsswb m3, [r2+mmsize*14+%%i] +%else packsswb m0, [r2+mmsize*1+%%i] packsswb m1, [r2+mmsize*3+%%i] packsswb m2, [r2+mmsize*5+%%i] packsswb m3, [r2+mmsize*7+%%i] +%endif paddb m0, m4 paddb m1, m4 paddb m2, m4 @@ -561,7 +568,7 @@ cglobal vp3_idct_put, 3, 4, 9 movq [r0+r1*2], m3 movhps [r0+r3 ], m3 %endif -%assign %%i %%i+64 +%assign %%i %%i+8 %endrep pxor m0, m0 @@ -575,47 +582,81 @@ cglobal vp3_idct_put, 3, 4, 9 cglobal vp3_idct_add, 3, 4, 9 VP3_IDCT r2 - mov r3, 4 - pxor m4, m4 movsxdifnidn r1, r1d -.loop: + lea r3, [r1*3] + pxor m4, m4 +%if mmsize == 16 +%assign %%i 0 +%rep 2 movq m0, [r0] movq m1, [r0+r1] -%if mmsize == 8 - mova m2, m0 - mova m3, m1 -%endif + movq m2, [r0+r1*2] + movq m3, [r0+r3] punpcklbw m0, m4 punpcklbw m1, m4 -%if mmsize == 8 - punpckhbw m2, m4 - punpckhbw m3, m4 -%endif - paddsw m0, [r2+ 0] - paddsw m1, [r2+16] -%if mmsize == 8 - paddsw m2, [r2+ 8] - paddsw m3, [r2+24] - packuswb m0, m2 - packuswb m1, m3 -%else ; mmsize == 16 + punpcklbw m2, m4 + punpcklbw m3, m4 + paddsw m0, [r2+ 0+%%i] + paddsw m1, [r2+16+%%i] + paddsw m2, [r2+32+%%i] + paddsw m3, [r2+48+%%i] packuswb m0, m1 + packuswb m2, m3 + movq [r0 ], m0 + movhps [r0+r1 ], m0 + movq [r0+r1*2], m2 + movhps [r0+r3 ], m2 +%if %%i == 0 + lea r0, [r0+r1*4] %endif - movq [r0 ], m0 -%if mmsize == 8 - movq [r0+r1], m1 -%else ; mmsize == 16 - movhps [r0+r1], m0 +%assign %%i %%i+64 +%endrep +%else +%assign %%i 0 +%rep 2 + movq m0, [r0] + movq m1, [r0+r1] + movq m2, [r0+r1*2] + movq m3, [r0+r3] + movq m5, m0 + movq m6, m1 + movq m7, m2 + punpcklbw m0, m4 + punpcklbw m1, m4 + punpcklbw m2, m4 + punpckhbw m5, m4 + punpckhbw m6, m4 + punpckhbw m7, m4 + paddsw m0, [r2+ 0+%%i] + paddsw m1, [r2+16+%%i] + paddsw m2, [r2+32+%%i] + paddsw m5, [r2+64+%%i] + paddsw m6, [r2+80+%%i] + paddsw m7, [r2+96+%%i] + packuswb m0, m5 + movq m5, m3 + punpcklbw m3, m4 + punpckhbw m5, m4 + packuswb m1, m6 + paddsw m3, [r2+48+%%i] + paddsw m5, [r2+112+%%i] + packuswb m2, m7 + packuswb m3, m5 + movq [r0 ], m0 + movq [r0+r1 ], m1 + movq [r0+r1*2], m2 + movq [r0+r3 ], m3 +%if %%i == 0 + lea r0, [r0+r1*4] %endif - lea r0, [r0+r1*2] -%assign %%offset 0 -%rep 32/mmsize - mova [r2+%%offset], m4 -%assign %%offset %%offset+mmsize +%assign %%i %%i+8 +%endrep +%endif +%assign %%i 0 +%rep 128/mmsize + mova [r2+%%i], m4 +%assign %%i %%i+mmsize %endrep - add r2, 32 - dec r3 - jg .loop RET %endmacro diff --git a/libavcodec/x86/vp3dsp_init.c b/libavcodec/x86/vp3dsp_init.c index 2668bcf5b9..cc52fbccdd 100644 --- a/libavcodec/x86/vp3dsp_init.c +++ b/libavcodec/x86/vp3dsp_init.c @@ -48,7 +48,6 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags) if (EXTERNAL_MMX(cpuflags)) { c->idct_put = ff_vp3_idct_put_mmx; c->idct_add = ff_vp3_idct_add_mmx; - c->idct_perm = FF_PARTTRANS_IDCT_PERM; } #endif @@ -64,6 +63,5 @@ av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags) if (EXTERNAL_SSE2(cpuflags)) { c->idct_put = ff_vp3_idct_put_sse2; c->idct_add = ff_vp3_idct_add_sse2; - c->idct_perm = FF_TRANSPOSE_IDCT_PERM; } }