~ubuntu-branches/ubuntu/lucid/ffmpeg-extra/lucid : revision 7

1

This patch is currently being worked on to resolve all non-PIC issues with

2

the libswscale library.

3

For inquiries about this patch, please see bug #528080.

4

==========================================================================

5

--- a/libswscale/rgb2rgb_template.c

6

+++ b/libswscale/rgb2rgb_template.c

7

@@ -1424,9 +1424,12 @@

8

__asm__ volatile (

9

"test %%"REG_a", %%"REG_a" \n\t"

10

"jns 2f \n\t"

11

- "movq "MANGLE(mask24r)", %%mm5 \n\t"

12

- "movq "MANGLE(mask24g)", %%mm6 \n\t"

13

- "movq "MANGLE(mask24b)", %%mm7 \n\t"

14

+ "movq %0, %%mm5 \n\t"

15

+ "movq %1, %%mm6 \n\t"

16

+ "movq %2, %%mm7 \n\t"

17

+ : : "m"(mask24r), "m"(mask24g), "m"(mask24b)

18

+ );

19

+ __asm__ volatile (

20

ASMALIGN(4)

21

"1: \n\t"

22

PREFETCH" 32(%1, %%"REG_a") \n\t"

23

@@ -2147,8 +2150,8 @@

24

{

25

__asm__ volatile(

26

"mov %2, %%"REG_a" \n\t"

27

- "movq "MANGLE(ff_bgr2YCoeff)", %%mm6 \n\t"

28

- "movq "MANGLE(ff_w1111)", %%mm5 \n\t"

29

+ "movq %3, %%mm6 \n\t"

30

+ "movq %4, %%mm5 \n\t"

31

"pxor %%mm7, %%mm7 \n\t"

32

"lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t"

33

ASMALIGN(4)

34

@@ -2206,12 +2209,13 @@

35

"psraw $7, %%mm4 \n\t"

36

37

"packuswb %%mm4, %%mm0 \n\t"

38

- "paddusb "MANGLE(ff_bgr2YOffset)", %%mm0 \n\t"

39

+ "paddusb %5, %%mm0 \n\t"

40

41

MOVNTQ" %%mm0, (%1, %%"REG_a") \n\t"

42

"add $8, %%"REG_a" \n\t"

43

" js 1b \n\t"

44

- : : "r" (src+width*3), "r" (ydst+width), "g" (-width)

45

+ : : "r" (src+width*3), "r" (ydst+width), "g" (-width),

46

+ "m"(ff_bgr2YCoeff), "m"(ff_w1111), "m"(ff_bgr2YOffset)

47

: "%"REG_a, "%"REG_d

48

);

49

ydst += lumStride;

50

@@ -2220,8 +2224,8 @@

51

src -= srcStride*2;

52

__asm__ volatile(

53

"mov %4, %%"REG_a" \n\t"

54

- "movq "MANGLE(ff_w1111)", %%mm5 \n\t"

55

- "movq "MANGLE(ff_bgr2UCoeff)", %%mm6 \n\t"

56

+ "movq %5, %%mm5 \n\t"

57

+ "movq %6, %%mm6 \n\t"

58

"pxor %%mm7, %%mm7 \n\t"

59

"lea (%%"REG_a", %%"REG_a", 2), %%"REG_d" \n\t"

60

"add %%"REG_d", %%"REG_d" \n\t"

61

@@ -2270,8 +2274,8 @@

62

"psrlw $2, %%mm0 \n\t"

63

"psrlw $2, %%mm2 \n\t"

64

#endif

65

- "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t"

66

- "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t"

67

+ "movq %7, %%mm1 \n\t"

68

+ "movq %7, %%mm3 \n\t"

69

70

"pmaddwd %%mm0, %%mm1 \n\t"

71

"pmaddwd %%mm2, %%mm3 \n\t"

72

@@ -2328,12 +2332,12 @@

73

"paddw %%mm1, %%mm5 \n\t"

74

"paddw %%mm3, %%mm2 \n\t"

75

"paddw %%mm5, %%mm2 \n\t"

76

- "movq "MANGLE(ff_w1111)", %%mm5 \n\t"

77

+ "movq %5, %%mm5 \n\t"

78

"psrlw $2, %%mm4 \n\t"

79

"psrlw $2, %%mm2 \n\t"

80

#endif

81

- "movq "MANGLE(ff_bgr2VCoeff)", %%mm1 \n\t"

82

- "movq "MANGLE(ff_bgr2VCoeff)", %%mm3 \n\t"

83

+ "movq %7, %%mm1 \n\t"

84

+ "movq %7, %%mm3 \n\t"

85

86

"pmaddwd %%mm4, %%mm1 \n\t"

87

"pmaddwd %%mm2, %%mm3 \n\t"

88

@@ -2357,13 +2361,16 @@

89

"punpckldq %%mm4, %%mm0 \n\t"

90

"punpckhdq %%mm4, %%mm1 \n\t"

91

"packsswb %%mm1, %%mm0 \n\t"

92

- "paddb "MANGLE(ff_bgr2UVOffset)", %%mm0 \n\t"

93

+ "paddb %8, %%mm0 \n\t"

94

"movd %%mm0, (%2, %%"REG_a") \n\t"

95

"punpckhdq %%mm0, %%mm0 \n\t"

96

"movd %%mm0, (%3, %%"REG_a") \n\t"

97

"add $4, %%"REG_a" \n\t"

98

" js 1b \n\t"

99

- : : "r" (src+chromWidth*6), "r" (src+srcStride+chromWidth*6), "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth)

100

+ : : "r" (src+chromWidth*6), "r" (src+srcStride+chromWidth*6),

101

+ "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth),

102

+ "m"(ff_w1111), "m"(ff_bgr2UCoeff), "m"(ff_bgr2VCoeff),

103

+ "m"(ff_bgr2UVOffset)

104

: "%"REG_a, "%"REG_d

105

);

106

107

--- a/libswscale/rgb2rgb.c

108

+++ b/libswscale/rgb2rgb.c

109

@@ -123,6 +123,18 @@

110

DECLARE_ASM_CONST(8, uint64_t, red_15mask) = 0x00007c0000007c00ULL;

111

DECLARE_ASM_CONST(8, uint64_t, green_15mask) = 0x000003e0000003e0ULL;

112

DECLARE_ASM_CONST(8, uint64_t, blue_15mask) = 0x0000001f0000001fULL;

113

+

114

+// Some constants from swscale.c that are used here

115

+extern const uint64_t __attribute__((visibility("hidden"))) ff_bgr2YCoeff;

116

+extern const uint64_t __attribute__((visibility("hidden"))) ff_bgr2UCoeff;

117

+extern const uint64_t __attribute__((visibility("hidden"))) ff_bgr2VCoeff;

118

+extern const uint64_t __attribute__((visibility("hidden"))) ff_bgr2YOffset;

119

+extern const uint64_t __attribute__((visibility("hidden"))) ff_bgr2UVOffset;

120

+extern const uint64_t __attribute__((visibility("hidden"))) ff_w1111;

121

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24A;

122

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24B;

123

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24C;

124

+

125

#endif /* ARCH_X86 */

126

127

#define RGB2YUV_SHIFT 8

128

--- a/libswscale/swscale_template.c

129

+++ b/libswscale/swscale_template.c

130

@@ -669,9 +669,9 @@

131

#define WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t)

132

133

#define REAL_WRITERGB16(dst, dstw, index) \

134

- "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\

135

- "pand "MANGLE(bFC)", %%mm4 \n\t" /* G */\

136

- "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\

137

+ "pand %6, %%mm2 \n\t" /* B */\

138

+ "pand %7, %%mm4 \n\t" /* G */\

139

+ "pand %6, %%mm5 \n\t" /* R */\

140

"psrlq $3, %%mm2 \n\t"\

141

\

142

"movq %%mm2, %%mm1 \n\t"\

143

@@ -695,11 +695,18 @@

144

"cmp "#dstw", "#index" \n\t"\

145

" jb 1b \n\t"

146

#define WRITERGB16(dst, dstw, index) REAL_WRITERGB16(dst, dstw, index)

147

+#define WRITERGB16_END \

148

+ :: "r" (&c->redDither), \

149

+ "m" (dummy), "m" (dummy), "m" (dummy),\

150

+ "r" (dest), "m" (dstW), "m" (bF8), \

151

+ "m" (bFC) \

152

+ : "%"REG_a, "%"REG_d, "%"REG_S \

153

+ );

154

155

#define REAL_WRITERGB15(dst, dstw, index) \

156

- "pand "MANGLE(bF8)", %%mm2 \n\t" /* B */\

157

- "pand "MANGLE(bF8)", %%mm4 \n\t" /* G */\

158

- "pand "MANGLE(bF8)", %%mm5 \n\t" /* R */\

159

+ "pand %6, %%mm2 \n\t" /* B */\

160

+ "pand %6, %%mm4 \n\t" /* G */\

161

+ "pand %6, %%mm5 \n\t" /* R */\

162

"psrlq $3, %%mm2 \n\t"\

163

"psrlq $1, %%mm5 \n\t"\

164

\

165

@@ -724,6 +731,12 @@

166

"cmp "#dstw", "#index" \n\t"\

167

" jb 1b \n\t"

168

#define WRITERGB15(dst, dstw, index) REAL_WRITERGB15(dst, dstw, index)

169

+#define WRITERGB15_END \

170

+ :: "r" (&c->redDither), \

171

+ "m" (dummy), "m" (dummy), "m" (dummy),\

172

+ "r" (dest), "m" (dstW), "m" (bF8) \

173

+ : "%"REG_a, "%"REG_d, "%"REG_S \

174

+ );

175

176

#define WRITEBGR24OLD(dst, dstw, index) \

177

/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\

178

@@ -1063,7 +1076,7 @@

179

#endif

180

181

WRITERGB15(%4, %5, %%REGa)

182

- YSCALEYUV2PACKEDX_END

183

+ WRITERGB15_END

184

return;

185

case PIX_FMT_RGB565:

186

YSCALEYUV2PACKEDX_ACCURATE

187

@@ -1077,7 +1090,7 @@

188

#endif

189

190

WRITERGB16(%4, %5, %%REGa)

191

- YSCALEYUV2PACKEDX_END

192

+ WRITERGB16_END

193

return;

194

case PIX_FMT_YUYV422:

195

YSCALEYUV2PACKEDX_ACCURATE

196

@@ -1127,7 +1140,7 @@

197

#endif

198

199

WRITERGB15(%4, %5, %%REGa)

200

- YSCALEYUV2PACKEDX_END

201

+ WRITERGB15_END

202

return;

203

case PIX_FMT_RGB565:

204

YSCALEYUV2PACKEDX

205

@@ -1141,7 +1154,7 @@

206

#endif

207

208

WRITERGB16(%4, %5, %%REGa)

209

- YSCALEYUV2PACKEDX_END

210

+ WRITERGB16_END

211

return;

212

case PIX_FMT_YUYV422:

213

YSCALEYUV2PACKEDX

214

@@ -1238,7 +1251,7 @@

215

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

216

217

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

218

- "a" (&c->redDither)

219

+ "a" (&c->redDither), "m" (bF8)

220

);

221

return;

222

case PIX_FMT_RGB565:

223

@@ -1259,7 +1272,7 @@

224

"pop %%"REG_BP" \n\t"

225

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

226

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

227

- "a" (&c->redDither)

228

+ "a" (&c->redDither), "m" (bF8), "m" (bFC)

229

);

230

return;

231

case PIX_FMT_YUYV422:

232

@@ -1354,7 +1367,7 @@

233

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

234

235

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

236

- "a" (&c->redDither)

237

+ "a" (&c->redDither), "m" (bF8)

238

);

239

return;

240

case PIX_FMT_RGB565:

241

@@ -1376,7 +1389,7 @@

242

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

243

244

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

245

- "a" (&c->redDither)

246

+ "a" (&c->redDither), "m" (bF8), "m" (bFC)

247

);

248

return;

249

case PIX_FMT_YUYV422:

250

@@ -1447,7 +1460,7 @@

251

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

252

253

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

254

- "a" (&c->redDither)

255

+ "a" (&c->redDither), "m" (bF8)

256

);

257

return;

258

case PIX_FMT_RGB565:

259

@@ -1469,7 +1482,7 @@

260

"mov "ESP_OFFSET"(%5), %%"REG_b" \n\t"

261

262

:: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

263

- "a" (&c->redDither)

264

+ "a" (&c->redDither), "m" (bF8), "m" (bFC)

265

);

266

return;

267

case PIX_FMT_YUYV422:

268

@@ -1504,7 +1517,7 @@

269

{

270

#if HAVE_MMX

271

__asm__ volatile(

272

- "movq "MANGLE(bm01010101)", %%mm2 \n\t"

273

+ "movq %3, %%mm2 \n\t"

274

"mov %0, %%"REG_a" \n\t"

275

"1: \n\t"

276

"movq (%1, %%"REG_a",2), %%mm0 \n\t"

277

@@ -1515,7 +1528,7 @@

278

"movq %%mm0, (%2, %%"REG_a") \n\t"

279

"add $8, %%"REG_a" \n\t"

280

" js 1b \n\t"

281

- : : "g" (-width), "r" (src+width*2), "r" (dst+width)

282

+ : : "g" (-width), "r" (src+width*2), "r" (dst+width), "m"(bm01010101)

283

: "%"REG_a

284

);

285

#else

286

@@ -1529,7 +1542,7 @@

287

{

288

#if HAVE_MMX

289

__asm__ volatile(

290

- "movq "MANGLE(bm01010101)", %%mm4 \n\t"

291

+ "movq %4, %%mm4 \n\t"

292

"mov %0, %%"REG_a" \n\t"

293

"1: \n\t"

294

"movq (%1, %%"REG_a",4), %%mm0 \n\t"

295

@@ -1546,7 +1559,8 @@

296

"movd %%mm1, (%2, %%"REG_a") \n\t"

297

"add $4, %%"REG_a" \n\t"

298

" js 1b \n\t"

299

- : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

300

+ : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width),

301

+ "m"(bm01010101)

302

: "%"REG_a

303

);

304

#else

305

@@ -1590,7 +1604,7 @@

306

{

307

#if HAVE_MMX

308

__asm__ volatile(

309

- "movq "MANGLE(bm01010101)", %%mm4 \n\t"

310

+ "movq %4, %%mm4 \n\t"

311

"mov %0, %%"REG_a" \n\t"

312

"1: \n\t"

313

"movq (%1, %%"REG_a",4), %%mm0 \n\t"

314

@@ -1607,7 +1621,8 @@

315

"movd %%mm1, (%2, %%"REG_a") \n\t"

316

"add $4, %%"REG_a" \n\t"

317

" js 1b \n\t"

318

- : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

319

+ : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width),

320

+ "m"(bm01010101)

321

: "%"REG_a

322

);

323

#else

324

@@ -1688,20 +1703,20 @@

325

326

if(srcFormat == PIX_FMT_BGR24){

327

__asm__ volatile(

328

- "movq "MANGLE(ff_bgr24toY1Coeff)", %%mm5 \n\t"

329

- "movq "MANGLE(ff_bgr24toY2Coeff)", %%mm6 \n\t"

330

- :

331

+ "movq %0, %%mm5 \n\t"

332

+ "movq %1, %%mm6 \n\t"

333

+ : : "m"(ff_bgr24toY1Coeff), "m"(ff_bgr24toY2Coeff)

334

);

335

}else{

336

__asm__ volatile(

337

- "movq "MANGLE(ff_rgb24toY1Coeff)", %%mm5 \n\t"

338

- "movq "MANGLE(ff_rgb24toY2Coeff)", %%mm6 \n\t"

339

- :

340

+ "movq %0, %%mm5 \n\t"

341

+ "movq %1, %%mm6 \n\t"

342

+ : : "m"(ff_rgb24toY1Coeff), "m"(ff_rgb24toY2Coeff)

343

);

344

}

345

346

__asm__ volatile(

347

- "movq "MANGLE(ff_bgr24toYOffset)", %%mm4 \n\t"

348

+ "movq %3, %%mm4 \n\t"

349

"mov %2, %%"REG_a" \n\t"

350

"pxor %%mm7, %%mm7 \n\t"

351

"1: \n\t"

352

@@ -1731,7 +1746,7 @@

353

"add $4, %%"REG_a" \n\t"

354

" js 1b \n\t"

355

: "+r" (src)

356

- : "r" (dst+width), "g" (-width)

357

+ : "r" (dst+width), "g" (-width), "m"(ff_bgr24toYOffset)

358

: "%"REG_a

359

);

360

}

361

@@ -1771,7 +1786,7 @@

362

"paddd %%mm3, %%mm1 \n\t"

363

"paddd %%mm5, %%mm4 \n\t"

364

365

- "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3 \n\t"

366

+ "movq %4, %%mm3 \n\t"

367

"paddd %%mm3, %%mm0 \n\t"

368

"paddd %%mm3, %%mm2 \n\t"

369

"paddd %%mm3, %%mm1 \n\t"

370

@@ -1789,7 +1804,8 @@

371

"add $4, %%"REG_a" \n\t"

372

" js 1b \n\t"

373

: "+r" (src)

374

- : "r" (dstU+width), "r" (dstV+width), "g" (-width), "m"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24][0])

375

+ : "r" (dstU+width), "r" (dstV+width), "g" (-width),

376

+ "m"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24][0]), "m"(ff_bgr24toUVOffset)

377

: "%"REG_a

378

);

379

}

380

--- a/libswscale/yuv2rgb_template.c

381

+++ b/libswscale/yuv2rgb_template.c

382

@@ -74,7 +74,7 @@

383

\

384

/* convert the luma part */\

385

"movq %%mm6, %%mm7;" /* Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */\

386

- "pand "MANGLE(mmx_00ffw)", %%mm6;" /* get Y even 00 Y6 00 Y4 00 Y2 00 Y0 */\

387

+ "pand %9, %%mm6;" /* get Y even 00 Y6 00 Y4 00 Y2 00 Y0 */\

388

\

389

"psrlw $8, %%mm7;" /* get Y odd 00 Y7 00 Y5 00 Y3 00 Y1 */\

390

\

391

@@ -158,21 +158,63 @@

392

PREFETCH" 64(%2) \n\t" \

393

*/ \

394

395

-#define YUV2RGB_ENDLOOP(depth) \

396

+#define YUV2RGB16_ENDLOOP(depth) \

397

"add $"AV_STRINGIFY(depth*8)", %1 \n\t" \

398

"add $4, %0 \n\t" \

399

" js 1b \n\t" \

400

\

401

: "+r" (index), "+r" (image) \

402

- : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index) \

403

+ : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index), \

404

+ "m"(mmx_redmask), "m"(mmx_grnmask), "m"(dummy), "m"(mmx_00ffw) \

405

); \

406

} \

407

__asm__ volatile (EMMS); \

408

return srcSliceH; \

409

410

+#define YUV2RGB15_ENDLOOP(depth) \

411

+ "add $"AV_STRINGIFY(depth*8)", %1 \n\t" \

412

+ "add $4, %0 \n\t" \

413

+ " js 1b \n\t" \

414

+\

415

+ : "+r" (index), "+r" (image) \

416

+ : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index), \

417

+ "m"(mmx_redmask), "m"(dummy), "m"(dummy), "m"(mmx_00ffw) \

418

+ ); \

419

+ } \

420

+ __asm__ volatile (EMMS); \

421

+ return srcSliceH; \

422

+

423

+#define YUV2RGB24_ENDLOOP(depth) \

424

+ "add $"AV_STRINGIFY(depth*8)", %1 \n\t" \

425

+ "add $4, %0 \n\t" \

426

+ " js 1b \n\t" \

427

+\

428

+ : "+r" (index), "+r" (image) \

429

+ : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index), \

430

+ "m"(ff_M24A), "m"(ff_M24C), "m"(ff_M24B), "m"(mmx_00ffw) \

431

+ ); \

432

+ } \

433

+ __asm__ volatile (EMMS); \

434

+ return srcSliceH; \

435

+

436

+#define YUV2RGB32_ENDLOOP(depth) \

437

+ "add $"AV_STRINGIFY(depth*8)", %1 \n\t" \

438

+ "add $4, %0 \n\t" \

439

+ " js 1b \n\t" \

440

+\

441

+ : "+r" (index), "+r" (image) \

442

+ : "r" (pu - index), "r" (pv - index), "r"(&c->redDither), "r" (py - 2*index), \

443

+ "m"(dummy), "m"(dummy), "m"(dummy), "m"(mmx_00ffw) \

444

+ ); \

445

+ } \

446

+ __asm__ volatile (EMMS); \

447

+ return srcSliceH; \

448

+

449

+

450

static inline int RENAME(yuv420_rgb16)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

451

int srcSliceH, uint8_t* dst[], int dstStride[]){

452

int y, h_size;

453

+ long dummy=0;

454

455

YUV422_UNSHIFT

456

YUV2RGB_LOOP(2)

457

@@ -190,9 +232,9 @@

458

"paddusb "RED_DITHER"(%4), %%mm1;"

459

#endif

460

/* mask unneeded bits off */

461

- "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */

462

- "pand "MANGLE(mmx_grnmask)", %%mm2;" /* g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 */

463

- "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

464

+ "pand %6, %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */

465

+ "pand %7, %%mm2;" /* g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 */

466

+ "pand %6, %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

467

468

"psrlw $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */

469

"pxor %%mm4, %%mm4;" /* zero mm4 */

470

@@ -222,12 +264,13 @@

471

472

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

473

474

- YUV2RGB_ENDLOOP(2)

475

+ YUV2RGB16_ENDLOOP(2)

476

}

477

478

static inline int RENAME(yuv420_rgb15)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

479

int srcSliceH, uint8_t* dst[], int dstStride[]){

480

int y, h_size;

481

+ long dummy=0;

482

483

YUV422_UNSHIFT

484

YUV2RGB_LOOP(2)

485

@@ -246,9 +289,9 @@

486

#endif

487

488

/* mask unneeded bits off */

489

- "pand "MANGLE(mmx_redmask)", %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */

490

- "pand "MANGLE(mmx_redmask)", %%mm2;" /* g7g6g5g4 g3_0_0_0 g7g6g5g4 g3_0_0_0 */

491

- "pand "MANGLE(mmx_redmask)", %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

492

+ "pand %6, %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */

493

+ "pand %6, %%mm2;" /* g7g6g5g4 g3_0_0_0 g7g6g5g4 g3_0_0_0 */

494

+ "pand %6, %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

495

496

"psrlw $3, %%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */

497

"psrlw $1, %%mm1;" /* 0_r7r6r5 r4r3_0_0 0_r7r6r5 r4r3_0_0 */

498

@@ -279,12 +322,13 @@

499

500

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

501

502

- YUV2RGB_ENDLOOP(2)

503

+ YUV2RGB15_ENDLOOP(2)

504

}

505

506

static inline int RENAME(yuv420_rgb24)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

507

int srcSliceH, uint8_t* dst[], int dstStride[]){

508

int y, h_size;

509

+ long dummy=0;

510

511

YUV422_UNSHIFT

512

YUV2RGB_LOOP(3)

513

@@ -293,8 +337,8 @@

514

YUV2RGB

515

/* mm0=B, %%mm2=G, %%mm1=R */

516

#if HAVE_MMX2

517

- "movq "MANGLE(ff_M24A)", %%mm4 \n\t"

518

- "movq "MANGLE(ff_M24C)", %%mm7 \n\t"

519

+ "movq %6, %%mm4 \n\t"

520

+ "movq %7, %%mm7 \n\t"

521

"pshufw $0x50, %%mm0, %%mm5 \n\t" /* B3 B2 B3 B2 B1 B0 B1 B0 */

522

"pshufw $0x50, %%mm2, %%mm3 \n\t" /* G3 G2 G3 G2 G1 G0 G1 G0 */

523

"pshufw $0x00, %%mm1, %%mm6 \n\t" /* R1 R0 R1 R0 R1 R0 R1 R0 */

524

@@ -313,7 +357,7 @@

525

"pshufw $0x55, %%mm2, %%mm3 \n\t" /* G4 G3 G4 G3 G4 G3 G4 G3 */

526

"pshufw $0xA5, %%mm1, %%mm6 \n\t" /* R5 R4 R5 R4 R3 R2 R3 R2 */

527

528

- "pand "MANGLE(ff_M24B)", %%mm5 \n\t" /* B5 B4 B3 */

529

+ "pand %8, %%mm5 \n\t" /* B5 B4 B3 */

530

"pand %%mm7, %%mm3 \n\t" /* G4 G3 */

531

"pand %%mm4, %%mm6 \n\t" /* R4 R3 R2 */

532

533

@@ -328,7 +372,7 @@

534

535

"pand %%mm7, %%mm5 \n\t" /* B7 B6 */

536

"pand %%mm4, %%mm3 \n\t" /* G7 G6 G5 */

537

- "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7 R6 R5 */

538

+ "pand %8, %%mm6 \n\t" /* R7 R6 R5 */

539

"movd 4 (%3, %0), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */

540

\

541

"por %%mm5, %%mm3 \n\t"

542

@@ -393,7 +437,7 @@

543

"pxor %%mm4, %%mm4 \n\t"

544

#endif

545

546

- YUV2RGB_ENDLOOP(3)

547

+ YUV2RGB24_ENDLOOP(3)

548

}

549

550

#define RGB_PLANAR2PACKED32 \

551

@@ -440,6 +484,7 @@

552

static inline int RENAME(yuv420_rgb32)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

553

int srcSliceH, uint8_t* dst[], int dstStride[]){

554

int y, h_size;

555

+ long dummy=0;

556

557

YUV422_UNSHIFT

558

YUV2RGB_LOOP(4)

559

@@ -449,5 +494,5 @@

560

"pcmpeqd %%mm3, %%mm3;" /* fill mm3 */

561

RGB_PLANAR2PACKED32

562

563

- YUV2RGB_ENDLOOP(4)

564

+ YUV2RGB32_ENDLOOP(4)

565

}

566

--- a/libswscale/yuv2rgb.c

567

+++ b/libswscale/yuv2rgb.c

568

@@ -47,6 +47,10 @@

569

DECLARE_ASM_CONST(8, uint64_t, mmx_redmask) = 0xf8f8f8f8f8f8f8f8ULL;

570

DECLARE_ASM_CONST(8, uint64_t, mmx_grnmask) = 0xfcfcfcfcfcfcfcfcULL;

571

572

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24A;

573

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24B;

574

+extern const uint64_t __attribute__((visibility("hidden"))) ff_M24C;

575

+

576

//MMX versions

577

#undef RENAME

578

#undef HAVE_MMX2