April 2012 - ffmpeg-cvslog - lists.ffmpeg.org

rv40: change a logical test into a bitwise one.
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Christophe Gisquet <christophe.gisquet(a)gmail.com> | Tue Apr 17 18:44:50 2012 +0000| [0195ab5f5ff216a8797623e62e2703f3df2624c2] | committer: Ronald S. Bultje rv40: change a logical test into a bitwise one. Signed-off-by: Ronald S. Bultje <rsbultje(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=0195ab5f5ff216a87… --- libavcodec/rv34.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c index 4c546c2..a3d3fcd 100644 --- a/libavcodec/rv34.c +++ b/libavcodec/rv34.c @@ -313,7 +313,7 @@ static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rv code = get_vlc2(gb, rvlc->third_pattern[sc].table, 9, 2); decode_subblock(dst + 4*2+2, code, 0, gb, &rvlc->coefficient, q_ac2); } - return has_ac || pattern; + return has_ac | pattern; } /**

1 0

rv40: don't always do the full prev_type search
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Christophe Gisquet <christophe.gisquet(a)gmail.com> | Tue Apr 17 18:44:48 2012 +0000| [34e1b0754638ea6e0356239aad01e74de3a582a5] | committer: Ronald S. Bultje rv40: don't always do the full prev_type search 120->100 cycles. Signed-off-by: Ronald S. Bultje <rsbultje(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=34e1b0754638ea6e0… --- libavcodec/rv40.c | 25 ++++++++++++++----------- 1 files changed, 14 insertions(+), 11 deletions(-) diff --git a/libavcodec/rv40.c b/libavcodec/rv40.c index b65a200..18dd51e 100644 --- a/libavcodec/rv40.c +++ b/libavcodec/rv40.c @@ -228,8 +228,6 @@ static int rv40_decode_mb_info(RV34DecContext *r) int q, i; int prev_type = 0; int mb_pos = s->mb_x + s->mb_y * s->mb_stride; - int blocks[RV34_MB_TYPES] = {0}; - int count = 0; if(!r->s.mb_skip_run) r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1; @@ -237,22 +235,27 @@ static int rv40_decode_mb_info(RV34DecContext *r) if(--r->s.mb_skip_run) return RV34_MB_SKIP; - if(r->avail_cache[6-1]) - blocks[r->mb_type[mb_pos - 1]]++; if(r->avail_cache[6-4]){ + int blocks[RV34_MB_TYPES] = {0}; + int count = 0; + if(r->avail_cache[6-1]) + blocks[r->mb_type[mb_pos - 1]]++; blocks[r->mb_type[mb_pos - s->mb_stride]]++; if(r->avail_cache[6-2]) blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++; if(r->avail_cache[6-5]) blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++; - } - - for(i = 0; i < RV34_MB_TYPES; i++){ - if(blocks[i] > count){ - count = blocks[i]; - prev_type = i; + for(i = 0; i < RV34_MB_TYPES; i++){ + if(blocks[i] > count){ + count = blocks[i]; + prev_type = i; + if(count>1) + break; + } } - } + } else if (r->avail_cache[6-1]) + prev_type = r->mb_type[mb_pos - 1]; + if(s->pict_type == AV_PICTURE_TYPE_P){ prev_type = block_num_to_ptype_vlc_num[prev_type]; q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);

1 0

rv34: remove constant parameter
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Christophe Gisquet <christophe.gisquet(a)gmail.com> | Wed Apr 4 10:22:28 2012 +0200| [88027999919332b2e6b27f357b3eb12d7e0abdad] | committer: Ronald S. Bultje rv34: remove constant parameter is_block2 was always 0, so just remove it, and change accordingly the code. Signed-off-by: Ronald S. Bultje <rsbultje(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=88027999919332b2e… --- libavcodec/rv34.c | 17 ++++++----------- 1 files changed, 6 insertions(+), 11 deletions(-) diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c index 1247569..4c546c2 100644 --- a/libavcodec/rv34.c +++ b/libavcodec/rv34.c @@ -260,20 +260,15 @@ static inline void decode_subblock1(DCTELEM *dst, int code, GetBitContext *gb, V decode_coeff(dst, coeff, 3, gb, vlc, q); } -static inline void decode_subblock3(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc, +static inline void decode_subblock3(DCTELEM *dst, int code, GetBitContext *gb, VLC *vlc, int q_dc, int q_ac1, int q_ac2) { int flags = modulo_three_table[code]; - decode_coeff( dst+0*4+0, (flags >> 6) , 3, gb, vlc, q_dc); - if(is_block2){ - decode_coeff(dst+1*4+0, (flags >> 4) & 3, 2, gb, vlc, q_ac1); - decode_coeff(dst+0*4+1, (flags >> 2) & 3, 2, gb, vlc, q_ac1); - }else{ - decode_coeff(dst+0*4+1, (flags >> 4) & 3, 2, gb, vlc, q_ac1); - decode_coeff(dst+1*4+0, (flags >> 2) & 3, 2, gb, vlc, q_ac1); - } - decode_coeff( dst+1*4+1, (flags >> 0) & 3, 2, gb, vlc, q_ac2); + decode_coeff(dst+0*4+0, (flags >> 6) , 3, gb, vlc, q_dc); + decode_coeff(dst+0*4+1, (flags >> 4) & 3, 2, gb, vlc, q_ac1); + decode_coeff(dst+1*4+0, (flags >> 2) & 3, 2, gb, vlc, q_ac1); + decode_coeff(dst+1*4+1, (flags >> 0) & 3, 2, gb, vlc, q_ac2); } /** @@ -298,7 +293,7 @@ static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rv code >>= 3; if (modulo_three_table[code] & 0x3F) { - decode_subblock3(dst, code, 0, gb, &rvlc->coefficient, q_dc, q_ac1, q_ac2); + decode_subblock3(dst, code, gb, &rvlc->coefficient, q_dc, q_ac1, q_ac2); } else { decode_subblock1(dst, code, gb, &rvlc->coefficient, q_dc); if (!pattern)

1 0

rv34dsp x86: implement MMX2 inverse transform
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Christophe Gisquet <christophe.gisquet(a)gmail.com> | Thu Mar 29 22:37:41 2012 +0200| [fe5ed69dc7f6acf6928c27d614f2be3fc89ed176] | committer: Ronald S. Bultje rv34dsp x86: implement MMX2 inverse transform 141 cycles down to 51. Signed-off-by: Ronald S. Bultje <rsbultje(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fe5ed69dc7f6acf69… --- libavcodec/x86/rv34dsp.asm | 78 +++++++++++++++++++++++++++++++++++++++++ libavcodec/x86/rv34dsp_init.c | 2 + 2 files changed, 80 insertions(+), 0 deletions(-) diff --git a/libavcodec/x86/rv34dsp.asm b/libavcodec/x86/rv34dsp.asm index 2d2f6e1..32bcdce 100644 --- a/libavcodec/x86/rv34dsp.asm +++ b/libavcodec/x86/rv34dsp.asm @@ -22,6 +22,16 @@ %include "x86inc.asm" %include "x86util.asm" +SECTION_RODATA +pw_row_coeffs: times 4 dw 13 + times 4 dw 17 + times 4 dw 7 +pd_512: times 2 dd 0x200 +pw_col_coeffs: dw 13, 13, 13, -13 + dw 17, 7, 7, -17 + dw 13, -13, 13, 13 + dw -7, 17, -17, -7 + SECTION .text %macro IDCT_DC_NOROUND 1 @@ -88,6 +98,74 @@ cglobal rv34_idct_dc_add_mmx, 3, 3 movh [r2+r1], m5 RET +; Load coeffs and perform row transform +; Output: coeffs in mm[0467], rounder in mm5 +%macro ROW_TRANSFORM 1 + pxor mm7, mm7 + mova mm0, [%1+ 0*8] + mova mm1, [%1+ 1*8] + mova mm2, [%1+ 2*8] + mova mm3, [%1+ 3*8] + mova [%1+ 0*8], mm7 + mova [%1+ 1*8], mm7 + mova [%1+ 2*8], mm7 + mova [%1+ 3*8], mm7 + mova mm4, mm0 + mova mm6, [pw_row_coeffs+ 0] + paddsw mm0, mm2 ; b0 + b2 + psubsw mm4, mm2 ; b0 - b2 + pmullw mm0, mm6 ; *13 = z0 + pmullw mm4, mm6 ; *13 = z1 + mova mm5, mm1 + pmullw mm1, [pw_row_coeffs+ 8] ; b1*17 + pmullw mm5, [pw_row_coeffs+16] ; b1* 7 + mova mm7, mm3 + pmullw mm3, [pw_row_coeffs+ 8] ; b3*17 + pmullw mm7, [pw_row_coeffs+16] ; b3* 7 + paddsw mm1, mm7 ; z3 = b1*17 + b3* 7 + psubsw mm5, mm3 ; z2 = b1* 7 - b3*17 + mova mm7, mm0 + mova mm6, mm4 + paddsw mm0, mm1 ; z0 + z3 + psubsw mm7, mm1 ; z0 - z3 + paddsw mm4, mm5 ; z1 + z2 + psubsw mm6, mm5 ; z1 - z2 + mova mm5, [pd_512] ; 0x200 +%endmacro + +; ff_rv34_idct_add_mmx2(uint8_t *dst, ptrdiff_t stride, DCTELEM *block); +%macro COL_TRANSFORM 4 + pshufw mm3, %2, 0xDD ; col. 1,3,1,3 + pshufw %2, %2, 0x88 ; col. 0,2,0,2 + pmaddwd %2, %3 ; 13*c0+13*c2 | 13*c0-13*c2 = z0 | z1 + pmaddwd mm3, %4 ; 17*c1+ 7*c3 | 7*c1-17*c3 = z3 | z2 + paddd %2, mm5 + pshufw mm1, %2, 01001110b ; z1 | z0 + pshufw mm2, mm3, 01001110b ; z2 | z3 + paddd %2, mm3 ; z0+z3 | z1+z2 + psubd mm1, mm2 ; z1-z2 | z0-z3 + movd mm3, %1 + psrad %2, 10 + pxor mm2, mm2 + psrad mm1, 10 + punpcklbw mm3, mm2 + packssdw %2, mm1 + paddw %2, mm3 + packuswb %2, %2 + movd %1, %2 +%endmacro +INIT_MMX mmx2 +cglobal rv34_idct_add, 3,3,0, d, s, b + ROW_TRANSFORM bq + COL_TRANSFORM [dq], mm0, [pw_col_coeffs+ 0], [pw_col_coeffs+ 8] + mova mm0, [pw_col_coeffs+ 0] + COL_TRANSFORM [dq+sq], mm4, mm0, [pw_col_coeffs+ 8] + mova mm4, [pw_col_coeffs+ 8] + lea dq, [dq + 2*sq] + COL_TRANSFORM [dq], mm6, mm0, mm4 + COL_TRANSFORM [dq+sq], mm7, mm0, mm4 + ret + ; ff_rv34_idct_dc_add_sse4(uint8_t *dst, int stride, int dc); INIT_XMM cglobal rv34_idct_dc_add_sse4, 3, 3, 6 diff --git a/libavcodec/x86/rv34dsp_init.c b/libavcodec/x86/rv34dsp_init.c index 3883125..d91818c 100644 --- a/libavcodec/x86/rv34dsp_init.c +++ b/libavcodec/x86/rv34dsp_init.c @@ -28,6 +28,7 @@ void ff_rv34_idct_dc_mmx2(DCTELEM *block); void ff_rv34_idct_dc_noround_mmx2(DCTELEM *block); void ff_rv34_idct_dc_add_mmx(uint8_t *dst, ptrdiff_t stride, int dc); void ff_rv34_idct_dc_add_sse4(uint8_t *dst, ptrdiff_t stride, int dc); +void ff_rv34_idct_add_mmx2(uint8_t *dst, ptrdiff_t stride, DCTELEM *block); av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp) { @@ -38,6 +39,7 @@ av_cold void ff_rv34dsp_init_x86(RV34DSPContext* c, DSPContext *dsp) c->rv34_idct_dc_add = ff_rv34_idct_dc_add_mmx; if (mm_flags & AV_CPU_FLAG_MMX2) { c->rv34_inv_transform_dc = ff_rv34_idct_dc_noround_mmx2; + c->rv34_idct_add = ff_rv34_idct_add_mmx2; } if (mm_flags & AV_CPU_FLAG_SSE4) c->rv34_idct_dc_add = ff_rv34_idct_dc_add_sse4;

1 0

dsputil x86: revert a test back to its previous value
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Christophe GISQUET <christophe.gisquet(a)gmail.com> | Wed Mar 28 23:22:22 2012 +0200| [e75d1d4f7321b097f4b09799bf28a22ad259fea6] | committer: Ronald S. Bultje dsputil x86: revert a test back to its previous value Commit 356ee8d caused the initial inversion. Signed-off-by: Ronald S. Bultje <rsbultje(a)gmail.com> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e75d1d4f7321b097f… --- libavcodec/x86/dsputil_mmx.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 89be328..3ef19c5 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -2961,7 +2961,7 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx, const int bit_depth = avctx->bits_per_raw_sample; const int high_bit_depth = bit_depth > 8; - if (mm_flags & AV_CPU_FLAG_3DNOW) { + if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) { // these functions are slower than mmx on AMD, but faster on Intel if (!high_bit_depth) { c->put_pixels_tab[0][0] = put_pixels16_sse2;

1 0

lavfi/buffersink: fix header inclusion guard name
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Stefano Sabatini <stefasab(a)gmail.com> | Sun Apr 29 18:43:32 2012 +0200| [46eba43e0ef28d9d08ed41774e6bd1ce94388c35] | committer: Stefano Sabatini lavfi/buffersink: fix header inclusion guard name > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=46eba43e0ef28d9d0… --- libavfilter/buffersink.h | 6 +++--- 1 files changed, 3 insertions(+), 3 deletions(-) diff --git a/libavfilter/buffersink.h b/libavfilter/buffersink.h index cf6bd69..4c13f0d 100644 --- a/libavfilter/buffersink.h +++ b/libavfilter/buffersink.h @@ -16,8 +16,8 @@ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ -#ifndef AVFILTER_VSINK_BUFFER_H -#define AVFILTER_VSINK_BUFFER_H +#ifndef AVFILTER_BUFFERSINK_H +#define AVFILTER_BUFFERSINK_H /** * @file @@ -98,4 +98,4 @@ int av_vsink_buffer_get_video_buffer_ref(AVFilterContext *buffer_sink, AVFilterBufferRef **picref, int flags); #endif -#endif /* AVFILTER_VSINK_BUFFER_H */ +#endif /* AVFILTER_BUFFERSINK_H */

1 0

Simplify camstudio decoding and fix odd 16bit decoding.
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Carl Eugen Hoyos <cehoyos(a)ag.or.at> | Sun Apr 29 18:35:39 2012 +0200| [3dafde0b961023692b4c295ffba2061a84af13af] | committer: Carl Eugen Hoyos Simplify camstudio decoding and fix odd 16bit decoding. Fixes ticket #1220. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3dafde0b961023692… --- libavcodec/cscd.c | 113 ++++---------------------------------------------- libavcodec/version.h | 2 +- 2 files changed, 10 insertions(+), 105 deletions(-) diff --git a/libavcodec/cscd.c b/libavcodec/cscd.c index 032b58a..01abeda 100644 --- a/libavcodec/cscd.c +++ b/libavcodec/cscd.c @@ -35,9 +35,9 @@ typedef struct { unsigned char* decomp_buf; } CamStudioContext; -static void copy_frame_default(AVFrame *f, const uint8_t *src, int src_stride, +static void copy_frame_default(AVFrame *f, const uint8_t *src, int linelen, int height) { - int i; + int i, src_stride = FFALIGN(linelen, 4); uint8_t *dst = f->data[0]; dst += (height - 1) * f->linesize[0]; for (i = height; i; i--) { @@ -47,9 +47,9 @@ static void copy_frame_default(AVFrame *f, const uint8_t *src, int src_stride, } } -static void add_frame_default(AVFrame *f, const uint8_t *src, int src_stride, +static void add_frame_default(AVFrame *f, const uint8_t *src, int linelen, int height) { - int i, j; + int i, j, src_stride = FFALIGN(linelen, 4); uint8_t *dst = f->data[0]; dst += (height - 1) * f->linesize[0]; for (i = height; i; i--) { @@ -60,81 +60,6 @@ static void add_frame_default(AVFrame *f, const uint8_t *src, int src_stride, } } -#if !HAVE_BIGENDIAN -#define copy_frame_16(f, s, l, h) copy_frame_default(f, s, l, l, h) -#define copy_frame_32(f, s, l, h) copy_frame_default(f, s, l, l, h) -#define add_frame_16(f, s, l, h) add_frame_default(f, s, l, l, h) -#define add_frame_32(f, s, l, h) add_frame_default(f, s, l, l, h) -#else -static void copy_frame_16(AVFrame *f, const uint8_t *src, - int linelen, int height) { - int i, j; - uint8_t *dst = f->data[0]; - dst += (height - 1) * f->linesize[0]; - for (i = height; i; i--) { - for (j = linelen / 2; j; j--) { - dst[0] = src[1]; - dst[1] = src[0]; - src += 2; - dst += 2; - } - dst -= f->linesize[0] + linelen; - } -} - -static void copy_frame_32(AVFrame *f, const uint8_t *src, - int linelen, int height) { - int i, j; - uint8_t *dst = f->data[0]; - dst += (height - 1) * f->linesize[0]; - for (i = height; i; i--) { - for (j = linelen / 4; j; j--) { - dst[0] = src[3]; - dst[1] = src[2]; - dst[2] = src[1]; - dst[3] = src[0]; - src += 4; - dst += 4; - } - dst -= f->linesize[0] + linelen; - } -} - -static void add_frame_16(AVFrame *f, const uint8_t *src, - int linelen, int height) { - int i, j; - uint8_t *dst = f->data[0]; - dst += (height - 1) * f->linesize[0]; - for (i = height; i; i--) { - for (j = linelen / 2; j; j--) { - dst[0] += src[1]; - dst[1] += src[0]; - src += 2; - dst += 2; - } - dst -= f->linesize[0] + linelen; - } -} - -static void add_frame_32(AVFrame *f, const uint8_t *src, - int linelen, int height) { - int i, j; - uint8_t *dst = f->data[0]; - dst += (height - 1) * f->linesize[0]; - for (i = height; i; i--) { - for (j = linelen / 4; j; j--) { - dst[0] += src[3]; - dst[1] += src[2]; - dst[2] += src[1]; - dst[3] += src[0]; - src += 4; - dst += 4; - } - dst -= f->linesize[0] + linelen; - } -} -#endif - static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; @@ -185,31 +110,13 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, if (buf[0] & 1) { // keyframe c->pic.pict_type = AV_PICTURE_TYPE_I; c->pic.key_frame = 1; - switch (c->bpp) { - case 16: - copy_frame_16(&c->pic, c->decomp_buf, c->linelen, c->height); - break; - case 32: - copy_frame_32(&c->pic, c->decomp_buf, c->linelen, c->height); - break; - default: - copy_frame_default(&c->pic, c->decomp_buf, FFALIGN(c->linelen, 4), + copy_frame_default(&c->pic, c->decomp_buf, c->linelen, c->height); - } } else { c->pic.pict_type = AV_PICTURE_TYPE_P; c->pic.key_frame = 0; - switch (c->bpp) { - case 16: - add_frame_16(&c->pic, c->decomp_buf, c->linelen, c->height); - break; - case 32: - add_frame_32(&c->pic, c->decomp_buf, c->linelen, c->height); - break; - default: - add_frame_default(&c->pic, c->decomp_buf, FFALIGN(c->linelen, 4), + add_frame_default(&c->pic, c->decomp_buf, c->linelen, c->height); - } } *picture = c->pic; @@ -221,9 +128,9 @@ static av_cold int decode_init(AVCodecContext *avctx) { CamStudioContext *c = avctx->priv_data; int stride; switch (avctx->bits_per_coded_sample) { - case 16: avctx->pix_fmt = PIX_FMT_RGB555; break; + case 16: avctx->pix_fmt = PIX_FMT_RGB555LE; break; case 24: avctx->pix_fmt = PIX_FMT_BGR24; break; - case 32: avctx->pix_fmt = PIX_FMT_RGB32; break; + case 32: avctx->pix_fmt = PIX_FMT_BGRA; break; default: av_log(avctx, AV_LOG_ERROR, "CamStudio codec error: invalid depth %i bpp\n", @@ -235,9 +142,7 @@ static av_cold int decode_init(AVCodecContext *avctx) { c->pic.data[0] = NULL; c->linelen = avctx->width * avctx->bits_per_coded_sample / 8; c->height = avctx->height; - stride = c->linelen; - if (avctx->bits_per_coded_sample == 24) - stride = FFALIGN(stride, 4); + stride = FFALIGN(c->linelen, 4); c->decomp_size = c->height * stride; c->decomp_buf = av_malloc(c->decomp_size + AV_LZO_OUTPUT_PADDING); if (!c->decomp_buf) { diff --git a/libavcodec/version.h b/libavcodec/version.h index 24e6837..23ce1db 100644 --- a/libavcodec/version.h +++ b/libavcodec/version.h @@ -28,7 +28,7 @@ #define LIBAVCODEC_VERSION_MAJOR 54 #define LIBAVCODEC_VERSION_MINOR 17 -#define LIBAVCODEC_VERSION_MICRO 100 +#define LIBAVCODEC_VERSION_MICRO 101 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ LIBAVCODEC_VERSION_MINOR, \

1 0

swr: use memcpy when input and output match
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Michael Niedermayer <michaelni(a)gmx.at> | Sun Apr 29 15:28:16 2012 +0200| [9f98a8e9bc9b669f348834908dce927a2a5d9b90] | committer: Michael Niedermayer swr: use memcpy when input and output match Signed-off-by: Michael Niedermayer <michaelni(a)gmx.at> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=9f98a8e9bc9b669f3… --- libswresample/audioconvert.c | 7 +++++++ 1 files changed, 7 insertions(+), 0 deletions(-) diff --git a/libswresample/audioconvert.c b/libswresample/audioconvert.c index 529c911..1cdd80f 100644 --- a/libswresample/audioconvert.c +++ b/libswresample/audioconvert.c @@ -107,6 +107,10 @@ static conv_func_type * const fmt_pair_to_conv_functions[AV_SAMPLE_FMT_NB*AV_SAM FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBL), }; +static void cpy(uint8_t **dst, const uint8_t **src, int len){ + memcpy(*dst, *src, len); +} + AudioConvert *swri_audio_convert_alloc(enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, const int *ch_map, @@ -126,6 +130,9 @@ AudioConvert *swri_audio_convert_alloc(enum AVSampleFormat out_fmt, if (in_fmt == AV_SAMPLE_FMT_U8) memset(ctx->silence, 0x80, sizeof(ctx->silence)); + if(out_fmt == in_fmt && !ch_map) + ctx->simd_f = cpy; + if(HAVE_YASM && HAVE_MMX) swri_audio_convert_init_x86(ctx, out_fmt, in_fmt, channels); return ctx;

1 0

swr: add set_audiodata_fmt() and use it to simplify code
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Michael Niedermayer <michaelni(a)gmx.at> | Sun Apr 29 15:29:28 2012 +0200| [2d6c29f5663afecec250f81a74c612bbadee1c71] | committer: Michael Niedermayer swr: add set_audiodata_fmt() and use it to simplify code Signed-off-by: Michael Niedermayer <michaelni(a)gmx.at> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2d6c29f5663afecec… --- libswresample/swresample.c | 21 +++++++++++---------- libswresample/swresample_internal.h | 1 - 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/libswresample/swresample.c b/libswresample/swresample.c index 8e3d3f5..9297a44 100644 --- a/libswresample/swresample.c +++ b/libswresample/swresample.c @@ -153,6 +153,10 @@ struct SwrContext *swr_alloc_set_opts(struct SwrContext *s, return s; } +static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){ + a->bps = av_get_bytes_per_sample(fmt); + a->planar= av_sample_fmt_is_planar(fmt); +} static void free_temp(AudioData *a){ av_free(a->data); @@ -191,9 +195,6 @@ int swr_init(struct SwrContext *s){ s->flushed = 0; - s-> in.planar= av_sample_fmt_is_planar(s-> in_sample_fmt); - s->out.planar= av_sample_fmt_is_planar(s->out_sample_fmt); - if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){ av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt); return AVERROR(EINVAL); @@ -216,6 +217,9 @@ int swr_init(struct SwrContext *s){ return AVERROR(EINVAL); } + set_audiodata_fmt(&s-> in, s-> in_sample_fmt); + set_audiodata_fmt(&s->out, s->out_sample_fmt); + if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){ s->resample = swri_resample_init(s->resample, s->out_sample_rate, s->in_sample_rate, s->filter_size, s->phase_shift, s->linear_interp, s->cutoff, s->int_sample_fmt); }else @@ -267,9 +271,6 @@ av_assert0(s->used_ch_count); av_assert0(s->out.ch_count); s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0; - s-> in.bps= av_get_bytes_per_sample(s-> in_sample_fmt); - s->int_bps= av_get_bytes_per_sample(s->int_sample_fmt); - s->out.bps= av_get_bytes_per_sample(s->out_sample_fmt); s->in_buffer= s->in; if(!s->resample && !s->rematrix && !s->channel_map){ @@ -300,12 +301,12 @@ av_assert0(s->out.ch_count); s->in_buffer.ch_count = s->out.ch_count; } - s->postin.bps = s->midbuf.bps = s->preout.bps = s->int_bps; - s->postin.planar = s->midbuf.planar = s->preout.planar = 1; + set_audiodata_fmt(&s->postin, s->int_sample_fmt); + set_audiodata_fmt(&s->midbuf, s->int_sample_fmt); + set_audiodata_fmt(&s->preout, s->int_sample_fmt); if(s->resample){ - s->in_buffer.bps = s->int_bps; - s->in_buffer.planar = 1; + set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt); } s->dither = s->preout; diff --git a/libswresample/swresample_internal.h b/libswresample/swresample_internal.h index 293075b..4f2e9ee 100644 --- a/libswresample/swresample_internal.h +++ b/libswresample/swresample_internal.h @@ -57,7 +57,6 @@ struct SwrContext { int linear_interp; /**< if 1 then the resampling FIR filter will be linearly interpolated */ double cutoff; /**< resampling cutoff frequency. 1.0 corresponds to half the output sample rate */ - int int_bps; ///< internal bytes per sample int resample_first; ///< 1 if resampling must come first, 0 if rematrixing int rematrix; ///< flag to indicate if rematrixing is needed (basically if input and output layouts mismatch) int rematrix_custom; ///< flag to indicate that a custom matrix has been defined

1 0

swr: add sample format to AudioData
by git＠videolan.org 29 Apr '12

29 Apr '12

ffmpeg | branch: master | Michael Niedermayer <michaelni(a)gmx.at> | Sun Apr 29 15:30:07 2012 +0200| [106789dfa0cb19142cf8d8dfd38bd006b393a3c2] | committer: Michael Niedermayer swr: add sample format to AudioData Signed-off-by: Michael Niedermayer <michaelni(a)gmx.at> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=106789dfa0cb19142… --- libswresample/swresample.c | 1 + libswresample/swresample_internal.h | 1 + 2 files changed, 2 insertions(+), 0 deletions(-) diff --git a/libswresample/swresample.c b/libswresample/swresample.c index 9297a44..7f545e0 100644 --- a/libswresample/swresample.c +++ b/libswresample/swresample.c @@ -154,6 +154,7 @@ struct SwrContext *swr_alloc_set_opts(struct SwrContext *s, } static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){ + a->fmt = fmt; a->bps = av_get_bytes_per_sample(fmt); a->planar= av_sample_fmt_is_planar(fmt); } diff --git a/libswresample/swresample_internal.h b/libswresample/swresample_internal.h index 4f2e9ee..bd46976 100644 --- a/libswresample/swresample_internal.h +++ b/libswresample/swresample_internal.h @@ -30,6 +30,7 @@ typedef struct AudioData{ int bps; ///< bytes per sample int count; ///< number of samples int planar; ///< 1 if planar audio, 0 otherwise + enum AVSampleFormat fmt; ///< sample format } AudioData; struct SwrContext {

1 0