diff options
40 files changed, 971 insertions, 680 deletions
@@ -1638,7 +1638,7 @@ wmv3_vdpau_decoder_select="vc1_vdpau_decoder" wmv3_vdpau_hwaccel_select="vc1_vdpau_hwaccel" # parsers -h264_parser_select="error_resilience golomb h264dsp h264pred h264qpel mpegvideo" +h264_parser_select="error_resilience golomb h264chroma h264dsp h264pred h264qpel mpegvideo" mpeg4video_parser_select="error_resilience mpegvideo" mpegvideo_parser_select="error_resilience mpegvideo" vc1_parser_select="error_resilience mpegvideo" diff --git a/libavcodec/Makefile b/libavcodec/Makefile index 774ea57..91b5d88 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -44,6 +44,7 @@ FFT-OBJS-$(CONFIG_HARDCODED_TABLES) += cos_tables.o cos_fixed_tables.o OBJS-$(CONFIG_FFT) += avfft.o fft_fixed.o fft_float.o \ $(FFT-OBJS-yes) OBJS-$(CONFIG_GOLOMB) += golomb.o +OBJS-$(CONFIG_H264CHROMA) += h264chroma.o OBJS-$(CONFIG_H264DSP) += h264dsp.o h264idct.o OBJS-$(CONFIG_H264PRED) += h264pred.o OBJS-$(CONFIG_H264QPEL) += h264qpel.o diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile index 0c372a4..1c91d62 100644 --- a/libavcodec/arm/Makefile +++ b/libavcodec/arm/Makefile @@ -26,6 +26,7 @@ ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o \ arm/vp8dsp_init_armv6.o \ arm/vp8dsp_armv6.o +OBJS-$(CONFIG_H264CHROMA) += arm/h264chroma_init_arm.o OBJS-$(CONFIG_H264DSP) += arm/h264dsp_init_arm.o OBJS-$(CONFIG_H264PRED) += arm/h264pred_init_arm.o OBJS-$(CONFIG_H264QPEL) += arm/h264qpel_init_arm.o @@ -67,9 +68,9 @@ NEON-OBJS-$(CONFIG_MDCT) += arm/mdct_neon.o \ NEON-OBJS-$(CONFIG_RDFT) += arm/rdft_neon.o \ +NEON-OBJS-$(CONFIG_H264CHROMA) += arm/h264cmc_neon.o NEON-OBJS-$(CONFIG_H264DSP) += arm/h264dsp_neon.o \ arm/h264idct_neon.o \ - arm/h264cmc_neon.o \ NEON-OBJS-$(CONFIG_H264PRED) += arm/h264pred_neon.o \ @@ -87,7 +88,6 @@ NEON-OBJS-$(CONFIG_MPEGVIDEO) += arm/mpegvideo_neon.o NEON-OBJS-$(CONFIG_RV30_DECODER) += arm/rv34dsp_neon.o NEON-OBJS-$(CONFIG_RV40_DECODER) += arm/rv34dsp_neon.o \ arm/rv40dsp_neon.o \ - arm/h264cmc_neon.o \ NEON-OBJS-$(CONFIG_VORBIS_DECODER) += arm/vorbisdsp_neon.o diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c index b8d1578..2c2d114 100644 --- a/libavcodec/arm/dsputil_init_neon.c +++ b/libavcodec/arm/dsputil_init_neon.c @@ -64,14 +64,6 @@ void ff_add_pixels_clamped_neon(const int16_t *, uint8_t *, int); void ff_put_pixels_clamped_neon(const int16_t *, uint8_t *, int); void ff_put_signed_pixels_clamped_neon(const int16_t *, uint8_t *, int); -void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int); -void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int); -void ff_put_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int); - -void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int); -void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int); -void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int); - void ff_vector_clipf_neon(float *dst, const float *src, float min, float max, int len); void ff_vector_clip_int32_neon(int32_t *dst, const int32_t *src, int32_t min, @@ -139,16 +131,6 @@ av_cold void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx) c->put_pixels_clamped = ff_put_pixels_clamped_neon; c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_neon; - if (CONFIG_H264_DECODER && !high_bit_depth) { - c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon; - c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon; - c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_neon; - - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon; - c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon; - c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_neon; - } - c->vector_clipf = ff_vector_clipf_neon; c->vector_clip_int32 = ff_vector_clip_int32_neon; diff --git a/libavcodec/arm/h264chroma_init_arm.c b/libavcodec/arm/h264chroma_init_arm.c new file mode 100644 index 0000000..6f36553 --- /dev/null +++ b/libavcodec/arm/h264chroma_init_arm.c @@ -0,0 +1,51 @@ +/* + * ARM NEON optimised H.264 chroma functions + * Copyright (c) 2008 Mans Rullgard <mans@mansr.com> + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdint.h> + +#include "libavutil/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/arm/cpu.h" +#include "libavcodec/h264chroma.h" + +void ff_put_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int); +void ff_put_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int); +void ff_put_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int); + +void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int); +void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int); +void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int); + +av_cold void ff_h264chroma_init_arm(H264ChromaContext *c, int bit_depth) +{ + const int high_bit_depth = bit_depth > 8; + int cpu_flags = av_get_cpu_flags(); + + if (have_neon(cpu_flags) && !high_bit_depth) { + c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_neon; + c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_neon; + c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_neon; + + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_neon; + c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_neon; + c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_neon; + } +} diff --git a/libavcodec/cavs.c b/libavcodec/cavs.c index 0e80fa2..16b6323 100644 --- a/libavcodec/cavs.c +++ b/libavcodec/cavs.c @@ -28,6 +28,7 @@ #include "avcodec.h" #include "get_bits.h" #include "golomb.h" +#include "h264chroma.h" #include "mathops.h" #include "cavs.h" @@ -464,30 +465,35 @@ void ff_cavs_inter(AVSContext *h, enum cavs_mb mb_type) { if(ff_cavs_partition_flags[mb_type] == 0){ // 16x16 mc_part_std(h, 8, 0, h->cy, h->cu, h->cv, 0, 0, h->cdsp.put_cavs_qpel_pixels_tab[0], - h->dsp.put_h264_chroma_pixels_tab[0], + h->h264chroma.put_h264_chroma_pixels_tab[0], h->cdsp.avg_cavs_qpel_pixels_tab[0], - h->dsp.avg_h264_chroma_pixels_tab[0],&h->mv[MV_FWD_X0]); + h->h264chroma.avg_h264_chroma_pixels_tab[0], + &h->mv[MV_FWD_X0]); }else{ mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 0, h->cdsp.put_cavs_qpel_pixels_tab[1], - h->dsp.put_h264_chroma_pixels_tab[1], + h->h264chroma.put_h264_chroma_pixels_tab[1], h->cdsp.avg_cavs_qpel_pixels_tab[1], - h->dsp.avg_h264_chroma_pixels_tab[1],&h->mv[MV_FWD_X0]); + h->h264chroma.avg_h264_chroma_pixels_tab[1], + &h->mv[MV_FWD_X0]); mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 0, h->cdsp.put_cavs_qpel_pixels_tab[1], - h->dsp.put_h264_chroma_pixels_tab[1], + h->h264chroma.put_h264_chroma_pixels_tab[1], h->cdsp.avg_cavs_qpel_pixels_tab[1], - h->dsp.avg_h264_chroma_pixels_tab[1],&h->mv[MV_FWD_X1]); + h->h264chroma.avg_h264_chroma_pixels_tab[1], + &h->mv[MV_FWD_X1]); mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 4, h->cdsp.put_cavs_qpel_pixels_tab[1], - h->dsp.put_h264_chroma_pixels_tab[1], + h->h264chroma.put_h264_chroma_pixels_tab[1], h->cdsp.avg_cavs_qpel_pixels_tab[1], - h->dsp.avg_h264_chroma_pixels_tab[1],&h->mv[MV_FWD_X2]); + h->h264chroma.avg_h264_chroma_pixels_tab[1], + &h->mv[MV_FWD_X2]); mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 4, h->cdsp.put_cavs_qpel_pixels_tab[1], - h->dsp.put_h264_chroma_pixels_tab[1], + h->h264chroma.put_h264_chroma_pixels_tab[1], h->cdsp.avg_cavs_qpel_pixels_tab[1], - h->dsp.avg_h264_chroma_pixels_tab[1],&h->mv[MV_FWD_X3]); + h->h264chroma.avg_h264_chroma_pixels_tab[1], + &h->mv[MV_FWD_X3]); } } @@ -720,6 +726,7 @@ av_cold int ff_cavs_init(AVCodecContext *avctx) { AVSContext *h = avctx->priv_data; ff_dsputil_init(&h->dsp, avctx); + ff_h264chroma_init(&h->h264chroma, 8); ff_videodsp_init(&h->vdsp, 8); ff_cavsdsp_init(&h->cdsp, avctx); ff_init_scantable_permutation(h->dsp.idct_permutation, diff --git a/libavcodec/cavs.h b/libavcodec/cavs.h index 401ec12..d33ff0a 100644 --- a/libavcodec/cavs.h +++ b/libavcodec/cavs.h @@ -24,6 +24,7 @@ #include "cavsdsp.h" #include "dsputil.h" +#include "h264chroma.h" #include "get_bits.h" #include "videodsp.h" @@ -161,6 +162,7 @@ typedef struct AVSFrame { typedef struct AVSContext { AVCodecContext *avctx; DSPContext dsp; + H264ChromaContext h264chroma; VideoDSPContext vdsp; CAVSDSPContext cdsp; GetBitContext gb; diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c index eddf4bd..1c8aae7 100644 --- a/libavcodec/dsputil.c +++ b/libavcodec/dsputil.c @@ -2719,13 +2719,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx) c->clear_blocks = FUNCC(clear_blocks ## dct , depth);\ c->add_pixels8 = FUNCC(add_pixels8 ## dct , depth);\ c->add_pixels4 = FUNCC(add_pixels4 ## dct , depth);\ -\ - c->put_h264_chroma_pixels_tab[0] = FUNCC(put_h264_chroma_mc8 , depth);\ - c->put_h264_chroma_pixels_tab[1] = FUNCC(put_h264_chroma_mc4 , depth);\ - c->put_h264_chroma_pixels_tab[2] = FUNCC(put_h264_chroma_mc2 , depth);\ - c->avg_h264_chroma_pixels_tab[0] = FUNCC(avg_h264_chroma_mc8 , depth);\ - c->avg_h264_chroma_pixels_tab[1] = FUNCC(avg_h264_chroma_mc4 , depth);\ - c->avg_h264_chroma_pixels_tab[2] = FUNCC(avg_h264_chroma_mc2 , depth) switch (avctx->bits_per_raw_sample) { case 9: diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h index f6dbae7..ef3b39b 100644 --- a/libavcodec/dsputil.h +++ b/libavcodec/dsputil.h @@ -140,7 +140,6 @@ void clear_blocks_c(int16_t *blocks); typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, ptrdiff_t line_size, int h); typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h); typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride); -typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y); typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h); @@ -308,12 +307,6 @@ typedef struct DSPContext { qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16]; qpel_mc_func put_mspel_pixels_tab[8]; - /** - * h264 Chroma MC - */ - h264_chroma_mc_func put_h264_chroma_pixels_tab[3]; - h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]; - me_cmp_func pix_abs[2][4]; /* huffyuv specific */ diff --git a/libavcodec/dsputil_template.c b/libavcodec/dsputil_template.c index 1dbeb71..f9cecd1 100644 --- a/libavcodec/dsputil_template.c +++ b/libavcodec/dsputil_template.c @@ -463,124 +463,6 @@ PIXOP2(put, op_put) #undef op_avg #undef op_put -#define H264_CHROMA_MC(OPNAME, OP)\ -static void FUNCC(OPNAME ## h264_chroma_mc2)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ - pixel *dst = (pixel*)_dst;\ - pixel *src = (pixel*)_src;\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - int i;\ - stride /= sizeof(pixel);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - if(D){\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ - OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ - dst+= stride;\ - src+= stride;\ - }\ - }else{\ - const int E= B+C;\ - const int step= C ? stride : 1;\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + E*src[step+0]));\ - OP(dst[1], (A*src[1] + E*src[step+1]));\ - dst+= stride;\ - src+= stride;\ - }\ - }\ -}\ -\ -static void FUNCC(OPNAME ## h264_chroma_mc4)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ - pixel *dst = (pixel*)_dst;\ - pixel *src = (pixel*)_src;\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - int i;\ - stride /= sizeof(pixel);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - if(D){\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ - OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ - OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ - OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ - dst+= stride;\ - src+= stride;\ - }\ - }else{\ - const int E= B+C;\ - const int step= C ? stride : 1;\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + E*src[step+0]));\ - OP(dst[1], (A*src[1] + E*src[step+1]));\ - OP(dst[2], (A*src[2] + E*src[step+2]));\ - OP(dst[3], (A*src[3] + E*src[step+3]));\ - dst+= stride;\ - src+= stride;\ - }\ - }\ -}\ -\ -static void FUNCC(OPNAME ## h264_chroma_mc8)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ - pixel *dst = (pixel*)_dst;\ - pixel *src = (pixel*)_src;\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - int i;\ - stride /= sizeof(pixel);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - if(D){\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ - OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ - OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ - OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ - OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5]));\ - OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6]));\ - OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7]));\ - OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8]));\ - dst+= stride;\ - src+= stride;\ - }\ - }else{\ - const int E= B+C;\ - const int step= C ? stride : 1;\ - for(i=0; i<h; i++){\ - OP(dst[0], (A*src[0] + E*src[step+0]));\ - OP(dst[1], (A*src[1] + E*src[step+1]));\ - OP(dst[2], (A*src[2] + E*src[step+2]));\ - OP(dst[3], (A*src[3] + E*src[step+3]));\ - OP(dst[4], (A*src[4] + E*src[step+4]));\ - OP(dst[5], (A*src[5] + E*src[step+5]));\ - OP(dst[6], (A*src[6] + E*src[step+6]));\ - OP(dst[7], (A*src[7] + E*src[step+7]));\ - dst+= stride;\ - src+= stride;\ - }\ - }\ -} - -#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1) -#define op_put(a, b) a = (((b) + 32)>>6) - -H264_CHROMA_MC(put_ , op_put) -H264_CHROMA_MC(avg_ , op_avg) -#undef op_avg -#undef op_put - void FUNCC(ff_put_pixels8x8)(uint8_t *dst, uint8_t *src, int stride) { FUNCC(put_pixels8)(dst, src, stride, 8); } diff --git a/libavcodec/h264.c b/libavcodec/h264.c index a903b7b..d52b932 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -34,6 +34,7 @@ #include "mpegvideo.h" #include "h264.h" #include "h264data.h" +#include "h264chroma.h" #include "h264_mvpred.h" #include "golomb.h" #include "mathops.h" @@ -976,6 +977,7 @@ static av_cold void common_init(H264Context *h) s->codec_id = s->avctx->codec->id; ff_h264dsp_init(&h->h264dsp, 8, 1); + ff_h264chroma_init(&h->h264chroma, h->sps.bit_depth_chroma); ff_h264qpel_init(&h->h264qpel, 8); ff_h264_pred_init(&h->hpc, s->codec_id, 8, 1); @@ -2445,6 +2447,7 @@ static int h264_set_parameter_from_sps(H264Context *h) ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc); + ff_h264chroma_init(&h->h264chroma, h->sps.bit_depth_chroma); ff_h264qpel_init(&h->h264qpel, h->sps.bit_depth_luma); ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc); diff --git a/libavcodec/h264.h b/libavcodec/h264.h index 323324d..2317d6a 100644 --- a/libavcodec/h264.h +++ b/libavcodec/h264.h @@ -31,6 +31,7 @@ #include "libavutil/intreadwrite.h" #include "cabac.h" #include "mpegvideo.h" +#include "h264chroma.h" #include "h264dsp.h" #include "h264pred.h" #include "h264qpel.h" @@ -254,6 +255,7 @@ typedef struct MMCO { typedef struct H264Context { MpegEncContext s; H264DSPContext h264dsp; + H264ChromaContext h264chroma; H264QpelContext h264qpel; int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264 int chroma_qp[2]; // QPc diff --git a/libavcodec/h264_mb_template.c b/libavcodec/h264_mb_template.c index 8b3e88c..167509e 100644 --- a/libavcodec/h264_mb_template.c +++ b/libavcodec/h264_mb_template.c @@ -176,14 +176,14 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h) } else if (is_h264) { if (chroma422) { FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr, - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, + s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, + s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, h->h264dsp.weight_h264_pixels_tab, h->h264dsp.biweight_h264_pixels_tab); } else { FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr, - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, + s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, + s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, h->h264dsp.weight_h264_pixels_tab, h->h264dsp.biweight_h264_pixels_tab); } @@ -360,8 +360,8 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h) linesize, 0, 1, SIMPLE, PIXEL_SHIFT); } else { FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2], - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, + s->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, + s->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, h->h264dsp.weight_h264_pixels_tab, h->h264dsp.biweight_h264_pixels_tab); } diff --git a/libavcodec/h264chroma.c b/libavcodec/h264chroma.c new file mode 100644 index 0000000..2bd2df3 --- /dev/null +++ b/libavcodec/h264chroma.c @@ -0,0 +1,64 @@ +/* + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" +#include "h264chroma.h" + +#define BIT_DEPTH 8 +#include "h264chroma_template.c" +#undef BIT_DEPTH + +#define BIT_DEPTH 9 +#include "h264chroma_template.c" +#undef BIT_DEPTH + +#define BIT_DEPTH 10 +#include "h264chroma_template.c" +#undef BIT_DEPTH + +#define SET_CHROMA(depth) \ + c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_ ## depth ## _c; \ + c->put_h264_chroma_pixels_tab[1] = put_h264_chroma_mc4_ ## depth ## _c; \ + c->put_h264_chroma_pixels_tab[2] = put_h264_chroma_mc2_ ## depth ## _c; \ + c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_ ## depth ## _c; \ + c->avg_h264_chroma_pixels_tab[1] = avg_h264_chroma_mc4_ ## depth ## _c; \ + c->avg_h264_chroma_pixels_tab[2] = avg_h264_chroma_mc2_ ## depth ## _c; \ + +void ff_h264chroma_init(H264ChromaContext *c, int bit_depth) +{ + switch (bit_depth) { + case 10: + SET_CHROMA(10); + break; + case 9: + SET_CHROMA(9); + break; + default: + SET_CHROMA(8); + break; + } + + if (ARCH_ARM) + ff_h264chroma_init_arm(c, bit_depth); + if (ARCH_PPC) + ff_h264chroma_init_ppc(c, bit_depth); + if (ARCH_SH4) + ff_h264chroma_init_sh4(c, bit_depth); + if (ARCH_X86) + ff_h264chroma_init_x86(c, bit_depth); +} diff --git a/libavcodec/h264chroma.h b/libavcodec/h264chroma.h new file mode 100644 index 0000000..7e1e2ed --- /dev/null +++ b/libavcodec/h264chroma.h @@ -0,0 +1,38 @@ +/* + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVCODEC_H264CHROMA_H +#define AVCODEC_H264CHROMA_H + +#include <stdint.h> + +typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y); + +typedef struct H264ChromaContext { + h264_chroma_mc_func put_h264_chroma_pixels_tab[3]; + h264_chroma_mc_func avg_h264_chroma_pixels_tab[3]; +} H264ChromaContext; + +void ff_h264chroma_init(H264ChromaContext *c, int bit_depth); + +void ff_h264chroma_init_arm(H264ChromaContext *c, int bit_depth); +void ff_h264chroma_init_ppc(H264ChromaContext *c, int bit_depth); +void ff_h264chroma_init_sh4(H264ChromaContext *c, int bit_depth); +void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth); + +#endif /* AVCODEC_H264CHROMA_H */ diff --git a/libavcodec/h264chroma_template.c b/libavcodec/h264chroma_template.c new file mode 100644 index 0000000..351d9d2 --- /dev/null +++ b/libavcodec/h264chroma_template.c @@ -0,0 +1,142 @@ +/* + * Copyright (c) 2000, 2001 Fabrice Bellard + * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <assert.h> + +#include "bit_depth_template.c" + +#define H264_CHROMA_MC(OPNAME, OP)\ +static void FUNCC(OPNAME ## h264_chroma_mc2)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ + pixel *dst = (pixel*)_dst;\ + pixel *src = (pixel*)_src;\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + int i;\ + stride /= sizeof(pixel);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + if(D){\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ + dst+= stride;\ + src+= stride;\ + }\ + }else{\ + const int E= B+C;\ + const int step= C ? stride : 1;\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + E*src[step+0]));\ + OP(dst[1], (A*src[1] + E*src[step+1]));\ + dst+= stride;\ + src+= stride;\ + }\ + }\ +}\ +\ +static void FUNCC(OPNAME ## h264_chroma_mc4)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ + pixel *dst = (pixel*)_dst;\ + pixel *src = (pixel*)_src;\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + int i;\ + stride /= sizeof(pixel);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + if(D){\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ + OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ + OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ + dst+= stride;\ + src+= stride;\ + }\ + }else{\ + const int E= B+C;\ + const int step= C ? stride : 1;\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + E*src[step+0]));\ + OP(dst[1], (A*src[1] + E*src[step+1]));\ + OP(dst[2], (A*src[2] + E*src[step+2]));\ + OP(dst[3], (A*src[3] + E*src[step+3]));\ + dst+= stride;\ + src+= stride;\ + }\ + }\ +}\ +\ +static void FUNCC(OPNAME ## h264_chroma_mc8)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\ + pixel *dst = (pixel*)_dst;\ + pixel *src = (pixel*)_src;\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + int i;\ + stride /= sizeof(pixel);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + if(D){\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ + OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ + OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ + OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5]));\ + OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6]));\ + OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7]));\ + OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8]));\ + dst+= stride;\ + src+= stride;\ + }\ + }else{\ + const int E= B+C;\ + const int step= C ? stride : 1;\ + for(i=0; i<h; i++){\ + OP(dst[0], (A*src[0] + E*src[step+0]));\ + OP(dst[1], (A*src[1] + E*src[step+1]));\ + OP(dst[2], (A*src[2] + E*src[step+2]));\ + OP(dst[3], (A*src[3] + E*src[step+3]));\ + OP(dst[4], (A*src[4] + E*src[step+4]));\ + OP(dst[5], (A*src[5] + E*src[step+5]));\ + OP(dst[6], (A*src[6] + E*src[step+6]));\ + OP(dst[7], (A*src[7] + E*src[step+7]));\ + dst+= stride;\ + src+= stride;\ + }\ + }\ +} + +#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1) +#define op_put(a, b) a = (((b) + 32)>>6) + +H264_CHROMA_MC(put_ , op_put) +H264_CHROMA_MC(avg_ , op_avg) +#undef op_avg +#undef op_put diff --git a/libavcodec/ppc/Makefile b/libavcodec/ppc/Makefile index e152483..a2ce9eb 100644 --- a/libavcodec/ppc/Makefile +++ b/libavcodec/ppc/Makefile @@ -1,6 +1,7 @@ OBJS += ppc/dsputil_ppc.o \ ppc/videodsp_ppc.o \ +OBJS-$(CONFIG_H264CHROMA) += ppc/h264chroma_init.o OBJS-$(CONFIG_H264QPEL) += ppc/h264_qpel.o OBJS-$(CONFIG_VORBIS_DECODER) += ppc/vorbisdsp_altivec.o OBJS-$(CONFIG_VP3DSP) += ppc/vp3dsp_altivec.o diff --git a/libavcodec/ppc/dsputil_altivec.h b/libavcodec/ppc/dsputil_altivec.h index e972344..de5054b 100644 --- a/libavcodec/ppc/dsputil_altivec.h +++ b/libavcodec/ppc/dsputil_altivec.h @@ -36,8 +36,6 @@ void ff_gmc1_altivec(uint8_t *dst, uint8_t *src, int stride, int h, void ff_idct_put_altivec(uint8_t *dest, int line_size, int16_t *block); void ff_idct_add_altivec(uint8_t *dest, int line_size, int16_t *block); -void ff_dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx); - void ff_dsputil_init_altivec(DSPContext* c, AVCodecContext *avctx); void ff_float_init_altivec(DSPContext* c, AVCodecContext *avctx); void ff_int_init_altivec(DSPContext* c, AVCodecContext *avctx); diff --git a/libavcodec/ppc/dsputil_ppc.c b/libavcodec/ppc/dsputil_ppc.c index 1174a7c..1eebafa 100644 --- a/libavcodec/ppc/dsputil_ppc.c +++ b/libavcodec/ppc/dsputil_ppc.c @@ -157,8 +157,6 @@ av_cold void ff_dsputil_init_ppc(DSPContext *c, AVCodecContext *avctx) } #if HAVE_ALTIVEC - if(CONFIG_H264_DECODER) ff_dsputil_h264_init_ppc(c, avctx); - if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) { ff_dsputil_init_altivec(c, avctx); ff_int_init_altivec(c, avctx); diff --git a/libavcodec/ppc/h264_qpel.c b/libavcodec/ppc/h264_qpel.c index b70d4a9..2e68a06 100644 --- a/libavcodec/ppc/h264_qpel.c +++ b/libavcodec/ppc/h264_qpel.c @@ -33,8 +33,6 @@ #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s) #define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC -#define PREFIX_h264_chroma_mc8_altivec put_h264_chroma_mc8_altivec -#define PREFIX_h264_chroma_mc8_num altivec_put_h264_chroma_mc8_num #define PREFIX_h264_qpel16_h_lowpass_altivec put_h264_qpel16_h_lowpass_altivec #define PREFIX_h264_qpel16_h_lowpass_num altivec_put_h264_qpel16_h_lowpass_num #define PREFIX_h264_qpel16_v_lowpass_altivec put_h264_qpel16_v_lowpass_altivec @@ -43,8 +41,6 @@ #define PREFIX_h264_qpel16_hv_lowpass_num altivec_put_h264_qpel16_hv_lowpass_num #include "h264_qpel_template.c" #undef OP_U8_ALTIVEC -#undef PREFIX_h264_chroma_mc8_altivec -#undef PREFIX_h264_chroma_mc8_num #undef PREFIX_h264_qpel16_h_lowpass_altivec #undef PREFIX_h264_qpel16_h_lowpass_num #undef PREFIX_h264_qpel16_v_lowpass_altivec @@ -53,8 +49,6 @@ #undef PREFIX_h264_qpel16_hv_lowpass_num #define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC -#define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec -#define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num #define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec #define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num #define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec @@ -63,8 +57,6 @@ #define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num #include "h264_qpel_template.c" #undef OP_U8_ALTIVEC -#undef PREFIX_h264_chroma_mc8_altivec -#undef PREFIX_h264_chroma_mc8_num #undef PREFIX_h264_qpel16_h_lowpass_altivec #undef PREFIX_h264_qpel16_h_lowpass_num #undef PREFIX_h264_qpel16_v_lowpass_altivec @@ -273,18 +265,6 @@ static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1, H264_MC(put_, 16, altivec) H264_MC(avg_, 16, altivec) - -void ff_dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) -{ - const int high_bit_depth = avctx->bits_per_raw_sample > 8; - - if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) { - if (!high_bit_depth) { - c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec; - c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec; - } - } -} #endif /* HAVE_ALTIVEC */ av_cold void ff_h264qpel_init_ppc(H264QpelContext *c, int bit_depth) diff --git a/libavcodec/ppc/h264_qpel_template.c b/libavcodec/ppc/h264_qpel_template.c index b445f92..4df1d09 100644 --- a/libavcodec/ppc/h264_qpel_template.c +++ b/libavcodec/ppc/h264_qpel_template.c @@ -26,274 +26,6 @@ #define ASSERT_ALIGNED(ptr) ; #endif -/* this code assume that stride % 16 == 0 */ - -#define CHROMA_MC8_ALTIVEC_CORE(BIAS1, BIAS2) \ - vsrc2ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc2uc);\ - vsrc3ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc3uc);\ -\ - psum = vec_mladd(vA, vsrc0ssH, BIAS1);\ - psum = vec_mladd(vB, vsrc1ssH, psum);\ - psum = vec_mladd(vC, vsrc2ssH, psum);\ - psum = vec_mladd(vD, vsrc3ssH, psum);\ - psum = BIAS2(psum);\ - psum = vec_sr(psum, v6us);\ -\ - vdst = vec_ld(0, dst);\ - ppsum = (vec_u8)vec_pack(psum, psum);\ - vfdst = vec_perm(vdst, ppsum, fperm);\ -\ - OP_U8_ALTIVEC(fsum, vfdst, vdst);\ -\ - vec_st(fsum, 0, dst);\ -\ - vsrc0ssH = vsrc2ssH;\ - vsrc1ssH = vsrc3ssH;\ -\ - dst += stride;\ - src += stride; - -#define CHROMA_MC8_ALTIVEC_CORE_SIMPLE \ -\ - vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc0uc);\ - vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc1uc);\ -\ - psum = vec_mladd(vA, vsrc0ssH, v32ss);\ - psum = vec_mladd(vE, vsrc1ssH, psum);\ - psum = vec_sr(psum, v6us);\ -\ - vdst = vec_ld(0, dst);\ - ppsum = (vec_u8)vec_pack(psum, psum);\ - vfdst = vec_perm(vdst, ppsum, fperm);\ -\ - OP_U8_ALTIVEC(fsum, vfdst, vdst);\ -\ - vec_st(fsum, 0, dst);\ -\ - dst += stride;\ - src += stride; - -#define noop(a) a -#define add28(a) vec_add(v28ss, a) - -#ifdef PREFIX_h264_chroma_mc8_altivec -static void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, - int stride, int h, int x, int y) { - DECLARE_ALIGNED(16, signed int, ABCD)[4] = - {((8 - x) * (8 - y)), - (( x) * (8 - y)), - ((8 - x) * ( y)), - (( x) * ( y))}; - register int i; - vec_u8 fperm; - const vec_s32 vABCD = vec_ld(0, ABCD); - const vec_s16 vA = vec_splat((vec_s16)vABCD, 1); - const vec_s16 vB = vec_splat((vec_s16)vABCD, 3); - const vec_s16 vC = vec_splat((vec_s16)vABCD, 5); - const vec_s16 vD = vec_splat((vec_s16)vABCD, 7); - LOAD_ZERO; - const vec_s16 v32ss = vec_sl(vec_splat_s16(1),vec_splat_u16(5)); - const vec_u16 v6us = vec_splat_u16(6); - register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1; - register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0; - - vec_u8 vsrcAuc, av_uninit(vsrcBuc), vsrcperm0, vsrcperm1; - vec_u8 vsrc0uc, vsrc1uc; - vec_s16 vsrc0ssH, vsrc1ssH; - vec_u8 vsrcCuc, vsrc2uc, vsrc3uc; - vec_s16 vsrc2ssH, vsrc3ssH, psum; - vec_u8 vdst, ppsum, vfdst, fsum; - - if (((unsigned long)dst) % 16 == 0) { - fperm = (vec_u8){0x10, 0x11, 0x12, 0x13, - 0x14, 0x15, 0x16, 0x17, - 0x08, 0x09, 0x0A, 0x0B, - 0x0C, 0x0D, 0x0E, 0x0F}; - } else { - fperm = (vec_u8){0x00, 0x01, 0x02, 0x03, - 0x04, 0x05, 0x06, 0x07, - 0x18, 0x19, 0x1A, 0x1B, - 0x1C, 0x1D, 0x1E, 0x1F}; - } - - vsrcAuc = vec_ld(0, src); - - if (loadSecond) - vsrcBuc = vec_ld(16, src); - vsrcperm0 = vec_lvsl(0, src); - vsrcperm1 = vec_lvsl(1, src); - - vsrc0uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm0); - if (reallyBadAlign) - vsrc1uc = vsrcBuc; - else - vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1); - - vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc0uc); - vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc1uc); - - if (ABCD[3]) { - if (!loadSecond) {// -> !reallyBadAlign - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(stride + 0, src); - vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); - vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE(v32ss, noop) - } - } else { - vec_u8 vsrcDuc; - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(stride + 0, src); - vsrcDuc = vec_ld(stride + 16, src); - vsrc2uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); - if (reallyBadAlign) - vsrc3uc = vsrcDuc; - else - vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE(v32ss, noop) - } - } - } else { - const vec_s16 vE = vec_add(vB, vC); - if (ABCD[2]) { // x == 0 B == 0 - if (!loadSecond) {// -> !reallyBadAlign - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(stride + 0, src); - vsrc1uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); - CHROMA_MC8_ALTIVEC_CORE_SIMPLE - - vsrc0uc = vsrc1uc; - } - } else { - vec_u8 vsrcDuc; - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(stride + 0, src); - vsrcDuc = vec_ld(stride + 15, src); - vsrc1uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); - CHROMA_MC8_ALTIVEC_CORE_SIMPLE - - vsrc0uc = vsrc1uc; - } - } - } else { // y == 0 C == 0 - if (!loadSecond) {// -> !reallyBadAlign - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(0, src); - vsrc0uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); - vsrc1uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE_SIMPLE - } - } else { - vec_u8 vsrcDuc; - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(0, src); - vsrcDuc = vec_ld(15, src); - vsrc0uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); - if (reallyBadAlign) - vsrc1uc = vsrcDuc; - else - vsrc1uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE_SIMPLE - } - } - } - } -} -#endif - -/* this code assume that stride % 16 == 0 */ -#ifdef PREFIX_no_rnd_vc1_chroma_mc8_altivec -static void PREFIX_no_rnd_vc1_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, int h, int x, int y) { - DECLARE_ALIGNED(16, signed int, ABCD)[4] = - {((8 - x) * (8 - y)), - (( x) * (8 - y)), - ((8 - x) * ( y)), - (( x) * ( y))}; - register int i; - vec_u8 fperm; - const vec_s32 vABCD = vec_ld(0, ABCD); - const vec_s16 vA = vec_splat((vec_s16)vABCD, 1); - const vec_s16 vB = vec_splat((vec_s16)vABCD, 3); - const vec_s16 vC = vec_splat((vec_s16)vABCD, 5); - const vec_s16 vD = vec_splat((vec_s16)vABCD, 7); - LOAD_ZERO; - const vec_s16 v28ss = vec_sub(vec_sl(vec_splat_s16(1),vec_splat_u16(5)),vec_splat_s16(4)); - const vec_u16 v6us = vec_splat_u16(6); - register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1; - register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0; - - vec_u8 vsrcAuc, av_uninit(vsrcBuc), vsrcperm0, vsrcperm1; - vec_u8 vsrc0uc, vsrc1uc; - vec_s16 vsrc0ssH, vsrc1ssH; - vec_u8 vsrcCuc, vsrc2uc, vsrc3uc; - vec_s16 vsrc2ssH, vsrc3ssH, psum; - vec_u8 vdst, ppsum, vfdst, fsum; - - if (((unsigned long)dst) % 16 == 0) { - fperm = (vec_u8){0x10, 0x11, 0x12, 0x13, - 0x14, 0x15, 0x16, 0x17, - 0x08, 0x09, 0x0A, 0x0B, - 0x0C, 0x0D, 0x0E, 0x0F}; - } else { - fperm = (vec_u8){0x00, 0x01, 0x02, 0x03, - 0x04, 0x05, 0x06, 0x07, - 0x18, 0x19, 0x1A, 0x1B, - 0x1C, 0x1D, 0x1E, 0x1F}; - } - - vsrcAuc = vec_ld(0, src); - - if (loadSecond) - vsrcBuc = vec_ld(16, src); - vsrcperm0 = vec_lvsl(0, src); - vsrcperm1 = vec_lvsl(1, src); - - vsrc0uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm0); - if (reallyBadAlign) - vsrc1uc = vsrcBuc; - else - vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1); - - vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v, (vec_u8)vsrc0uc); - vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v, (vec_u8)vsrc1uc); - - if (!loadSecond) {// -> !reallyBadAlign - for (i = 0 ; i < h ; i++) { - - - vsrcCuc = vec_ld(stride + 0, src); - - vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); - vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE(vec_splat_s16(0), add28) - } - } else { - vec_u8 vsrcDuc; - for (i = 0 ; i < h ; i++) { - vsrcCuc = vec_ld(stride + 0, src); - vsrcDuc = vec_ld(stride + 16, src); - - vsrc2uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); - if (reallyBadAlign) - vsrc3uc = vsrcDuc; - else - vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); - - CHROMA_MC8_ALTIVEC_CORE(vec_splat_s16(0), add28) - } - } -} -#endif - -#undef noop -#undef add28 -#undef CHROMA_MC8_ALTIVEC_CORE - /* this code assume stride % 16 == 0 */ #ifdef PREFIX_h264_qpel16_h_lowpass_altivec static void PREFIX_h264_qpel16_h_lowpass_altivec(uint8_t * dst, uint8_t * src, int dstStride, int srcStride) { diff --git a/libavcodec/ppc/h264chroma_init.c b/libavcodec/ppc/h264chroma_init.c new file mode 100644 index 0000000..06456ec --- /dev/null +++ b/libavcodec/ppc/h264chroma_init.c @@ -0,0 +1,64 @@ +/* + * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org> + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" +#include "libavutil/attributes.h" +#include "libavcodec/h264chroma.h" + +#if HAVE_ALTIVEC +#include "libavutil/cpu.h" +#include "libavutil/intreadwrite.h" +#include "libavutil/ppc/types_altivec.h" +#include "libavutil/ppc/util_altivec.h" +#include "dsputil_altivec.h" + +#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s +#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s) + +#define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC +#define PREFIX_h264_chroma_mc8_altivec put_h264_chroma_mc8_altivec +#define PREFIX_h264_chroma_mc8_num altivec_put_h264_chroma_mc8_num +#include "h264chroma_template.c" +#undef OP_U8_ALTIVEC +#undef PREFIX_h264_chroma_mc8_altivec +#undef PREFIX_h264_chroma_mc8_num + +#define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC +#define PREFIX_h264_chroma_mc8_altivec avg_h264_chroma_mc8_altivec +#define PREFIX_h264_chroma_mc8_num altivec_avg_h264_chroma_mc8_num +#include "h264chroma_template.c" +#undef OP_U8_ALTIVEC +#undef PREFIX_h264_chroma_mc8_altivec +#undef PREFIX_h264_chroma_mc8_num +#endif /* HAVE_ALTIVEC */ + +av_cold void ff_h264chroma_init_ppc(H264ChromaContext *c, int bit_depth) +{ +#if HAVE_ALTIVEC + const int high_bit_depth = bit_depth > 8; + + if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) { + if (!high_bit_depth) { + c->put_h264_chroma_pixels_tab[0] = put_h264_chroma_mc8_altivec; + c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec; + } + } +#endif /* HAVE_ALTIVEC */ +} diff --git a/libavcodec/ppc/h264chroma_template.c b/libavcodec/ppc/h264chroma_template.c new file mode 100644 index 0000000..293fef5 --- /dev/null +++ b/libavcodec/ppc/h264chroma_template.c @@ -0,0 +1,289 @@ +/* + * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org> + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/mem.h" + +/* this code assume that stride % 16 == 0 */ + +#define CHROMA_MC8_ALTIVEC_CORE(BIAS1, BIAS2) \ + vsrc2ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc2uc);\ + vsrc3ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc3uc);\ +\ + psum = vec_mladd(vA, vsrc0ssH, BIAS1);\ + psum = vec_mladd(vB, vsrc1ssH, psum);\ + psum = vec_mladd(vC, vsrc2ssH, psum);\ + psum = vec_mladd(vD, vsrc3ssH, psum);\ + psum = BIAS2(psum);\ + psum = vec_sr(psum, v6us);\ +\ + vdst = vec_ld(0, dst);\ + ppsum = (vec_u8)vec_pack(psum, psum);\ + vfdst = vec_perm(vdst, ppsum, fperm);\ +\ + OP_U8_ALTIVEC(fsum, vfdst, vdst);\ +\ + vec_st(fsum, 0, dst);\ +\ + vsrc0ssH = vsrc2ssH;\ + vsrc1ssH = vsrc3ssH;\ +\ + dst += stride;\ + src += stride; + +#define CHROMA_MC8_ALTIVEC_CORE_SIMPLE \ +\ + vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc0uc);\ + vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc1uc);\ +\ + psum = vec_mladd(vA, vsrc0ssH, v32ss);\ + psum = vec_mladd(vE, vsrc1ssH, psum);\ + psum = vec_sr(psum, v6us);\ +\ + vdst = vec_ld(0, dst);\ + ppsum = (vec_u8)vec_pack(psum, psum);\ + vfdst = vec_perm(vdst, ppsum, fperm);\ +\ + OP_U8_ALTIVEC(fsum, vfdst, vdst);\ +\ + vec_st(fsum, 0, dst);\ +\ + dst += stride;\ + src += stride; + +#define noop(a) a +#define add28(a) vec_add(v28ss, a) + +#ifdef PREFIX_h264_chroma_mc8_altivec +static void PREFIX_h264_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, + int stride, int h, int x, int y) { + DECLARE_ALIGNED(16, signed int, ABCD)[4] = + {((8 - x) * (8 - y)), + (( x) * (8 - y)), + ((8 - x) * ( y)), + (( x) * ( y))}; + register int i; + vec_u8 fperm; + const vec_s32 vABCD = vec_ld(0, ABCD); + const vec_s16 vA = vec_splat((vec_s16)vABCD, 1); + const vec_s16 vB = vec_splat((vec_s16)vABCD, 3); + const vec_s16 vC = vec_splat((vec_s16)vABCD, 5); + const vec_s16 vD = vec_splat((vec_s16)vABCD, 7); + LOAD_ZERO; + const vec_s16 v32ss = vec_sl(vec_splat_s16(1),vec_splat_u16(5)); + const vec_u16 v6us = vec_splat_u16(6); + register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1; + register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0; + + vec_u8 vsrcAuc, av_uninit(vsrcBuc), vsrcperm0, vsrcperm1; + vec_u8 vsrc0uc, vsrc1uc; + vec_s16 vsrc0ssH, vsrc1ssH; + vec_u8 vsrcCuc, vsrc2uc, vsrc3uc; + vec_s16 vsrc2ssH, vsrc3ssH, psum; + vec_u8 vdst, ppsum, vfdst, fsum; + + if (((unsigned long)dst) % 16 == 0) { + fperm = (vec_u8){0x10, 0x11, 0x12, 0x13, + 0x14, 0x15, 0x16, 0x17, + 0x08, 0x09, 0x0A, 0x0B, + 0x0C, 0x0D, 0x0E, 0x0F}; + } else { + fperm = (vec_u8){0x00, 0x01, 0x02, 0x03, + 0x04, 0x05, 0x06, 0x07, + 0x18, 0x19, 0x1A, 0x1B, + 0x1C, 0x1D, 0x1E, 0x1F}; + } + + vsrcAuc = vec_ld(0, src); + + if (loadSecond) + vsrcBuc = vec_ld(16, src); + vsrcperm0 = vec_lvsl(0, src); + vsrcperm1 = vec_lvsl(1, src); + + vsrc0uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm0); + if (reallyBadAlign) + vsrc1uc = vsrcBuc; + else + vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1); + + vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc0uc); + vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v,(vec_u8)vsrc1uc); + + if (ABCD[3]) { + if (!loadSecond) {// -> !reallyBadAlign + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(stride + 0, src); + vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); + vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE(v32ss, noop) + } + } else { + vec_u8 vsrcDuc; + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(stride + 0, src); + vsrcDuc = vec_ld(stride + 16, src); + vsrc2uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); + if (reallyBadAlign) + vsrc3uc = vsrcDuc; + else + vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE(v32ss, noop) + } + } + } else { + const vec_s16 vE = vec_add(vB, vC); + if (ABCD[2]) { // x == 0 B == 0 + if (!loadSecond) {// -> !reallyBadAlign + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(stride + 0, src); + vsrc1uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); + CHROMA_MC8_ALTIVEC_CORE_SIMPLE + + vsrc0uc = vsrc1uc; + } + } else { + vec_u8 vsrcDuc; + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(stride + 0, src); + vsrcDuc = vec_ld(stride + 15, src); + vsrc1uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); + CHROMA_MC8_ALTIVEC_CORE_SIMPLE + + vsrc0uc = vsrc1uc; + } + } + } else { // y == 0 C == 0 + if (!loadSecond) {// -> !reallyBadAlign + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(0, src); + vsrc0uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); + vsrc1uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE_SIMPLE + } + } else { + vec_u8 vsrcDuc; + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(0, src); + vsrcDuc = vec_ld(15, src); + vsrc0uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); + if (reallyBadAlign) + vsrc1uc = vsrcDuc; + else + vsrc1uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE_SIMPLE + } + } + } + } +} +#endif + +/* this code assume that stride % 16 == 0 */ +#ifdef PREFIX_no_rnd_vc1_chroma_mc8_altivec +static void PREFIX_no_rnd_vc1_chroma_mc8_altivec(uint8_t * dst, uint8_t * src, int stride, int h, int x, int y) { + DECLARE_ALIGNED(16, signed int, ABCD)[4] = + {((8 - x) * (8 - y)), + (( x) * (8 - y)), + ((8 - x) * ( y)), + (( x) * ( y))}; + register int i; + vec_u8 fperm; + const vec_s32 vABCD = vec_ld(0, ABCD); + const vec_s16 vA = vec_splat((vec_s16)vABCD, 1); + const vec_s16 vB = vec_splat((vec_s16)vABCD, 3); + const vec_s16 vC = vec_splat((vec_s16)vABCD, 5); + const vec_s16 vD = vec_splat((vec_s16)vABCD, 7); + LOAD_ZERO; + const vec_s16 v28ss = vec_sub(vec_sl(vec_splat_s16(1),vec_splat_u16(5)),vec_splat_s16(4)); + const vec_u16 v6us = vec_splat_u16(6); + register int loadSecond = (((unsigned long)src) % 16) <= 7 ? 0 : 1; + register int reallyBadAlign = (((unsigned long)src) % 16) == 15 ? 1 : 0; + + vec_u8 vsrcAuc, av_uninit(vsrcBuc), vsrcperm0, vsrcperm1; + vec_u8 vsrc0uc, vsrc1uc; + vec_s16 vsrc0ssH, vsrc1ssH; + vec_u8 vsrcCuc, vsrc2uc, vsrc3uc; + vec_s16 vsrc2ssH, vsrc3ssH, psum; + vec_u8 vdst, ppsum, vfdst, fsum; + + if (((unsigned long)dst) % 16 == 0) { + fperm = (vec_u8){0x10, 0x11, 0x12, 0x13, + 0x14, 0x15, 0x16, 0x17, + 0x08, 0x09, 0x0A, 0x0B, + 0x0C, 0x0D, 0x0E, 0x0F}; + } else { + fperm = (vec_u8){0x00, 0x01, 0x02, 0x03, + 0x04, 0x05, 0x06, 0x07, + 0x18, 0x19, 0x1A, 0x1B, + 0x1C, 0x1D, 0x1E, 0x1F}; + } + + vsrcAuc = vec_ld(0, src); + + if (loadSecond) + vsrcBuc = vec_ld(16, src); + vsrcperm0 = vec_lvsl(0, src); + vsrcperm1 = vec_lvsl(1, src); + + vsrc0uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm0); + if (reallyBadAlign) + vsrc1uc = vsrcBuc; + else + vsrc1uc = vec_perm(vsrcAuc, vsrcBuc, vsrcperm1); + + vsrc0ssH = (vec_s16)vec_mergeh(zero_u8v, (vec_u8)vsrc0uc); + vsrc1ssH = (vec_s16)vec_mergeh(zero_u8v, (vec_u8)vsrc1uc); + + if (!loadSecond) {// -> !reallyBadAlign + for (i = 0 ; i < h ; i++) { + + + vsrcCuc = vec_ld(stride + 0, src); + + vsrc2uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm0); + vsrc3uc = vec_perm(vsrcCuc, vsrcCuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE(vec_splat_s16(0), add28) + } + } else { + vec_u8 vsrcDuc; + for (i = 0 ; i < h ; i++) { + vsrcCuc = vec_ld(stride + 0, src); + vsrcDuc = vec_ld(stride + 16, src); + + vsrc2uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm0); + if (reallyBadAlign) + vsrc3uc = vsrcDuc; + else + vsrc3uc = vec_perm(vsrcCuc, vsrcDuc, vsrcperm1); + + CHROMA_MC8_ALTIVEC_CORE(vec_splat_s16(0), add28) + } + } +} +#endif + +#undef noop +#undef add28 +#undef CHROMA_MC8_ALTIVEC_CORE diff --git a/libavcodec/ppc/vc1dsp_altivec.c b/libavcodec/ppc/vc1dsp_altivec.c index e6ebb9c..18ec50f 100644 --- a/libavcodec/ppc/vc1dsp_altivec.c +++ b/libavcodec/ppc/vc1dsp_altivec.c @@ -326,13 +326,13 @@ static void vc1_inv_trans_8x4_altivec(uint8_t *dest, int stride, int16_t *block) #define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC #define PREFIX_no_rnd_vc1_chroma_mc8_altivec put_no_rnd_vc1_chroma_mc8_altivec -#include "h264_qpel_template.c" +#include "h264chroma_template.c" #undef OP_U8_ALTIVEC #undef PREFIX_no_rnd_vc1_chroma_mc8_altivec #define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC #define PREFIX_no_rnd_vc1_chroma_mc8_altivec avg_no_rnd_vc1_chroma_mc8_altivec -#include "h264_qpel_template.c" +#include "h264chroma_template.c" #undef OP_U8_ALTIVEC #undef PREFIX_no_rnd_vc1_chroma_mc8_altivec diff --git a/libavcodec/rv30dsp.c b/libavcodec/rv30dsp.c index ba5ae0f..394aeca 100644 --- a/libavcodec/rv30dsp.c +++ b/libavcodec/rv30dsp.c @@ -26,6 +26,7 @@ #include "avcodec.h" #include "dsputil.h" +#include "h264chroma.h" #include "h264qpel.h" #include "rv34dsp.h" @@ -254,9 +255,11 @@ RV30_MC(avg_, 8) RV30_MC(avg_, 16) av_cold void ff_rv30dsp_init(RV34DSPContext *c, DSPContext* dsp) { + H264ChromaContext h264chroma; H264QpelContext qpel; ff_rv34dsp_init(c, dsp); + ff_h264chroma_init(&h264chroma, 8); ff_h264qpel_init(&qpel, 8); c->put_pixels_tab[0][ 0] = qpel.put_h264_qpel_pixels_tab[0][0]; @@ -296,8 +299,8 @@ av_cold void ff_rv30dsp_init(RV34DSPContext *c, DSPContext* dsp) { c->avg_pixels_tab[1][ 9] = avg_rv30_tpel8_mc12_c; c->avg_pixels_tab[1][10] = avg_rv30_tpel8_mc22_c; - c->put_chroma_pixels_tab[0] = dsp->put_h264_chroma_pixels_tab[0]; - c->put_chroma_pixels_tab[1] = dsp->put_h264_chroma_pixels_tab[1]; - c->avg_chroma_pixels_tab[0] = dsp->avg_h264_chroma_pixels_tab[0]; - c->avg_chroma_pixels_tab[1] = dsp->avg_h264_chroma_pixels_tab[1]; + c->put_chroma_pixels_tab[0] = h264chroma.put_h264_chroma_pixels_tab[0]; + c->put_chroma_pixels_tab[1] = h264chroma.put_h264_chroma_pixels_tab[1]; + c->avg_chroma_pixels_tab[0] = h264chroma.avg_h264_chroma_pixels_tab[0]; + c->avg_chroma_pixels_tab[1] = h264chroma.avg_h264_chroma_pixels_tab[1]; } diff --git a/libavcodec/rv34dsp.h b/libavcodec/rv34dsp.h index e04a47d..6d2f834 100644 --- a/libavcodec/rv34dsp.h +++ b/libavcodec/rv34dsp.h @@ -28,6 +28,7 @@ #define AVCODEC_RV34DSP_H #include "dsputil.h" +#include "h264chroma.h" typedef void (*rv40_weight_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src1/*align width (8 or 16)*/, diff --git a/libavcodec/sh4/Makefile b/libavcodec/sh4/Makefile index aa17eab..f907408 100644 --- a/libavcodec/sh4/Makefile +++ b/libavcodec/sh4/Makefile @@ -1,3 +1,5 @@ OBJS += sh4/dsputil_align.o \ sh4/dsputil_sh4.o \ sh4/idct_sh4.o \ + +OBJS-$(CONFIG_H264CHROMA) += sh4/h264chroma_init.o \ diff --git a/libavcodec/sh4/dsputil_align.c b/libavcodec/sh4/dsputil_align.c index 91b34a3..bf9bff2 100644 --- a/libavcodec/sh4/dsputil_align.c +++ b/libavcodec/sh4/dsputil_align.c @@ -369,14 +369,6 @@ av_cold void ff_dsputil_init_align(DSPContext *c, AVCodecContext *avctx) /* dspfunc(avg_no_rnd_qpel, 1, 8); */ #undef dspfunc - if (!high_bit_depth) { - c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_sh4; - c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_sh4; - c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_sh4; - c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_sh4; - c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_sh4; - c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_sh4; - } c->put_mspel_pixels_tab[0]= put_mspel8_mc00_sh4; c->put_mspel_pixels_tab[1]= put_mspel8_mc10_sh4; diff --git a/libavcodec/sh4/h264chroma_init.c b/libavcodec/sh4/h264chroma_init.c new file mode 100644 index 0000000..5681a24 --- /dev/null +++ b/libavcodec/sh4/h264chroma_init.c @@ -0,0 +1,132 @@ +/* + * aligned/packed access motion + * + * Copyright (c) 2001-2003 BERO <bero@geocities.co.jp> + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <assert.h> +#include <stdint.h> + +#include "libavutil/attributes.h" +#include "libavcodec/h264chroma.h" + +#define H264_CHROMA_MC(OPNAME, OP)\ +static void OPNAME ## h264_chroma_mc2_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + do {\ + int t0,t1,t2,t3; \ + uint8_t *s0 = src; \ + uint8_t *s1 = src+stride; \ + t0 = *s0++; t2 = *s1++; \ + t1 = *s0++; t3 = *s1++; \ + OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ + dst+= stride;\ + src+= stride;\ + }while(--h);\ +}\ +\ +static void OPNAME ## h264_chroma_mc4_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + do {\ + int t0,t1,t2,t3; \ + uint8_t *s0 = src; \ + uint8_t *s1 = src+stride; \ + t0 = *s0++; t2 = *s1++; \ + t1 = *s0++; t3 = *s1++; \ + OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ + t1 = *s0++; t3 = *s1++; \ + OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\ + dst+= stride;\ + src+= stride;\ + }while(--h);\ +}\ +\ +static void OPNAME ## h264_chroma_mc8_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + do {\ + int t0,t1,t2,t3; \ + uint8_t *s0 = src; \ + uint8_t *s1 = src+stride; \ + t0 = *s0++; t2 = *s1++; \ + t1 = *s0++; t3 = *s1++; \ + OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ + t1 = *s0++; t3 = *s1++; \ + OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\ + t1 = *s0++; t3 = *s1++; \ + OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\ + t1 = *s0++; t3 = *s1++; \ + OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\ + t0 = *s0++; t2 = *s1++; \ + OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\ + dst+= stride;\ + src+= stride;\ + }while(--h);\ +} + +#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1) +#define op_put(a, b) a = (((b) + 32)>>6) + +H264_CHROMA_MC(put_ , op_put) +H264_CHROMA_MC(avg_ , op_avg) +#undef op_avg +#undef op_put + +av_cold void ff_h264chroma_init_sh4(H264ChromaContext *c, int bit_depth) +{ + const int high_bit_depth = bit_depth > 8; + + if (!high_bit_depth) { + c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_sh4; + c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_sh4; + c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_sh4; + c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_sh4; + c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_sh4; + c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_sh4; + } +} diff --git a/libavcodec/sh4/qpel.c b/libavcodec/sh4/qpel.c index 055d184..27d56a15 100644 --- a/libavcodec/sh4/qpel.c +++ b/libavcodec/sh4/qpel.c @@ -359,97 +359,6 @@ static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y }while(--h); } -#define H264_CHROMA_MC(OPNAME, OP)\ -static void OPNAME ## h264_chroma_mc2_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - do {\ - int t0,t1,t2,t3; \ - uint8_t *s0 = src; \ - uint8_t *s1 = src+stride; \ - t0 = *s0++; t2 = *s1++; \ - t1 = *s0++; t3 = *s1++; \ - OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ - dst+= stride;\ - src+= stride;\ - }while(--h);\ -}\ -\ -static void OPNAME ## h264_chroma_mc4_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - do {\ - int t0,t1,t2,t3; \ - uint8_t *s0 = src; \ - uint8_t *s1 = src+stride; \ - t0 = *s0++; t2 = *s1++; \ - t1 = *s0++; t3 = *s1++; \ - OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ - t1 = *s0++; t3 = *s1++; \ - OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\ - dst+= stride;\ - src+= stride;\ - }while(--h);\ -}\ -\ -static void OPNAME ## h264_chroma_mc8_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ - const int A=(8-x)*(8-y);\ - const int B=( x)*(8-y);\ - const int C=(8-x)*( y);\ - const int D=( x)*( y);\ - \ - assert(x<8 && y<8 && x>=0 && y>=0);\ -\ - do {\ - int t0,t1,t2,t3; \ - uint8_t *s0 = src; \ - uint8_t *s1 = src+stride; \ - t0 = *s0++; t2 = *s1++; \ - t1 = *s0++; t3 = *s1++; \ - OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\ - t1 = *s0++; t3 = *s1++; \ - OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\ - t1 = *s0++; t3 = *s1++; \ - OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\ - t1 = *s0++; t3 = *s1++; \ - OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\ - t0 = *s0++; t2 = *s1++; \ - OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\ - dst+= stride;\ - src+= stride;\ - }while(--h);\ -} - -#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1) -#define op_put(a, b) a = (((b) + 32)>>6) - -H264_CHROMA_MC(put_ , op_put) -H264_CHROMA_MC(avg_ , op_avg) -#undef op_avg -#undef op_put - #define QPEL_MC(r, OPNAME, RND, OP) \ static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\ uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ diff --git a/libavcodec/vc1.h b/libavcodec/vc1.h index f3285f0..466a58e 100644 --- a/libavcodec/vc1.h +++ b/libavcodec/vc1.h @@ -24,6 +24,7 @@ #define AVCODEC_VC1_H #include "avcodec.h" +#include "h264chroma.h" #include "mpegvideo.h" #include "intrax8.h" #include "vc1dsp.h" @@ -181,6 +182,7 @@ enum FrameCodingMode { typedef struct VC1Context{ MpegEncContext s; IntraX8Context x8; + H264ChromaContext h264chroma; VC1DSPContext vc1dsp; int bits; diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c index 81b0966..38b8216 100644 --- a/libavcodec/vc1dec.c +++ b/libavcodec/vc1dec.c @@ -31,6 +31,7 @@ #include "avcodec.h" #include "mpegvideo.h" #include "h263.h" +#include "h264chroma.h" #include "vc1.h" #include "vc1data.h" #include "vc1acdata.h" @@ -331,6 +332,7 @@ static void vc1_mc_1mv(VC1Context *v, int dir) { MpegEncContext *s = &v->s; DSPContext *dsp = &v->s.dsp; + H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcY, *srcU, *srcV; int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y; int off, off_uv; @@ -519,8 +521,8 @@ static void vc1_mc_1mv(VC1Context *v, int dir) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->put_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); } else { v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); @@ -769,7 +771,7 @@ static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, static void vc1_mc_4mv_chroma(VC1Context *v, int dir) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcU, *srcV; int uvmx, uvmy, uvsrc_x, uvsrc_y; int k, tx = 0, ty = 0; @@ -915,8 +917,8 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->put_h264_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy); } else { v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy); v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy); @@ -928,7 +930,7 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) static void vc1_mc_4mv_chroma4(VC1Context *v) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcU, *srcV; int uvsrc_x, uvsrc_y; int uvmx_field[4], uvmy_field[4]; @@ -1000,8 +1002,8 @@ static void vc1_mc_4mv_chroma4(VC1Context *v) } } if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); - dsp->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + h264chroma->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + h264chroma->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); } else { v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); @@ -1828,6 +1830,7 @@ static void vc1_interp_mc(VC1Context *v) { MpegEncContext *s = &v->s; DSPContext *dsp = &v->s.dsp; + H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcY, *srcU, *srcV; int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y; int off, off_uv; @@ -1957,8 +1960,8 @@ static void vc1_interp_mc(VC1Context *v) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); } else { v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); @@ -5164,6 +5167,7 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) if (ff_vc1_init_common(v) < 0) return -1; + ff_h264chroma_init(&v->h264chroma, 8); ff_vc1dsp_init(&v->vc1dsp); if (avctx->codec_id == AV_CODEC_ID_WMV3 || avctx->codec_id == AV_CODEC_ID_WMV3IMAGE) { diff --git a/libavcodec/vc1dsp.c b/libavcodec/vc1dsp.c index 24ed900..99ede18 100644 --- a/libavcodec/vc1dsp.c +++ b/libavcodec/vc1dsp.c @@ -25,8 +25,9 @@ * */ -#include "vc1dsp.h" #include "libavutil/common.h" +#include "h264chroma.h" +#include "vc1dsp.h" /** Apply overlap transform to horizontal edge diff --git a/libavcodec/vc1dsp.h b/libavcodec/vc1dsp.h index 26b4a06..63256e5 100644 --- a/libavcodec/vc1dsp.h +++ b/libavcodec/vc1dsp.h @@ -29,6 +29,7 @@ #define AVCODEC_VC1DSP_H #include "dsputil.h" +#include "h264chroma.h" typedef struct VC1DSPContext { /* vc1 functions */ diff --git a/libavcodec/vp56.c b/libavcodec/vp56.c index 6421088..dec7869 100644 --- a/libavcodec/vp56.c +++ b/libavcodec/vp56.c @@ -26,7 +26,7 @@ #include "avcodec.h" #include "bytestream.h" #include "internal.h" - +#include "h264chroma.h" #include "vp56.h" #include "vp56data.h" @@ -674,6 +674,7 @@ av_cold void ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha) avctx->pix_fmt = has_alpha ? AV_PIX_FMT_YUVA420P : AV_PIX_FMT_YUV420P; ff_dsputil_init(&s->dsp, avctx); + ff_h264chroma_init(&s->h264chroma, 8); ff_videodsp_init(&s->vdsp, 8); ff_vp3dsp_init(&s->vp3dsp, avctx->flags); ff_vp56dsp_init(&s->vp56dsp, avctx->codec->id); diff --git a/libavcodec/vp56.h b/libavcodec/vp56.h index 61325e4..8c6ba14 100644 --- a/libavcodec/vp56.h +++ b/libavcodec/vp56.h @@ -30,6 +30,7 @@ #include "dsputil.h" #include "get_bits.h" #include "bytestream.h" +#include "h264chroma.h" #include "videodsp.h" #include "vp3dsp.h" #include "vp56dsp.h" @@ -95,6 +96,7 @@ typedef struct VP56Model { struct vp56_context { AVCodecContext *avctx; DSPContext dsp; + H264ChromaContext h264chroma; VideoDSPContext vdsp; VP3DSPContext vp3dsp; VP56DSPContext vp56dsp; diff --git a/libavcodec/vp6.c b/libavcodec/vp6.c index c433acb..87c887b 100644 --- a/libavcodec/vp6.c +++ b/libavcodec/vp6.c @@ -536,8 +536,8 @@ static void vp6_filter_diag2(VP56Context *s, uint8_t *dst, uint8_t *src, int stride, int h_weight, int v_weight) { uint8_t *tmp = s->edge_emu_buffer+16; - s->dsp.put_h264_chroma_pixels_tab[0](tmp, src, stride, 9, h_weight, 0); - s->dsp.put_h264_chroma_pixels_tab[0](dst, tmp, stride, 8, 0, v_weight); + s->h264chroma.put_h264_chroma_pixels_tab[0](tmp, src, stride, 9, h_weight, 0); + s->h264chroma.put_h264_chroma_pixels_tab[0](dst, tmp, stride, 8, 0, v_weight); } static void vp6_filter(VP56Context *s, uint8_t *dst, uint8_t *src, @@ -583,7 +583,7 @@ static void vp6_filter(VP56Context *s, uint8_t *dst, uint8_t *src, } } else { if (!x8 || !y8) { - s->dsp.put_h264_chroma_pixels_tab[0](dst, src+offset1, stride, 8, x8, y8); + s->h264chroma.put_h264_chroma_pixels_tab[0](dst, src + offset1, stride, 8, x8, y8); } else { vp6_filter_diag2(s, dst, src+offset1 + ((mv.x^mv.y)>>31), stride, x8, y8); } diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile index 1feb060..f72765a 100644 --- a/libavcodec/x86/Makefile +++ b/libavcodec/x86/Makefile @@ -43,7 +43,8 @@ YASM-OBJS-$(CONFIG_AC3DSP) += x86/ac3dsp.o YASM-OBJS-$(CONFIG_DCT) += x86/dct32.o YASM-OBJS-$(CONFIG_ENCODERS) += x86/dsputilenc.o YASM-OBJS-$(CONFIG_FFT) += x86/fft.o -YASM-OBJS-$(CONFIG_H264CHROMA) += x86/h264_chromamc.o \ +YASM-OBJS-$(CONFIG_H264CHROMA) += x86/h264chroma_init.o \ + x86/h264_chromamc.o \ x86/h264_chromamc_10bit.o YASM-OBJS-$(CONFIG_H264DSP) += x86/h264_deblock.o \ x86/h264_deblock_10bit.o \ diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 92f532a..3938386 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -1460,49 +1460,6 @@ void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h); -void ff_put_h264_chroma_mc8_rnd_mmx (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc8_rnd_mmxext(uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc8_rnd_3dnow(uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); - -void ff_put_h264_chroma_mc4_mmx (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc4_mmxext (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc4_3dnow (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); - -void ff_put_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); - -void ff_put_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_put_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); - -void ff_avg_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); -void ff_avg_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src, - int stride, int h, int x, int y); - -#define CHROMA_MC(OP, NUM, DEPTH, OPT) \ -void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \ - (uint8_t *dst, uint8_t *src, \ - int stride, int h, int x, int y); - -CHROMA_MC(put, 2, 10, mmxext) -CHROMA_MC(avg, 2, 10, mmxext) -CHROMA_MC(put, 4, 10, mmxext) -CHROMA_MC(avg, 4, 10, mmxext) -CHROMA_MC(put, 8, 10, sse2) -CHROMA_MC(avg, 8, 10, sse2) -CHROMA_MC(put, 8, 10, avx) -CHROMA_MC(avg, 8, 10, avx) - #if HAVE_INLINE_ASM /* CAVS-specific */ @@ -1704,11 +1661,6 @@ static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, #endif /* HAVE_INLINE_ASM */ #if HAVE_YASM - if (!high_bit_depth && CONFIG_H264CHROMA) { - c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_mmx; - c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_mmx; - } - c->vector_clip_int32 = ff_vector_clip_int32_mmx; #endif @@ -1773,19 +1725,6 @@ static av_cold void dsputil_init_mmxext(DSPContext *c, AVCodecContext *avctx, c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_mmxext; } - if (!high_bit_depth && CONFIG_H264CHROMA) { - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext; - c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext; - c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmxext; - c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmxext; - } - if (bit_depth == 10 && CONFIG_H264CHROMA) { - c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmxext; - c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmxext; - c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmxext; - c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext; - } - /* slower than cmov version on AMD */ if (!(mm_flags & AV_CPU_FLAG_3DNOW)) c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmxext; @@ -1838,11 +1777,6 @@ static av_cold void dsputil_init_3dnow(DSPContext *c, AVCodecContext *avctx, c->put_no_rnd_pixels_tab[1][1] = ff_put_no_rnd_pixels8_x2_exact_3dnow; c->put_no_rnd_pixels_tab[1][2] = ff_put_no_rnd_pixels8_y2_exact_3dnow; } - - if (!high_bit_depth && CONFIG_H264CHROMA) { - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_3dnow; - c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_3dnow; - } #endif /* HAVE_YASM */ } @@ -1889,13 +1823,6 @@ static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx, } } - if (bit_depth == 10) { - if (CONFIG_H264CHROMA) { - c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2; - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2; - } - } - c->scalarproduct_int16 = ff_scalarproduct_int16_sse2; c->scalarproduct_and_madd_int16 = ff_scalarproduct_and_madd_int16_sse2; if (mm_flags & AV_CPU_FLAG_ATOM) { @@ -1916,14 +1843,6 @@ static av_cold void dsputil_init_ssse3(DSPContext *c, AVCodecContext *avctx, int mm_flags) { #if HAVE_SSSE3_EXTERNAL - const int high_bit_depth = avctx->bits_per_raw_sample > 8; - - if (!high_bit_depth && CONFIG_H264CHROMA) { - c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_ssse3; - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_ssse3; - c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_ssse3; - c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_ssse3; - } c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_ssse3; if (mm_flags & AV_CPU_FLAG_SSE4) // not really sse4, just slow on Conroe c->add_hfyu_left_prediction = ff_add_hfyu_left_prediction_sse4; @@ -1946,20 +1865,6 @@ static av_cold void dsputil_init_sse4(DSPContext *c, AVCodecContext *avctx, #endif /* HAVE_SSE4_EXTERNAL */ } -static av_cold void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags) -{ -#if HAVE_AVX_EXTERNAL - const int bit_depth = avctx->bits_per_raw_sample; - - if (bit_depth == 10) { - if (CONFIG_H264CHROMA) { - c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx; - c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_avx; - } - } -#endif /* HAVE_AVX_EXTERNAL */ -} - av_cold void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx) { int mm_flags = av_get_cpu_flags(); @@ -1990,9 +1895,6 @@ av_cold void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx) if (mm_flags & AV_CPU_FLAG_SSE4) dsputil_init_sse4(c, avctx, mm_flags); - if (mm_flags & AV_CPU_FLAG_AVX) - dsputil_init_avx(c, avctx, mm_flags); - if (CONFIG_ENCODERS) ff_dsputilenc_init_mmx(c, avctx); } diff --git a/libavcodec/x86/h264chroma_init.c b/libavcodec/x86/h264chroma_init.c new file mode 100644 index 0000000..f1f3a19 --- /dev/null +++ b/libavcodec/x86/h264chroma_init.c @@ -0,0 +1,116 @@ +/* + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdint.h> + +#include "config.h" +#include "libavutil/cpu.h" +#include "libavutil/x86/cpu.h" +#include "libavcodec/h264chroma.h" + +void ff_put_h264_chroma_mc8_rnd_mmx (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc8_rnd_mmxext(uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc8_rnd_3dnow(uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); + +void ff_put_h264_chroma_mc4_mmx (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc4_mmxext (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc4_3dnow (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); + +void ff_put_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc2_mmxext (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); + +void ff_put_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_put_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); + +void ff_avg_h264_chroma_mc8_rnd_ssse3(uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); +void ff_avg_h264_chroma_mc4_ssse3 (uint8_t *dst, uint8_t *src, + int stride, int h, int x, int y); + +#define CHROMA_MC(OP, NUM, DEPTH, OPT) \ +void ff_ ## OP ## _h264_chroma_mc ## NUM ## _ ## DEPTH ## _ ## OPT \ + (uint8_t *dst, uint8_t *src, \ + int stride, int h, int x, int y); + +CHROMA_MC(put, 2, 10, mmxext) +CHROMA_MC(avg, 2, 10, mmxext) +CHROMA_MC(put, 4, 10, mmxext) +CHROMA_MC(avg, 4, 10, mmxext) +CHROMA_MC(put, 8, 10, sse2) +CHROMA_MC(avg, 8, 10, sse2) +CHROMA_MC(put, 8, 10, avx) +CHROMA_MC(avg, 8, 10, avx) + +void ff_h264chroma_init_x86(H264ChromaContext *c, int bit_depth) +{ + int high_bit_depth = bit_depth > 8; + int mm_flags = av_get_cpu_flags(); + + if (EXTERNAL_MMX(mm_flags) && !high_bit_depth) { + c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_mmx; + c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_mmx; + } + + if (EXTERNAL_AMD3DNOW(mm_flags) && !high_bit_depth) { + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_3dnow; + c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_3dnow; + } + + if (EXTERNAL_MMXEXT(mm_flags) && !high_bit_depth) { + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_mmxext; + c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_mmxext; + c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_mmxext; + c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_mmxext; + } + + if (EXTERNAL_MMXEXT(mm_flags) && bit_depth == 10) { + c->put_h264_chroma_pixels_tab[2] = ff_put_h264_chroma_mc2_10_mmxext; + c->avg_h264_chroma_pixels_tab[2] = ff_avg_h264_chroma_mc2_10_mmxext; + c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_10_mmxext; + c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_10_mmxext; + } + + if (EXTERNAL_SSE2(mm_flags) && bit_depth == 10) { + c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_sse2; + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_sse2; + } + + if (EXTERNAL_SSSE3(mm_flags) && !high_bit_depth) { + c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_rnd_ssse3; + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_rnd_ssse3; + c->put_h264_chroma_pixels_tab[1] = ff_put_h264_chroma_mc4_ssse3; + c->avg_h264_chroma_pixels_tab[1] = ff_avg_h264_chroma_mc4_ssse3; + } + + if (EXTERNAL_AVX(mm_flags) && bit_depth == 10) { + // AVX implies !cache64. + // TODO: Port cache(32|64) detection from x264. + c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx; + c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_avx; + } +} |