2 * Copyright (C) 2004 The FFmpeg project
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 * 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) ((int)((SUINT)(a) * (b)) >> 16)
46 static av_always_inline void idct(uint8_t *dst, ptrdiff_t stride,
47 int16_t *input, int type)
51 int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
52 int Ed, Gd, Add, Bdd, Fd, Hd;
56 /* Inverse DCT on the rows now */
57 for (i = 0; i < 8; i++) {
58 /* Check for non-zero values */
59 if (ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
60 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8]) {
61 A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
62 B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
63 C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
64 D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
66 Ad = M(xC4S4, (A - C));
67 Bd = M(xC4S4, (B - D));
72 E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
73 F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
75 G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
76 H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
87 /* Final sequence of operations over-write original inputs. */
101 ip += 1; /* next row */
106 for (i = 0; i < 8; i++) {
107 /* Check for non-zero values (bitwise or faster than ||) */
108 if (ip[1] | ip[2] | ip[3] |
109 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;
124 if (type == 1) { // HACK
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));
176 dst[7*stride] = av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20));
179 int v = (xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20;
180 dst[0 * stride] = av_clip_uint8(dst[0 * stride] + v);
181 dst[1 * stride] = av_clip_uint8(dst[1 * stride] + v);
182 dst[2 * stride] = av_clip_uint8(dst[2 * stride] + v);
183 dst[3 * stride] = av_clip_uint8(dst[3 * stride] + v);
184 dst[4 * stride] = av_clip_uint8(dst[4 * stride] + v);
185 dst[5 * stride] = av_clip_uint8(dst[5 * stride] + v);
186 dst[6 * stride] = av_clip_uint8(dst[6 * stride] + v);
187 dst[7 * stride] = av_clip_uint8(dst[7 * stride] + v);
192 ip += 8; /* next column */
197 static void vp3_idct_put_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
198 int16_t *block /* align 16 */)
200 idct(dest, stride, block, 1);
201 memset(block, 0, sizeof(*block) * 64);
204 static void vp3_idct_add_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
205 int16_t *block /* align 16 */)
207 idct(dest, stride, block, 2);
208 memset(block, 0, sizeof(*block) * 64);
211 static void vp3_idct_dc_add_c(uint8_t *dest /* align 8 */, ptrdiff_t stride,
212 int16_t *block /* align 16 */)
214 int i, dc = (block[0] + 15) >> 5;
216 for (i = 0; i < 8; i++) {
217 dest[0] = av_clip_uint8(dest[0] + dc);
218 dest[1] = av_clip_uint8(dest[1] + dc);
219 dest[2] = av_clip_uint8(dest[2] + dc);
220 dest[3] = av_clip_uint8(dest[3] + dc);
221 dest[4] = av_clip_uint8(dest[4] + dc);
222 dest[5] = av_clip_uint8(dest[5] + dc);
223 dest[6] = av_clip_uint8(dest[6] + dc);
224 dest[7] = av_clip_uint8(dest[7] + dc);
230 static void vp3_v_loop_filter_c(uint8_t *first_pixel, ptrdiff_t stride,
231 int *bounding_values)
235 const ptrdiff_t nstride = -stride;
237 for (end = first_pixel + 8; first_pixel < end; first_pixel++) {
238 filter_value = (first_pixel[2 * nstride] - first_pixel[stride]) +
239 (first_pixel[0] - first_pixel[nstride]) * 3;
240 filter_value = bounding_values[(filter_value + 4) >> 3];
242 first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
243 first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
247 static void vp3_h_loop_filter_c(uint8_t *first_pixel, ptrdiff_t stride,
248 int *bounding_values)
253 for (end = first_pixel + 8 * stride; first_pixel != end; first_pixel += stride) {
254 filter_value = (first_pixel[-2] - first_pixel[1]) +
255 (first_pixel[ 0] - first_pixel[-1]) * 3;
256 filter_value = bounding_values[(filter_value + 4) >> 3];
258 first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
259 first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
263 static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
264 const uint8_t *src2, ptrdiff_t stride, int h)
268 for (i = 0; i < h; i++) {
271 a = AV_RN32(&src1[i * stride]);
272 b = AV_RN32(&src2[i * stride]);
273 AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
274 a = AV_RN32(&src1[i * stride + 4]);
275 b = AV_RN32(&src2[i * stride + 4]);
276 AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
280 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
282 c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
284 c->idct_put = vp3_idct_put_c;
285 c->idct_add = vp3_idct_add_c;
286 c->idct_dc_add = vp3_idct_dc_add_c;
287 c->v_loop_filter = vp3_v_loop_filter_c;
288 c->h_loop_filter = vp3_h_loop_filter_c;
291 ff_vp3dsp_init_arm(c, flags);
293 ff_vp3dsp_init_ppc(c, flags);
295 ff_vp3dsp_init_x86(c, flags);