2 * Copyright (C) 2009 David Conrad
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "libavcodec/x86/diracdsp_mmx.h"
25 #define FILTER(src, stride) \
26 ((21*((src)[ 0*stride] + (src)[1*stride]) \
27 -7*((src)[-1*stride] + (src)[2*stride]) \
28 +3*((src)[-2*stride] + (src)[3*stride]) \
29 -1*((src)[-3*stride] + (src)[4*stride]) + 16) >> 5)
31 static void dirac_hpel_filter(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, const uint8_t *src,
32 int stride, int width, int height)
36 for (y = 0; y < height; y++) {
37 for (x = -3; x < width+5; x++)
38 dstv[x] = av_clip_uint8(FILTER(src+x, stride));
40 for (x = 0; x < width; x++)
41 dstc[x] = av_clip_uint8(FILTER(dstv+x, 1));
43 for (x = 0; x < width; x++)
44 dsth[x] = av_clip_uint8(FILTER(src+x, 1));
53 #define PIXOP_BILINEAR(PFX, OP, WIDTH) \
54 static void ff_ ## PFX ## _dirac_pixels ## WIDTH ## _bilinear_c(uint8_t *dst, const uint8_t *src[5], int stride, int h) \
57 const uint8_t *s0 = src[0]; \
58 const uint8_t *s1 = src[1]; \
59 const uint8_t *s2 = src[2]; \
60 const uint8_t *s3 = src[3]; \
61 const uint8_t *w = src[4]; \
64 for (x = 0; x < WIDTH; x++) { \
65 OP(dst[x], (s0[x]*w[0] + s1[x]*w[1] + s2[x]*w[2] + s3[x]*w[3] + 8) >> 4); \
76 #define OP_PUT(dst, val) (dst) = (val)
77 #define OP_AVG(dst, val) (dst) = (((dst) + (val) + 1)>>1)
79 PIXOP_BILINEAR(put, OP_PUT, 8)
80 PIXOP_BILINEAR(put, OP_PUT, 16)
81 PIXOP_BILINEAR(put, OP_PUT, 32)
82 PIXOP_BILINEAR(avg, OP_AVG, 8)
83 PIXOP_BILINEAR(avg, OP_AVG, 16)
84 PIXOP_BILINEAR(avg, OP_AVG, 32)
86 #define op_scale1(x) block[x] = av_clip_uint8( (block[x]*weight + (1<<(log2_denom-1))) >> log2_denom)
87 #define op_scale2(x) dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + (1<<(log2_denom-1))) >> log2_denom)
89 #define DIRAC_WEIGHT(W) \
90 static void weight_dirac_pixels ## W ## _c(uint8_t *block, int stride, int log2_denom, \
91 int weight, int h) { \
94 for (x = 0; x < W; x++) { \
101 static void biweight_dirac_pixels ## W ## _c(uint8_t *dst, const uint8_t *src, int stride, int log2_denom, \
102 int weightd, int weights, int h) { \
105 for (x = 0; x < W; x++) { \
118 #define ADD_OBMC(xblen) \
119 static void add_obmc ## xblen ## _c(uint16_t *dst, const uint8_t *src, int stride, \
120 const uint8_t *obmc_weight, int yblen) \
124 for (x = 0; x < xblen; x += 2) { \
125 dst[x ] += src[x ] * obmc_weight[x ]; \
126 dst[x+1] += src[x+1] * obmc_weight[x+1]; \
138 static void put_signed_rect_clamped_8bit_c(uint8_t *dst, int dst_stride, const uint8_t *_src, int src_stride, int width, int height)
141 int16_t *src = (int16_t *)_src;
142 for (y = 0; y < height; y++) {
143 for (x = 0; x < width; x+=4) {
144 dst[x ] = av_clip_uint8(src[x ] + 128);
145 dst[x+1] = av_clip_uint8(src[x+1] + 128);
146 dst[x+2] = av_clip_uint8(src[x+2] + 128);
147 dst[x+3] = av_clip_uint8(src[x+3] + 128);
150 src += src_stride >> 1;
154 #define PUT_SIGNED_RECT_CLAMPED(PX) \
155 static void put_signed_rect_clamped_ ## PX ## bit_c(uint8_t *_dst, int dst_stride, const uint8_t *_src, \
156 int src_stride, int width, int height) \
159 uint16_t *dst = (uint16_t *)_dst; \
160 int32_t *src = (int32_t *)_src; \
161 for (y = 0; y < height; y++) { \
162 for (x = 0; x < width; x+=4) { \
163 dst[x ] = av_clip_uintp2(src[x ] + (1 << (PX - 1)), PX); \
164 dst[x+1] = av_clip_uintp2(src[x+1] + (1 << (PX - 1)), PX); \
165 dst[x+2] = av_clip_uintp2(src[x+2] + (1 << (PX - 1)), PX); \
166 dst[x+3] = av_clip_uintp2(src[x+3] + (1 << (PX - 1)), PX); \
168 dst += dst_stride >> 1; \
169 src += src_stride >> 2; \
173 PUT_SIGNED_RECT_CLAMPED(10)
174 PUT_SIGNED_RECT_CLAMPED(12)
176 static void add_rect_clamped_c(uint8_t *dst, const uint16_t *src, int stride,
177 const int16_t *idwt, int idwt_stride,
178 int width, int height)
182 for (y = 0; y < height; y++) {
183 for (x = 0; x < width; x+=2) {
184 dst[x ] = av_clip_uint8(((src[x ]+32)>>6) + idwt[x ]);
185 dst[x+1] = av_clip_uint8(((src[x+1]+32)>>6) + idwt[x+1]);
193 #define PIXFUNC(PFX, WIDTH) \
194 c->PFX ## _dirac_pixels_tab[WIDTH>>4][0] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _c; \
195 c->PFX ## _dirac_pixels_tab[WIDTH>>4][1] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _l2_c; \
196 c->PFX ## _dirac_pixels_tab[WIDTH>>4][2] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _l4_c; \
197 c->PFX ## _dirac_pixels_tab[WIDTH>>4][3] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _bilinear_c
199 av_cold void ff_diracdsp_init(DiracDSPContext *c)
201 c->dirac_hpel_filter = dirac_hpel_filter;
202 c->add_rect_clamped = add_rect_clamped_c;
203 c->put_signed_rect_clamped[0] = put_signed_rect_clamped_8bit_c;
204 c->put_signed_rect_clamped[1] = put_signed_rect_clamped_10bit_c;
205 c->put_signed_rect_clamped[2] = put_signed_rect_clamped_12bit_c;
207 c->add_dirac_obmc[0] = add_obmc8_c;
208 c->add_dirac_obmc[1] = add_obmc16_c;
209 c->add_dirac_obmc[2] = add_obmc32_c;
211 c->weight_dirac_pixels_tab[0] = weight_dirac_pixels8_c;
212 c->weight_dirac_pixels_tab[1] = weight_dirac_pixels16_c;
213 c->weight_dirac_pixels_tab[2] = weight_dirac_pixels32_c;
214 c->biweight_dirac_pixels_tab[0] = biweight_dirac_pixels8_c;
215 c->biweight_dirac_pixels_tab[1] = biweight_dirac_pixels16_c;
216 c->biweight_dirac_pixels_tab[2] = biweight_dirac_pixels32_c;
225 if (HAVE_MMX && HAVE_YASM) ff_diracdsp_init_mmx(c);