| 1 | /* |
| 2 | * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at> |
| 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 <inttypes.h> |
| 22 | #include "config.h" |
| 23 | #include "libswscale/swscale.h" |
| 24 | #include "libswscale/swscale_internal.h" |
| 25 | #include "libavutil/attributes.h" |
| 26 | #include "libavutil/avassert.h" |
| 27 | #include "libavutil/intreadwrite.h" |
| 28 | #include "libavutil/x86/asm.h" |
| 29 | #include "libavutil/x86/cpu.h" |
| 30 | #include "libavutil/cpu.h" |
| 31 | #include "libavutil/pixdesc.h" |
| 32 | |
| 33 | #if HAVE_INLINE_ASM |
| 34 | |
| 35 | #define DITHER1XBPP |
| 36 | |
| 37 | DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL; |
| 38 | DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL; |
| 39 | DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL; |
| 40 | DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL; |
| 41 | |
| 42 | const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = { |
| 43 | 0x0103010301030103LL, |
| 44 | 0x0200020002000200LL,}; |
| 45 | |
| 46 | const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = { |
| 47 | 0x0602060206020602LL, |
| 48 | 0x0004000400040004LL,}; |
| 49 | |
| 50 | DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL; |
| 51 | DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL; |
| 52 | DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL; |
| 53 | DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL; |
| 54 | DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL; |
| 55 | DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL; |
| 56 | |
| 57 | DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL; |
| 58 | DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL; |
| 59 | DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL; |
| 60 | |
| 61 | DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL; |
| 62 | DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL; |
| 63 | DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL; |
| 64 | |
| 65 | |
| 66 | //MMX versions |
| 67 | #if HAVE_MMX_INLINE |
| 68 | #undef RENAME |
| 69 | #define COMPILE_TEMPLATE_MMXEXT 0 |
| 70 | #define RENAME(a) a ## _mmx |
| 71 | #include "swscale_template.c" |
| 72 | #endif |
| 73 | |
| 74 | // MMXEXT versions |
| 75 | #if HAVE_MMXEXT_INLINE |
| 76 | #undef RENAME |
| 77 | #undef COMPILE_TEMPLATE_MMXEXT |
| 78 | #define COMPILE_TEMPLATE_MMXEXT 1 |
| 79 | #define RENAME(a) a ## _mmxext |
| 80 | #include "swscale_template.c" |
| 81 | #endif |
| 82 | |
| 83 | void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex, |
| 84 | int lastInLumBuf, int lastInChrBuf) |
| 85 | { |
| 86 | const int dstH= c->dstH; |
| 87 | const int flags= c->flags; |
| 88 | int16_t **lumPixBuf= c->lumPixBuf; |
| 89 | int16_t **chrUPixBuf= c->chrUPixBuf; |
| 90 | int16_t **alpPixBuf= c->alpPixBuf; |
| 91 | const int vLumBufSize= c->vLumBufSize; |
| 92 | const int vChrBufSize= c->vChrBufSize; |
| 93 | int32_t *vLumFilterPos= c->vLumFilterPos; |
| 94 | int32_t *vChrFilterPos= c->vChrFilterPos; |
| 95 | int16_t *vLumFilter= c->vLumFilter; |
| 96 | int16_t *vChrFilter= c->vChrFilter; |
| 97 | int32_t *lumMmxFilter= c->lumMmxFilter; |
| 98 | int32_t *chrMmxFilter= c->chrMmxFilter; |
| 99 | int32_t av_unused *alpMmxFilter= c->alpMmxFilter; |
| 100 | const int vLumFilterSize= c->vLumFilterSize; |
| 101 | const int vChrFilterSize= c->vChrFilterSize; |
| 102 | const int chrDstY= dstY>>c->chrDstVSubSample; |
| 103 | const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input |
| 104 | const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input |
| 105 | |
| 106 | c->blueDither= ff_dither8[dstY&1]; |
| 107 | if (c->dstFormat == AV_PIX_FMT_RGB555 || c->dstFormat == AV_PIX_FMT_BGR555) |
| 108 | c->greenDither= ff_dither8[dstY&1]; |
| 109 | else |
| 110 | c->greenDither= ff_dither4[dstY&1]; |
| 111 | c->redDither= ff_dither8[(dstY+1)&1]; |
| 112 | if (dstY < dstH - 2) { |
| 113 | const int16_t **lumSrcPtr= (const int16_t **)(void*) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize; |
| 114 | const int16_t **chrUSrcPtr= (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize; |
| 115 | const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL; |
| 116 | int i; |
| 117 | |
| 118 | if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) { |
| 119 | const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * vLumBufSize; |
| 120 | int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - firstLumSrcY, vLumFilterSize); |
| 121 | for (i = 0; i < neg; i++) |
| 122 | tmpY[i] = lumSrcPtr[neg]; |
| 123 | for ( ; i < end; i++) |
| 124 | tmpY[i] = lumSrcPtr[i]; |
| 125 | for ( ; i < vLumFilterSize; i++) |
| 126 | tmpY[i] = tmpY[i-1]; |
| 127 | lumSrcPtr = tmpY; |
| 128 | |
| 129 | if (alpSrcPtr) { |
| 130 | const int16_t **tmpA = (const int16_t **) alpPixBuf + 2 * vLumBufSize; |
| 131 | for (i = 0; i < neg; i++) |
| 132 | tmpA[i] = alpSrcPtr[neg]; |
| 133 | for ( ; i < end; i++) |
| 134 | tmpA[i] = alpSrcPtr[i]; |
| 135 | for ( ; i < vLumFilterSize; i++) |
| 136 | tmpA[i] = tmpA[i - 1]; |
| 137 | alpSrcPtr = tmpA; |
| 138 | } |
| 139 | } |
| 140 | if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->chrSrcH) { |
| 141 | const int16_t **tmpU = (const int16_t **) chrUPixBuf + 2 * vChrBufSize; |
| 142 | int neg = -firstChrSrcY, i, end = FFMIN(c->chrSrcH - firstChrSrcY, vChrFilterSize); |
| 143 | for (i = 0; i < neg; i++) { |
| 144 | tmpU[i] = chrUSrcPtr[neg]; |
| 145 | } |
| 146 | for ( ; i < end; i++) { |
| 147 | tmpU[i] = chrUSrcPtr[i]; |
| 148 | } |
| 149 | for ( ; i < vChrFilterSize; i++) { |
| 150 | tmpU[i] = tmpU[i - 1]; |
| 151 | } |
| 152 | chrUSrcPtr = tmpU; |
| 153 | } |
| 154 | |
| 155 | if (flags & SWS_ACCURATE_RND) { |
| 156 | int s= APCK_SIZE / 8; |
| 157 | for (i=0; i<vLumFilterSize; i+=2) { |
| 158 | *(const void**)&lumMmxFilter[s*i ]= lumSrcPtr[i ]; |
| 159 | *(const void**)&lumMmxFilter[s*i+APCK_PTR2/4 ]= lumSrcPtr[i+(vLumFilterSize>1)]; |
| 160 | lumMmxFilter[s*i+APCK_COEF/4 ]= |
| 161 | lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i ] |
| 162 | + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0); |
| 163 | if (CONFIG_SWSCALE_ALPHA && alpPixBuf) { |
| 164 | *(const void**)&alpMmxFilter[s*i ]= alpSrcPtr[i ]; |
| 165 | *(const void**)&alpMmxFilter[s*i+APCK_PTR2/4 ]= alpSrcPtr[i+(vLumFilterSize>1)]; |
| 166 | alpMmxFilter[s*i+APCK_COEF/4 ]= |
| 167 | alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4 ]; |
| 168 | } |
| 169 | } |
| 170 | for (i=0; i<vChrFilterSize; i+=2) { |
| 171 | *(const void**)&chrMmxFilter[s*i ]= chrUSrcPtr[i ]; |
| 172 | *(const void**)&chrMmxFilter[s*i+APCK_PTR2/4 ]= chrUSrcPtr[i+(vChrFilterSize>1)]; |
| 173 | chrMmxFilter[s*i+APCK_COEF/4 ]= |
| 174 | chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i ] |
| 175 | + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0); |
| 176 | } |
| 177 | } else { |
| 178 | for (i=0; i<vLumFilterSize; i++) { |
| 179 | *(const void**)&lumMmxFilter[4*i+0]= lumSrcPtr[i]; |
| 180 | lumMmxFilter[4*i+2]= |
| 181 | lumMmxFilter[4*i+3]= |
| 182 | ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001U; |
| 183 | if (CONFIG_SWSCALE_ALPHA && alpPixBuf) { |
| 184 | *(const void**)&alpMmxFilter[4*i+0]= alpSrcPtr[i]; |
| 185 | alpMmxFilter[4*i+2]= |
| 186 | alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2]; |
| 187 | } |
| 188 | } |
| 189 | for (i=0; i<vChrFilterSize; i++) { |
| 190 | *(const void**)&chrMmxFilter[4*i+0]= chrUSrcPtr[i]; |
| 191 | chrMmxFilter[4*i+2]= |
| 192 | chrMmxFilter[4*i+3]= |
| 193 | ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001U; |
| 194 | } |
| 195 | } |
| 196 | } |
| 197 | } |
| 198 | |
| 199 | #if HAVE_MMXEXT |
| 200 | static void yuv2yuvX_sse3(const int16_t *filter, int filterSize, |
| 201 | const int16_t **src, uint8_t *dest, int dstW, |
| 202 | const uint8_t *dither, int offset) |
| 203 | { |
| 204 | if(((int)dest) & 15){ |
| 205 | yuv2yuvX_mmxext(filter, filterSize, src, dest, dstW, dither, offset); |
| 206 | return; |
| 207 | } |
| 208 | filterSize--; |
| 209 | #define MAIN_FUNCTION \ |
| 210 | "pxor %%xmm0, %%xmm0 \n\t" \ |
| 211 | "punpcklbw %%xmm0, %%xmm3 \n\t" \ |
| 212 | "movd %4, %%xmm1 \n\t" \ |
| 213 | "punpcklwd %%xmm1, %%xmm1 \n\t" \ |
| 214 | "punpckldq %%xmm1, %%xmm1 \n\t" \ |
| 215 | "punpcklqdq %%xmm1, %%xmm1 \n\t" \ |
| 216 | "psllw $3, %%xmm1 \n\t" \ |
| 217 | "paddw %%xmm1, %%xmm3 \n\t" \ |
| 218 | "psraw $4, %%xmm3 \n\t" \ |
| 219 | "movdqa %%xmm3, %%xmm4 \n\t" \ |
| 220 | "movdqa %%xmm3, %%xmm7 \n\t" \ |
| 221 | "movl %3, %%ecx \n\t" \ |
| 222 | "mov %0, %%"REG_d" \n\t"\ |
| 223 | "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
| 224 | ".p2align 4 \n\t" /* FIXME Unroll? */\ |
| 225 | "1: \n\t"\ |
| 226 | "movddup 8(%%"REG_d"), %%xmm0 \n\t" /* filterCoeff */\ |
| 227 | "movdqa (%%"REG_S", %%"REG_c", 2), %%xmm2 \n\t" /* srcData */\ |
| 228 | "movdqa 16(%%"REG_S", %%"REG_c", 2), %%xmm5 \n\t" /* srcData */\ |
| 229 | "add $16, %%"REG_d" \n\t"\ |
| 230 | "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
| 231 | "test %%"REG_S", %%"REG_S" \n\t"\ |
| 232 | "pmulhw %%xmm0, %%xmm2 \n\t"\ |
| 233 | "pmulhw %%xmm0, %%xmm5 \n\t"\ |
| 234 | "paddw %%xmm2, %%xmm3 \n\t"\ |
| 235 | "paddw %%xmm5, %%xmm4 \n\t"\ |
| 236 | " jnz 1b \n\t"\ |
| 237 | "psraw $3, %%xmm3 \n\t"\ |
| 238 | "psraw $3, %%xmm4 \n\t"\ |
| 239 | "packuswb %%xmm4, %%xmm3 \n\t"\ |
| 240 | "movntdq %%xmm3, (%1, %%"REG_c")\n\t"\ |
| 241 | "add $16, %%"REG_c" \n\t"\ |
| 242 | "cmp %2, %%"REG_c" \n\t"\ |
| 243 | "movdqa %%xmm7, %%xmm3 \n\t" \ |
| 244 | "movdqa %%xmm7, %%xmm4 \n\t" \ |
| 245 | "mov %0, %%"REG_d" \n\t"\ |
| 246 | "mov (%%"REG_d"), %%"REG_S" \n\t"\ |
| 247 | "jb 1b \n\t" |
| 248 | |
| 249 | if (offset) { |
| 250 | __asm__ volatile( |
| 251 | "movq %5, %%xmm3 \n\t" |
| 252 | "movdqa %%xmm3, %%xmm4 \n\t" |
| 253 | "psrlq $24, %%xmm3 \n\t" |
| 254 | "psllq $40, %%xmm4 \n\t" |
| 255 | "por %%xmm4, %%xmm3 \n\t" |
| 256 | MAIN_FUNCTION |
| 257 | :: "g" (filter), |
| 258 | "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset), |
| 259 | "m"(filterSize), "m"(((uint64_t *) dither)[0]) |
| 260 | : XMM_CLOBBERS("%xmm0" , "%xmm1" , "%xmm2" , "%xmm3" , "%xmm4" , "%xmm5" , "%xmm7" ,) |
| 261 | "%"REG_d, "%"REG_S, "%"REG_c |
| 262 | ); |
| 263 | } else { |
| 264 | __asm__ volatile( |
| 265 | "movq %5, %%xmm3 \n\t" |
| 266 | MAIN_FUNCTION |
| 267 | :: "g" (filter), |
| 268 | "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset), |
| 269 | "m"(filterSize), "m"(((uint64_t *) dither)[0]) |
| 270 | : XMM_CLOBBERS("%xmm0" , "%xmm1" , "%xmm2" , "%xmm3" , "%xmm4" , "%xmm5" , "%xmm7" ,) |
| 271 | "%"REG_d, "%"REG_S, "%"REG_c |
| 272 | ); |
| 273 | } |
| 274 | } |
| 275 | #endif |
| 276 | |
| 277 | #endif /* HAVE_INLINE_ASM */ |
| 278 | |
| 279 | #define SCALE_FUNC(filter_n, from_bpc, to_bpc, opt) \ |
| 280 | void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt( \ |
| 281 | SwsContext *c, int16_t *data, \ |
| 282 | int dstW, const uint8_t *src, \ |
| 283 | const int16_t *filter, \ |
| 284 | const int32_t *filterPos, int filterSize) |
| 285 | |
| 286 | #define SCALE_FUNCS(filter_n, opt) \ |
| 287 | SCALE_FUNC(filter_n, 8, 15, opt); \ |
| 288 | SCALE_FUNC(filter_n, 9, 15, opt); \ |
| 289 | SCALE_FUNC(filter_n, 10, 15, opt); \ |
| 290 | SCALE_FUNC(filter_n, 12, 15, opt); \ |
| 291 | SCALE_FUNC(filter_n, 14, 15, opt); \ |
| 292 | SCALE_FUNC(filter_n, 16, 15, opt); \ |
| 293 | SCALE_FUNC(filter_n, 8, 19, opt); \ |
| 294 | SCALE_FUNC(filter_n, 9, 19, opt); \ |
| 295 | SCALE_FUNC(filter_n, 10, 19, opt); \ |
| 296 | SCALE_FUNC(filter_n, 12, 19, opt); \ |
| 297 | SCALE_FUNC(filter_n, 14, 19, opt); \ |
| 298 | SCALE_FUNC(filter_n, 16, 19, opt) |
| 299 | |
| 300 | #define SCALE_FUNCS_MMX(opt) \ |
| 301 | SCALE_FUNCS(4, opt); \ |
| 302 | SCALE_FUNCS(8, opt); \ |
| 303 | SCALE_FUNCS(X, opt) |
| 304 | |
| 305 | #define SCALE_FUNCS_SSE(opt) \ |
| 306 | SCALE_FUNCS(4, opt); \ |
| 307 | SCALE_FUNCS(8, opt); \ |
| 308 | SCALE_FUNCS(X4, opt); \ |
| 309 | SCALE_FUNCS(X8, opt) |
| 310 | |
| 311 | #if ARCH_X86_32 |
| 312 | SCALE_FUNCS_MMX(mmx); |
| 313 | #endif |
| 314 | SCALE_FUNCS_SSE(sse2); |
| 315 | SCALE_FUNCS_SSE(ssse3); |
| 316 | SCALE_FUNCS_SSE(sse4); |
| 317 | |
| 318 | #define VSCALEX_FUNC(size, opt) \ |
| 319 | void ff_yuv2planeX_ ## size ## _ ## opt(const int16_t *filter, int filterSize, \ |
| 320 | const int16_t **src, uint8_t *dest, int dstW, \ |
| 321 | const uint8_t *dither, int offset) |
| 322 | #define VSCALEX_FUNCS(opt) \ |
| 323 | VSCALEX_FUNC(8, opt); \ |
| 324 | VSCALEX_FUNC(9, opt); \ |
| 325 | VSCALEX_FUNC(10, opt) |
| 326 | |
| 327 | #if ARCH_X86_32 |
| 328 | VSCALEX_FUNCS(mmxext); |
| 329 | #endif |
| 330 | VSCALEX_FUNCS(sse2); |
| 331 | VSCALEX_FUNCS(sse4); |
| 332 | VSCALEX_FUNC(16, sse4); |
| 333 | VSCALEX_FUNCS(avx); |
| 334 | |
| 335 | #define VSCALE_FUNC(size, opt) \ |
| 336 | void ff_yuv2plane1_ ## size ## _ ## opt(const int16_t *src, uint8_t *dst, int dstW, \ |
| 337 | const uint8_t *dither, int offset) |
| 338 | #define VSCALE_FUNCS(opt1, opt2) \ |
| 339 | VSCALE_FUNC(8, opt1); \ |
| 340 | VSCALE_FUNC(9, opt2); \ |
| 341 | VSCALE_FUNC(10, opt2); \ |
| 342 | VSCALE_FUNC(16, opt1) |
| 343 | |
| 344 | #if ARCH_X86_32 |
| 345 | VSCALE_FUNCS(mmx, mmxext); |
| 346 | #endif |
| 347 | VSCALE_FUNCS(sse2, sse2); |
| 348 | VSCALE_FUNC(16, sse4); |
| 349 | VSCALE_FUNCS(avx, avx); |
| 350 | |
| 351 | #define INPUT_Y_FUNC(fmt, opt) \ |
| 352 | void ff_ ## fmt ## ToY_ ## opt(uint8_t *dst, const uint8_t *src, \ |
| 353 | const uint8_t *unused1, const uint8_t *unused2, \ |
| 354 | int w, uint32_t *unused) |
| 355 | #define INPUT_UV_FUNC(fmt, opt) \ |
| 356 | void ff_ ## fmt ## ToUV_ ## opt(uint8_t *dstU, uint8_t *dstV, \ |
| 357 | const uint8_t *unused0, \ |
| 358 | const uint8_t *src1, \ |
| 359 | const uint8_t *src2, \ |
| 360 | int w, uint32_t *unused) |
| 361 | #define INPUT_FUNC(fmt, opt) \ |
| 362 | INPUT_Y_FUNC(fmt, opt); \ |
| 363 | INPUT_UV_FUNC(fmt, opt) |
| 364 | #define INPUT_FUNCS(opt) \ |
| 365 | INPUT_FUNC(uyvy, opt); \ |
| 366 | INPUT_FUNC(yuyv, opt); \ |
| 367 | INPUT_UV_FUNC(nv12, opt); \ |
| 368 | INPUT_UV_FUNC(nv21, opt); \ |
| 369 | INPUT_FUNC(rgba, opt); \ |
| 370 | INPUT_FUNC(bgra, opt); \ |
| 371 | INPUT_FUNC(argb, opt); \ |
| 372 | INPUT_FUNC(abgr, opt); \ |
| 373 | INPUT_FUNC(rgb24, opt); \ |
| 374 | INPUT_FUNC(bgr24, opt) |
| 375 | |
| 376 | #if ARCH_X86_32 |
| 377 | INPUT_FUNCS(mmx); |
| 378 | #endif |
| 379 | INPUT_FUNCS(sse2); |
| 380 | INPUT_FUNCS(ssse3); |
| 381 | INPUT_FUNCS(avx); |
| 382 | |
| 383 | av_cold void ff_sws_init_swscale_x86(SwsContext *c) |
| 384 | { |
| 385 | int cpu_flags = av_get_cpu_flags(); |
| 386 | |
| 387 | #if HAVE_MMX_INLINE |
| 388 | if (INLINE_MMX(cpu_flags)) |
| 389 | sws_init_swscale_mmx(c); |
| 390 | #endif |
| 391 | #if HAVE_MMXEXT_INLINE |
| 392 | if (INLINE_MMXEXT(cpu_flags)) |
| 393 | sws_init_swscale_mmxext(c); |
| 394 | if (cpu_flags & AV_CPU_FLAG_SSE3){ |
| 395 | if(c->use_mmx_vfilter && !(c->flags & SWS_ACCURATE_RND)) |
| 396 | c->yuv2planeX = yuv2yuvX_sse3; |
| 397 | } |
| 398 | #endif |
| 399 | |
| 400 | #define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \ |
| 401 | if (c->srcBpc == 8) { \ |
| 402 | hscalefn = c->dstBpc <= 14 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \ |
| 403 | ff_hscale8to19_ ## filtersize ## _ ## opt1; \ |
| 404 | } else if (c->srcBpc == 9) { \ |
| 405 | hscalefn = c->dstBpc <= 14 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \ |
| 406 | ff_hscale9to19_ ## filtersize ## _ ## opt1; \ |
| 407 | } else if (c->srcBpc == 10) { \ |
| 408 | hscalefn = c->dstBpc <= 14 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \ |
| 409 | ff_hscale10to19_ ## filtersize ## _ ## opt1; \ |
| 410 | } else if (c->srcBpc == 12) { \ |
| 411 | hscalefn = c->dstBpc <= 14 ? ff_hscale12to15_ ## filtersize ## _ ## opt2 : \ |
| 412 | ff_hscale12to19_ ## filtersize ## _ ## opt1; \ |
| 413 | } else if (c->srcBpc == 14 || ((c->srcFormat==AV_PIX_FMT_PAL8||isAnyRGB(c->srcFormat)) && av_pix_fmt_desc_get(c->srcFormat)->comp[0].depth_minus1<15)) { \ |
| 414 | hscalefn = c->dstBpc <= 14 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \ |
| 415 | ff_hscale14to19_ ## filtersize ## _ ## opt1; \ |
| 416 | } else { /* c->srcBpc == 16 */ \ |
| 417 | av_assert0(c->srcBpc == 16);\ |
| 418 | hscalefn = c->dstBpc <= 14 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \ |
| 419 | ff_hscale16to19_ ## filtersize ## _ ## opt1; \ |
| 420 | } \ |
| 421 | } while (0) |
| 422 | #define ASSIGN_MMX_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \ |
| 423 | switch (filtersize) { \ |
| 424 | case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \ |
| 425 | case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \ |
| 426 | default: ASSIGN_SCALE_FUNC2(hscalefn, X, opt1, opt2); break; \ |
| 427 | } |
| 428 | #define ASSIGN_VSCALEX_FUNC(vscalefn, opt, do_16_case, condition_8bit) \ |
| 429 | switch(c->dstBpc){ \ |
| 430 | case 16: do_16_case; break; \ |
| 431 | case 10: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_10_ ## opt; break; \ |
| 432 | case 9: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2planeX_9_ ## opt; break; \ |
| 433 | case 8: if ((condition_8bit) && !c->use_mmx_vfilter) vscalefn = ff_yuv2planeX_8_ ## opt; break; \ |
| 434 | } |
| 435 | #define ASSIGN_VSCALE_FUNC(vscalefn, opt1, opt2, opt2chk) \ |
| 436 | switch(c->dstBpc){ \ |
| 437 | case 16: if (!isBE(c->dstFormat)) vscalefn = ff_yuv2plane1_16_ ## opt1; break; \ |
| 438 | case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \ |
| 439 | case 9: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_ ## opt2; break; \ |
| 440 | case 8: vscalefn = ff_yuv2plane1_8_ ## opt1; break; \ |
| 441 | default: av_assert0(c->dstBpc>8); \ |
| 442 | } |
| 443 | #define case_rgb(x, X, opt) \ |
| 444 | case AV_PIX_FMT_ ## X: \ |
| 445 | c->lumToYV12 = ff_ ## x ## ToY_ ## opt; \ |
| 446 | if (!c->chrSrcHSubSample) \ |
| 447 | c->chrToYV12 = ff_ ## x ## ToUV_ ## opt; \ |
| 448 | break |
| 449 | #if ARCH_X86_32 |
| 450 | if (EXTERNAL_MMX(cpu_flags)) { |
| 451 | ASSIGN_MMX_SCALE_FUNC(c->hyScale, c->hLumFilterSize, mmx, mmx); |
| 452 | ASSIGN_MMX_SCALE_FUNC(c->hcScale, c->hChrFilterSize, mmx, mmx); |
| 453 | ASSIGN_VSCALE_FUNC(c->yuv2plane1, mmx, mmxext, cpu_flags & AV_CPU_FLAG_MMXEXT); |
| 454 | |
| 455 | switch (c->srcFormat) { |
| 456 | case AV_PIX_FMT_YA8: |
| 457 | c->lumToYV12 = ff_yuyvToY_mmx; |
| 458 | if (c->alpPixBuf) |
| 459 | c->alpToYV12 = ff_uyvyToY_mmx; |
| 460 | break; |
| 461 | case AV_PIX_FMT_YUYV422: |
| 462 | c->lumToYV12 = ff_yuyvToY_mmx; |
| 463 | c->chrToYV12 = ff_yuyvToUV_mmx; |
| 464 | break; |
| 465 | case AV_PIX_FMT_UYVY422: |
| 466 | c->lumToYV12 = ff_uyvyToY_mmx; |
| 467 | c->chrToYV12 = ff_uyvyToUV_mmx; |
| 468 | break; |
| 469 | case AV_PIX_FMT_NV12: |
| 470 | c->chrToYV12 = ff_nv12ToUV_mmx; |
| 471 | break; |
| 472 | case AV_PIX_FMT_NV21: |
| 473 | c->chrToYV12 = ff_nv21ToUV_mmx; |
| 474 | break; |
| 475 | case_rgb(rgb24, RGB24, mmx); |
| 476 | case_rgb(bgr24, BGR24, mmx); |
| 477 | case_rgb(bgra, BGRA, mmx); |
| 478 | case_rgb(rgba, RGBA, mmx); |
| 479 | case_rgb(abgr, ABGR, mmx); |
| 480 | case_rgb(argb, ARGB, mmx); |
| 481 | default: |
| 482 | break; |
| 483 | } |
| 484 | } |
| 485 | if (EXTERNAL_MMXEXT(cpu_flags)) { |
| 486 | ASSIGN_VSCALEX_FUNC(c->yuv2planeX, mmxext, , 1); |
| 487 | } |
| 488 | #endif /* ARCH_X86_32 */ |
| 489 | #define ASSIGN_SSE_SCALE_FUNC(hscalefn, filtersize, opt1, opt2) \ |
| 490 | switch (filtersize) { \ |
| 491 | case 4: ASSIGN_SCALE_FUNC2(hscalefn, 4, opt1, opt2); break; \ |
| 492 | case 8: ASSIGN_SCALE_FUNC2(hscalefn, 8, opt1, opt2); break; \ |
| 493 | default: if (filtersize & 4) ASSIGN_SCALE_FUNC2(hscalefn, X4, opt1, opt2); \ |
| 494 | else ASSIGN_SCALE_FUNC2(hscalefn, X8, opt1, opt2); \ |
| 495 | break; \ |
| 496 | } |
| 497 | if (EXTERNAL_SSE2(cpu_flags)) { |
| 498 | ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse2, sse2); |
| 499 | ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse2, sse2); |
| 500 | ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse2, , |
| 501 | HAVE_ALIGNED_STACK || ARCH_X86_64); |
| 502 | ASSIGN_VSCALE_FUNC(c->yuv2plane1, sse2, sse2, 1); |
| 503 | |
| 504 | switch (c->srcFormat) { |
| 505 | case AV_PIX_FMT_YA8: |
| 506 | c->lumToYV12 = ff_yuyvToY_sse2; |
| 507 | if (c->alpPixBuf) |
| 508 | c->alpToYV12 = ff_uyvyToY_sse2; |
| 509 | break; |
| 510 | case AV_PIX_FMT_YUYV422: |
| 511 | c->lumToYV12 = ff_yuyvToY_sse2; |
| 512 | c->chrToYV12 = ff_yuyvToUV_sse2; |
| 513 | break; |
| 514 | case AV_PIX_FMT_UYVY422: |
| 515 | c->lumToYV12 = ff_uyvyToY_sse2; |
| 516 | c->chrToYV12 = ff_uyvyToUV_sse2; |
| 517 | break; |
| 518 | case AV_PIX_FMT_NV12: |
| 519 | c->chrToYV12 = ff_nv12ToUV_sse2; |
| 520 | break; |
| 521 | case AV_PIX_FMT_NV21: |
| 522 | c->chrToYV12 = ff_nv21ToUV_sse2; |
| 523 | break; |
| 524 | case_rgb(rgb24, RGB24, sse2); |
| 525 | case_rgb(bgr24, BGR24, sse2); |
| 526 | case_rgb(bgra, BGRA, sse2); |
| 527 | case_rgb(rgba, RGBA, sse2); |
| 528 | case_rgb(abgr, ABGR, sse2); |
| 529 | case_rgb(argb, ARGB, sse2); |
| 530 | default: |
| 531 | break; |
| 532 | } |
| 533 | } |
| 534 | if (EXTERNAL_SSSE3(cpu_flags)) { |
| 535 | ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, ssse3, ssse3); |
| 536 | ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, ssse3, ssse3); |
| 537 | switch (c->srcFormat) { |
| 538 | case_rgb(rgb24, RGB24, ssse3); |
| 539 | case_rgb(bgr24, BGR24, ssse3); |
| 540 | default: |
| 541 | break; |
| 542 | } |
| 543 | } |
| 544 | if (EXTERNAL_SSE4(cpu_flags)) { |
| 545 | /* Xto15 don't need special sse4 functions */ |
| 546 | ASSIGN_SSE_SCALE_FUNC(c->hyScale, c->hLumFilterSize, sse4, ssse3); |
| 547 | ASSIGN_SSE_SCALE_FUNC(c->hcScale, c->hChrFilterSize, sse4, ssse3); |
| 548 | ASSIGN_VSCALEX_FUNC(c->yuv2planeX, sse4, |
| 549 | if (!isBE(c->dstFormat)) c->yuv2planeX = ff_yuv2planeX_16_sse4, |
| 550 | HAVE_ALIGNED_STACK || ARCH_X86_64); |
| 551 | if (c->dstBpc == 16 && !isBE(c->dstFormat)) |
| 552 | c->yuv2plane1 = ff_yuv2plane1_16_sse4; |
| 553 | } |
| 554 | |
| 555 | if (EXTERNAL_AVX(cpu_flags)) { |
| 556 | ASSIGN_VSCALEX_FUNC(c->yuv2planeX, avx, , |
| 557 | HAVE_ALIGNED_STACK || ARCH_X86_64); |
| 558 | ASSIGN_VSCALE_FUNC(c->yuv2plane1, avx, avx, 1); |
| 559 | |
| 560 | switch (c->srcFormat) { |
| 561 | case AV_PIX_FMT_YUYV422: |
| 562 | c->chrToYV12 = ff_yuyvToUV_avx; |
| 563 | break; |
| 564 | case AV_PIX_FMT_UYVY422: |
| 565 | c->chrToYV12 = ff_uyvyToUV_avx; |
| 566 | break; |
| 567 | case AV_PIX_FMT_NV12: |
| 568 | c->chrToYV12 = ff_nv12ToUV_avx; |
| 569 | break; |
| 570 | case AV_PIX_FMT_NV21: |
| 571 | c->chrToYV12 = ff_nv21ToUV_avx; |
| 572 | break; |
| 573 | case_rgb(rgb24, RGB24, avx); |
| 574 | case_rgb(bgr24, BGR24, avx); |
| 575 | case_rgb(bgra, BGRA, avx); |
| 576 | case_rgb(rgba, RGBA, avx); |
| 577 | case_rgb(abgr, ABGR, avx); |
| 578 | case_rgb(argb, ARGB, avx); |
| 579 | default: |
| 580 | break; |
| 581 | } |
| 582 | } |
| 583 | } |