X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdct.c;h=83ea00f9cbdc4f77ab2d60fafb442be49ea36cf5;hb=f4e043ff63935a71b98a36bc98b501c36ceadb92;hp=7cc383379b477680eb8538fe30128eb82506b862;hpb=e30ab38e94252398e4390e12e1a883d24cc5afa5;p=ffmpeg diff --git a/libavcodec/dct.c b/libavcodec/dct.c index 7cc383379b4..83ea00f9cbd 100644 --- a/libavcodec/dct.c +++ b/libavcodec/dct.c @@ -4,32 +4,35 @@ * Copyright (c) 2010 Alex Converse * Copyright (c) 2010 Vitor Sessak * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg is free software; you can redistribute it and/or + * 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. * - * FFmpeg is distributed in the hope that it will be useful, + * 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 FFmpeg; if not, write to the Free Software + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ /** - * @file libavcodec/dct.c + * @file * (Inverse) Discrete Cosine Transforms. These are also known as the * type II and type III DCTs respectively. */ #include #include "libavutil/mathematics.h" -#include "fft.h" +#include "dct.h" + +#define DCT32_FLOAT +#include "dct32.c" /* sin((M_PI * x / (2*n)) */ #define SIN(s,n,x) (s->costab[(n) - (x)]) @@ -37,78 +40,142 @@ /* cos((M_PI * x / (2*n)) */ #define COS(s,n,x) (s->costab[x]) -static void ff_dct_calc_c(DCTContext *ctx, FFTSample *data) +static void ff_dst_calc_I_c(DCTContext *ctx, FFTSample *data) +{ + int n = 1 << ctx->nbits; + int i; + + data[0] = 0; + for(i = 1; i < n/2; i++) { + float tmp1 = data[i ]; + float tmp2 = data[n - i]; + float s = SIN(ctx, n, 2*i); + + s *= tmp1 + tmp2; + tmp1 = (tmp1 - tmp2) * 0.5f; + data[i ] = s + tmp1; + data[n - i] = s - tmp1; + } + + data[n/2] *= 2; + ctx->rdft.rdft_calc(&ctx->rdft, data); + + data[0] *= 0.5f; + + for(i = 1; i < n-2; i += 2) { + data[i + 1] += data[i - 1]; + data[i ] = -data[i + 2]; + } + + data[n-1] = 0; +} + +static void ff_dct_calc_I_c(DCTContext *ctx, FFTSample *data) +{ + int n = 1 << ctx->nbits; + int i; + float next = -0.5f * (data[0] - data[n]); + + for(i = 0; i < n/2; i++) { + float tmp1 = data[i ]; + float tmp2 = data[n - i]; + float s = SIN(ctx, n, 2*i); + float c = COS(ctx, n, 2*i); + + c *= tmp1 - tmp2; + s *= tmp1 - tmp2; + + next += c; + + tmp1 = (tmp1 + tmp2) * 0.5f; + data[i ] = tmp1 - s; + data[n - i] = tmp1 + s; + } + + ctx->rdft.rdft_calc(&ctx->rdft, data); + data[n] = data[1]; + data[1] = next; + + for(i = 3; i <= n; i += 2) + data[i] = data[i - 2] - data[i]; +} + +static void ff_dct_calc_III_c(DCTContext *ctx, FFTSample *data) { int n = 1 << ctx->nbits; int i; - if (ctx->inverse) { - float next = data[n - 1]; - float inv_n = 1.0f / n; + float next = data[n - 1]; + float inv_n = 1.0f / n; - for (i = n - 2; i >= 2; i -= 2) { - float val1 = data[i ]; - float val2 = data[i - 1] - data[i + 1]; - float c = COS(ctx, n, i); - float s = SIN(ctx, n, i); + for (i = n - 2; i >= 2; i -= 2) { + float val1 = data[i ]; + float val2 = data[i - 1] - data[i + 1]; + float c = COS(ctx, n, i); + float s = SIN(ctx, n, i); - data[i ] = c * val1 + s * val2; - data[i + 1] = s * val1 - c * val2; - } + data[i ] = c * val1 + s * val2; + data[i + 1] = s * val1 - c * val2; + } - data[1] = 2 * next; + data[1] = 2 * next; - ff_rdft_calc(&ctx->rdft, data); + ctx->rdft.rdft_calc(&ctx->rdft, data); - for (i = 0; i < n / 2; i++) { - float tmp1 = data[i ] * inv_n; - float tmp2 = data[n - i - 1] * inv_n; - float csc = ctx->csc2[i] * (tmp1 - tmp2); + for (i = 0; i < n / 2; i++) { + float tmp1 = data[i ] * inv_n; + float tmp2 = data[n - i - 1] * inv_n; + float csc = ctx->csc2[i] * (tmp1 - tmp2); - tmp1 += tmp2; - data[i ] = tmp1 + csc; - data[n - i - 1] = tmp1 - csc; - } - } else { - float next; - for (i=0; i < n/2; i++) { - float tmp1 = data[i ]; - float tmp2 = data[n - i - 1]; - float s = SIN(ctx, n, 2*i + 1); + tmp1 += tmp2; + data[i ] = tmp1 + csc; + data[n - i - 1] = tmp1 - csc; + } +} - s *= tmp1 - tmp2; - tmp1 = (tmp1 + tmp2) * 0.5f; +static void ff_dct_calc_II_c(DCTContext *ctx, FFTSample *data) +{ + int n = 1 << ctx->nbits; + int i; + float next; - data[i ] = tmp1 + s; - data[n-i-1] = tmp1 - s; - } + for (i=0; i < n/2; i++) { + float tmp1 = data[i ]; + float tmp2 = data[n - i - 1]; + float s = SIN(ctx, n, 2*i + 1); - ff_rdft_calc(&ctx->rdft, data); + s *= tmp1 - tmp2; + tmp1 = (tmp1 + tmp2) * 0.5f; + + data[i ] = tmp1 + s; + data[n-i-1] = tmp1 - s; + } - next = data[1] * 0.5; - data[1] *= -1; + ctx->rdft.rdft_calc(&ctx->rdft, data); - for (i = n - 2; i >= 0; i -= 2) { - float inr = data[i ]; - float ini = data[i + 1]; - float c = COS(ctx, n, i); - float s = SIN(ctx, n, i); + next = data[1] * 0.5; + data[1] *= -1; - data[i ] = c * inr + s * ini; + for (i = n - 2; i >= 0; i -= 2) { + float inr = data[i ]; + float ini = data[i + 1]; + float c = COS(ctx, n, i); + float s = SIN(ctx, n, i); - data[i+1] = next; + data[i ] = c * inr + s * ini; - next += s * inr - c * ini; - } + data[i+1] = next; + + next += s * inr - c * ini; } } -void ff_dct_calc(DCTContext *s, FFTSample *data) +static void dct32_func(DCTContext *ctx, FFTSample *data) { - s->dct_calc(s, data); + ctx->dct32(data, data); } -av_cold int ff_dct_init(DCTContext *s, int nbits, int inverse) +av_cold int ff_dct_init(DCTContext *s, int nbits, enum DCTTransformType inverse) { int n = 1 << nbits; int i; @@ -122,7 +189,7 @@ av_cold int ff_dct_init(DCTContext *s, int nbits, int inverse) s->csc2 = av_malloc(n/2 * sizeof(FFTSample)); - if (ff_rdft_init(&s->rdft, nbits, inverse) < 0) { + if (ff_rdft_init(&s->rdft, nbits, inverse == DCT_III) < 0) { av_free(s->csc2); return -1; } @@ -130,7 +197,18 @@ av_cold int ff_dct_init(DCTContext *s, int nbits, int inverse) for (i = 0; i < n/2; i++) s->csc2[i] = 0.5 / sin((M_PI / (2*n) * (2*i + 1))); - s->dct_calc = ff_dct_calc_c; + switch(inverse) { + case DCT_I : s->dct_calc = ff_dct_calc_I_c; break; + case DCT_II : s->dct_calc = ff_dct_calc_II_c ; break; + case DCT_III: s->dct_calc = ff_dct_calc_III_c; break; + case DST_I : s->dct_calc = ff_dst_calc_I_c; break; + } + + if (inverse == DCT_II && nbits == 5) + s->dct_calc = dct32_func; + + s->dct32 = dct32; + if (HAVE_MMX) ff_dct_init_mmx(s); return 0; }