1 /*
2 * Copyright (C) 2009 David Conrad
3 *
4 * This file is part of FFmpeg.
5 *
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.
10 *
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.
15 *
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
19 */
20
21 #include "avcodec.h"
22 #include "diracdsp.h"
23
24 #define FILTER(src, stride) \
25 ((21*((src)[ 0*stride] + (src)[1*stride]) \
26 -7*((src)[-1*stride] + (src)[2*stride]) \
27 +3*((src)[-2*stride] + (src)[3*stride]) \
28 -1*((src)[-3*stride] + (src)[4*stride]) + 16) >> 5)
29
dirac_hpel_filter(uint8_t * dsth,uint8_t * dstv,uint8_t * dstc,const uint8_t * src,int stride,int width,int height)30 static void dirac_hpel_filter(uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, const uint8_t *src,
31 int stride, int width, int height)
32 {
33 int x, y;
34
35 for (y = 0; y < height; y++) {
36 for (x = -3; x < width+5; x++)
37 dstv[x] = av_clip_uint8(FILTER(src+x, stride));
38
39 for (x = 0; x < width; x++)
40 dstc[x] = av_clip_uint8(FILTER(dstv+x, 1));
41
42 for (x = 0; x < width; x++)
43 dsth[x] = av_clip_uint8(FILTER(src+x, 1));
44
45 src += stride;
46 dsth += stride;
47 dstv += stride;
48 dstc += stride;
49 }
50 }
51
52 #define PIXOP_BILINEAR(PFX, OP, WIDTH) \
53 static void ff_ ## PFX ## _dirac_pixels ## WIDTH ## _bilinear_c(uint8_t *dst, const uint8_t *src[5], int stride, int h) \
54 { \
55 int x; \
56 const uint8_t *s0 = src[0]; \
57 const uint8_t *s1 = src[1]; \
58 const uint8_t *s2 = src[2]; \
59 const uint8_t *s3 = src[3]; \
60 const uint8_t *w = src[4]; \
61 \
62 while (h--) { \
63 for (x = 0; x < WIDTH; x++) { \
64 OP(dst[x], (s0[x]*w[0] + s1[x]*w[1] + s2[x]*w[2] + s3[x]*w[3] + 8) >> 4); \
65 } \
66 \
67 dst += stride; \
68 s0 += stride; \
69 s1 += stride; \
70 s2 += stride; \
71 s3 += stride; \
72 } \
73 }
74
75 #define OP_PUT(dst, val) (dst) = (val)
76 #define OP_AVG(dst, val) (dst) = (((dst) + (val) + 1)>>1)
77
78 PIXOP_BILINEAR(put, OP_PUT, 8)
79 PIXOP_BILINEAR(put, OP_PUT, 16)
80 PIXOP_BILINEAR(put, OP_PUT, 32)
81 PIXOP_BILINEAR(avg, OP_AVG, 8)
82 PIXOP_BILINEAR(avg, OP_AVG, 16)
83 PIXOP_BILINEAR(avg, OP_AVG, 32)
84
85 #define op_scale1(x) block[x] = av_clip_uint8( (block[x]*weight + (1<<(log2_denom-1))) >> log2_denom)
86 #define op_scale2(x) dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + (1<<(log2_denom-1))) >> log2_denom)
87
88 #define DIRAC_WEIGHT(W) \
89 static void weight_dirac_pixels ## W ## _c(uint8_t *block, int stride, int log2_denom, \
90 int weight, int h) { \
91 int x; \
92 while (h--) { \
93 for (x = 0; x < W; x++) { \
94 op_scale1(x); \
95 op_scale1(x+1); \
96 } \
97 block += stride; \
98 } \
99 } \
100 static void biweight_dirac_pixels ## W ## _c(uint8_t *dst, const uint8_t *src, int stride, int log2_denom, \
101 int weightd, int weights, int h) { \
102 int x; \
103 while (h--) { \
104 for (x = 0; x < W; x++) { \
105 op_scale2(x); \
106 op_scale2(x+1); \
107 } \
108 dst += stride; \
109 src += stride; \
110 } \
111 }
112
113 DIRAC_WEIGHT(8)
114 DIRAC_WEIGHT(16)
115 DIRAC_WEIGHT(32)
116
117 #define ADD_OBMC(xblen) \
118 static void add_obmc ## xblen ## _c(uint16_t *dst, const uint8_t *src, int stride, \
119 const uint8_t *obmc_weight, int yblen) \
120 { \
121 int x; \
122 while (yblen--) { \
123 for (x = 0; x < xblen; x += 2) { \
124 dst[x ] += src[x ] * obmc_weight[x ]; \
125 dst[x+1] += src[x+1] * obmc_weight[x+1]; \
126 } \
127 dst += stride; \
128 src += stride; \
129 obmc_weight += 32; \
130 } \
131 }
132
133 ADD_OBMC(8)
134 ADD_OBMC(16)
135 ADD_OBMC(32)
136
put_signed_rect_clamped_8bit_c(uint8_t * dst,int dst_stride,const uint8_t * _src,int src_stride,int width,int height)137 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)
138 {
139 int x, y;
140 int16_t *src = (int16_t *)_src;
141 for (y = 0; y < height; y++) {
142 for (x = 0; x < width; x+=4) {
143 dst[x ] = av_clip_uint8(src[x ] + 128);
144 dst[x+1] = av_clip_uint8(src[x+1] + 128);
145 dst[x+2] = av_clip_uint8(src[x+2] + 128);
146 dst[x+3] = av_clip_uint8(src[x+3] + 128);
147 }
148 dst += dst_stride;
149 src += src_stride >> 1;
150 }
151 }
152
153 #define PUT_SIGNED_RECT_CLAMPED(PX) \
154 static void put_signed_rect_clamped_ ## PX ## bit_c(uint8_t *_dst, int dst_stride, const uint8_t *_src, \
155 int src_stride, int width, int height) \
156 { \
157 int x, y; \
158 uint16_t *dst = (uint16_t *)_dst; \
159 int32_t *src = (int32_t *)_src; \
160 for (y = 0; y < height; y++) { \
161 for (x = 0; x < width; x+=4) { \
162 dst[x ] = av_clip_uintp2(src[x ] + (1U << (PX - 1)), PX); \
163 dst[x+1] = av_clip_uintp2(src[x+1] + (1U << (PX - 1)), PX); \
164 dst[x+2] = av_clip_uintp2(src[x+2] + (1U << (PX - 1)), PX); \
165 dst[x+3] = av_clip_uintp2(src[x+3] + (1U << (PX - 1)), PX); \
166 } \
167 dst += dst_stride >> 1; \
168 src += src_stride >> 2; \
169 } \
170 }
171
172 PUT_SIGNED_RECT_CLAMPED(10)
173 PUT_SIGNED_RECT_CLAMPED(12)
174
add_rect_clamped_c(uint8_t * dst,const uint16_t * src,int stride,const int16_t * idwt,int idwt_stride,int width,int height)175 static void add_rect_clamped_c(uint8_t *dst, const uint16_t *src, int stride,
176 const int16_t *idwt, int idwt_stride,
177 int width, int height)
178 {
179 int x, y;
180
181 for (y = 0; y < height; y++) {
182 for (x = 0; x < width; x+=2) {
183 dst[x ] = av_clip_uint8(((src[x ]+32)>>6) + idwt[x ]);
184 dst[x+1] = av_clip_uint8(((src[x+1]+32)>>6) + idwt[x+1]);
185 }
186 dst += stride;
187 src += stride;
188 idwt += idwt_stride;
189 }
190 }
191
192 #define DEQUANT_SUBBAND(PX) \
193 static void dequant_subband_ ## PX ## _c(uint8_t *src, uint8_t *dst, ptrdiff_t stride, \
194 const int qf, const int qs, int tot_v, int tot_h) \
195 { \
196 int i, y; \
197 for (y = 0; y < tot_v; y++) { \
198 PX c, sign, *src_r = (PX *)src, *dst_r = (PX *)dst; \
199 for (i = 0; i < tot_h; i++) { \
200 c = *src_r++; \
201 sign = FFSIGN(c)*(!!c); \
202 c = (FFABS(c)*(unsigned)qf + qs) >> 2; \
203 *dst_r++ = c*sign; \
204 } \
205 src += tot_h << (sizeof(PX) >> 1); \
206 dst += stride; \
207 } \
208 }
209
210 DEQUANT_SUBBAND(int16_t)
DEQUANT_SUBBAND(int32_t)211 DEQUANT_SUBBAND(int32_t)
212
213 #define PIXFUNC(PFX, WIDTH) \
214 c->PFX ## _dirac_pixels_tab[WIDTH>>4][0] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _c; \
215 c->PFX ## _dirac_pixels_tab[WIDTH>>4][1] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _l2_c; \
216 c->PFX ## _dirac_pixels_tab[WIDTH>>4][2] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _l4_c; \
217 c->PFX ## _dirac_pixels_tab[WIDTH>>4][3] = ff_ ## PFX ## _dirac_pixels ## WIDTH ## _bilinear_c
218
219 av_cold void ff_diracdsp_init(DiracDSPContext *c)
220 {
221 c->dirac_hpel_filter = dirac_hpel_filter;
222 c->add_rect_clamped = add_rect_clamped_c;
223 c->put_signed_rect_clamped[0] = put_signed_rect_clamped_8bit_c;
224 c->put_signed_rect_clamped[1] = put_signed_rect_clamped_10bit_c;
225 c->put_signed_rect_clamped[2] = put_signed_rect_clamped_12bit_c;
226
227 c->add_dirac_obmc[0] = add_obmc8_c;
228 c->add_dirac_obmc[1] = add_obmc16_c;
229 c->add_dirac_obmc[2] = add_obmc32_c;
230
231 c->weight_dirac_pixels_tab[0] = weight_dirac_pixels8_c;
232 c->weight_dirac_pixels_tab[1] = weight_dirac_pixels16_c;
233 c->weight_dirac_pixels_tab[2] = weight_dirac_pixels32_c;
234 c->biweight_dirac_pixels_tab[0] = biweight_dirac_pixels8_c;
235 c->biweight_dirac_pixels_tab[1] = biweight_dirac_pixels16_c;
236 c->biweight_dirac_pixels_tab[2] = biweight_dirac_pixels32_c;
237
238 c->dequant_subband[0] = c->dequant_subband[2] = dequant_subband_int16_t_c;
239 c->dequant_subband[1] = c->dequant_subband[3] = dequant_subband_int32_t_c;
240
241 PIXFUNC(put, 8);
242 PIXFUNC(put, 16);
243 PIXFUNC(put, 32);
244 PIXFUNC(avg, 8);
245 PIXFUNC(avg, 16);
246 PIXFUNC(avg, 32);
247
248 if (ARCH_X86)
249 ff_diracdsp_init_x86(c);
250 }
251