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"
34 #define IdctAdjustBeforeShift 8
43 #define M(a,b) (((a) * (b))>>16)
45 static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
49 int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
50 int Ed, Gd, Add, Bdd, Fd, Hd;
54 /* Inverse DCT on the rows now */
55 for (i = 0; i < 8; i++) {
56 /* Check for non-zero values */
57 if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
58 A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
59 B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
60 C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
61 D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
63 Ad = M(xC4S4, (A - C));
64 Bd = M(xC4S4, (B - D));
69 E = M(xC4S4, (ip[0] + ip[4]));
70 F = M(xC4S4, (ip[0] - ip[4]));
72 G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
73 H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
84 /* Final sequence of operations over-write original inputs. */
98 ip += 8; /* next row */
103 for ( i = 0; i < 8; i++) {
104 /* Check for non-zero values (bitwise or faster than ||) */
105 if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
106 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
108 A = M(xC1S7, ip[1*8]) + M(xC7S1, ip[7*8]);
109 B = M(xC7S1, ip[1*8]) - M(xC1S7, ip[7*8]);
110 C = M(xC3S5, ip[3*8]) + M(xC5S3, ip[5*8]);
111 D = M(xC3S5, ip[5*8]) - M(xC5S3, ip[3*8]);
113 Ad = M(xC4S4, (A - C));
114 Bd = M(xC4S4, (B - D));
119 E = M(xC4S4, (ip[0*8] + ip[4*8])) + 8;
120 F = M(xC4S4, (ip[0*8] - ip[4*8])) + 8;
127 G = M(xC2S6, ip[2*8]) + M(xC6S2, ip[6*8]);
128 H = M(xC6S2, ip[2*8]) - M(xC2S6, ip[6*8]);
139 /* Final sequence of operations over-write original inputs. */
141 ip[0*8] = (Gd + Cd ) >> 4;
142 ip[7*8] = (Gd - Cd ) >> 4;
144 ip[1*8] = (Add + Hd ) >> 4;
145 ip[2*8] = (Add - Hd ) >> 4;
147 ip[3*8] = (Ed + Dd ) >> 4;
148 ip[4*8] = (Ed - Dd ) >> 4;
150 ip[5*8] = (Fd + Bdd ) >> 4;
151 ip[6*8] = (Fd - Bdd ) >> 4;
153 dst[0*stride] = av_clip_uint8((Gd + Cd ) >> 4);
154 dst[7*stride] = av_clip_uint8((Gd - Cd ) >> 4);
156 dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
157 dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
159 dst[3*stride] = av_clip_uint8((Ed + Dd ) >> 4);
160 dst[4*stride] = av_clip_uint8((Ed - Dd ) >> 4);
162 dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
163 dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
165 dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd ) >> 4));
166 dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd ) >> 4));
168 dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
169 dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
171 dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd ) >> 4));
172 dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd ) >> 4));
174 dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
175 dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
187 ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
196 dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20));
199 int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
200 dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
201 dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
202 dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
203 dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
204 dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
205 dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
206 dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
207 dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
212 ip++; /* next column */
217 static void vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size,
218 int16_t *block/*align 16*/)
220 idct(dest, line_size, block, 1);
221 memset(block, 0, sizeof(*block) * 64);
224 static void vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size,
225 int16_t *block/*align 16*/)
227 idct(dest, line_size, block, 2);
228 memset(block, 0, sizeof(*block) * 64);
231 static void vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size,
232 int16_t *block/*align 16*/)
234 int i, dc = (block[0] + 15) >> 5;
236 for(i = 0; i < 8; i++){
237 dest[0] = av_clip_uint8(dest[0] + dc);
238 dest[1] = av_clip_uint8(dest[1] + dc);
239 dest[2] = av_clip_uint8(dest[2] + dc);
240 dest[3] = av_clip_uint8(dest[3] + dc);
241 dest[4] = av_clip_uint8(dest[4] + dc);
242 dest[5] = av_clip_uint8(dest[5] + dc);
243 dest[6] = av_clip_uint8(dest[6] + dc);
244 dest[7] = av_clip_uint8(dest[7] + dc);
250 static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
251 int *bounding_values)
255 const int nstride= -stride;
257 for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
259 (first_pixel[2 * nstride] - first_pixel[ stride])
260 +3*(first_pixel[0 ] - first_pixel[nstride]);
261 filter_value = bounding_values[(filter_value + 4) >> 3];
262 first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
263 first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
267 static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
268 int *bounding_values)
273 for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
275 (first_pixel[-2] - first_pixel[ 1])
276 +3*(first_pixel[ 0] - first_pixel[-1]);
277 filter_value = bounding_values[(filter_value + 4) >> 3];
278 first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
279 first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
283 static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
284 const uint8_t *src2, ptrdiff_t stride, int h)
288 for (i = 0; i < h; i++) {
291 a = AV_RN32(&src1[i * stride]);
292 b = AV_RN32(&src2[i * stride]);
293 AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
294 a = AV_RN32(&src1[i * stride + 4]);
295 b = AV_RN32(&src2[i * stride + 4]);
296 AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
300 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
302 c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
304 c->idct_put = vp3_idct_put_c;
305 c->idct_add = vp3_idct_add_c;
306 c->idct_dc_add = vp3_idct_dc_add_c;
307 c->v_loop_filter = vp3_v_loop_filter_c;
308 c->h_loop_filter = vp3_h_loop_filter_c;
310 c->idct_perm = FF_NO_IDCT_PERM;
313 ff_vp3dsp_init_arm(c, flags);
315 ff_vp3dsp_init_bfin(c, flags);
317 ff_vp3dsp_init_ppc(c, flags);
319 ff_vp3dsp_init_x86(c, flags);