Vikas Arora | af51b94 | 2014-08-28 10:51:12 -0700 | [diff] [blame] | 1 | // Copyright 2014 Google Inc. All Rights Reserved. |
| 2 | // |
| 3 | // Use of this source code is governed by a BSD-style license |
| 4 | // that can be found in the COPYING file in the root of the source |
| 5 | // tree. An additional intellectual property rights grant can be found |
| 6 | // in the file PATENTS. All contributing project authors may |
| 7 | // be found in the AUTHORS file in the root of the source tree. |
| 8 | // ----------------------------------------------------------------------------- |
| 9 | // |
| 10 | // MIPS version of dsp functions |
| 11 | // |
| 12 | // Author(s): Djordje Pesut (djordje.pesut@imgtec.com) |
| 13 | // Jovan Zelincevic (jovan.zelincevic@imgtec.com) |
| 14 | |
| 15 | #include "./dsp.h" |
| 16 | |
| 17 | #if defined(WEBP_USE_MIPS32) |
| 18 | |
| 19 | static const int kC1 = 20091 + (1 << 16); |
| 20 | static const int kC2 = 35468; |
| 21 | |
| 22 | static WEBP_INLINE int abs_mips32(int x) { |
| 23 | const int sign = x >> 31; |
| 24 | return (x ^ sign) - sign; |
| 25 | } |
| 26 | |
| 27 | // 4 pixels in, 2 pixels out |
| 28 | static WEBP_INLINE void do_filter2(uint8_t* p, int step) { |
| 29 | const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step]; |
| 30 | const int a = 3 * (q0 - p0) + VP8ksclip1[p1 - q1]; |
| 31 | const int a1 = VP8ksclip2[(a + 4) >> 3]; |
| 32 | const int a2 = VP8ksclip2[(a + 3) >> 3]; |
| 33 | p[-step] = VP8kclip1[p0 + a2]; |
| 34 | p[ 0] = VP8kclip1[q0 - a1]; |
| 35 | } |
| 36 | |
| 37 | // 4 pixels in, 4 pixels out |
| 38 | static WEBP_INLINE void do_filter4(uint8_t* p, int step) { |
| 39 | const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step]; |
| 40 | const int a = 3 * (q0 - p0); |
| 41 | const int a1 = VP8ksclip2[(a + 4) >> 3]; |
| 42 | const int a2 = VP8ksclip2[(a + 3) >> 3]; |
| 43 | const int a3 = (a1 + 1) >> 1; |
| 44 | p[-2 * step] = VP8kclip1[p1 + a3]; |
| 45 | p[- step] = VP8kclip1[p0 + a2]; |
| 46 | p[ 0] = VP8kclip1[q0 - a1]; |
| 47 | p[ step] = VP8kclip1[q1 - a3]; |
| 48 | } |
| 49 | |
| 50 | // 6 pixels in, 6 pixels out |
| 51 | static WEBP_INLINE void do_filter6(uint8_t* p, int step) { |
| 52 | const int p2 = p[-3 * step], p1 = p[-2 * step], p0 = p[-step]; |
| 53 | const int q0 = p[0], q1 = p[step], q2 = p[2 * step]; |
| 54 | const int a = VP8ksclip1[3 * (q0 - p0) + VP8ksclip1[p1 - q1]]; |
| 55 | const int a1 = (27 * a + 63) >> 7; // eq. to ((3 * a + 7) * 9) >> 7 |
| 56 | const int a2 = (18 * a + 63) >> 7; // eq. to ((2 * a + 7) * 9) >> 7 |
| 57 | const int a3 = (9 * a + 63) >> 7; // eq. to ((1 * a + 7) * 9) >> 7 |
| 58 | p[-3 * step] = VP8kclip1[p2 + a3]; |
| 59 | p[-2 * step] = VP8kclip1[p1 + a2]; |
| 60 | p[- step] = VP8kclip1[p0 + a1]; |
| 61 | p[ 0] = VP8kclip1[q0 - a1]; |
| 62 | p[ step] = VP8kclip1[q1 - a2]; |
| 63 | p[ 2 * step] = VP8kclip1[q2 - a3]; |
| 64 | } |
| 65 | |
| 66 | static WEBP_INLINE int hev(const uint8_t* p, int step, int thresh) { |
| 67 | const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step]; |
| 68 | return (abs_mips32(p1 - p0) > thresh) || (abs_mips32(q1 - q0) > thresh); |
| 69 | } |
| 70 | |
| 71 | static WEBP_INLINE int needs_filter(const uint8_t* p, int step, int thresh) { |
| 72 | const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step]; |
| 73 | return ((2 * abs_mips32(p0 - q0) + (abs_mips32(p1 - q1) >> 1)) <= thresh); |
| 74 | } |
| 75 | |
| 76 | static WEBP_INLINE int needs_filter2(const uint8_t* p, |
| 77 | int step, int t, int it) { |
| 78 | const int p3 = p[-4 * step], p2 = p[-3 * step]; |
| 79 | const int p1 = p[-2 * step], p0 = p[-step]; |
| 80 | const int q0 = p[0], q1 = p[step], q2 = p[2 * step], q3 = p[3 * step]; |
| 81 | if ((2 * abs_mips32(p0 - q0) + (abs_mips32(p1 - q1) >> 1)) > t) { |
| 82 | return 0; |
| 83 | } |
| 84 | return abs_mips32(p3 - p2) <= it && abs_mips32(p2 - p1) <= it && |
| 85 | abs_mips32(p1 - p0) <= it && abs_mips32(q3 - q2) <= it && |
| 86 | abs_mips32(q2 - q1) <= it && abs_mips32(q1 - q0) <= it; |
| 87 | } |
| 88 | |
| 89 | static WEBP_INLINE void FilterLoop26(uint8_t* p, |
| 90 | int hstride, int vstride, int size, |
| 91 | int thresh, int ithresh, int hev_thresh) { |
| 92 | while (size-- > 0) { |
| 93 | if (needs_filter2(p, hstride, thresh, ithresh)) { |
| 94 | if (hev(p, hstride, hev_thresh)) { |
| 95 | do_filter2(p, hstride); |
| 96 | } else { |
| 97 | do_filter6(p, hstride); |
| 98 | } |
| 99 | } |
| 100 | p += vstride; |
| 101 | } |
| 102 | } |
| 103 | |
| 104 | static WEBP_INLINE void FilterLoop24(uint8_t* p, |
| 105 | int hstride, int vstride, int size, |
| 106 | int thresh, int ithresh, int hev_thresh) { |
| 107 | while (size-- > 0) { |
| 108 | if (needs_filter2(p, hstride, thresh, ithresh)) { |
| 109 | if (hev(p, hstride, hev_thresh)) { |
| 110 | do_filter2(p, hstride); |
| 111 | } else { |
| 112 | do_filter4(p, hstride); |
| 113 | } |
| 114 | } |
| 115 | p += vstride; |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | // on macroblock edges |
| 120 | static void VFilter16(uint8_t* p, int stride, |
| 121 | int thresh, int ithresh, int hev_thresh) { |
| 122 | FilterLoop26(p, stride, 1, 16, thresh, ithresh, hev_thresh); |
| 123 | } |
| 124 | |
| 125 | static void HFilter16(uint8_t* p, int stride, |
| 126 | int thresh, int ithresh, int hev_thresh) { |
| 127 | FilterLoop26(p, 1, stride, 16, thresh, ithresh, hev_thresh); |
| 128 | } |
| 129 | |
| 130 | // 8-pixels wide variant, for chroma filtering |
| 131 | static void VFilter8(uint8_t* u, uint8_t* v, int stride, |
| 132 | int thresh, int ithresh, int hev_thresh) { |
| 133 | FilterLoop26(u, stride, 1, 8, thresh, ithresh, hev_thresh); |
| 134 | FilterLoop26(v, stride, 1, 8, thresh, ithresh, hev_thresh); |
| 135 | } |
| 136 | |
| 137 | static void HFilter8(uint8_t* u, uint8_t* v, int stride, |
| 138 | int thresh, int ithresh, int hev_thresh) { |
| 139 | FilterLoop26(u, 1, stride, 8, thresh, ithresh, hev_thresh); |
| 140 | FilterLoop26(v, 1, stride, 8, thresh, ithresh, hev_thresh); |
| 141 | } |
| 142 | |
| 143 | static void VFilter8i(uint8_t* u, uint8_t* v, int stride, |
| 144 | int thresh, int ithresh, int hev_thresh) { |
| 145 | FilterLoop24(u + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh); |
| 146 | FilterLoop24(v + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh); |
| 147 | } |
| 148 | |
| 149 | static void HFilter8i(uint8_t* u, uint8_t* v, int stride, |
| 150 | int thresh, int ithresh, int hev_thresh) { |
| 151 | FilterLoop24(u + 4, 1, stride, 8, thresh, ithresh, hev_thresh); |
| 152 | FilterLoop24(v + 4, 1, stride, 8, thresh, ithresh, hev_thresh); |
| 153 | } |
| 154 | |
| 155 | // on three inner edges |
| 156 | static void VFilter16i(uint8_t* p, int stride, |
| 157 | int thresh, int ithresh, int hev_thresh) { |
| 158 | int k; |
| 159 | for (k = 3; k > 0; --k) { |
| 160 | p += 4 * stride; |
| 161 | FilterLoop24(p, stride, 1, 16, thresh, ithresh, hev_thresh); |
| 162 | } |
| 163 | } |
| 164 | |
| 165 | static void HFilter16i(uint8_t* p, int stride, |
| 166 | int thresh, int ithresh, int hev_thresh) { |
| 167 | int k; |
| 168 | for (k = 3; k > 0; --k) { |
| 169 | p += 4; |
| 170 | FilterLoop24(p, 1, stride, 16, thresh, ithresh, hev_thresh); |
| 171 | } |
| 172 | } |
| 173 | |
| 174 | //------------------------------------------------------------------------------ |
| 175 | // Simple In-loop filtering (Paragraph 15.2) |
| 176 | |
| 177 | static void SimpleVFilter16(uint8_t* p, int stride, int thresh) { |
| 178 | int i; |
| 179 | for (i = 0; i < 16; ++i) { |
| 180 | if (needs_filter(p + i, stride, thresh)) { |
| 181 | do_filter2(p + i, stride); |
| 182 | } |
| 183 | } |
| 184 | } |
| 185 | |
| 186 | static void SimpleHFilter16(uint8_t* p, int stride, int thresh) { |
| 187 | int i; |
| 188 | for (i = 0; i < 16; ++i) { |
| 189 | if (needs_filter(p + i * stride, 1, thresh)) { |
| 190 | do_filter2(p + i * stride, 1); |
| 191 | } |
| 192 | } |
| 193 | } |
| 194 | |
| 195 | static void SimpleVFilter16i(uint8_t* p, int stride, int thresh) { |
| 196 | int k; |
| 197 | for (k = 3; k > 0; --k) { |
| 198 | p += 4 * stride; |
| 199 | SimpleVFilter16(p, stride, thresh); |
| 200 | } |
| 201 | } |
| 202 | |
| 203 | static void SimpleHFilter16i(uint8_t* p, int stride, int thresh) { |
| 204 | int k; |
| 205 | for (k = 3; k > 0; --k) { |
| 206 | p += 4; |
| 207 | SimpleHFilter16(p, stride, thresh); |
| 208 | } |
| 209 | } |
| 210 | |
| 211 | static void TransformOne(const int16_t* in, uint8_t* dst) { |
| 212 | int temp0, temp1, temp2, temp3, temp4; |
| 213 | int temp5, temp6, temp7, temp8, temp9; |
| 214 | int temp10, temp11, temp12, temp13, temp14; |
| 215 | int temp15, temp16, temp17, temp18; |
| 216 | int16_t* p_in = (int16_t*)in; |
| 217 | |
| 218 | // loops unrolled and merged to avoid usage of tmp buffer |
| 219 | // and to reduce number of stalls. MUL macro is written |
| 220 | // in assembler and inlined |
| 221 | __asm__ volatile( |
| 222 | "lh %[temp0], 0(%[in]) \n\t" |
| 223 | "lh %[temp8], 16(%[in]) \n\t" |
| 224 | "lh %[temp4], 8(%[in]) \n\t" |
| 225 | "lh %[temp12], 24(%[in]) \n\t" |
| 226 | "addu %[temp16], %[temp0], %[temp8] \n\t" |
| 227 | "subu %[temp0], %[temp0], %[temp8] \n\t" |
| 228 | "mul %[temp8], %[temp4], %[kC2] \n\t" |
| 229 | "mul %[temp17], %[temp12], %[kC1] \n\t" |
| 230 | "mul %[temp4], %[temp4], %[kC1] \n\t" |
| 231 | "mul %[temp12], %[temp12], %[kC2] \n\t" |
| 232 | "lh %[temp1], 2(%[in]) \n\t" |
| 233 | "lh %[temp5], 10(%[in]) \n\t" |
| 234 | "lh %[temp9], 18(%[in]) \n\t" |
| 235 | "lh %[temp13], 26(%[in]) \n\t" |
| 236 | "sra %[temp8], %[temp8], 16 \n\t" |
| 237 | "sra %[temp17], %[temp17], 16 \n\t" |
| 238 | "sra %[temp4], %[temp4], 16 \n\t" |
| 239 | "sra %[temp12], %[temp12], 16 \n\t" |
| 240 | "lh %[temp2], 4(%[in]) \n\t" |
| 241 | "lh %[temp6], 12(%[in]) \n\t" |
| 242 | "lh %[temp10], 20(%[in]) \n\t" |
| 243 | "lh %[temp14], 28(%[in]) \n\t" |
| 244 | "subu %[temp17], %[temp8], %[temp17] \n\t" |
| 245 | "addu %[temp4], %[temp4], %[temp12] \n\t" |
| 246 | "addu %[temp8], %[temp16], %[temp4] \n\t" |
| 247 | "subu %[temp4], %[temp16], %[temp4] \n\t" |
| 248 | "addu %[temp16], %[temp1], %[temp9] \n\t" |
| 249 | "subu %[temp1], %[temp1], %[temp9] \n\t" |
| 250 | "lh %[temp3], 6(%[in]) \n\t" |
| 251 | "lh %[temp7], 14(%[in]) \n\t" |
| 252 | "lh %[temp11], 22(%[in]) \n\t" |
| 253 | "lh %[temp15], 30(%[in]) \n\t" |
| 254 | "addu %[temp12], %[temp0], %[temp17] \n\t" |
| 255 | "subu %[temp0], %[temp0], %[temp17] \n\t" |
| 256 | "mul %[temp9], %[temp5], %[kC2] \n\t" |
| 257 | "mul %[temp17], %[temp13], %[kC1] \n\t" |
| 258 | "mul %[temp5], %[temp5], %[kC1] \n\t" |
| 259 | "mul %[temp13], %[temp13], %[kC2] \n\t" |
| 260 | "sra %[temp9], %[temp9], 16 \n\t" |
| 261 | "sra %[temp17], %[temp17], 16 \n\t" |
| 262 | "subu %[temp17], %[temp9], %[temp17] \n\t" |
| 263 | "sra %[temp5], %[temp5], 16 \n\t" |
| 264 | "sra %[temp13], %[temp13], 16 \n\t" |
| 265 | "addu %[temp5], %[temp5], %[temp13] \n\t" |
| 266 | "addu %[temp13], %[temp1], %[temp17] \n\t" |
| 267 | "subu %[temp1], %[temp1], %[temp17] \n\t" |
| 268 | "mul %[temp17], %[temp14], %[kC1] \n\t" |
| 269 | "mul %[temp14], %[temp14], %[kC2] \n\t" |
| 270 | "addu %[temp9], %[temp16], %[temp5] \n\t" |
| 271 | "subu %[temp5], %[temp16], %[temp5] \n\t" |
| 272 | "addu %[temp16], %[temp2], %[temp10] \n\t" |
| 273 | "subu %[temp2], %[temp2], %[temp10] \n\t" |
| 274 | "mul %[temp10], %[temp6], %[kC2] \n\t" |
| 275 | "mul %[temp6], %[temp6], %[kC1] \n\t" |
| 276 | "sra %[temp17], %[temp17], 16 \n\t" |
| 277 | "sra %[temp14], %[temp14], 16 \n\t" |
| 278 | "sra %[temp10], %[temp10], 16 \n\t" |
| 279 | "sra %[temp6], %[temp6], 16 \n\t" |
| 280 | "subu %[temp17], %[temp10], %[temp17] \n\t" |
| 281 | "addu %[temp6], %[temp6], %[temp14] \n\t" |
| 282 | "addu %[temp10], %[temp16], %[temp6] \n\t" |
| 283 | "subu %[temp6], %[temp16], %[temp6] \n\t" |
| 284 | "addu %[temp14], %[temp2], %[temp17] \n\t" |
| 285 | "subu %[temp2], %[temp2], %[temp17] \n\t" |
| 286 | "mul %[temp17], %[temp15], %[kC1] \n\t" |
| 287 | "mul %[temp15], %[temp15], %[kC2] \n\t" |
| 288 | "addu %[temp16], %[temp3], %[temp11] \n\t" |
| 289 | "subu %[temp3], %[temp3], %[temp11] \n\t" |
| 290 | "mul %[temp11], %[temp7], %[kC2] \n\t" |
| 291 | "mul %[temp7], %[temp7], %[kC1] \n\t" |
| 292 | "addiu %[temp8], %[temp8], 4 \n\t" |
| 293 | "addiu %[temp12], %[temp12], 4 \n\t" |
| 294 | "addiu %[temp0], %[temp0], 4 \n\t" |
| 295 | "addiu %[temp4], %[temp4], 4 \n\t" |
| 296 | "sra %[temp17], %[temp17], 16 \n\t" |
| 297 | "sra %[temp15], %[temp15], 16 \n\t" |
| 298 | "sra %[temp11], %[temp11], 16 \n\t" |
| 299 | "sra %[temp7], %[temp7], 16 \n\t" |
| 300 | "subu %[temp17], %[temp11], %[temp17] \n\t" |
| 301 | "addu %[temp7], %[temp7], %[temp15] \n\t" |
| 302 | "addu %[temp15], %[temp3], %[temp17] \n\t" |
| 303 | "subu %[temp3], %[temp3], %[temp17] \n\t" |
| 304 | "addu %[temp11], %[temp16], %[temp7] \n\t" |
| 305 | "subu %[temp7], %[temp16], %[temp7] \n\t" |
| 306 | "addu %[temp16], %[temp8], %[temp10] \n\t" |
| 307 | "subu %[temp8], %[temp8], %[temp10] \n\t" |
| 308 | "mul %[temp10], %[temp9], %[kC2] \n\t" |
| 309 | "mul %[temp17], %[temp11], %[kC1] \n\t" |
| 310 | "mul %[temp9], %[temp9], %[kC1] \n\t" |
| 311 | "mul %[temp11], %[temp11], %[kC2] \n\t" |
| 312 | "sra %[temp10], %[temp10], 16 \n\t" |
| 313 | "sra %[temp17], %[temp17], 16 \n\t" |
| 314 | "sra %[temp9], %[temp9], 16 \n\t" |
| 315 | "sra %[temp11], %[temp11], 16 \n\t" |
| 316 | "subu %[temp17], %[temp10], %[temp17] \n\t" |
| 317 | "addu %[temp11], %[temp9], %[temp11] \n\t" |
| 318 | "addu %[temp10], %[temp12], %[temp14] \n\t" |
| 319 | "subu %[temp12], %[temp12], %[temp14] \n\t" |
| 320 | "mul %[temp14], %[temp13], %[kC2] \n\t" |
| 321 | "mul %[temp9], %[temp15], %[kC1] \n\t" |
| 322 | "mul %[temp13], %[temp13], %[kC1] \n\t" |
| 323 | "mul %[temp15], %[temp15], %[kC2] \n\t" |
| 324 | "sra %[temp14], %[temp14], 16 \n\t" |
| 325 | "sra %[temp9], %[temp9], 16 \n\t" |
| 326 | "sra %[temp13], %[temp13], 16 \n\t" |
| 327 | "sra %[temp15], %[temp15], 16 \n\t" |
| 328 | "subu %[temp9], %[temp14], %[temp9] \n\t" |
| 329 | "addu %[temp15], %[temp13], %[temp15] \n\t" |
| 330 | "addu %[temp14], %[temp0], %[temp2] \n\t" |
| 331 | "subu %[temp0], %[temp0], %[temp2] \n\t" |
| 332 | "mul %[temp2], %[temp1], %[kC2] \n\t" |
| 333 | "mul %[temp13], %[temp3], %[kC1] \n\t" |
| 334 | "mul %[temp1], %[temp1], %[kC1] \n\t" |
| 335 | "mul %[temp3], %[temp3], %[kC2] \n\t" |
| 336 | "sra %[temp2], %[temp2], 16 \n\t" |
| 337 | "sra %[temp13], %[temp13], 16 \n\t" |
| 338 | "sra %[temp1], %[temp1], 16 \n\t" |
| 339 | "sra %[temp3], %[temp3], 16 \n\t" |
| 340 | "subu %[temp13], %[temp2], %[temp13] \n\t" |
| 341 | "addu %[temp3], %[temp1], %[temp3] \n\t" |
| 342 | "addu %[temp2], %[temp4], %[temp6] \n\t" |
| 343 | "subu %[temp4], %[temp4], %[temp6] \n\t" |
| 344 | "mul %[temp6], %[temp5], %[kC2] \n\t" |
| 345 | "mul %[temp1], %[temp7], %[kC1] \n\t" |
| 346 | "mul %[temp5], %[temp5], %[kC1] \n\t" |
| 347 | "mul %[temp7], %[temp7], %[kC2] \n\t" |
| 348 | "sra %[temp6], %[temp6], 16 \n\t" |
| 349 | "sra %[temp1], %[temp1], 16 \n\t" |
| 350 | "sra %[temp5], %[temp5], 16 \n\t" |
| 351 | "sra %[temp7], %[temp7], 16 \n\t" |
| 352 | "subu %[temp1], %[temp6], %[temp1] \n\t" |
| 353 | "addu %[temp7], %[temp5], %[temp7] \n\t" |
| 354 | "addu %[temp5], %[temp16], %[temp11] \n\t" |
| 355 | "subu %[temp16], %[temp16], %[temp11] \n\t" |
| 356 | "addu %[temp11], %[temp8], %[temp17] \n\t" |
| 357 | "subu %[temp8], %[temp8], %[temp17] \n\t" |
| 358 | "sra %[temp5], %[temp5], 3 \n\t" |
| 359 | "sra %[temp16], %[temp16], 3 \n\t" |
| 360 | "sra %[temp11], %[temp11], 3 \n\t" |
| 361 | "sra %[temp8], %[temp8], 3 \n\t" |
| 362 | "addu %[temp17], %[temp10], %[temp15] \n\t" |
| 363 | "subu %[temp10], %[temp10], %[temp15] \n\t" |
| 364 | "addu %[temp15], %[temp12], %[temp9] \n\t" |
| 365 | "subu %[temp12], %[temp12], %[temp9] \n\t" |
| 366 | "sra %[temp17], %[temp17], 3 \n\t" |
| 367 | "sra %[temp10], %[temp10], 3 \n\t" |
| 368 | "sra %[temp15], %[temp15], 3 \n\t" |
| 369 | "sra %[temp12], %[temp12], 3 \n\t" |
| 370 | "addu %[temp9], %[temp14], %[temp3] \n\t" |
| 371 | "subu %[temp14], %[temp14], %[temp3] \n\t" |
| 372 | "addu %[temp3], %[temp0], %[temp13] \n\t" |
| 373 | "subu %[temp0], %[temp0], %[temp13] \n\t" |
| 374 | "sra %[temp9], %[temp9], 3 \n\t" |
| 375 | "sra %[temp14], %[temp14], 3 \n\t" |
| 376 | "sra %[temp3], %[temp3], 3 \n\t" |
| 377 | "sra %[temp0], %[temp0], 3 \n\t" |
| 378 | "addu %[temp13], %[temp2], %[temp7] \n\t" |
| 379 | "subu %[temp2], %[temp2], %[temp7] \n\t" |
| 380 | "addu %[temp7], %[temp4], %[temp1] \n\t" |
| 381 | "subu %[temp4], %[temp4], %[temp1] \n\t" |
| 382 | "sra %[temp13], %[temp13], 3 \n\t" |
| 383 | "sra %[temp2], %[temp2], 3 \n\t" |
| 384 | "sra %[temp7], %[temp7], 3 \n\t" |
| 385 | "sra %[temp4], %[temp4], 3 \n\t" |
| 386 | "addiu %[temp6], $zero, 255 \n\t" |
| 387 | "lbu %[temp1], 0(%[dst]) \n\t" |
| 388 | "addu %[temp1], %[temp1], %[temp5] \n\t" |
| 389 | "sra %[temp5], %[temp1], 8 \n\t" |
| 390 | "sra %[temp18], %[temp1], 31 \n\t" |
| 391 | "beqz %[temp5], 1f \n\t" |
| 392 | "xor %[temp1], %[temp1], %[temp1] \n\t" |
| 393 | "movz %[temp1], %[temp6], %[temp18] \n\t" |
| 394 | "1: \n\t" |
| 395 | "lbu %[temp18], 1(%[dst]) \n\t" |
| 396 | "sb %[temp1], 0(%[dst]) \n\t" |
| 397 | "addu %[temp18], %[temp18], %[temp11] \n\t" |
| 398 | "sra %[temp11], %[temp18], 8 \n\t" |
| 399 | "sra %[temp1], %[temp18], 31 \n\t" |
| 400 | "beqz %[temp11], 2f \n\t" |
| 401 | "xor %[temp18], %[temp18], %[temp18] \n\t" |
| 402 | "movz %[temp18], %[temp6], %[temp1] \n\t" |
| 403 | "2: \n\t" |
| 404 | "lbu %[temp1], 2(%[dst]) \n\t" |
| 405 | "sb %[temp18], 1(%[dst]) \n\t" |
| 406 | "addu %[temp1], %[temp1], %[temp8] \n\t" |
| 407 | "sra %[temp8], %[temp1], 8 \n\t" |
| 408 | "sra %[temp18], %[temp1], 31 \n\t" |
| 409 | "beqz %[temp8], 3f \n\t" |
| 410 | "xor %[temp1], %[temp1], %[temp1] \n\t" |
| 411 | "movz %[temp1], %[temp6], %[temp18] \n\t" |
| 412 | "3: \n\t" |
| 413 | "lbu %[temp18], 3(%[dst]) \n\t" |
| 414 | "sb %[temp1], 2(%[dst]) \n\t" |
| 415 | "addu %[temp18], %[temp18], %[temp16] \n\t" |
| 416 | "sra %[temp16], %[temp18], 8 \n\t" |
| 417 | "sra %[temp1], %[temp18], 31 \n\t" |
| 418 | "beqz %[temp16], 4f \n\t" |
| 419 | "xor %[temp18], %[temp18], %[temp18] \n\t" |
| 420 | "movz %[temp18], %[temp6], %[temp1] \n\t" |
| 421 | "4: \n\t" |
| 422 | "sb %[temp18], 3(%[dst]) \n\t" |
| 423 | "lbu %[temp5], 32(%[dst]) \n\t" |
| 424 | "lbu %[temp8], 33(%[dst]) \n\t" |
| 425 | "lbu %[temp11], 34(%[dst]) \n\t" |
| 426 | "lbu %[temp16], 35(%[dst]) \n\t" |
| 427 | "addu %[temp5], %[temp5], %[temp17] \n\t" |
| 428 | "addu %[temp8], %[temp8], %[temp15] \n\t" |
| 429 | "addu %[temp11], %[temp11], %[temp12] \n\t" |
| 430 | "addu %[temp16], %[temp16], %[temp10] \n\t" |
| 431 | "sra %[temp18], %[temp5], 8 \n\t" |
| 432 | "sra %[temp1], %[temp5], 31 \n\t" |
| 433 | "beqz %[temp18], 5f \n\t" |
| 434 | "xor %[temp5], %[temp5], %[temp5] \n\t" |
| 435 | "movz %[temp5], %[temp6], %[temp1] \n\t" |
| 436 | "5: \n\t" |
| 437 | "sra %[temp18], %[temp8], 8 \n\t" |
| 438 | "sra %[temp1], %[temp8], 31 \n\t" |
| 439 | "beqz %[temp18], 6f \n\t" |
| 440 | "xor %[temp8], %[temp8], %[temp8] \n\t" |
| 441 | "movz %[temp8], %[temp6], %[temp1] \n\t" |
| 442 | "6: \n\t" |
| 443 | "sra %[temp18], %[temp11], 8 \n\t" |
| 444 | "sra %[temp1], %[temp11], 31 \n\t" |
| 445 | "sra %[temp17], %[temp16], 8 \n\t" |
| 446 | "sra %[temp15], %[temp16], 31 \n\t" |
| 447 | "beqz %[temp18], 7f \n\t" |
| 448 | "xor %[temp11], %[temp11], %[temp11] \n\t" |
| 449 | "movz %[temp11], %[temp6], %[temp1] \n\t" |
| 450 | "7: \n\t" |
| 451 | "beqz %[temp17], 8f \n\t" |
| 452 | "xor %[temp16], %[temp16], %[temp16] \n\t" |
| 453 | "movz %[temp16], %[temp6], %[temp15] \n\t" |
| 454 | "8: \n\t" |
| 455 | "sb %[temp5], 32(%[dst]) \n\t" |
| 456 | "sb %[temp8], 33(%[dst]) \n\t" |
| 457 | "sb %[temp11], 34(%[dst]) \n\t" |
| 458 | "sb %[temp16], 35(%[dst]) \n\t" |
| 459 | "lbu %[temp5], 64(%[dst]) \n\t" |
| 460 | "lbu %[temp8], 65(%[dst]) \n\t" |
| 461 | "lbu %[temp11], 66(%[dst]) \n\t" |
| 462 | "lbu %[temp16], 67(%[dst]) \n\t" |
| 463 | "addu %[temp5], %[temp5], %[temp9] \n\t" |
| 464 | "addu %[temp8], %[temp8], %[temp3] \n\t" |
| 465 | "addu %[temp11], %[temp11], %[temp0] \n\t" |
| 466 | "addu %[temp16], %[temp16], %[temp14] \n\t" |
| 467 | "sra %[temp18], %[temp5], 8 \n\t" |
| 468 | "sra %[temp1], %[temp5], 31 \n\t" |
| 469 | "sra %[temp17], %[temp8], 8 \n\t" |
| 470 | "sra %[temp15], %[temp8], 31 \n\t" |
| 471 | "sra %[temp12], %[temp11], 8 \n\t" |
| 472 | "sra %[temp10], %[temp11], 31 \n\t" |
| 473 | "sra %[temp9], %[temp16], 8 \n\t" |
| 474 | "sra %[temp3], %[temp16], 31 \n\t" |
| 475 | "beqz %[temp18], 9f \n\t" |
| 476 | "xor %[temp5], %[temp5], %[temp5] \n\t" |
| 477 | "movz %[temp5], %[temp6], %[temp1] \n\t" |
| 478 | "9: \n\t" |
| 479 | "beqz %[temp17], 10f \n\t" |
| 480 | "xor %[temp8], %[temp8], %[temp8] \n\t" |
| 481 | "movz %[temp8], %[temp6], %[temp15] \n\t" |
| 482 | "10: \n\t" |
| 483 | "beqz %[temp12], 11f \n\t" |
| 484 | "xor %[temp11], %[temp11], %[temp11] \n\t" |
| 485 | "movz %[temp11], %[temp6], %[temp10] \n\t" |
| 486 | "11: \n\t" |
| 487 | "beqz %[temp9], 12f \n\t" |
| 488 | "xor %[temp16], %[temp16], %[temp16] \n\t" |
| 489 | "movz %[temp16], %[temp6], %[temp3] \n\t" |
| 490 | "12: \n\t" |
| 491 | "sb %[temp5], 64(%[dst]) \n\t" |
| 492 | "sb %[temp8], 65(%[dst]) \n\t" |
| 493 | "sb %[temp11], 66(%[dst]) \n\t" |
| 494 | "sb %[temp16], 67(%[dst]) \n\t" |
| 495 | "lbu %[temp5], 96(%[dst]) \n\t" |
| 496 | "lbu %[temp8], 97(%[dst]) \n\t" |
| 497 | "lbu %[temp11], 98(%[dst]) \n\t" |
| 498 | "lbu %[temp16], 99(%[dst]) \n\t" |
| 499 | "addu %[temp5], %[temp5], %[temp13] \n\t" |
| 500 | "addu %[temp8], %[temp8], %[temp7] \n\t" |
| 501 | "addu %[temp11], %[temp11], %[temp4] \n\t" |
| 502 | "addu %[temp16], %[temp16], %[temp2] \n\t" |
| 503 | "sra %[temp18], %[temp5], 8 \n\t" |
| 504 | "sra %[temp1], %[temp5], 31 \n\t" |
| 505 | "sra %[temp17], %[temp8], 8 \n\t" |
| 506 | "sra %[temp15], %[temp8], 31 \n\t" |
| 507 | "sra %[temp12], %[temp11], 8 \n\t" |
| 508 | "sra %[temp10], %[temp11], 31 \n\t" |
| 509 | "sra %[temp9], %[temp16], 8 \n\t" |
| 510 | "sra %[temp3], %[temp16], 31 \n\t" |
| 511 | "beqz %[temp18], 13f \n\t" |
| 512 | "xor %[temp5], %[temp5], %[temp5] \n\t" |
| 513 | "movz %[temp5], %[temp6], %[temp1] \n\t" |
| 514 | "13: \n\t" |
| 515 | "beqz %[temp17], 14f \n\t" |
| 516 | "xor %[temp8], %[temp8], %[temp8] \n\t" |
| 517 | "movz %[temp8], %[temp6], %[temp15] \n\t" |
| 518 | "14: \n\t" |
| 519 | "beqz %[temp12], 15f \n\t" |
| 520 | "xor %[temp11], %[temp11], %[temp11] \n\t" |
| 521 | "movz %[temp11], %[temp6], %[temp10] \n\t" |
| 522 | "15: \n\t" |
| 523 | "beqz %[temp9], 16f \n\t" |
| 524 | "xor %[temp16], %[temp16], %[temp16] \n\t" |
| 525 | "movz %[temp16], %[temp6], %[temp3] \n\t" |
| 526 | "16: \n\t" |
| 527 | "sb %[temp5], 96(%[dst]) \n\t" |
| 528 | "sb %[temp8], 97(%[dst]) \n\t" |
| 529 | "sb %[temp11], 98(%[dst]) \n\t" |
| 530 | "sb %[temp16], 99(%[dst]) \n\t" |
| 531 | |
| 532 | : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), |
| 533 | [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 534 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8), |
| 535 | [temp9]"=&r"(temp9), [temp10]"=&r"(temp10), [temp11]"=&r"(temp11), |
| 536 | [temp12]"=&r"(temp12), [temp13]"=&r"(temp13), [temp14]"=&r"(temp14), |
| 537 | [temp15]"=&r"(temp15), [temp16]"=&r"(temp16), [temp17]"=&r"(temp17), |
| 538 | [temp18]"=&r"(temp18) |
| 539 | : [in]"r"(p_in), [kC1]"r"(kC1), [kC2]"r"(kC2), [dst]"r"(dst) |
| 540 | : "memory", "hi", "lo" |
| 541 | ); |
| 542 | } |
| 543 | |
| 544 | static void TransformTwo(const int16_t* in, uint8_t* dst, int do_two) { |
| 545 | TransformOne(in, dst); |
| 546 | if (do_two) { |
| 547 | TransformOne(in + 16, dst + 4); |
| 548 | } |
| 549 | } |
| 550 | |
| 551 | #endif // WEBP_USE_MIPS32 |
| 552 | |
| 553 | //------------------------------------------------------------------------------ |
| 554 | // Entry point |
| 555 | |
| 556 | extern void VP8DspInitMIPS32(void); |
| 557 | |
| 558 | void VP8DspInitMIPS32(void) { |
| 559 | #if defined(WEBP_USE_MIPS32) |
| 560 | VP8InitClipTables(); |
| 561 | |
| 562 | VP8Transform = TransformTwo; |
| 563 | |
| 564 | VP8VFilter16 = VFilter16; |
| 565 | VP8HFilter16 = HFilter16; |
| 566 | VP8VFilter8 = VFilter8; |
| 567 | VP8HFilter8 = HFilter8; |
| 568 | VP8VFilter16i = VFilter16i; |
| 569 | VP8HFilter16i = HFilter16i; |
| 570 | VP8VFilter8i = VFilter8i; |
| 571 | VP8HFilter8i = HFilter8i; |
| 572 | |
| 573 | VP8SimpleVFilter16 = SimpleVFilter16; |
| 574 | VP8SimpleHFilter16 = SimpleHFilter16; |
| 575 | VP8SimpleVFilter16i = SimpleVFilter16i; |
| 576 | VP8SimpleHFilter16i = SimpleHFilter16i; |
| 577 | #endif // WEBP_USE_MIPS32 |
| 578 | } |