2 * Copyright (C) 2004 the ffmpeg project
4 * This file is part of Libav.
6 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * Standard C DSP-oriented functions cribbed from the original VP3
27 #include "libavutil/attributes.h"
28 #include "libavutil/common.h"
29 #include "libavutil/intreadwrite.h"
35 #define IdctAdjustBeforeShift 8
44 #define M(a,b) (((a) * (b))>>16)
46 static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
50 int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
51 int Ed, Gd, Add, Bdd, Fd, Hd;
55 /* Inverse DCT on the rows now */
56 for (i = 0; i < 8; i++) {
57 /* Check for non-zero values */
58 if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
59 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
60 A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
61 B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
62 C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
63 D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
65 Ad = M(xC4S4, (A - C));
66 Bd = M(xC4S4, (B - D));
71 E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
72 F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
74 G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
75 H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
86 /* Final sequence of operations over-write original inputs. */
100 ip += 1; /* next row */
105 for ( i = 0; i < 8; i++) {
106 /* Check for non-zero values (bitwise or faster than ||) */
107 if ( ip[1] | ip[2] | ip[3] |
108 ip[4] | ip[5] | ip[6] | ip[7] ) {
110 A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
111 B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
112 C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
113 D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
115 Ad = M(xC4S4, (A - C));
116 Bd = M(xC4S4, (B - D));
121 E = M(xC4S4, (ip[0] + ip[4])) + 8;
122 F = M(xC4S4, (ip[0] - ip[4])) + 8;
129 G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
130 H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
141 /* Final sequence of operations over-write original inputs. */
143 dst[0*stride] = av_clip_uint8((Gd + Cd ) >> 4);
144 dst[7*stride] = av_clip_uint8((Gd - Cd ) >> 4);
146 dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
147 dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
149 dst[3*stride] = av_clip_uint8((Ed + Dd ) >> 4);
150 dst[4*stride] = av_clip_uint8((Ed - Dd ) >> 4);
152 dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
153 dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
155 dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd ) >> 4));
156 dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd ) >> 4));
158 dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
159 dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
161 dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd ) >> 4));
162 dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd ) >> 4));
164 dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
165 dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
177 dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20));
180 int v= ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20);
181 dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
182 dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
183 dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
184 dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
185 dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
186 dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
187 dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
188 dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
193 ip += 8; /* next column */
198 static void vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size,
199 int16_t *block/*align 16*/)
201 idct(dest, line_size, block, 1);
202 memset(block, 0, sizeof(*block) * 64);
205 static void vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size,
206 int16_t *block/*align 16*/)
208 idct(dest, line_size, block, 2);
209 memset(block, 0, sizeof(*block) * 64);
212 static void vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size,
213 int16_t *block/*align 16*/)
215 int i, dc = (block[0] + 15) >> 5;
217 for(i = 0; i < 8; i++){
218 dest[0] = av_clip_uint8(dest[0] + dc);
219 dest[1] = av_clip_uint8(dest[1] + dc);
220 dest[2] = av_clip_uint8(dest[2] + dc);
221 dest[3] = av_clip_uint8(dest[3] + dc);
222 dest[4] = av_clip_uint8(dest[4] + dc);
223 dest[5] = av_clip_uint8(dest[5] + dc);
224 dest[6] = av_clip_uint8(dest[6] + dc);
225 dest[7] = av_clip_uint8(dest[7] + dc);
231 static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
232 int *bounding_values)
236 const int nstride= -stride;
238 for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
240 (first_pixel[2 * nstride] - first_pixel[ stride])
241 +3*(first_pixel[0 ] - first_pixel[nstride]);
242 filter_value = bounding_values[(filter_value + 4) >> 3];
243 first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
244 first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
248 static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
249 int *bounding_values)
254 for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
256 (first_pixel[-2] - first_pixel[ 1])
257 +3*(first_pixel[ 0] - first_pixel[-1]);
258 filter_value = bounding_values[(filter_value + 4) >> 3];
259 first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
260 first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
264 static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
265 const uint8_t *src2, ptrdiff_t stride, int h)
269 for (i = 0; i < h; i++) {
272 a = AV_RN32(&src1[i * stride]);
273 b = AV_RN32(&src2[i * stride]);
274 AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
275 a = AV_RN32(&src1[i * stride + 4]);
276 b = AV_RN32(&src2[i * stride + 4]);
277 AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
281 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
283 c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
285 c->idct_put = vp3_idct_put_c;
286 c->idct_add = vp3_idct_add_c;
287 c->idct_dc_add = vp3_idct_dc_add_c;
288 c->v_loop_filter = vp3_v_loop_filter_c;
289 c->h_loop_filter = vp3_h_loop_filter_c;
292 ff_vp3dsp_init_arm(c, flags);
294 ff_vp3dsp_init_bfin(c, flags);
296 ff_vp3dsp_init_ppc(c, flags);
298 ff_vp3dsp_init_x86(c, flags);