Statistics
| Branch: | Revision:

ffmpeg / libswscale / yuv2rgb_template.c @ c1578479

History | View | Annotate | Download (21.6 KB)

1
/*
2
 * yuv2rgb_mmx.c, Software YUV to RGB coverter with Intel MMX "technology"
3
 *
4
 * Copyright (C) 2000, Silicon Integrated System Corp.
5
 * All Rights Reserved.
6
 *
7
 * Author: Olie Lho <ollie@sis.com.tw>
8
 *
9
 * 15,24 bpp and dithering from Michael Niedermayer (michaelni@gmx.at)
10
 * MMX/MMX2 Template stuff from Michael Niedermayer (needed for fast movntq support)
11
 * context / deglobalize stuff by Michael Niedermayer
12
 *
13
 * This file is part of mpeg2dec, a free MPEG-2 video decoder
14
 *
15
 * mpeg2dec is free software; you can redistribute it and/or modify
16
 * it under the terms of the GNU General Public License as published by
17
 * the Free Software Foundation; either version 2, or (at your option)
18
 * any later version.
19
 *
20
 * mpeg2dec is distributed in the hope that it will be useful,
21
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23
 * GNU General Public License for more details.
24
 *
25
 * You should have received a copy of the GNU General Public License
26
 * along with mpeg2dec; if not, write to the Free Software
27
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
28
 */
29

    
30
#undef MOVNTQ
31
#undef EMMS
32
#undef SFENCE
33

    
34
#ifdef HAVE_3DNOW
35
/* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */
36
#define EMMS     "femms"
37
#else
38
#define EMMS     "emms"
39
#endif
40

    
41
#ifdef HAVE_MMX2
42
#define MOVNTQ "movntq"
43
#define SFENCE "sfence"
44
#else
45
#define MOVNTQ "movq"
46
#define SFENCE "/nop"
47
#endif
48

    
49
#define YUV2RGB \
50
    /* Do the multiply part of the conversion for even and odd pixels,
51
       register usage:
52
       mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
53
       mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
54
       mm6 -> Y even, mm7 -> Y odd */\
55
    /* convert the chroma part */\
56
    "punpcklbw %%mm4, %%mm0;" /* scatter 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
57
    "punpcklbw %%mm4, %%mm1;" /* scatter 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
58
\
59
    "psllw $3, %%mm0;" /* Promote precision */ \
60
    "psllw $3, %%mm1;" /* Promote precision */ \
61
\
62
    "psubsw "U_OFFSET"(%4), %%mm0;" /* Cb -= 128 */ \
63
    "psubsw "V_OFFSET"(%4), %%mm1;" /* Cr -= 128 */ \
64
\
65
    "movq %%mm0, %%mm2;" /* Copy 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
66
    "movq %%mm1, %%mm3;" /* Copy 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
67
\
68
    "pmulhw "UG_COEFF"(%4), %%mm2;" /* Mul Cb with green coeff -> Cb green */ \
69
    "pmulhw "VG_COEFF"(%4), %%mm3;" /* Mul Cr with green coeff -> Cr green */ \
70
\
71
    "pmulhw "UB_COEFF"(%4), %%mm0;" /* Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0 */\
72
    "pmulhw "VR_COEFF"(%4), %%mm1;" /* Mul Cr -> Cred 00 r3 00 r2 00 r1 00 r0 */\
73
\
74
    "paddsw %%mm3, %%mm2;" /* Cb green + Cr green -> Cgreen */\
75
\
76
    /* convert the luma part */\
77
    "movq %%mm6, %%mm7;" /* Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */\
78
    "pand "MANGLE(mmx_00ffw)", %%mm6;" /* get Y even 00 Y6 00 Y4 00 Y2 00 Y0 */\
79
\
80
    "psrlw $8, %%mm7;" /* get Y odd 00 Y7 00 Y5 00 Y3 00 Y1 */\
81
\
82
    "psllw $3, %%mm6;" /* Promote precision */\
83
    "psllw $3, %%mm7;" /* Promote precision */\
84
\
85
    "psubw "Y_OFFSET"(%4), %%mm6;" /* Y -= 16 */\
86
    "psubw "Y_OFFSET"(%4), %%mm7;" /* Y -= 16 */\
87
\
88
    "pmulhw "Y_COEFF"(%4), %%mm6;" /* Mul 4 Y even 00 y6 00 y4 00 y2 00 y0 */\
89
    "pmulhw "Y_COEFF"(%4), %%mm7;" /* Mul 4 Y odd 00 y7 00 y5 00 y3 00 y1 */\
90
\
91
    /* Do the addition part of the conversion for even and odd pixels,
92
       register usage:
93
       mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
94
       mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
95
       mm6 -> Y even, mm7 -> Y odd */\
96
    "movq %%mm0, %%mm3;" /* Copy Cblue */\
97
    "movq %%mm1, %%mm4;" /* Copy Cred */\
98
    "movq %%mm2, %%mm5;" /* Copy Cgreen */\
99
\
100
    "paddsw %%mm6, %%mm0;" /* Y even + Cblue 00 B6 00 B4 00 B2 00 B0 */\
101
    "paddsw %%mm7, %%mm3;" /* Y odd + Cblue 00 B7 00 B5 00 B3 00 B1 */\
102
\
103
    "paddsw %%mm6, %%mm1;" /* Y even + Cred 00 R6 00 R4 00 R2 00 R0 */\
104
    "paddsw %%mm7, %%mm4;" /* Y odd + Cred 00 R7 00 R5 00 R3 00 R1 */\
105
\
106
    "paddsw %%mm6, %%mm2;" /* Y even + Cgreen 00 G6 00 G4 00 G2 00 G0 */\
107
    "paddsw %%mm7, %%mm5;" /* Y odd + Cgreen 00 G7 00 G5 00 G3 00 G1 */\
108
\
109
    /* Limit RGB even to 0..255 */\
110
    "packuswb %%mm0, %%mm0;" /* B6 B4 B2 B0  B6 B4 B2 B0 */\
111
    "packuswb %%mm1, %%mm1;" /* R6 R4 R2 R0  R6 R4 R2 R0 */\
112
    "packuswb %%mm2, %%mm2;" /* G6 G4 G2 G0  G6 G4 G2 G0 */\
113
\
114
    /* Limit RGB odd to 0..255 */\
115
    "packuswb %%mm3, %%mm3;" /* B7 B5 B3 B1  B7 B5 B3 B1 */\
116
    "packuswb %%mm4, %%mm4;" /* R7 R5 R3 R1  R7 R5 R3 R1 */\
117
    "packuswb %%mm5, %%mm5;" /* G7 G5 G3 G1  G7 G5 G3 G1 */\
118
\
119
    /* Interleave RGB even and odd */\
120
    "punpcklbw %%mm3, %%mm0;" /* B7 B6 B5 B4 B3 B2 B1 B0 */\
121
    "punpcklbw %%mm4, %%mm1;" /* R7 R6 R5 R4 R3 R2 R1 R0 */\
122
    "punpcklbw %%mm5, %%mm2;" /* G7 G6 G5 G4 G3 G2 G1 G0 */\
123

    
124

    
125
static inline int RENAME(yuv420_rgb16)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
126
                                       int srcSliceH, uint8_t* dst[], int dstStride[]){
127
    int y, h_size;
128

    
129
    if(c->srcFormat == PIX_FMT_YUV422P){
130
        srcStride[1] *= 2;
131
        srcStride[2] *= 2;
132
    }
133

    
134
    h_size= (c->dstW+7)&~7;
135
    if(h_size*2 > FFABS(dstStride[0])) h_size-=8;
136

    
137
    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );
138
    //printf("%X %X %X %X %X %X %X %X %X %X\n", (int)&c->redDither, (int)&b5Dither, (int)src[0], (int)src[1], (int)src[2], (int)dst[0],
139
    //srcStride[0],srcStride[1],srcStride[2],dstStride[0]);
140
    for (y= 0; y<srcSliceH; y++ ) {
141
        uint8_t *_image = dst[0] + (y+srcSliceY)*dstStride[0];
142
        uint8_t *_py = src[0] + y*srcStride[0];
143
        uint8_t *_pu = src[1] + (y>>1)*srcStride[1];
144
        uint8_t *_pv = src[2] + (y>>1)*srcStride[2];
145
        long index= -h_size/2;
146

    
147
        b5Dither= dither8[y&1];
148
        g6Dither= dither4[y&1];
149
        g5Dither= dither8[y&1];
150
        r5Dither= dither8[(y+1)&1];
151
        /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
152
           pixels in each iteration */
153
        __asm__ __volatile__ (
154
        /* load data for start of next scan line */
155
        "movd    (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
156
        "movd    (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
157
        "movq (%5, %0, 2), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
158
        //".balign 16     \n\t"
159
        "1:             \n\t"
160
        /* no speed diference on my p3@500 with prefetch,
161
         * if it is faster for anyone with -benchmark then tell me
162
        PREFETCH" 64(%0) \n\t"
163
        PREFETCH" 64(%1) \n\t"
164
        PREFETCH" 64(%2) \n\t"
165
        */
166
YUV2RGB
167

    
168
#ifdef DITHER1XBPP
169
        "paddusb "MANGLE(b5Dither)", %%mm0;"
170
        "paddusb "MANGLE(g6Dither)", %%mm2;"
171
        "paddusb "MANGLE(r5Dither)", %%mm1;"
172
#endif
173
        /* mask unneeded bits off */
174
        "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
175
        "pand "MANGLE(mmx_grnmask)", %%mm2;" /* g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 */
176
        "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */
177

    
178
        "psrlw   $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
179
        "pxor %%mm4, %%mm4;" /* zero mm4 */
180

    
181
        "movq %%mm0, %%mm5;" /* Copy B7-B0 */
182
        "movq %%mm2, %%mm7;" /* Copy G7-G0 */
183

    
184
        /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
185
        "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
186
        "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
187

    
188
        "psllw  $3, %%mm2;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
189
        "por %%mm2, %%mm0;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */
190

    
191
        "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
192
        MOVNTQ "      %%mm0, (%1);" /* store pixel 0-3 */
193

    
194
        /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
195
        "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
196
        "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
197

    
198
        "psllw        $3, %%mm7;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
199
        "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
200

    
201
        "por       %%mm7, %%mm5;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */
202
        "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
203

    
204
        MOVNTQ "   %%mm5, 8 (%1);" /* store pixel 4-7 */
205

    
206
        "add $16, %1    \n\t"
207
        "add  $4, %0    \n\t"
208
        " js  1b        \n\t"
209

    
210
        : "+r" (index), "+r" (_image)
211
        : "r" (_pu - index), "r" (_pv - index), "r"(&c->redDither), "r" (_py - 2*index)
212
        );
213
    }
214

    
215
    __asm__ __volatile__ (EMMS);
216

    
217
    return srcSliceH;
218
}
219

    
220
static inline int RENAME(yuv420_rgb15)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
221
                                       int srcSliceH, uint8_t* dst[], int dstStride[]){
222
    int y, h_size;
223

    
224
    if(c->srcFormat == PIX_FMT_YUV422P){
225
        srcStride[1] *= 2;
226
        srcStride[2] *= 2;
227
    }
228

    
229
    h_size= (c->dstW+7)&~7;
230
    if(h_size*2 > FFABS(dstStride[0])) h_size-=8;
231

    
232
    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );
233
    //printf("%X %X %X %X %X %X %X %X %X %X\n", (int)&c->redDither, (int)&b5Dither, (int)src[0], (int)src[1], (int)src[2], (int)dst[0],
234
    //srcStride[0],srcStride[1],srcStride[2],dstStride[0]);
235
    for (y= 0; y<srcSliceH; y++ ) {
236
        uint8_t *_image = dst[0] + (y+srcSliceY)*dstStride[0];
237
        uint8_t *_py = src[0] + y*srcStride[0];
238
        uint8_t *_pu = src[1] + (y>>1)*srcStride[1];
239
        uint8_t *_pv = src[2] + (y>>1)*srcStride[2];
240
        long index= -h_size/2;
241

    
242
        b5Dither= dither8[y&1];
243
        g6Dither= dither4[y&1];
244
        g5Dither= dither8[y&1];
245
        r5Dither= dither8[(y+1)&1];
246
        /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
247
           pixels in each iteration */
248
        __asm__ __volatile__ (
249
        /* load data for start of next scan line */
250
        "movd    (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
251
        "movd    (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
252
        "movq (%5, %0, 2), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
253
        //".balign 16     \n\t"
254
        "1:             \n\t"
255
YUV2RGB
256

    
257
#ifdef DITHER1XBPP
258
        "paddusb "MANGLE(b5Dither)", %%mm0  \n\t"
259
        "paddusb "MANGLE(g5Dither)", %%mm2  \n\t"
260
        "paddusb "MANGLE(r5Dither)", %%mm1  \n\t"
261
#endif
262

    
263
        /* mask unneeded bits off */
264
        "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
265
        "pand "MANGLE(mmx_redmask)", %%mm2;" /* g7g6g5g4 g3_0_0_0 g7g6g5g4 g3_0_0_0 */
266
        "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */
267

    
268
        "psrlw   $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
269
        "psrlw   $1, %%mm1;" /* 0_r7r6r5  r4r3_0_0 0_r7r6r5 r4r3_0_0 */
270
        "pxor %%mm4, %%mm4;" /* zero mm4 */
271

    
272
        "movq %%mm0, %%mm5;" /* Copy B7-B0 */
273
        "movq %%mm2, %%mm7;" /* Copy G7-G0 */
274

    
275
        /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
276
        "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3_0_0_0 */
277
        "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
278

    
279
        "psllw  $2, %%mm2;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
280
        "por %%mm2, %%mm0;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */
281

    
282
        "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
283
        MOVNTQ "      %%mm0, (%1);"  /* store pixel 0-3 */
284

    
285
        /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
286
        "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 0_g7g6g5 g4g3_0_0 */
287
        "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */
288

    
289
        "psllw        $2, %%mm7;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
290
        "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
291

    
292
        "por       %%mm7, %%mm5;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */
293
        "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
294

    
295
        MOVNTQ " %%mm5, 8 (%1);" /* store pixel 4-7 */
296

    
297
        "add $16, %1            \n\t"
298
        "add $4, %0             \n\t"
299
        " js 1b                 \n\t"
300
        : "+r" (index), "+r" (_image)
301
        : "r" (_pu - index), "r" (_pv - index), "r"(&c->redDither), "r" (_py - 2*index)
302
        );
303
    }
304

    
305
    __asm__ __volatile__ (EMMS);
306
    return srcSliceH;
307
}
308

    
309
static inline int RENAME(yuv420_rgb24)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
310
                                       int srcSliceH, uint8_t* dst[], int dstStride[]){
311
    int y, h_size;
312

    
313
    if(c->srcFormat == PIX_FMT_YUV422P){
314
        srcStride[1] *= 2;
315
        srcStride[2] *= 2;
316
    }
317

    
318
    h_size= (c->dstW+7)&~7;
319
    if(h_size*3 > FFABS(dstStride[0])) h_size-=8;
320

    
321
    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );
322

    
323
    for (y= 0; y<srcSliceH; y++ ) {
324
        uint8_t *_image = dst[0] + (y+srcSliceY)*dstStride[0];
325
        uint8_t *_py = src[0] + y*srcStride[0];
326
        uint8_t *_pu = src[1] + (y>>1)*srcStride[1];
327
        uint8_t *_pv = src[2] + (y>>1)*srcStride[2];
328
        long index= -h_size/2;
329

    
330
        /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
331
           pixels in each iteration */
332
        __asm__ __volatile__ (
333
        /* load data for start of next scan line */
334
        "movd    (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
335
        "movd    (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
336
        "movq (%5, %0, 2), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
337
        //".balign 16     \n\t"
338
        "1:             \n\t"
339
YUV2RGB
340
        /* mm0=B, %%mm2=G, %%mm1=R */
341
#ifdef HAVE_MMX2
342
        "movq "MANGLE(M24A)", %%mm4     \n\t"
343
        "movq "MANGLE(M24C)", %%mm7     \n\t"
344
        "pshufw $0x50, %%mm0, %%mm5     \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */
345
        "pshufw $0x50, %%mm2, %%mm3     \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */
346
        "pshufw $0x00, %%mm1, %%mm6     \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */
347

    
348
        "pand   %%mm4, %%mm5            \n\t" /*    B2        B1       B0 */
349
        "pand   %%mm4, %%mm3            \n\t" /*    G2        G1       G0 */
350
        "pand   %%mm7, %%mm6            \n\t" /*       R1        R0       */
351

    
352
        "psllq     $8, %%mm3            \n\t" /* G2        G1       G0    */
353
        "por    %%mm5, %%mm6            \n\t"
354
        "por    %%mm3, %%mm6            \n\t"
355
        MOVNTQ" %%mm6, (%1)             \n\t"
356

    
357
        "psrlq     $8, %%mm2            \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */
358
        "pshufw $0xA5, %%mm0, %%mm5     \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */
359
        "pshufw $0x55, %%mm2, %%mm3     \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */
360
        "pshufw $0xA5, %%mm1, %%mm6     \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */
361

    
362
        "pand "MANGLE(M24B)", %%mm5     \n\t" /* B5       B4        B3    */
363
        "pand          %%mm7, %%mm3     \n\t" /*       G4        G3       */
364
        "pand          %%mm4, %%mm6     \n\t" /*    R4        R3       R2 */
365

    
366
        "por    %%mm5, %%mm3            \n\t" /* B5    G4 B4     G3 B3    */
367
        "por    %%mm3, %%mm6            \n\t"
368
        MOVNTQ" %%mm6, 8(%1)            \n\t"
369

    
370
        "pshufw $0xFF, %%mm0, %%mm5     \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */
371
        "pshufw $0xFA, %%mm2, %%mm3     \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */
372
        "pshufw $0xFA, %%mm1, %%mm6     \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */
373
        "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
374

    
375
        "pand          %%mm7, %%mm5     \n\t" /*       B7        B6       */
376
        "pand          %%mm4, %%mm3     \n\t" /*    G7        G6       G5 */
377
        "pand "MANGLE(M24B)", %%mm6     \n\t" /* R7       R6        R5    */
378
        "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
379
\
380
        "por          %%mm5, %%mm3      \n\t"
381
        "por          %%mm3, %%mm6      \n\t"
382
        MOVNTQ"       %%mm6, 16(%1)     \n\t"
383
        "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
384
        "pxor         %%mm4, %%mm4      \n\t"
385

    
386
#else
387

    
388
        "pxor      %%mm4, %%mm4     \n\t"
389
        "movq      %%mm0, %%mm5     \n\t" /* B */
390
        "movq      %%mm1, %%mm6     \n\t" /* R */
391
        "punpcklbw %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */
392
        "punpcklbw %%mm4, %%mm1     \n\t" /* 0R0R0R0R 0 */
393
        "punpckhbw %%mm2, %%mm5     \n\t" /* GBGBGBGB 2 */
394
        "punpckhbw %%mm4, %%mm6     \n\t" /* 0R0R0R0R 2 */
395
        "movq      %%mm0, %%mm7     \n\t" /* GBGBGBGB 0 */
396
        "movq      %%mm5, %%mm3     \n\t" /* GBGBGBGB 2 */
397
        "punpcklwd %%mm1, %%mm7     \n\t" /* 0RGB0RGB 0 */
398
        "punpckhwd %%mm1, %%mm0     \n\t" /* 0RGB0RGB 1 */
399
        "punpcklwd %%mm6, %%mm5     \n\t" /* 0RGB0RGB 2 */
400
        "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */
401

    
402
        "movq      %%mm7, %%mm2     \n\t" /* 0RGB0RGB 0 */
403
        "movq      %%mm0, %%mm6     \n\t" /* 0RGB0RGB 1 */
404
        "movq      %%mm5, %%mm1     \n\t" /* 0RGB0RGB 2 */
405
        "movq      %%mm3, %%mm4     \n\t" /* 0RGB0RGB 3 */
406

    
407
        "psllq       $40, %%mm7     \n\t" /* RGB00000 0 */
408
        "psllq       $40, %%mm0     \n\t" /* RGB00000 1 */
409
        "psllq       $40, %%mm5     \n\t" /* RGB00000 2 */
410
        "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */
411

    
412
        "punpckhdq %%mm2, %%mm7     \n\t" /* 0RGBRGB0 0 */
413
        "punpckhdq %%mm6, %%mm0     \n\t" /* 0RGBRGB0 1 */
414
        "punpckhdq %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */
415
        "punpckhdq %%mm4, %%mm3     \n\t" /* 0RGBRGB0 3 */
416

    
417
        "psrlq        $8, %%mm7     \n\t" /* 00RGBRGB 0 */
418
        "movq      %%mm0, %%mm6     \n\t" /* 0RGBRGB0 1 */
419
        "psllq       $40, %%mm0     \n\t" /* GB000000 1 */
420
        "por       %%mm0, %%mm7     \n\t" /* GBRGBRGB 0 */
421
        MOVNTQ"    %%mm7, (%1)      \n\t"
422

    
423
        "movd 4 (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
424

    
425
        "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */
426
        "movq      %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */
427
        "psllq       $24, %%mm5     \n\t" /* BRGB0000 2 */
428
        "por       %%mm5, %%mm6     \n\t" /* BRGBRGBR 1 */
429
        MOVNTQ"    %%mm6, 8(%1)     \n\t"
430

    
431
        "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
432

    
433
        "psrlq       $40, %%mm1     \n\t" /* 000000RG 2 */
434
        "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */
435
        "por       %%mm3, %%mm1     \n\t" /* RGBRGBRG 2 */
436
        MOVNTQ"    %%mm1, 16(%1)    \n\t"
437

    
438
        "movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
439
        "pxor      %%mm4, %%mm4     \n\t"
440
#endif
441

    
442
        "add $24, %1    \n\t"
443
        "add  $4, %0    \n\t"
444
        " js  1b        \n\t"
445

    
446
        : "+r" (index), "+r" (_image)
447
        : "r" (_pu - index), "r" (_pv - index), "r"(&c->redDither), "r" (_py - 2*index)
448
        );
449
    }
450

    
451
    __asm__ __volatile__ (EMMS);
452
    return srcSliceH;
453
}
454

    
455
static inline int RENAME(yuv420_rgb32)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
456
                                       int srcSliceH, uint8_t* dst[], int dstStride[]){
457
    int y, h_size;
458

    
459
    if(c->srcFormat == PIX_FMT_YUV422P){
460
        srcStride[1] *= 2;
461
        srcStride[2] *= 2;
462
    }
463

    
464
    h_size= (c->dstW+7)&~7;
465
    if(h_size*4 > FFABS(dstStride[0])) h_size-=8;
466

    
467
    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );
468

    
469
    for (y= 0; y<srcSliceH; y++ ) {
470
        uint8_t *_image = dst[0] + (y+srcSliceY)*dstStride[0];
471
        uint8_t *_py = src[0] + y*srcStride[0];
472
        uint8_t *_pu = src[1] + (y>>1)*srcStride[1];
473
        uint8_t *_pv = src[2] + (y>>1)*srcStride[2];
474
        long index= -h_size/2;
475

    
476
        /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
477
           pixels in each iteration */
478
        __asm__ __volatile__ (
479
        /* load data for start of next scan line */
480
        "movd    (%2, %0), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
481
        "movd    (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
482
        "movq (%5, %0, 2), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
483
        //".balign 16     \n\t"
484
        "1:             \n\t"
485
YUV2RGB
486
        /* convert RGB plane to RGB packed format,
487
           mm0 ->  B, mm1 -> R, mm2 -> G, mm3 -> 0,
488
           mm4 -> GB, mm5 -> AR pixel 4-7,
489
           mm6 -> GB, mm7 -> AR pixel 0-3 */
490
        "pxor      %%mm3, %%mm3;"   /* zero mm3 */
491

    
492
        "movq      %%mm0, %%mm6;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */
493
        "movq      %%mm1, %%mm7;"   /* R7 R6 R5 R4 R3 R2 R1 R0 */
494

    
495
        "movq      %%mm0, %%mm4;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */
496
        "movq      %%mm1, %%mm5;"   /* R7 R6 R5 R4 R3 R2 R1 R0 */
497

    
498
        "punpcklbw %%mm2, %%mm6;"   /* G3 B3 G2 B2 G1 B1 G0 B0 */
499
        "punpcklbw %%mm3, %%mm7;"   /* 00 R3 00 R2 00 R1 00 R0 */
500

    
501
        "punpcklwd %%mm7, %%mm6;"   /* 00 R1 B1 G1 00 R0 B0 G0 */
502
        MOVNTQ "   %%mm6, (%1);"    /* Store ARGB1 ARGB0 */
503

    
504
        "movq      %%mm0, %%mm6;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */
505
        "punpcklbw %%mm2, %%mm6;"   /* G3 B3 G2 B2 G1 B1 G0 B0 */
506

    
507
        "punpckhwd %%mm7, %%mm6;"   /* 00 R3 G3 B3 00 R2 B3 G2 */
508
        MOVNTQ "   %%mm6, 8 (%1);"  /* Store ARGB3 ARGB2 */
509

    
510
        "punpckhbw %%mm2, %%mm4;"   /* G7 B7 G6 B6 G5 B5 G4 B4 */
511
        "punpckhbw %%mm3, %%mm5;"   /* 00 R7 00 R6 00 R5 00 R4 */
512

    
513
        "punpcklwd %%mm5, %%mm4;"   /* 00 R5 B5 G5 00 R4 B4 G4 */
514
        MOVNTQ "   %%mm4, 16 (%1);" /* Store ARGB5 ARGB4 */
515

    
516
        "movq      %%mm0, %%mm4;"   /* B7 B6 B5 B4 B3 B2 B1 B0 */
517
        "punpckhbw %%mm2, %%mm4;"   /* G7 B7 G6 B6 G5 B5 G4 B4 */
518

    
519
        "punpckhwd %%mm5, %%mm4;"   /* 00 R7 G7 B7 00 R6 B6 G6 */
520
        MOVNTQ "   %%mm4, 24 (%1);" /* Store ARGB7 ARGB6 */
521

    
522
        "movd 4 (%2, %0), %%mm0;"   /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
523
        "movd 4 (%3, %0), %%mm1;"   /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
524

    
525
        "pxor         %%mm4, %%mm4;" /* zero mm4 */
526
        "movq 8 (%5, %0, 2), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
527

    
528
        "add $32, %1    \n\t"
529
        "add  $4, %0    \n\t"
530
        " js  1b        \n\t"
531

    
532
        : "+r" (index), "+r" (_image)
533
        : "r" (_pu - index), "r" (_pv - index), "r"(&c->redDither), "r" (_py - 2*index)
534
        );
535
    }
536

    
537
    __asm__ __volatile__ (EMMS);
538
    return srcSliceH;
539
}