diff options
Diffstat (limited to 'libavcodec/dsputil.c')
| -rw-r--r-- | libavcodec/dsputil.c | 274 |
1 files changed, 225 insertions, 49 deletions
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c index 0ef9d8c8f7..db9f345b80 100644 --- a/libavcodec/dsputil.c +++ b/libavcodec/dsputil.c @@ -5,20 +5,20 @@ * * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at> * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * FFmpeg 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, + * FFmpeg 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 + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -29,6 +29,7 @@ #include "libavutil/attributes.h" #include "libavutil/imgutils.h" +#include "libavutil/internal.h" #include "avcodec.h" #include "copy_block.h" #include "dct.h" @@ -281,7 +282,7 @@ static int sse16_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, return s; } -static void diff_pixels_c(int16_t *restrict block, const uint8_t *s1, +static void diff_pixels_c(int16_t *av_restrict block, const uint8_t *s1, const uint8_t *s2, int stride) { int i; @@ -302,7 +303,7 @@ static void diff_pixels_c(int16_t *restrict block, const uint8_t *s1, } } -static void put_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, +static void put_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels, int line_size) { int i; @@ -323,8 +324,40 @@ static void put_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, } } +static void put_pixels_clamped4_c(const int16_t *block, uint8_t *av_restrict pixels, + int line_size) +{ + int i; + + /* read the pixels */ + for(i=0;i<4;i++) { + pixels[0] = av_clip_uint8(block[0]); + pixels[1] = av_clip_uint8(block[1]); + pixels[2] = av_clip_uint8(block[2]); + pixels[3] = av_clip_uint8(block[3]); + + pixels += line_size; + block += 8; + } +} + +static void put_pixels_clamped2_c(const int16_t *block, uint8_t *av_restrict pixels, + int line_size) +{ + int i; + + /* read the pixels */ + for(i=0;i<2;i++) { + pixels[0] = av_clip_uint8(block[0]); + pixels[1] = av_clip_uint8(block[1]); + + pixels += line_size; + block += 8; + } +} + static void put_signed_pixels_clamped_c(const int16_t *block, - uint8_t *restrict pixels, + uint8_t *av_restrict pixels, int line_size) { int i, j; @@ -344,7 +377,7 @@ static void put_signed_pixels_clamped_c(const int16_t *block, } } -static void add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, +static void add_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels, int line_size) { int i; @@ -364,6 +397,36 @@ static void add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, } } +static void add_pixels_clamped4_c(const int16_t *block, uint8_t *av_restrict pixels, + int line_size) +{ + int i; + + /* read the pixels */ + for(i=0;i<4;i++) { + pixels[0] = av_clip_uint8(pixels[0] + block[0]); + pixels[1] = av_clip_uint8(pixels[1] + block[1]); + pixels[2] = av_clip_uint8(pixels[2] + block[2]); + pixels[3] = av_clip_uint8(pixels[3] + block[3]); + pixels += line_size; + block += 8; + } +} + +static void add_pixels_clamped2_c(const int16_t *block, uint8_t *av_restrict pixels, + int line_size) +{ + int i; + + /* read the pixels */ + for(i=0;i<2;i++) { + pixels[0] = av_clip_uint8(pixels[0] + block[0]); + pixels[1] = av_clip_uint8(pixels[1] + block[1]); + pixels += line_size; + block += 8; + } +} + static int sum_abs_dctelem_c(int16_t *block) { int sum = 0, i; @@ -742,7 +805,7 @@ static int try_8x8basis_c(int16_t rem[64], int16_t weight[64], (BASIS_SHIFT - RECON_SHIFT)); int w = weight[i]; b >>= RECON_SHIFT; - assert(-512 < b && b < 512); + av_assert2(-512 < b && b < 512); sum += (w * b) * (w * b) >> 4; } @@ -812,6 +875,14 @@ void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type) case FF_CMP_NSSE: cmp[i] = c->nsse[i]; break; +#if CONFIG_DWT + case FF_CMP_W53: + cmp[i]= c->w53[i]; + break; + case FF_CMP_W97: + cmp[i]= c->w97[i]; + break; +#endif default: av_log(NULL, AV_LOG_ERROR, "internal error in cmp function selection\n"); @@ -839,7 +910,7 @@ static int hadamard8_diff8x8_c(MpegEncContext *s, uint8_t *dst, { int i, temp[64], sum = 0; - assert(h == 8); + av_assert2(h == 8); for (i = 0; i < 8; i++) { // FIXME: try pointer walks @@ -891,7 +962,7 @@ static int hadamard8_intra8x8_c(MpegEncContext *s, uint8_t *src, { int i, temp[64], sum = 0; - assert(h == 8); + av_assert2(h == 8); for (i = 0; i < 8; i++) { // FIXME: try pointer walks @@ -943,7 +1014,7 @@ static int dct_sad8x8_c(MpegEncContext *s, uint8_t *src1, { LOCAL_ALIGNED_16(int16_t, temp, [64]); - assert(h == 8); + av_assert2(h == 8); s->dsp.diff_pixels(temp, src1, src2, stride); s->dsp.fdct(temp); @@ -1010,7 +1081,7 @@ static int dct_max8x8_c(MpegEncContext *s, uint8_t *src1, LOCAL_ALIGNED_16(int16_t, temp, [64]); int sum = 0, i; - assert(h == 8); + av_assert2(h == 8); s->dsp.diff_pixels(temp, src1, src2, stride); s->dsp.fdct(temp); @@ -1028,7 +1099,7 @@ static int quant_psnr8x8_c(MpegEncContext *s, uint8_t *src1, int16_t *const bak = temp + 64; int sum = 0, i; - assert(h == 8); + av_assert2(h == 8); s->mb_intra = 0; s->dsp.diff_pixels(temp, src1, src2, stride); @@ -1057,7 +1128,7 @@ static int rd8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2, const int esc_length = s->ac_esc_length; uint8_t *length, *last_length; - assert(h == 8); + av_assert2(h == 8); copy_block8(lsrc1, src1, 8, stride, 8); copy_block8(lsrc2, src2, 8, stride, 8); @@ -1101,7 +1172,7 @@ static int rd8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2, level = temp[i] + 64; - assert(level - 64); + av_assert2(level - 64); if ((level & (~127)) == 0) { bits += last_length[UNI_AC_ENC_INDEX(run, level)]; @@ -1132,7 +1203,7 @@ static int bit8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2, const int esc_length = s->ac_esc_length; uint8_t *length, *last_length; - assert(h == 8); + av_assert2(h == 8); s->dsp.diff_pixels(temp, src1, src2, stride); @@ -1173,7 +1244,7 @@ static int bit8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2, level = temp[i] + 64; - assert(level - 64); + av_assert2(level - 64); if ((level & (~127)) == 0) bits += last_length[UNI_AC_ENC_INDEX(run, level)]; @@ -1206,20 +1277,24 @@ static int vsad_intra ## size ## _c(MpegEncContext *c, \ VSAD_INTRA(8) VSAD_INTRA(16) -static int vsad16_c(MpegEncContext *c, uint8_t *s1, uint8_t *s2, - int stride, int h) -{ - int score = 0, x, y; - - for (y = 1; y < h; y++) { - for (x = 0; x < 16; x++) - score += FFABS(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]); - s1 += stride; - s2 += stride; - } - - return score; +#define VSAD(size) \ +static int vsad ## size ## _c(MpegEncContext *c, \ + uint8_t *s1, uint8_t *s2, \ + int stride, int h) \ +{ \ + int score = 0, x, y; \ + \ + for (y = 1; y < h; y++) { \ + for (x = 0; x < size; x++) \ + score += FFABS(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]); \ + s1 += stride; \ + s2 += stride; \ + } \ + \ + return score; \ } +VSAD(8) +VSAD(16) #define SQ(a) ((a) * (a)) #define VSSE_INTRA(size) \ @@ -1244,20 +1319,23 @@ static int vsse_intra ## size ## _c(MpegEncContext *c, \ VSSE_INTRA(8) VSSE_INTRA(16) -static int vsse16_c(MpegEncContext *c, uint8_t *s1, uint8_t *s2, - int stride, int h) -{ - int score = 0, x, y; - - for (y = 1; y < h; y++) { - for (x = 0; x < 16; x++) - score += SQ(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]); - s1 += stride; - s2 += stride; - } - - return score; +#define VSSE(size) \ +static int vsse ## size ## _c(MpegEncContext *c, uint8_t *s1, uint8_t *s2, \ + int stride, int h) \ +{ \ + int score = 0, x, y; \ + \ + for (y = 1; y < h; y++) { \ + for (x = 0; x < size; x++) \ + score += SQ(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]); \ + s1 += stride; \ + s2 += stride; \ + } \ + \ + return score; \ } +VSSE(8) +VSSE(16) #define WRAPPER8_16_SQ(name8, name16) \ static int name16(MpegEncContext *s, uint8_t *dst, uint8_t *src, \ @@ -1380,6 +1458,37 @@ static void jref_idct_add(uint8_t *dest, int line_size, int16_t *block) add_pixels_clamped_c(block, dest, line_size); } +static void ff_jref_idct4_put(uint8_t *dest, int line_size, int16_t *block) +{ + ff_j_rev_dct4 (block); + put_pixels_clamped4_c(block, dest, line_size); +} +static void ff_jref_idct4_add(uint8_t *dest, int line_size, int16_t *block) +{ + ff_j_rev_dct4 (block); + add_pixels_clamped4_c(block, dest, line_size); +} + +static void ff_jref_idct2_put(uint8_t *dest, int line_size, int16_t *block) +{ + ff_j_rev_dct2 (block); + put_pixels_clamped2_c(block, dest, line_size); +} +static void ff_jref_idct2_add(uint8_t *dest, int line_size, int16_t *block) +{ + ff_j_rev_dct2 (block); + add_pixels_clamped2_c(block, dest, line_size); +} + +static void ff_jref_idct1_put(uint8_t *dest, int line_size, int16_t *block) +{ + dest[0] = av_clip_uint8((block[0] + 4)>>3); +} +static void ff_jref_idct1_add(uint8_t *dest, int line_size, int16_t *block) +{ + dest[0] = av_clip_uint8(dest[0] + ((block[0] + 4)>>3)); +} + /* draw the edges of width 'w' of an image of size width, height */ // FIXME: Check that this is OK for MPEG-4 interlaced. static void draw_edges_8_c(uint8_t *buf, int wrap, int width, int height, @@ -1427,10 +1536,33 @@ av_cold void ff_dsputil_static_init(void) ff_square_tab[i] = (i - 256) * (i - 256); } +int ff_check_alignment(void) +{ + static int did_fail = 0; + LOCAL_ALIGNED_16(int, aligned, [4]); + + if ((intptr_t)aligned & 15) { + if (!did_fail) { +#if HAVE_MMX || HAVE_ALTIVEC + av_log(NULL, AV_LOG_ERROR, + "Compiler did not align stack variables. Libavcodec has been miscompiled\n" + "and may be very slow or crash. This is not a bug in libavcodec,\n" + "but in the compiler. You may try recompiling using gcc >= 4.2.\n" + "Do not report crashes to FFmpeg developers.\n"); +#endif + did_fail=1; + } + return -1; + } + return 0; +} + av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) { const unsigned high_bit_depth = avctx->bits_per_raw_sample > 8; + ff_check_alignment(); + #if CONFIG_ENCODERS if (avctx->bits_per_raw_sample == 10) { c->fdct = ff_jpeg_fdct_islow_10; @@ -1449,12 +1581,33 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) } #endif /* CONFIG_ENCODERS */ - if (avctx->bits_per_raw_sample == 10) { - c->idct_put = ff_simple_idct_put_10; - c->idct_add = ff_simple_idct_add_10; - c->idct = ff_simple_idct_10; + if (avctx->lowres==1) { + c->idct_put = ff_jref_idct4_put; + c->idct_add = ff_jref_idct4_add; + c->idct = ff_j_rev_dct4; + c->idct_permutation_type = FF_NO_IDCT_PERM; + } else if (avctx->lowres==2) { + c->idct_put = ff_jref_idct2_put; + c->idct_add = ff_jref_idct2_add; + c->idct = ff_j_rev_dct2; + c->idct_permutation_type = FF_NO_IDCT_PERM; + } else if (avctx->lowres==3) { + c->idct_put = ff_jref_idct1_put; + c->idct_add = ff_jref_idct1_add; + c->idct = ff_j_rev_dct1; c->idct_permutation_type = FF_NO_IDCT_PERM; } else { + if (avctx->bits_per_raw_sample == 10) { + c->idct_put = ff_simple_idct_put_10; + c->idct_add = ff_simple_idct_add_10; + c->idct = ff_simple_idct_10; + c->idct_permutation_type = FF_NO_IDCT_PERM; + } else if (avctx->bits_per_raw_sample == 12) { + c->idct_put = ff_simple_idct_put_12; + c->idct_add = ff_simple_idct_add_12; + c->idct = ff_simple_idct_12; + c->idct_permutation_type = FF_NO_IDCT_PERM; + } else { if (avctx->idct_algo == FF_IDCT_INT) { c->idct_put = jref_idct_put; c->idct_add = jref_idct_add; @@ -1471,6 +1624,7 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) c->idct = ff_simple_idct_8; c->idct_permutation_type = FF_NO_IDCT_PERM; } + } } c->diff_pixels = diff_pixels_c; @@ -1521,13 +1675,18 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) SET_CMP_FUNC(rd) SET_CMP_FUNC(bit) c->vsad[0] = vsad16_c; + c->vsad[1] = vsad8_c; c->vsad[4] = vsad_intra16_c; c->vsad[5] = vsad_intra8_c; c->vsse[0] = vsse16_c; + c->vsse[1] = vsse8_c; c->vsse[4] = vsse_intra16_c; c->vsse[5] = vsse_intra8_c; c->nsse[0] = nsse16_c; c->nsse[1] = nsse8_c; +#if CONFIG_SNOW_DECODER || CONFIG_SNOW_ENCODER + ff_dsputil_init_dwt(c); +#endif c->bswap_buf = bswap_buf; c->bswap16_buf = bswap16_buf; @@ -1552,13 +1711,20 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) switch (avctx->bits_per_raw_sample) { case 9: case 10: + case 12: + case 14: c->get_pixels = get_pixels_16_c; break; default: - c->get_pixels = get_pixels_8_c; + if (avctx->bits_per_raw_sample<=8 || avctx->codec_type != AVMEDIA_TYPE_VIDEO) { + c->get_pixels = get_pixels_8_c; + } break; } + + if (ARCH_ALPHA) + ff_dsputil_init_alpha(c, avctx); if (ARCH_ARM) ff_dsputil_init_arm(c, avctx, high_bit_depth); if (ARCH_PPC) @@ -1569,3 +1735,13 @@ av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx) ff_init_scantable_permutation(c->idct_permutation, c->idct_permutation_type); } + +av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx) +{ + ff_dsputil_init(c, avctx); +} + +av_cold void avpriv_dsputil_init(DSPContext *c, AVCodecContext *avctx) +{ + ff_dsputil_init(c, avctx); +} |