Libav 0.7.1
|
00001 /* 00002 * software YUV to RGB converter 00003 * 00004 * Copyright (C) 2001-2007 Michael Niedermayer 00005 * (c) 2010 Konstantin Shishkov 00006 * 00007 * This file is part of Libav. 00008 * 00009 * Libav is free software; you can redistribute it and/or 00010 * modify it under the terms of the GNU Lesser General Public 00011 * License as published by the Free Software Foundation; either 00012 * version 2.1 of the License, or (at your option) any later version. 00013 * 00014 * Libav is distributed in the hope that it will be useful, 00015 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00016 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00017 * Lesser General Public License for more details. 00018 * 00019 * You should have received a copy of the GNU Lesser General Public 00020 * License along with Libav; if not, write to the Free Software 00021 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00022 */ 00023 00024 #undef MOVNTQ 00025 #undef EMMS 00026 #undef SFENCE 00027 00028 #if COMPILE_TEMPLATE_MMX2 00029 #define MOVNTQ "movntq" 00030 #define SFENCE "sfence" 00031 #else 00032 #define MOVNTQ "movq" 00033 #define SFENCE " # nop" 00034 #endif 00035 00036 #define REG_BLUE "0" 00037 #define REG_RED "1" 00038 #define REG_GREEN "2" 00039 #define REG_ALPHA "3" 00040 00041 #define YUV2RGB_LOOP(depth) \ 00042 h_size = (c->dstW + 7) & ~7; \ 00043 if (h_size * depth > FFABS(dstStride[0])) \ 00044 h_size -= 8; \ 00045 \ 00046 if (c->srcFormat == PIX_FMT_YUV422P) { \ 00047 srcStride[1] *= 2; \ 00048 srcStride[2] *= 2; \ 00049 } \ 00050 \ 00051 __asm__ volatile ("pxor %mm4, %mm4\n\t"); \ 00052 for (y = 0; y < srcSliceH; y++) { \ 00053 uint8_t *image = dst[0] + (y + srcSliceY) * dstStride[0]; \ 00054 const uint8_t *py = src[0] + y * srcStride[0]; \ 00055 const uint8_t *pu = src[1] + (y >> 1) * srcStride[1]; \ 00056 const uint8_t *pv = src[2] + (y >> 1) * srcStride[2]; \ 00057 x86_reg index = -h_size / 2; \ 00058 00059 #define YUV2RGB_INITIAL_LOAD \ 00060 __asm__ volatile ( \ 00061 "movq (%5, %0, 2), %%mm6\n\t" \ 00062 "movd (%2, %0), %%mm0\n\t" \ 00063 "movd (%3, %0), %%mm1\n\t" \ 00064 "1: \n\t" \ 00065 00066 /* YUV2RGB core 00067 * Conversion is performed in usual way: 00068 * R = Y' * Ycoef + Vred * V' 00069 * G = Y' * Ycoef + Vgreen * V' + Ugreen * U' 00070 * B = Y' * Ycoef + Ublue * U' 00071 * 00072 * where X' = X * 8 - Xoffset (multiplication is performed to increase 00073 * precision a bit). 00074 * Since it operates in YUV420 colorspace, Y component is additionally 00075 * split into Y1 and Y2 for even and odd pixels. 00076 * 00077 * Input: 00078 * mm0 - U (4 elems), mm1 - V (4 elems), mm6 - Y (8 elems), mm4 - zero register 00079 * Output: 00080 * mm1 - R, mm2 - G, mm0 - B 00081 */ 00082 #define YUV2RGB \ 00083 /* convert Y, U, V into Y1', Y2', U', V' */ \ 00084 "movq %%mm6, %%mm7\n\t" \ 00085 "punpcklbw %%mm4, %%mm0\n\t" \ 00086 "punpcklbw %%mm4, %%mm1\n\t" \ 00087 "pand "MANGLE(mmx_00ffw)", %%mm6\n\t" \ 00088 "psrlw $8, %%mm7\n\t" \ 00089 "psllw $3, %%mm0\n\t" \ 00090 "psllw $3, %%mm1\n\t" \ 00091 "psllw $3, %%mm6\n\t" \ 00092 "psllw $3, %%mm7\n\t" \ 00093 "psubsw "U_OFFSET"(%4), %%mm0\n\t" \ 00094 "psubsw "V_OFFSET"(%4), %%mm1\n\t" \ 00095 "psubw "Y_OFFSET"(%4), %%mm6\n\t" \ 00096 "psubw "Y_OFFSET"(%4), %%mm7\n\t" \ 00097 \ 00098 /* multiply by coefficients */ \ 00099 "movq %%mm0, %%mm2\n\t" \ 00100 "movq %%mm1, %%mm3\n\t" \ 00101 "pmulhw "UG_COEFF"(%4), %%mm2\n\t" \ 00102 "pmulhw "VG_COEFF"(%4), %%mm3\n\t" \ 00103 "pmulhw "Y_COEFF" (%4), %%mm6\n\t" \ 00104 "pmulhw "Y_COEFF" (%4), %%mm7\n\t" \ 00105 "pmulhw "UB_COEFF"(%4), %%mm0\n\t" \ 00106 "pmulhw "VR_COEFF"(%4), %%mm1\n\t" \ 00107 "paddsw %%mm3, %%mm2\n\t" \ 00108 /* now: mm0 = UB, mm1 = VR, mm2 = CG */ \ 00109 /* mm6 = Y1, mm7 = Y2 */ \ 00110 \ 00111 /* produce RGB */ \ 00112 "movq %%mm7, %%mm3\n\t" \ 00113 "movq %%mm7, %%mm5\n\t" \ 00114 "paddsw %%mm0, %%mm3\n\t" \ 00115 "paddsw %%mm1, %%mm5\n\t" \ 00116 "paddsw %%mm2, %%mm7\n\t" \ 00117 "paddsw %%mm6, %%mm0\n\t" \ 00118 "paddsw %%mm6, %%mm1\n\t" \ 00119 "paddsw %%mm6, %%mm2\n\t" \ 00120 00121 #define RGB_PACK_INTERLEAVE \ 00122 /* pack and interleave even/odd pixels */ \ 00123 "packuswb %%mm1, %%mm0\n\t" \ 00124 "packuswb %%mm5, %%mm3\n\t" \ 00125 "packuswb %%mm2, %%mm2\n\t" \ 00126 "movq %%mm0, %%mm1\n\n" \ 00127 "packuswb %%mm7, %%mm7\n\t" \ 00128 "punpcklbw %%mm3, %%mm0\n\t" \ 00129 "punpckhbw %%mm3, %%mm1\n\t" \ 00130 "punpcklbw %%mm7, %%mm2\n\t" \ 00131 00132 #define YUV2RGB_ENDLOOP(depth) \ 00133 "movq 8 (%5, %0, 2), %%mm6\n\t" \ 00134 "movd 4 (%3, %0), %%mm1\n\t" \ 00135 "movd 4 (%2, %0), %%mm0\n\t" \ 00136 "add $"AV_STRINGIFY(depth * 8)", %1\n\t" \ 00137 "add $4, %0\n\t" \ 00138 "js 1b\n\t" \ 00139 00140 #define YUV2RGB_OPERANDS \ 00141 : "+r" (index), "+r" (image) \ 00142 : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), \ 00143 "r" (py - 2*index) \ 00144 ); \ 00145 } \ 00146 00147 #define YUV2RGB_OPERANDS_ALPHA \ 00148 : "+r" (index), "+r" (image) \ 00149 : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), \ 00150 "r" (py - 2*index), "r" (pa - 2*index) \ 00151 ); \ 00152 } \ 00153 00154 #define YUV2RGB_ENDFUNC \ 00155 __asm__ volatile (SFENCE"\n\t" \ 00156 "emms \n\t"); \ 00157 return srcSliceH; \ 00158 00159 #define IF0(x) 00160 #define IF1(x) x 00161 00162 #define RGB_PACK16(gmask, is15) \ 00163 "pand "MANGLE(mmx_redmask)", %%mm0\n\t" \ 00164 "pand "MANGLE(mmx_redmask)", %%mm1\n\t" \ 00165 "movq %%mm2, %%mm3\n\t" \ 00166 "psllw $"AV_STRINGIFY(3-is15)", %%mm2\n\t" \ 00167 "psrlw $"AV_STRINGIFY(5+is15)", %%mm3\n\t" \ 00168 "psrlw $3, %%mm0\n\t" \ 00169 IF##is15("psrlw $1, %%mm1\n\t") \ 00170 "pand "MANGLE(pb_e0)", %%mm2\n\t" \ 00171 "pand "MANGLE(gmask)", %%mm3\n\t" \ 00172 "por %%mm2, %%mm0\n\t" \ 00173 "por %%mm3, %%mm1\n\t" \ 00174 "movq %%mm0, %%mm2\n\t" \ 00175 "punpcklbw %%mm1, %%mm0\n\t" \ 00176 "punpckhbw %%mm1, %%mm2\n\t" \ 00177 MOVNTQ " %%mm0, (%1)\n\t" \ 00178 MOVNTQ " %%mm2, 8(%1)\n\t" \ 00179 00180 #define DITHER_RGB \ 00181 "paddusb "BLUE_DITHER"(%4), %%mm0\n\t" \ 00182 "paddusb "GREEN_DITHER"(%4), %%mm2\n\t" \ 00183 "paddusb "RED_DITHER"(%4), %%mm1\n\t" \ 00184 00185 #if !COMPILE_TEMPLATE_MMX2 00186 static inline int RENAME(yuv420_rgb15)(SwsContext *c, const uint8_t *src[], 00187 int srcStride[], 00188 int srcSliceY, int srcSliceH, 00189 uint8_t *dst[], int dstStride[]) 00190 { 00191 int y, h_size; 00192 00193 YUV2RGB_LOOP(2) 00194 00195 #ifdef DITHER1XBPP 00196 c->blueDither = ff_dither8[y & 1]; 00197 c->greenDither = ff_dither8[y & 1]; 00198 c->redDither = ff_dither8[(y + 1) & 1]; 00199 #endif 00200 00201 YUV2RGB_INITIAL_LOAD 00202 YUV2RGB 00203 RGB_PACK_INTERLEAVE 00204 #ifdef DITHER1XBPP 00205 DITHER_RGB 00206 #endif 00207 RGB_PACK16(pb_03, 1) 00208 00209 YUV2RGB_ENDLOOP(2) 00210 YUV2RGB_OPERANDS 00211 YUV2RGB_ENDFUNC 00212 } 00213 00214 static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[], 00215 int srcStride[], 00216 int srcSliceY, int srcSliceH, 00217 uint8_t *dst[], int dstStride[]) 00218 { 00219 int y, h_size; 00220 00221 YUV2RGB_LOOP(2) 00222 00223 #ifdef DITHER1XBPP 00224 c->blueDither = ff_dither8[y & 1]; 00225 c->greenDither = ff_dither4[y & 1]; 00226 c->redDither = ff_dither8[(y + 1) & 1]; 00227 #endif 00228 00229 YUV2RGB_INITIAL_LOAD 00230 YUV2RGB 00231 RGB_PACK_INTERLEAVE 00232 #ifdef DITHER1XBPP 00233 DITHER_RGB 00234 #endif 00235 RGB_PACK16(pb_07, 0) 00236 00237 YUV2RGB_ENDLOOP(2) 00238 YUV2RGB_OPERANDS 00239 YUV2RGB_ENDFUNC 00240 } 00241 #endif /* !COMPILE_TEMPLATE_MMX2 */ 00242 00243 #define RGB_PACK24(blue, red)\ 00244 "packuswb %%mm3, %%mm0 \n" /* R0 R2 R4 R6 R1 R3 R5 R7 */\ 00245 "packuswb %%mm5, %%mm1 \n" /* B0 B2 B4 B6 B1 B3 B5 B7 */\ 00246 "packuswb %%mm7, %%mm2 \n" /* G0 G2 G4 G6 G1 G3 G5 G7 */\ 00247 "movq %%mm"red", %%mm3 \n"\ 00248 "movq %%mm"blue", %%mm6 \n"\ 00249 "psrlq $32, %%mm"red" \n" /* R1 R3 R5 R7 */\ 00250 "punpcklbw %%mm2, %%mm3 \n" /* R0 G0 R2 G2 R4 G4 R6 G6 */\ 00251 "punpcklbw %%mm"red", %%mm6 \n" /* B0 R1 B2 R3 B4 R5 B6 R7 */\ 00252 "movq %%mm3, %%mm5 \n"\ 00253 "punpckhbw %%mm"blue", %%mm2 \n" /* G1 B1 G3 B3 G5 B5 G7 B7 */\ 00254 "punpcklwd %%mm6, %%mm3 \n" /* R0 G0 B0 R1 R2 G2 B2 R3 */\ 00255 "punpckhwd %%mm6, %%mm5 \n" /* R4 G4 B4 R5 R6 G6 B6 R7 */\ 00256 RGB_PACK24_B 00257 00258 #if COMPILE_TEMPLATE_MMX2 00259 DECLARE_ASM_CONST(8, int16_t, mask1101[4]) = {-1,-1, 0,-1}; 00260 DECLARE_ASM_CONST(8, int16_t, mask0010[4]) = { 0, 0,-1, 0}; 00261 DECLARE_ASM_CONST(8, int16_t, mask0110[4]) = { 0,-1,-1, 0}; 00262 DECLARE_ASM_CONST(8, int16_t, mask1001[4]) = {-1, 0, 0,-1}; 00263 DECLARE_ASM_CONST(8, int16_t, mask0100[4]) = { 0,-1, 0, 0}; 00264 #undef RGB_PACK24_B 00265 #define RGB_PACK24_B\ 00266 "pshufw $0xc6, %%mm2, %%mm1 \n"\ 00267 "pshufw $0x84, %%mm3, %%mm6 \n"\ 00268 "pshufw $0x38, %%mm5, %%mm7 \n"\ 00269 "pand "MANGLE(mask1101)", %%mm6 \n" /* R0 G0 B0 R1 -- -- R2 G2 */\ 00270 "movq %%mm1, %%mm0 \n"\ 00271 "pand "MANGLE(mask0110)", %%mm7 \n" /* -- -- R6 G6 B6 R7 -- -- */\ 00272 "movq %%mm1, %%mm2 \n"\ 00273 "pand "MANGLE(mask0100)", %%mm1 \n" /* -- -- G3 B3 -- -- -- -- */\ 00274 "psrlq $48, %%mm3 \n" /* B2 R3 -- -- -- -- -- -- */\ 00275 "pand "MANGLE(mask0010)", %%mm0 \n" /* -- -- -- -- G1 B1 -- -- */\ 00276 "psllq $32, %%mm5 \n" /* -- -- -- -- R4 G4 B4 R5 */\ 00277 "pand "MANGLE(mask1001)", %%mm2 \n" /* G5 B5 -- -- -- -- G7 B7 */\ 00278 "por %%mm3, %%mm1 \n"\ 00279 "por %%mm6, %%mm0 \n"\ 00280 "por %%mm5, %%mm1 \n"\ 00281 "por %%mm7, %%mm2 \n"\ 00282 MOVNTQ" %%mm0, (%1) \n"\ 00283 MOVNTQ" %%mm1, 8(%1) \n"\ 00284 MOVNTQ" %%mm2, 16(%1) \n"\ 00285 00286 #else 00287 #undef RGB_PACK24_B 00288 #define RGB_PACK24_B\ 00289 "movd %%mm3, (%1) \n" /* R0 G0 B0 R1 */\ 00290 "movd %%mm2, 4(%1) \n" /* G1 B1 */\ 00291 "psrlq $32, %%mm3 \n"\ 00292 "psrlq $16, %%mm2 \n"\ 00293 "movd %%mm3, 6(%1) \n" /* R2 G2 B2 R3 */\ 00294 "movd %%mm2, 10(%1) \n" /* G3 B3 */\ 00295 "psrlq $16, %%mm2 \n"\ 00296 "movd %%mm5, 12(%1) \n" /* R4 G4 B4 R5 */\ 00297 "movd %%mm2, 16(%1) \n" /* G5 B5 */\ 00298 "psrlq $32, %%mm5 \n"\ 00299 "movd %%mm2, 20(%1) \n" /* -- -- G7 B7 */\ 00300 "movd %%mm5, 18(%1) \n" /* R6 G6 B6 R7 */\ 00301 00302 #endif 00303 00304 static inline int RENAME(yuv420_rgb24)(SwsContext *c, const uint8_t *src[], 00305 int srcStride[], 00306 int srcSliceY, int srcSliceH, 00307 uint8_t *dst[], int dstStride[]) 00308 { 00309 int y, h_size; 00310 00311 YUV2RGB_LOOP(3) 00312 00313 YUV2RGB_INITIAL_LOAD 00314 YUV2RGB 00315 RGB_PACK24(REG_BLUE, REG_RED) 00316 00317 YUV2RGB_ENDLOOP(3) 00318 YUV2RGB_OPERANDS 00319 YUV2RGB_ENDFUNC 00320 } 00321 00322 static inline int RENAME(yuv420_bgr24)(SwsContext *c, const uint8_t *src[], 00323 int srcStride[], 00324 int srcSliceY, int srcSliceH, 00325 uint8_t *dst[], int dstStride[]) 00326 { 00327 int y, h_size; 00328 00329 YUV2RGB_LOOP(3) 00330 00331 YUV2RGB_INITIAL_LOAD 00332 YUV2RGB 00333 RGB_PACK24(REG_RED, REG_BLUE) 00334 00335 YUV2RGB_ENDLOOP(3) 00336 YUV2RGB_OPERANDS 00337 YUV2RGB_ENDFUNC 00338 } 00339 00340 00341 #define SET_EMPTY_ALPHA \ 00342 "pcmpeqd %%mm"REG_ALPHA", %%mm"REG_ALPHA"\n\t" /* set alpha to 0xFF */ \ 00343 00344 #define LOAD_ALPHA \ 00345 "movq (%6, %0, 2), %%mm"REG_ALPHA"\n\t" \ 00346 00347 #define RGB_PACK32(red, green, blue, alpha) \ 00348 "movq %%mm"blue", %%mm5\n\t" \ 00349 "movq %%mm"red", %%mm6\n\t" \ 00350 "punpckhbw %%mm"green", %%mm5\n\t" \ 00351 "punpcklbw %%mm"green", %%mm"blue"\n\t" \ 00352 "punpckhbw %%mm"alpha", %%mm6\n\t" \ 00353 "punpcklbw %%mm"alpha", %%mm"red"\n\t" \ 00354 "movq %%mm"blue", %%mm"green"\n\t" \ 00355 "movq %%mm5, %%mm"alpha"\n\t" \ 00356 "punpcklwd %%mm"red", %%mm"blue"\n\t" \ 00357 "punpckhwd %%mm"red", %%mm"green"\n\t" \ 00358 "punpcklwd %%mm6, %%mm5\n\t" \ 00359 "punpckhwd %%mm6, %%mm"alpha"\n\t" \ 00360 MOVNTQ " %%mm"blue", 0(%1)\n\t" \ 00361 MOVNTQ " %%mm"green", 8(%1)\n\t" \ 00362 MOVNTQ " %%mm5, 16(%1)\n\t" \ 00363 MOVNTQ " %%mm"alpha", 24(%1)\n\t" \ 00364 00365 #if !COMPILE_TEMPLATE_MMX2 00366 static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[], 00367 int srcStride[], 00368 int srcSliceY, int srcSliceH, 00369 uint8_t *dst[], int dstStride[]) 00370 { 00371 int y, h_size; 00372 00373 YUV2RGB_LOOP(4) 00374 00375 YUV2RGB_INITIAL_LOAD 00376 YUV2RGB 00377 RGB_PACK_INTERLEAVE 00378 SET_EMPTY_ALPHA 00379 RGB_PACK32(REG_RED, REG_GREEN, REG_BLUE, REG_ALPHA) 00380 00381 YUV2RGB_ENDLOOP(4) 00382 YUV2RGB_OPERANDS 00383 YUV2RGB_ENDFUNC 00384 } 00385 00386 #if HAVE_7REGS && CONFIG_SWSCALE_ALPHA 00387 static inline int RENAME(yuva420_rgb32)(SwsContext *c, const uint8_t *src[], 00388 int srcStride[], 00389 int srcSliceY, int srcSliceH, 00390 uint8_t *dst[], int dstStride[]) 00391 { 00392 int y, h_size; 00393 00394 YUV2RGB_LOOP(4) 00395 00396 const uint8_t *pa = src[3] + y * srcStride[3]; 00397 YUV2RGB_INITIAL_LOAD 00398 YUV2RGB 00399 RGB_PACK_INTERLEAVE 00400 LOAD_ALPHA 00401 RGB_PACK32(REG_RED, REG_GREEN, REG_BLUE, REG_ALPHA) 00402 00403 YUV2RGB_ENDLOOP(4) 00404 YUV2RGB_OPERANDS_ALPHA 00405 YUV2RGB_ENDFUNC 00406 } 00407 #endif 00408 00409 static inline int RENAME(yuv420_bgr32)(SwsContext *c, const uint8_t *src[], 00410 int srcStride[], 00411 int srcSliceY, int srcSliceH, 00412 uint8_t *dst[], int dstStride[]) 00413 { 00414 int y, h_size; 00415 00416 YUV2RGB_LOOP(4) 00417 00418 YUV2RGB_INITIAL_LOAD 00419 YUV2RGB 00420 RGB_PACK_INTERLEAVE 00421 SET_EMPTY_ALPHA 00422 RGB_PACK32(REG_BLUE, REG_GREEN, REG_RED, REG_ALPHA) 00423 00424 YUV2RGB_ENDLOOP(4) 00425 YUV2RGB_OPERANDS 00426 YUV2RGB_ENDFUNC 00427 } 00428 00429 #if HAVE_7REGS && CONFIG_SWSCALE_ALPHA 00430 static inline int RENAME(yuva420_bgr32)(SwsContext *c, const uint8_t *src[], 00431 int srcStride[], 00432 int srcSliceY, int srcSliceH, 00433 uint8_t *dst[], int dstStride[]) 00434 { 00435 int y, h_size; 00436 00437 YUV2RGB_LOOP(4) 00438 00439 const uint8_t *pa = src[3] + y * srcStride[3]; 00440 YUV2RGB_INITIAL_LOAD 00441 YUV2RGB 00442 RGB_PACK_INTERLEAVE 00443 LOAD_ALPHA 00444 RGB_PACK32(REG_BLUE, REG_GREEN, REG_RED, REG_ALPHA) 00445 00446 YUV2RGB_ENDLOOP(4) 00447 YUV2RGB_OPERANDS_ALPHA 00448 YUV2RGB_ENDFUNC 00449 } 00450 #endif 00451 00452 #endif /* !COMPILE_TEMPLATE_MMX2 */