FFmpeg: libswscale/yuv2rgb_template.c Source File

00001 /*
00002  * yuv2rgb_mmx.c, software YUV to RGB converter with Intel MMX "technology"
00003  *
00004  * Copyright (C) 2000, Silicon Integrated System Corp
00005  *
00006  * Author: Olie Lho <ollie@sis.com.tw>
00007  *
00008  * 15,24 bpp and dithering from Michael Niedermayer (michaelni@gmx.at)
00009  * MMX/MMX2 Template stuff from Michael Niedermayer (needed for fast movntq support)
00010  * context / deglobalize stuff by Michael Niedermayer
00011  *
00012  * This file is part of mpeg2dec, a free MPEG-2 video decoder
00013  *
00014  * mpeg2dec is free software; you can redistribute it and/or modify
00015  * it under the terms of the GNU General Public License as published by
00016  * the Free Software Foundation; either version 2, or (at your option)
00017  * any later version.
00018  *
00019  * mpeg2dec is distributed in the hope that it will be useful,
00020  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00021  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
00022  * GNU General Public License for more details.
00023  *
00024  * You should have received a copy of the GNU General Public License
00025  * along with mpeg2dec; if not, write to the Free Software
00026  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00027  */
00028 
00029 #undef MOVNTQ
00030 #undef EMMS
00031 #undef SFENCE
00032 
00033 #if HAVE_AMD3DNOW
00034 /* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */
00035 #define EMMS     "femms"
00036 #else
00037 #define EMMS     "emms"
00038 #endif
00039 
00040 #if HAVE_MMX2
00041 #define MOVNTQ "movntq"
00042 #define SFENCE "sfence"
00043 #else
00044 #define MOVNTQ "movq"
00045 #define SFENCE "/nop"
00046 #endif
00047 
00048 #define YUV2RGB \
00049     /* Do the multiply part of the conversion for even and odd pixels,
00050        register usage:
00051        mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
00052        mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
00053        mm6 -> Y even, mm7 -> Y odd */\
00054     /* convert the chroma part */\
00055     "punpcklbw %%mm4, %%mm0;" /* scatter 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
00056     "punpcklbw %%mm4, %%mm1;" /* scatter 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
00057 \
00058     "psllw $3, %%mm0;" /* Promote precision */ \
00059     "psllw $3, %%mm1;" /* Promote precision */ \
00060 \
00061     "psubsw "U_OFFSET"(%4), %%mm0;" /* Cb -= 128 */ \
00062     "psubsw "V_OFFSET"(%4), %%mm1;" /* Cr -= 128 */ \
00063 \
00064     "movq %%mm0, %%mm2;" /* Copy 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
00065     "movq %%mm1, %%mm3;" /* Copy 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
00066 \
00067     "pmulhw "UG_COEFF"(%4), %%mm2;" /* Mul Cb with green coeff -> Cb green */ \
00068     "pmulhw "VG_COEFF"(%4), %%mm3;" /* Mul Cr with green coeff -> Cr green */ \
00069 \
00070     "pmulhw "UB_COEFF"(%4), %%mm0;" /* Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0 */\
00071     "pmulhw "VR_COEFF"(%4), %%mm1;" /* Mul Cr -> Cred 00 r3 00 r2 00 r1 00 r0 */\
00072 \
00073     "paddsw %%mm3, %%mm2;" /* Cb green + Cr green -> Cgreen */\
00074 \
00075     /* convert the luma part */\
00076     "movq %%mm6, %%mm7;" /* Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */\
00077     "pand "MANGLE(mmx_00ffw)", %%mm6;" /* get Y even 00 Y6 00 Y4 00 Y2 00 Y0 */\
00078 \
00079     "psrlw $8, %%mm7;" /* get Y odd 00 Y7 00 Y5 00 Y3 00 Y1 */\
00080 \
00081     "psllw $3, %%mm6;" /* Promote precision */\
00082     "psllw $3, %%mm7;" /* Promote precision */\
00083 \
00084     "psubw "Y_OFFSET"(%4), %%mm6;" /* Y -= 16 */\
00085     "psubw "Y_OFFSET"(%4), %%mm7;" /* Y -= 16 */\
00086 \
00087     "pmulhw "Y_COEFF"(%4), %%mm6;" /* Mul 4 Y even 00 y6 00 y4 00 y2 00 y0 */\
00088     "pmulhw "Y_COEFF"(%4), %%mm7;" /* Mul 4 Y odd 00 y7 00 y5 00 y3 00 y1 */\
00089 \
00090     /* Do the addition part of the conversion for even and odd pixels,
00091        register usage:
00092        mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
00093        mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
00094        mm6 -> Y even, mm7 -> Y odd */\
00095     "movq %%mm0, %%mm3;" /* Copy Cblue */\
00096     "movq %%mm1, %%mm4;" /* Copy Cred */\
00097     "movq %%mm2, %%mm5;" /* Copy Cgreen */\
00098 \
00099     "paddsw %%mm6, %%mm0;" /* Y even + Cblue 00 B6 00 B4 00 B2 00 B0 */\
00100     "paddsw %%mm7, %%mm3;" /* Y odd + Cblue 00 B7 00 B5 00 B3 00 B1 */\
00101 \
00102     "paddsw %%mm6, %%mm1;" /* Y even + Cred 00 R6 00 R4 00 R2 00 R0 */\
00103     "paddsw %%mm7, %%mm4;" /* Y odd + Cred 00 R7 00 R5 00 R3 00 R1 */\
00104 \
00105     "paddsw %%mm6, %%mm2;" /* Y even + Cgreen 00 G6 00 G4 00 G2 00 G0 */\
00106     "paddsw %%mm7, %%mm5;" /* Y odd + Cgreen 00 G7 00 G5 00 G3 00 G1 */\
00107 \
00108     /* Limit RGB even to 0..255 */\
00109     "packuswb %%mm0, %%mm0;" /* B6 B4 B2 B0  B6 B4 B2 B0 */\
00110     "packuswb %%mm1, %%mm1;" /* R6 R4 R2 R0  R6 R4 R2 R0 */\
00111     "packuswb %%mm2, %%mm2;" /* G6 G4 G2 G0  G6 G4 G2 G0 */\
00112 \
00113     /* Limit RGB odd to 0..255 */\
00114     "packuswb %%mm3, %%mm3;" /* B7 B5 B3 B1  B7 B5 B3 B1 */\
00115     "packuswb %%mm4, %%mm4;" /* R7 R5 R3 R1  R7 R5 R3 R1 */\
00116     "packuswb %%mm5, %%mm5;" /* G7 G5 G3 G1  G7 G5 G3 G1 */\
00117 \
00118     /* Interleave RGB even and odd */\
00119     "punpcklbw %%mm3, %%mm0;" /* B7 B6 B5 B4 B3 B2 B1 B0 */\
00120     "punpcklbw %%mm4, %%mm1;" /* R7 R6 R5 R4 R3 R2 R1 R0 */\
00121     "punpcklbw %%mm5, %%mm2;" /* G7 G6 G5 G4 G3 G2 G1 G0 */\
00122 
00123 
00124 #define YUV422_UNSHIFT                   \
00125     if(c->srcFormat == PIX_FMT_YUV422P){ \
00126         srcStride[1] *= 2;               \
00127         srcStride[2] *= 2;               \
00128     }                                    \
00129 
00130 #define YUV2RGB_LOOP(depth)                                   \
00131     h_size= (c->dstW+7)&~7;                                   \
00132     if(h_size*depth > FFABS(dstStride[0])) h_size-=8;         \
00133 \
00134     __asm__ volatile ("pxor %mm4, %mm4;" /* zero mm4 */ );    \
00135     for (y= 0; y<srcSliceH; y++ ) {                           \
00136         uint8_t *image = dst[0] + (y+srcSliceY)*dstStride[0]; \
00137         uint8_t *py = src[0] + y*srcStride[0];                \
00138         uint8_t *pu = src[1] + (y>>1)*srcStride[1];           \
00139         uint8_t *pv = src[2] + (y>>1)*srcStride[2];           \
00140         long index= -h_size/2;                                \
00141 
00142 #define YUV2RGB_INIT                                                       \
00143         /* This MMX assembly code deals with a SINGLE scan line at a time, \
00144          * it converts 8 pixels in each iteration. */                      \
00145         __asm__ volatile (                                                 \
00146         /* load data for start of next scan line */                        \
00147         "movd    (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */ \
00148         "movd    (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */ \
00149         "movq (%5, %0, 2), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */ \
00150         /*                                                                 \
00151         ".balign 16     \n\t"                                              \
00152         */                                                                 \
00153         "1:             \n\t"                                              \
00154         /* No speed difference on my p3@500 with prefetch,                 \
00155          * if it is faster for anyone with -benchmark then tell me.        \
00156         PREFETCH" 64(%0) \n\t"                                             \
00157         PREFETCH" 64(%1) \n\t"                                             \
00158         PREFETCH" 64(%2) \n\t"                                             \
00159         */                                                                 \
00160 
00161 #define YUV2RGB_ENDLOOP(depth) \
00162         "add $"AV_STRINGIFY(depth*8)", %1    \n\t" \
00163         "add                       $4, %0    \n\t" \
00164         " js                       1b        \n\t" \
00165 \
00166         : "+r" (index), "+r" (image) \
00167         : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index) \
00168         ); \
00169     } \
00170     __asm__ volatile (EMMS); \
00171     return srcSliceH; \
00172 
00173 static inline int RENAME(yuv420_rgb16)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
00174                                        int srcSliceH, uint8_t* dst[], int dstStride[]){
00175     int y, h_size;
00176 
00177     YUV422_UNSHIFT
00178     YUV2RGB_LOOP(2)
00179 
00180         c->blueDither= ff_dither8[y&1];
00181         c->greenDither= ff_dither4[y&1];
00182         c->redDither= ff_dither8[(y+1)&1];
00183 
00184         YUV2RGB_INIT
00185         YUV2RGB
00186 
00187 #ifdef DITHER1XBPP
00188         "paddusb "BLUE_DITHER"(%4), %%mm0;"
00189         "paddusb "GREEN_DITHER"(%4), %%mm2;"
00190         "paddusb "RED_DITHER"(%4), %%mm1;"
00191 #endif
00192         /* mask unneeded bits off */
00193         "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
00194         "pand "MANGLE(mmx_grnmask)", %%mm2;" /* g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 */
00195         "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */
00196 
00197         "psrlw   $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
00198         "pxor %%mm4, %%mm4;" /* zero mm4 */
00199 
00200         "movq %%mm0, %%mm5;" /* Copy B7-B0 */
00201         "movq %%mm2, %%mm7;" /* Copy G7-G0 */
00202 
00203         /* convert RGB24 plane to RGB16 pack for pixel 0-3 */
00204         "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
00205         "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
00206 
00207         "psllw  $3, %%mm2;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
00208         "por %%mm2, %%mm0;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */
00209 
00210         "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
00211         MOVNTQ "      %%mm0, (%1);" /* store pixel 0-3 */
00212 
00213         /* convert RGB24 plane to RGB16 pack for pixel 0-3 */
00214         "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
00215         "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
00216 
00217         "psllw        $3, %%mm7;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
00218         "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
00219 
00220         "por       %%mm7, %%mm5;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */
00221         "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
00222 
00223         MOVNTQ "   %%mm5, 8 (%1);" /* store pixel 4-7 */
00224 
00225     YUV2RGB_ENDLOOP(2)
00226 }
00227 
00228 static inline int RENAME(yuv420_rgb15)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
00229                                        int srcSliceH, uint8_t* dst[], int dstStride[]){
00230     int y, h_size;
00231 
00232     YUV422_UNSHIFT
00233     YUV2RGB_LOOP(2)
00234 
00235         c->blueDither= ff_dither8[y&1];
00236         c->greenDither= ff_dither8[y&1];
00237         c->redDither= ff_dither8[(y+1)&1];
00238 
00239         YUV2RGB_INIT
00240         YUV2RGB
00241 
00242 #ifdef DITHER1XBPP
00243         "paddusb "BLUE_DITHER"(%4), %%mm0  \n\t"
00244         "paddusb "GREEN_DITHER"(%4), %%mm2  \n\t"
00245         "paddusb "RED_DITHER"(%4), %%mm1  \n\t"
00246 #endif
00247 
00248         /* mask unneeded bits off */
00249         "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
00250         "pand "MANGLE(mmx_redmask)", %%mm2;" /* g7g6g5g4 g3_0_0_0 g7g6g5g4 g3_0_0_0 */
00251         "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */
00252 
00253         "psrlw   $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
00254         "psrlw   $1, %%mm1;" /* 0_r7r6r5  r4r3_0_0 0_r7r6r5 r4r3_0_0 */
00255         "pxor %%mm4, %%mm4;" /* zero mm4 */
00256 
00257         "movq %%mm0, %%mm5;" /* Copy B7-B0 */
00258         "movq %%mm2, %%mm7;" /* Copy G7-G0 */
00259 
00260         /* convert RGB24 plane to RGB16 pack for pixel 0-3 */
00261         "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3_0_0_0 */
00262         "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
00263 
00264         "psllw  $2, %%mm2;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
00265         "por %%mm2, %%mm0;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */
00266 
00267         "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
00268         MOVNTQ "      %%mm0, (%1);"  /* store pixel 0-3 */
00269 
00270         /* convert RGB24 plane to RGB16 pack for pixel 0-3 */
00271         "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 0_g7g6g5 g4g3_0_0 */
00272         "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
00273 
00274         "psllw        $2, %%mm7;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
00275         "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
00276 
00277         "por       %%mm7, %%mm5;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */
00278         "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
00279 
00280         MOVNTQ " %%mm5, 8 (%1);" /* store pixel 4-7 */
00281 
00282     YUV2RGB_ENDLOOP(2)
00283 }
00284 
00285 static inline int RENAME(yuv420_rgb24)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
00286                                        int srcSliceH, uint8_t* dst[], int dstStride[]){
00287     int y, h_size;
00288 
00289     YUV422_UNSHIFT
00290     YUV2RGB_LOOP(3)
00291 
00292         YUV2RGB_INIT
00293         YUV2RGB
00294         /* mm0=B, %%mm2=G, %%mm1=R */
00295 #if HAVE_MMX2
00296         "movq "MANGLE(ff_M24A)", %%mm4     \n\t"
00297         "movq "MANGLE(ff_M24C)", %%mm7     \n\t"
00298         "pshufw $0x50, %%mm0, %%mm5     \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */
00299         "pshufw $0x50, %%mm2, %%mm3     \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */
00300         "pshufw $0x00, %%mm1, %%mm6     \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */
00301 
00302         "pand   %%mm4, %%mm5            \n\t" /*    B2        B1       B0 */
00303         "pand   %%mm4, %%mm3            \n\t" /*    G2        G1       G0 */
00304         "pand   %%mm7, %%mm6            \n\t" /*       R1        R0       */
00305 
00306         "psllq     $8, %%mm3            \n\t" /* G2        G1       G0    */
00307         "por    %%mm5, %%mm6            \n\t"
00308         "por    %%mm3, %%mm6            \n\t"
00309         MOVNTQ" %%mm6, (%1)             \n\t"
00310 
00311         "psrlq     $8, %%mm2            \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */
00312         "pshufw $0xA5, %%mm0, %%mm5     \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */
00313         "pshufw $0x55, %%mm2, %%mm3     \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */
00314         "pshufw $0xA5, %%mm1, %%mm6     \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */
00315 
00316         "pand "MANGLE(ff_M24B)", %%mm5     \n\t" /* B5       B4        B3    */
00317         "pand          %%mm7, %%mm3     \n\t" /*       G4        G3       */
00318         "pand          %%mm4, %%mm6     \n\t" /*    R4        R3       R2 */
00319 
00320         "por    %%mm5, %%mm3            \n\t" /* B5    G4 B4     G3 B3    */
00321         "por    %%mm3, %%mm6            \n\t"
00322         MOVNTQ" %%mm6, 8(%1)            \n\t"
00323 
00324         "pshufw $0xFF, %%mm0, %%mm5     \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */
00325         "pshufw $0xFA, %%mm2, %%mm3     \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */
00326         "pshufw $0xFA, %%mm1, %%mm6     \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */
00327         "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
00328 
00329         "pand          %%mm7, %%mm5     \n\t" /*       B7        B6       */
00330         "pand          %%mm4, %%mm3     \n\t" /*    G7        G6       G5 */
00331         "pand "MANGLE(ff_M24B)", %%mm6     \n\t" /* R7       R6        R5    */
00332         "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
00333 \
00334         "por          %%mm5, %%mm3      \n\t"
00335         "por          %%mm3, %%mm6      \n\t"
00336         MOVNTQ"       %%mm6, 16(%1)     \n\t"
00337         "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
00338         "pxor         %%mm4, %%mm4      \n\t"
00339 
00340 #else
00341 
00342         "pxor      %%mm4, %%mm4     \n\t"
00343         "movq      %%mm0, %%mm5     \n\t" /* B */
00344         "movq      %%mm1, %%mm6     \n\t" /* R */
00345         "punpcklbw %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */
00346         "punpcklbw %%mm4, %%mm1     \n\t" /* 0R0R0R0R 0 */
00347         "punpckhbw %%mm2, %%mm5     \n\t" /* GBGBGBGB 2 */
00348         "punpckhbw %%mm4, %%mm6     \n\t" /* 0R0R0R0R 2 */
00349         "movq      %%mm0, %%mm7     \n\t" /* GBGBGBGB 0 */
00350         "movq      %%mm5, %%mm3     \n\t" /* GBGBGBGB 2 */
00351         "punpcklwd %%mm1, %%mm7     \n\t" /* 0RGB0RGB 0 */
00352         "punpckhwd %%mm1, %%mm0     \n\t" /* 0RGB0RGB 1 */
00353         "punpcklwd %%mm6, %%mm5     \n\t" /* 0RGB0RGB 2 */
00354         "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */
00355 
00356         "movq      %%mm7, %%mm2     \n\t" /* 0RGB0RGB 0 */
00357         "movq      %%mm0, %%mm6     \n\t" /* 0RGB0RGB 1 */
00358         "movq      %%mm5, %%mm1     \n\t" /* 0RGB0RGB 2 */
00359         "movq      %%mm3, %%mm4     \n\t" /* 0RGB0RGB 3 */
00360 
00361         "psllq       $40, %%mm7     \n\t" /* RGB00000 0 */
00362         "psllq       $40, %%mm0     \n\t" /* RGB00000 1 */
00363         "psllq       $40, %%mm5     \n\t" /* RGB00000 2 */
00364         "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */
00365 
00366         "punpckhdq %%mm2, %%mm7     \n\t" /* 0RGBRGB0 0 */
00367         "punpckhdq %%mm6, %%mm0     \n\t" /* 0RGBRGB0 1 */
00368         "punpckhdq %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */
00369         "punpckhdq %%mm4, %%mm3     \n\t" /* 0RGBRGB0 3 */
00370 
00371         "psrlq        $8, %%mm7     \n\t" /* 00RGBRGB 0 */
00372         "movq      %%mm0, %%mm6     \n\t" /* 0RGBRGB0 1 */
00373         "psllq       $40, %%mm0     \n\t" /* GB000000 1 */
00374         "por       %%mm0, %%mm7     \n\t" /* GBRGBRGB 0 */
00375         MOVNTQ"    %%mm7, (%1)      \n\t"
00376 
00377         "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
00378 
00379         "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */
00380         "movq      %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */
00381         "psllq       $24, %%mm5     \n\t" /* BRGB0000 2 */
00382         "por       %%mm5, %%mm6     \n\t" /* BRGBRGBR 1 */
00383         MOVNTQ"    %%mm6, 8(%1)     \n\t"
00384 
00385         "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
00386 
00387         "psrlq       $40, %%mm1     \n\t" /* 000000RG 2 */
00388         "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */
00389         "por       %%mm3, %%mm1     \n\t" /* RGBRGBRG 2 */
00390         MOVNTQ"    %%mm1, 16(%1)    \n\t"
00391 
00392         "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
00393         "pxor      %%mm4, %%mm4     \n\t"
00394 #endif
00395 
00396     YUV2RGB_ENDLOOP(3)
00397 }
00398 
00399 #define RGB_PLANAR2PACKED32                                             \
00400     /* convert RGB plane to RGB packed format,                          \
00401        mm0 ->  B, mm1 -> R, mm2 -> G, mm3 -> A,                         \
00402        mm4 -> GB, mm5 -> AR pixel 4-7,                                  \
00403        mm6 -> GB, mm7 -> AR pixel 0-3 */                                \
00404     "movq      %%mm0, %%mm6;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */           \
00405     "movq      %%mm1, %%mm7;"   /* R7 R6 R5 R4 R3 R2 R1 R0 */           \
00406 \
00407     "movq      %%mm0, %%mm4;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */           \
00408     "movq      %%mm1, %%mm5;"   /* R7 R6 R5 R4 R3 R2 R1 R0 */           \
00409 \
00410     "punpcklbw %%mm2, %%mm6;"   /* G3 B3 G2 B2 G1 B1 G0 B0 */           \
00411     "punpcklbw %%mm3, %%mm7;"   /* A3 R3 A2 R2 A1 R1 A0 R0 */           \
00412 \
00413     "punpcklwd %%mm7, %%mm6;"   /* A1 R1 B1 G1 A0 R0 B0 G0 */           \
00414     MOVNTQ "   %%mm6, (%1);"    /* Store ARGB1 ARGB0 */                 \
00415 \
00416     "movq      %%mm0, %%mm6;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */           \
00417     "punpcklbw %%mm2, %%mm6;"   /* G3 B3 G2 B2 G1 B1 G0 B0 */           \
00418 \
00419     "punpckhwd %%mm7, %%mm6;"   /* A3 R3 G3 B3 A2 R2 B3 G2 */           \
00420     MOVNTQ "   %%mm6, 8 (%1);"  /* Store ARGB3 ARGB2 */                 \
00421 \
00422     "punpckhbw %%mm2, %%mm4;"   /* G7 B7 G6 B6 G5 B5 G4 B4 */           \
00423     "punpckhbw %%mm3, %%mm5;"   /* A7 R7 A6 R6 A5 R5 A4 R4 */           \
00424 \
00425     "punpcklwd %%mm5, %%mm4;"   /* A5 R5 B5 G5 A4 R4 B4 G4 */           \
00426     MOVNTQ "   %%mm4, 16 (%1);" /* Store ARGB5 ARGB4 */                 \
00427 \
00428     "movq      %%mm0, %%mm4;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */           \
00429     "punpckhbw %%mm2, %%mm4;"   /* G7 B7 G6 B6 G5 B5 G4 B4 */           \
00430 \
00431     "punpckhwd %%mm5, %%mm4;"   /* A7 R7 G7 B7 A6 R6 B6 G6 */           \
00432     MOVNTQ "   %%mm4, 24 (%1);" /* Store ARGB7 ARGB6 */                 \
00433 \
00434     "movd 4 (%2, %0), %%mm0;"   /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */ \
00435     "movd 4 (%3, %0), %%mm1;"   /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */ \
00436 \
00437     "pxor         %%mm4, %%mm4;" /* zero mm4 */                         \
00438     "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */ \
00439 
00440 static inline int RENAME(yuv420_rgb32)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
00441                                        int srcSliceH, uint8_t* dst[], int dstStride[]){
00442     int y, h_size;
00443 
00444     YUV422_UNSHIFT
00445     YUV2RGB_LOOP(4)
00446 
00447         YUV2RGB_INIT
00448         YUV2RGB
00449         "pcmpeqd   %%mm3, %%mm3;"   /* fill mm3 */
00450         RGB_PLANAR2PACKED32
00451 
00452     YUV2RGB_ENDLOOP(4)
00453 }