Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

cosmetics attack, part I: Remove all tabs and prettyprint/reindent the code.

Browse Source 
Originally committed as revision 23158 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale
pull/126/head
Diego Biurrun 18 years ago
parent 1da87823db
commit 6e42e6c4b4
5 changed files with 3074 additions and 3081 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 6
      libswscale/cs_test.c
  2. 609
      libswscale/rgb2rgb.c
  3. 69
      libswscale/rgb2rgb.h
  4. 4619
      libswscale/rgb2rgb_template.c
  5. 852
      libswscale/swscale_altivec_template.c

6
libswscale/cs_test.c
Unescape View File

@ -142,7 +142,7 @@ int main(int argc, char **argv)
for(i=0; i<SIZE; i++){ for(i=0; i<SIZE; i++){
if(srcBuffer[i]!=srcByte){ if(srcBuffer[i]!=srcByte){
av_log(NULL, AV_LOG_INFO, "src damaged at %d w:%d src:%d dst:%d %s\n", av_log(NULL, AV_LOG_INFO, "src damaged at %d w:%d src:%d dst:%d %s\n",
i, width, srcOffset, dstOffset, name); i, width, srcOffset, dstOffset, name);
failed=1; failed=1;
break; break;
} }
@ -150,7 +150,7 @@ int main(int argc, char **argv)
for(i=0; i<dstOffset; i++){ for(i=0; i<dstOffset; i++){
if(dstBuffer[i]!=dstByte){ if(dstBuffer[i]!=dstByte){
av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n", av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n",
i, width, srcOffset, dstOffset, name); i, width, srcOffset, dstOffset, name);
failed=1; failed=1;
break; break;
} }
@ -158,7 +158,7 @@ int main(int argc, char **argv)
for(i=dstOffset + width*dstBpp; i<SIZE; i++){ for(i=dstOffset + width*dstBpp; i<SIZE; i++){
if(dstBuffer[i]!=dstByte){ if(dstBuffer[i]!=dstByte){
av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n", av_log(NULL, AV_LOG_INFO, "dst damaged at %d w:%d src:%d dst:%d %s\n",
i, width, srcOffset, dstOffset, name); i, width, srcOffset, dstOffset, name);
failed=1; failed=1;
break; break;
} }

609
libswscale/rgb2rgb.c
Unescape View File

@ -1,5 +1,4 @@
/* /*
*
* rgb2rgb.c, Software RGB to RGB convertor * rgb2rgb.c, Software RGB to RGB convertor
* pluralize by Software PAL8 to RGB convertor * pluralize by Software PAL8 to RGB convertor
* Software YUV to YUV convertor * Software YUV to YUV convertor
@ -61,83 +60,83 @@ void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
long width, long height, long width, long height,
long lumStride, long chromStride, long srcStride); long lumStride, long chromStride, long srcStride);
void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
long width, long height, long width, long height,
long lumStride, long chromStride, long srcStride); long lumStride, long chromStride, long srcStride);
void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height, void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
long srcStride, long dstStride); long srcStride, long dstStride);
void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst, void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
long width, long height, long src1Stride, long width, long height, long src1Stride,
long src2Stride, long dstStride); long src2Stride, long dstStride);
void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2, void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
uint8_t *dst1, uint8_t *dst2, uint8_t *dst1, uint8_t *dst2,
long width, long height, long width, long height,
long srcStride1, long srcStride2, long srcStride1, long srcStride2,
long dstStride1, long dstStride2); long dstStride1, long dstStride2);
void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
uint8_t *dst, uint8_t *dst,
long width, long height, long width, long height,
long srcStride1, long srcStride2, long srcStride1, long srcStride2,
long srcStride3, long dstStride); long srcStride3, long dstStride);
#if defined(ARCH_X86) && defined(CONFIG_GPL) #if defined(ARCH_X86) && defined(CONFIG_GPL)
static const uint64_t mmx_null __attribute__((aligned(8))) = 0x0000000000000000ULL; static const uint64_t mmx_null __attribute__((aligned(8))) = 0x0000000000000000ULL;
static const uint64_t mmx_one __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL; static const uint64_t mmx_one __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
static const uint64_t mask32b attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL; static const uint64_t mask32b attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL;
static const uint64_t mask32g attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL; static const uint64_t mask32g attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL;
static const uint64_t mask32r attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL; static const uint64_t mask32r attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL;
static const uint64_t mask32 __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL; static const uint64_t mask32 __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL;
static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL; static const uint64_t mask3216br __attribute__((aligned(8))) = 0x00F800F800F800F8ULL;
static const uint64_t mask3216g __attribute__((aligned(8)))=0x0000FC000000FC00ULL; static const uint64_t mask3216g __attribute__((aligned(8))) = 0x0000FC000000FC00ULL;
static const uint64_t mask3215g __attribute__((aligned(8)))=0x0000F8000000F800ULL; static const uint64_t mask3215g __attribute__((aligned(8))) = 0x0000F8000000F800ULL;
static const uint64_t mul3216 __attribute__((aligned(8))) = 0x2000000420000004ULL; static const uint64_t mul3216 __attribute__((aligned(8))) = 0x2000000420000004ULL;
static const uint64_t mul3215 __attribute__((aligned(8))) = 0x2000000820000008ULL; static const uint64_t mul3215 __attribute__((aligned(8))) = 0x2000000820000008ULL;
static const uint64_t mask24b attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL; static const uint64_t mask24b attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL;
static const uint64_t mask24g attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL; static const uint64_t mask24g attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL;
static const uint64_t mask24r attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL; static const uint64_t mask24r attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL;
static const uint64_t mask24l __attribute__((aligned(8))) = 0x0000000000FFFFFFULL; static const uint64_t mask24l __attribute__((aligned(8))) = 0x0000000000FFFFFFULL;
static const uint64_t mask24h __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL; static const uint64_t mask24h __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL;
static const uint64_t mask24hh __attribute__((aligned(8))) = 0xffff000000000000ULL; static const uint64_t mask24hh __attribute__((aligned(8))) = 0xffff000000000000ULL;
static const uint64_t mask24hhh __attribute__((aligned(8))) = 0xffffffff00000000ULL; static const uint64_t mask24hhh __attribute__((aligned(8))) = 0xffffffff00000000ULL;
static const uint64_t mask24hhhh __attribute__((aligned(8))) = 0xffffffffffff0000ULL; static const uint64_t mask24hhhh __attribute__((aligned(8))) = 0xffffffffffff0000ULL;
static const uint64_t mask15b __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */ static const uint64_t mask15b __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */
static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */ static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */
static const uint64_t mask15s __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL; static const uint64_t mask15s __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL;
static const uint64_t mask15g __attribute__((aligned(8))) = 0x03E003E003E003E0ULL; static const uint64_t mask15g __attribute__((aligned(8))) = 0x03E003E003E003E0ULL;
static const uint64_t mask15r __attribute__((aligned(8))) = 0x7C007C007C007C00ULL; static const uint64_t mask15r __attribute__((aligned(8))) = 0x7C007C007C007C00ULL;
#define mask16b mask15b #define mask16b mask15b
static const uint64_t mask16g __attribute__((aligned(8))) = 0x07E007E007E007E0ULL; static const uint64_t mask16g __attribute__((aligned(8))) = 0x07E007E007E007E0ULL;
static const uint64_t mask16r __attribute__((aligned(8))) = 0xF800F800F800F800ULL; static const uint64_t mask16r __attribute__((aligned(8))) = 0xF800F800F800F800ULL;
static const uint64_t red_16mask __attribute__((aligned(8))) = 0x0000f8000000f800ULL; static const uint64_t red_16mask __attribute__((aligned(8))) = 0x0000f8000000f800ULL;
static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL; static const uint64_t green_16mask __attribute__((aligned(8))) = 0x000007e0000007e0ULL;
static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL; static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
static const uint64_t red_15mask __attribute__((aligned(8))) = 0x00007c0000007c00ULL; static const uint64_t red_15mask __attribute__((aligned(8))) = 0x00007c0000007c00ULL;
static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000003e0ULL; static const uint64_t green_15mask __attribute__((aligned(8))) = 0x000003e0000003e0ULL;
static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL; static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL;
#ifdef FAST_BGR2YV12 #ifdef FAST_BGR2YV12
static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL; static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL; static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL; static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
#else #else
static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL; static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL; static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL; static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
#endif #endif
static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL; static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL; static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8))) = 0x8080808080808080ULL;
static const uint64_t w1111 attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL; static const uint64_t w1111 attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
#if 0 #if 0
static volatile uint64_t __attribute__((aligned(8))) b5Dither; static volatile uint64_t __attribute__((aligned(8))) b5Dither;
@ -146,12 +145,12 @@ static volatile uint64_t __attribute__((aligned(8))) g6Dither;
static volatile uint64_t __attribute__((aligned(8))) r5Dither; static volatile uint64_t __attribute__((aligned(8))) r5Dither;
static uint64_t __attribute__((aligned(8))) dither4[2]={ static uint64_t __attribute__((aligned(8))) dither4[2]={
0x0103010301030103LL, 0x0103010301030103LL,
0x0200020002000200LL,}; 0x0200020002000200LL,};
static uint64_t __attribute__((aligned(8))) dither8[2]={ static uint64_t __attribute__((aligned(8))) dither8[2]={
0x0602060206020602LL, 0x0602060206020602LL,
0x0004000400040004LL,}; 0x0004000400040004LL,};
#endif #endif
#endif /* defined(ARCH_X86) */ #endif /* defined(ARCH_X86) */
@ -215,15 +214,15 @@ static uint64_t __attribute__((aligned(8))) dither8[2]={
void sws_rgb2rgb_init(int flags){ void sws_rgb2rgb_init(int flags){
#if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX)) && defined(CONFIG_GPL) #if (defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX)) && defined(CONFIG_GPL)
if(flags & SWS_CPU_CAPS_MMX2) if (flags & SWS_CPU_CAPS_MMX2)
rgb2rgb_init_MMX2(); rgb2rgb_init_MMX2();
else if(flags & SWS_CPU_CAPS_3DNOW) else if (flags & SWS_CPU_CAPS_3DNOW)
rgb2rgb_init_3DNOW(); rgb2rgb_init_3DNOW();
else if(flags & SWS_CPU_CAPS_MMX) else if (flags & SWS_CPU_CAPS_MMX)
rgb2rgb_init_MMX(); rgb2rgb_init_MMX();
else else
#endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */ #endif /* defined(HAVE_MMX2) || defined(HAVE_3DNOW) || defined(HAVE_MMX) */
rgb2rgb_init_C(); rgb2rgb_init_C();
} }
/** /**
@ -231,49 +230,49 @@ void sws_rgb2rgb_init(int flags){
*/ */
void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
/* /*
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ]; ((unsigned *)dst)[i] = ((unsigned *)palette)[ src[i] ];
*/ */
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
dst[3]= palette[ src[i]*4+2 ]; dst[3]= palette[ src[i]*4+2 ];
dst[2]= palette[ src[i]*4+1 ]; dst[2]= palette[ src[i]*4+1 ];
dst[1]= palette[ src[i]*4+0 ]; dst[1]= palette[ src[i]*4+0 ];
#else #else
//FIXME slow? //FIXME slow?
dst[0]= palette[ src[i]*4+2 ]; dst[0]= palette[ src[i]*4+2 ];
dst[1]= palette[ src[i]*4+1 ]; dst[1]= palette[ src[i]*4+1 ];
dst[2]= palette[ src[i]*4+0 ]; dst[2]= palette[ src[i]*4+0 ];
//dst[3]= 0; /* do we need this cleansing? */ //dst[3]= 0; /* do we need this cleansing? */
#endif #endif
dst+= 4; dst+= 4;
} }
} }
void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
dst[3]= palette[ src[i]*4+0 ]; dst[3]= palette[ src[i]*4+0 ];
dst[2]= palette[ src[i]*4+1 ]; dst[2]= palette[ src[i]*4+1 ];
dst[1]= palette[ src[i]*4+2 ]; dst[1]= palette[ src[i]*4+2 ];
#else #else
//FIXME slow? //FIXME slow?
dst[0]= palette[ src[i]*4+0 ]; dst[0]= palette[ src[i]*4+0 ];
dst[1]= palette[ src[i]*4+1 ]; dst[1]= palette[ src[i]*4+1 ];
dst[2]= palette[ src[i]*4+2 ]; dst[2]= palette[ src[i]*4+2 ];
//dst[3]= 0; /* do we need this cleansing? */ //dst[3]= 0; /* do we need this cleansing? */
#endif #endif
dst+= 4; dst+= 4;
} }
} }
/** /**
@ -281,38 +280,38 @@ void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const ui
*/ */
void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
/* /*
writes 1 byte o much and might cause alignment issues on some architectures? writes 1 byte o much and might cause alignment issues on some architectures?
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ]; ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
*/ */
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
//FIXME slow? //FIXME slow?
dst[0]= palette[ src[i]*4+2 ]; dst[0]= palette[ src[i]*4+2 ];
dst[1]= palette[ src[i]*4+1 ]; dst[1]= palette[ src[i]*4+1 ];
dst[2]= palette[ src[i]*4+0 ]; dst[2]= palette[ src[i]*4+0 ];
dst+= 3; dst+= 3;
} }
} }
void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
/* /*
writes 1 byte o much and might cause alignment issues on some architectures? writes 1 byte o much and might cause alignment issues on some architectures?
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ]; ((unsigned *)(&dst[i*3])) = ((unsigned *)palette)[ src[i] ];
*/ */
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
//FIXME slow? //FIXME slow?
dst[0]= palette[ src[i]*4+0 ]; dst[0]= palette[ src[i]*4+0 ];
dst[1]= palette[ src[i]*4+1 ]; dst[1]= palette[ src[i]*4+1 ];
dst[2]= palette[ src[i]*4+2 ]; dst[2]= palette[ src[i]*4+2 ];
dst+= 3; dst+= 3;
} }
} }
/** /**
@ -320,15 +319,15 @@ void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const ui
*/ */
void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ]; ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
} }
void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]); ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
} }
/** /**
@ -336,216 +335,216 @@ void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const ui
*/ */
void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ]; ((uint16_t *)dst)[i] = ((uint16_t *)palette)[ src[i] ];
} }
void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
{ {
long i; long i;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]); ((uint16_t *)dst)[i] = bswap_16(((uint16_t *)palette)[ src[i] ]);
} }
void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size) void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size >> 2; long num_pixels = src_size >> 2;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
/* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */ /* RGB32 (= A,B,G,R) -> BGR24 (= B,G,R) */
dst[3*i + 0] = src[4*i + 1]; dst[3*i + 0] = src[4*i + 1];
dst[3*i + 1] = src[4*i + 2]; dst[3*i + 1] = src[4*i + 2];
dst[3*i + 2] = src[4*i + 3]; dst[3*i + 2] = src[4*i + 3];
#else #else
dst[3*i + 0] = src[4*i + 2]; dst[3*i + 0] = src[4*i + 2];
dst[3*i + 1] = src[4*i + 1]; dst[3*i + 1] = src[4*i + 1];
dst[3*i + 2] = src[4*i + 0]; dst[3*i + 2] = src[4*i + 0];
#endif #endif
} }
} }
void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size) void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
for(i=0; 3*i<src_size; i++) for (i=0; 3*i<src_size; i++)
{ {
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
/* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */ /* RGB24 (= R,G,B) -> BGR32 (= A,R,G,B) */
dst[4*i + 0] = 0; dst[4*i + 0] = 0;
dst[4*i + 1] = src[3*i + 0]; dst[4*i + 1] = src[3*i + 0];
dst[4*i + 2] = src[3*i + 1]; dst[4*i + 2] = src[3*i + 1];
dst[4*i + 3] = src[3*i + 2]; dst[4*i + 3] = src[3*i + 2];
#else #else
dst[4*i + 0] = src[3*i + 2]; dst[4*i + 0] = src[3*i + 2];
dst[4*i + 1] = src[3*i + 1]; dst[4*i + 1] = src[3*i + 1];
dst[4*i + 2] = src[3*i + 0]; dst[4*i + 2] = src[3*i + 0];
dst[4*i + 3] = 0; dst[4*i + 3] = 0;
#endif #endif
} }
} }
void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size) void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
{ {
const uint16_t *end; const uint16_t *end;
uint8_t *d = (uint8_t *)dst; uint8_t *d = (uint8_t *)dst;
const uint16_t *s = (uint16_t *)src; const uint16_t *s = (uint16_t *)src;
end = s + src_size/2; end = s + src_size/2;
while(s < end) while (s < end)
{ {
register uint16_t bgr; register uint16_t bgr;
bgr = *s++; bgr = *s++;
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
*d++ = 0; *d++ = 0;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x7E0)>>3; *d++ = (bgr&0x7E0)>>3;
*d++ = (bgr&0xF800)>>8; *d++ = (bgr&0xF800)>>8;
#else #else
*d++ = (bgr&0xF800)>>8; *d++ = (bgr&0xF800)>>8;
*d++ = (bgr&0x7E0)>>3; *d++ = (bgr&0x7E0)>>3;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
*d++ = 0; *d++ = 0;
#endif #endif
} }
} }
void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size) void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
{ {
const uint16_t *end; const uint16_t *end;
uint8_t *d = (uint8_t *)dst; uint8_t *d = (uint8_t *)dst;
const uint16_t *s = (const uint16_t *)src; const uint16_t *s = (const uint16_t *)src;
end = s + src_size/2; end = s + src_size/2;
while(s < end) while (s < end)
{ {
register uint16_t bgr; register uint16_t bgr;
bgr = *s++; bgr = *s++;
*d++ = (bgr&0xF800)>>8; *d++ = (bgr&0xF800)>>8;
*d++ = (bgr&0x7E0)>>3; *d++ = (bgr&0x7E0)>>3;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
} }
} }
void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size) void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size >> 1; long num_pixels = src_size >> 1;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
unsigned b,g,r; unsigned b,g,r;
register uint16_t rgb; register uint16_t rgb;
rgb = src[2*i]; rgb = src[2*i];
r = rgb&0x1F; r = rgb&0x1F;
g = (rgb&0x7E0)>>5; g = (rgb&0x7E0)>>5;
b = (rgb&0xF800)>>11; b = (rgb&0xF800)>>11;
dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11); dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
} }
} }
void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size) void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size >> 1; long num_pixels = src_size >> 1;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
unsigned b,g,r; unsigned b,g,r;
register uint16_t rgb; register uint16_t rgb;
rgb = src[2*i]; rgb = src[2*i];
r = rgb&0x1F; r = rgb&0x1F;
g = (rgb&0x7E0)>>5; g = (rgb&0x7E0)>>5;
b = (rgb&0xF800)>>11; b = (rgb&0xF800)>>11;
dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10); dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
} }
} }
void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size) void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size)
{ {
const uint16_t *end; const uint16_t *end;
uint8_t *d = (uint8_t *)dst; uint8_t *d = (uint8_t *)dst;
const uint16_t *s = (const uint16_t *)src; const uint16_t *s = (const uint16_t *)src;
end = s + src_size/2; end = s + src_size/2;
while(s < end) while (s < end)
{ {
register uint16_t bgr; register uint16_t bgr;
bgr = *s++; bgr = *s++;
#ifdef WORDS_BIGENDIAN #ifdef WORDS_BIGENDIAN
*d++ = 0; *d++ = 0;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
*d++ = (bgr&0x3E0)>>2; *d++ = (bgr&0x3E0)>>2;
*d++ = (bgr&0x7C00)>>7; *d++ = (bgr&0x7C00)>>7;
#else #else
*d++ = (bgr&0x7C00)>>7; *d++ = (bgr&0x7C00)>>7;
*d++ = (bgr&0x3E0)>>2; *d++ = (bgr&0x3E0)>>2;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
*d++ = 0; *d++ = 0;
#endif #endif
} }
} }
void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size) void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size)
{ {
const uint16_t *end; const uint16_t *end;
uint8_t *d = (uint8_t *)dst; uint8_t *d = (uint8_t *)dst;
const uint16_t *s = (uint16_t *)src; const uint16_t *s = (uint16_t *)src;
end = s + src_size/2; end = s + src_size/2;
while(s < end) while (s < end)
{ {
register uint16_t bgr; register uint16_t bgr;
bgr = *s++; bgr = *s++;
*d++ = (bgr&0x7C00)>>7; *d++ = (bgr&0x7C00)>>7;
*d++ = (bgr&0x3E0)>>2; *d++ = (bgr&0x3E0)>>2;
*d++ = (bgr&0x1F)<<3; *d++ = (bgr&0x1F)<<3;
} }
} }
void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size) void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size >> 1; long num_pixels = src_size >> 1;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
unsigned b,g,r; unsigned b,g,r;
register uint16_t rgb; register uint16_t rgb;
rgb = src[2*i]; rgb = src[2*i];
r = rgb&0x1F; r = rgb&0x1F;
g = (rgb&0x3E0)>>5; g = (rgb&0x3E0)>>5;
b = (rgb&0x7C00)>>10; b = (rgb&0x7C00)>>10;
dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11); dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11);
} }
} }
void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size) void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size >> 1; long num_pixels = src_size >> 1;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
unsigned b,g,r; unsigned b,g,r;
register uint16_t rgb; register uint16_t rgb;
rgb = src[2*i]; rgb = src[2*i];
r = rgb&0x1F; r = rgb&0x1F;
g = (rgb&0x3E0)>>5; g = (rgb&0x3E0)>>5;
b = (rgb&0x7C00)>>10; b = (rgb&0x7C00)>>10;
dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10); dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10);
} }
} }
void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size) void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size)
{ {
long i; long i;
long num_pixels = src_size; long num_pixels = src_size;
for(i=0; i<num_pixels; i++) for (i=0; i<num_pixels; i++)
{ {
unsigned b,g,r; unsigned b,g,r;
register uint8_t rgb; register uint8_t rgb;
rgb = src[i]; rgb = src[i];
r = (rgb&0x07); r = (rgb&0x07);
g = (rgb&0x38)>>3; g = (rgb&0x38)>>3;
b = (rgb&0xC0)>>6; b = (rgb&0xC0)>>6;
dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6); dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6);
} }
} }

69
libswscale/rgb2rgb.h
Unescape View File

@ -1,5 +1,4 @@
/* /*
*
* rgb2rgb.h, Software RGB to RGB convertor * rgb2rgb.h, Software RGB to RGB convertor
* pluralize by Software PAL8 to RGB convertor * pluralize by Software PAL8 to RGB convertor
* Software YUV to YUV convertor * Software YUV to YUV convertor
@ -30,18 +29,18 @@
// Note: do not fix the dependence on stdio.h // Note: do not fix the dependence on stdio.h
/* A full collection of rgb to rgb(bgr) convertors */ /* A full collection of rgb to rgb(bgr) convertors */
extern void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb24to32) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb24to16) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb24to15) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb32to24) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb32to16) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb32to15) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb15to16) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb15to24) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb15to32) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb16to15) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb16to24) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size); extern void (*rgb16to32) (const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size); extern void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size); extern void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size);
extern void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); extern void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size);
@ -59,7 +58,7 @@ extern void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size);
extern void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size); extern void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size);
extern void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size); extern void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size);
extern void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size); extern void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size);
extern void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size); extern void rgb8tobgr8 (const uint8_t *src, uint8_t *dst, long src_size);
extern void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); extern void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette);
@ -85,16 +84,16 @@ extern void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, c
* problem for anyone then tell me, and ill fix it) * problem for anyone then tell me, and ill fix it)
*/ */
extern void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, extern void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
/** /**
* *
* width should be a multiple of 16 * width should be a multiple of 16
*/ */
extern void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, extern void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
/** /**
* *
@ -102,8 +101,8 @@ extern void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uin
* problem for anyone then tell me, and ill fix it) * problem for anyone then tell me, and ill fix it)
*/ */
extern void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, extern void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
long width, long height, long width, long height,
long lumStride, long chromStride, long srcStride); long lumStride, long chromStride, long srcStride);
/** /**
* *
@ -111,8 +110,8 @@ extern void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint
* problem for anyone then tell me, and ill fix it) * problem for anyone then tell me, and ill fix it)
*/ */
extern void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, extern void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
long width, long height, long width, long height,
long lumStride, long chromStride, long dstStride); long lumStride, long chromStride, long dstStride);
/** /**
* *
@ -121,26 +120,26 @@ extern void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_
* chrominance data is only taken from every secound line others are ignored FIXME write HQ version * chrominance data is only taken from every secound line others are ignored FIXME write HQ version
*/ */
extern void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, extern void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
long width, long height, long width, long height,
long lumStride, long chromStride, long srcStride); long lumStride, long chromStride, long srcStride);
extern void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height, extern void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height,
long srcStride, long dstStride); long srcStride, long dstStride);
extern void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst, extern void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst,
long width, long height, long src1Stride, long width, long height, long src1Stride,
long src2Stride, long dstStride); long src2Stride, long dstStride);
extern void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2, extern void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,
uint8_t *dst1, uint8_t *dst2, uint8_t *dst1, uint8_t *dst2,
long width, long height, long width, long height,
long srcStride1, long srcStride2, long srcStride1, long srcStride2,
long dstStride1, long dstStride2); long dstStride1, long dstStride2);
extern void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, extern void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
uint8_t *dst, uint8_t *dst,
long width, long height, long width, long height,
long srcStride1, long srcStride2, long srcStride1, long srcStride2,
long srcStride3, long dstStride); long srcStride3, long dstStride);
void sws_rgb2rgb_init(int flags); void sws_rgb2rgb_init(int flags);

4619
libswscale/rgb2rgb_template.c
View File

File diff suppressed because it is too large Load Diff

852
libswscale/swscale_altivec_template.c
Unescape View File

@ -31,518 +31,514 @@
static inline void static inline void
altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW) { altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW) {
register int i; register int i;
vector unsigned int altivec_vectorShiftInt19 = vector unsigned int altivec_vectorShiftInt19 =
vec_add(vec_splat_u32(10),vec_splat_u32(9)); vec_add(vec_splat_u32(10),vec_splat_u32(9));
if ((unsigned long)dest % 16) { if ((unsigned long)dest % 16) {
/* badly aligned store, we force store alignement */ /* badly aligned store, we force store alignement */
/* and will handle load misalignement on val w/ vec_perm */ /* and will handle load misalignement on val w/ vec_perm */
vector unsigned char perm1; vector unsigned char perm1;
vector signed int v1; vector signed int v1;
for (i = 0 ; (i < dstW) && for (i = 0 ; (i < dstW) &&
(((unsigned long)dest + i) % 16) ; i++) { (((unsigned long)dest + i) % 16) ; i++) {
int t = val[i] >> 19; int t = val[i] >> 19;
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t); dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
} }
perm1 = vec_lvsl(i << 2, val); perm1 = vec_lvsl(i << 2, val);
v1 = vec_ld(i << 2, val); v1 = vec_ld(i << 2, val);
for ( ; i < (dstW - 15); i+=16) { for ( ; i < (dstW - 15); i+=16) {
int offset = i << 2; int offset = i << 2;
vector signed int v2 = vec_ld(offset + 16, val); vector signed int v2 = vec_ld(offset + 16, val);
vector signed int v3 = vec_ld(offset + 32, val); vector signed int v3 = vec_ld(offset + 32, val);
vector signed int v4 = vec_ld(offset + 48, val); vector signed int v4 = vec_ld(offset + 48, val);
vector signed int v5 = vec_ld(offset + 64, val); vector signed int v5 = vec_ld(offset + 64, val);
vector signed int v12 = vec_perm(v1,v2,perm1); vector signed int v12 = vec_perm(v1,v2,perm1);
vector signed int v23 = vec_perm(v2,v3,perm1); vector signed int v23 = vec_perm(v2,v3,perm1);
vector signed int v34 = vec_perm(v3,v4,perm1); vector signed int v34 = vec_perm(v3,v4,perm1);
vector signed int v45 = vec_perm(v4,v5,perm1); vector signed int v45 = vec_perm(v4,v5,perm1);
vector signed int vA = vec_sra(v12, altivec_vectorShiftInt19); vector signed int vA = vec_sra(v12, altivec_vectorShiftInt19);
vector signed int vB = vec_sra(v23, altivec_vectorShiftInt19); vector signed int vB = vec_sra(v23, altivec_vectorShiftInt19);
vector signed int vC = vec_sra(v34, altivec_vectorShiftInt19); vector signed int vC = vec_sra(v34, altivec_vectorShiftInt19);
vector signed int vD = vec_sra(v45, altivec_vectorShiftInt19); vector signed int vD = vec_sra(v45, altivec_vectorShiftInt19);
vector unsigned short vs1 = vec_packsu(vA, vB); vector unsigned short vs1 = vec_packsu(vA, vB);
vector unsigned short vs2 = vec_packsu(vC, vD); vector unsigned short vs2 = vec_packsu(vC, vD);
vector unsigned char vf = vec_packsu(vs1, vs2); vector unsigned char vf = vec_packsu(vs1, vs2);
vec_st(vf, i, dest); vec_st(vf, i, dest);
v1 = v5; v1 = v5;
}
} else { // dest is properly aligned, great
for (i = 0; i < (dstW - 15); i+=16) {
int offset = i << 2;
vector signed int v1 = vec_ld(offset, val);
vector signed int v2 = vec_ld(offset + 16, val);
vector signed int v3 = vec_ld(offset + 32, val);
vector signed int v4 = vec_ld(offset + 48, val);
vector signed int v5 = vec_sra(v1, altivec_vectorShiftInt19);
vector signed int v6 = vec_sra(v2, altivec_vectorShiftInt19);
vector signed int v7 = vec_sra(v3, altivec_vectorShiftInt19);
vector signed int v8 = vec_sra(v4, altivec_vectorShiftInt19);
vector unsigned short vs1 = vec_packsu(v5, v6);
vector unsigned short vs2 = vec_packsu(v7, v8);
vector unsigned char vf = vec_packsu(vs1, vs2);
vec_st(vf, i, dest);
}
} }
} else { // dest is properly aligned, great for ( ; i < dstW ; i++) {
for (i = 0; i < (dstW - 15); i+=16) { int t = val[i] >> 19;
int offset = i << 2; dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
vector signed int v1 = vec_ld(offset, val);
vector signed int v2 = vec_ld(offset + 16, val);
vector signed int v3 = vec_ld(offset + 32, val);
vector signed int v4 = vec_ld(offset + 48, val);
vector signed int v5 = vec_sra(v1, altivec_vectorShiftInt19);
vector signed int v6 = vec_sra(v2, altivec_vectorShiftInt19);
vector signed int v7 = vec_sra(v3, altivec_vectorShiftInt19);
vector signed int v8 = vec_sra(v4, altivec_vectorShiftInt19);
vector unsigned short vs1 = vec_packsu(v5, v6);
vector unsigned short vs2 = vec_packsu(v7, v8);
vector unsigned char vf = vec_packsu(vs1, vs2);
vec_st(vf, i, dest);
} }
}
for ( ; i < dstW ; i++) {
int t = val[i] >> 19;
dest[i] = (t < 0) ? 0 : ((t > 255) ? 255 : t);
}
} }
static inline void static inline void
yuv2yuvX_altivec_real(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, yuv2yuvX_altivec_real(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW) uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
{ {
const vector signed int vini = {(1 << 18), (1 << 18), (1 << 18), (1 << 18)}; const vector signed int vini = {(1 << 18), (1 << 18), (1 << 18), (1 << 18)};
register int i, j; register int i, j;
{ {
int __attribute__ ((aligned (16))) val[dstW]; int __attribute__ ((aligned (16))) val[dstW];
for (i = 0; i < (dstW -7); i+=4) { for (i = 0; i < (dstW -7); i+=4) {
vec_st(vini, i << 2, val); vec_st(vini, i << 2, val);
} }
for (; i < dstW; i++) { for (; i < dstW; i++) {
val[i] = (1 << 18); val[i] = (1 << 18);
} }
for (j = 0; j < lumFilterSize; j++) { for (j = 0; j < lumFilterSize; j++) {
vector signed short l1, vLumFilter = vec_ld(j << 1, lumFilter); vector signed short l1, vLumFilter = vec_ld(j << 1, lumFilter);
vector unsigned char perm, perm0 = vec_lvsl(j << 1, lumFilter); vector unsigned char perm, perm0 = vec_lvsl(j << 1, lumFilter);
vLumFilter = vec_perm(vLumFilter, vLumFilter, perm0); vLumFilter = vec_perm(vLumFilter, vLumFilter, perm0);
vLumFilter = vec_splat(vLumFilter, 0); // lumFilter[j] is loaded 8 times in vLumFilter vLumFilter = vec_splat(vLumFilter, 0); // lumFilter[j] is loaded 8 times in vLumFilter
perm = vec_lvsl(0, lumSrc[j]); perm = vec_lvsl(0, lumSrc[j]);
l1 = vec_ld(0, lumSrc[j]); l1 = vec_ld(0, lumSrc[j]);
for (i = 0; i < (dstW - 7); i+=8) { for (i = 0; i < (dstW - 7); i+=8) {
int offset = i << 2; int offset = i << 2;
vector signed short l2 = vec_ld((i << 1) + 16, lumSrc[j]); vector signed short l2 = vec_ld((i << 1) + 16, lumSrc[j]);
vector signed int v1 = vec_ld(offset, val); vector signed int v1 = vec_ld(offset, val);
vector signed int v2 = vec_ld(offset + 16, val); vector signed int v2 = vec_ld(offset + 16, val);
vector signed short ls = vec_perm(l1, l2, perm); // lumSrc[j][i] ... lumSrc[j][i+7] vector signed short ls = vec_perm(l1, l2, perm); // lumSrc[j][i] ... lumSrc[j][i+7]
vector signed int i1 = vec_mule(vLumFilter, ls); vector signed int i1 = vec_mule(vLumFilter, ls);
vector signed int i2 = vec_mulo(vLumFilter, ls); vector signed int i2 = vec_mulo(vLumFilter, ls);
vector signed int vf1 = vec_mergeh(i1, i2); vector signed int vf1 = vec_mergeh(i1, i2);
vector signed int vf2 = vec_mergel(i1, i2); // lumSrc[j][i] * lumFilter[j] ... lumSrc[j][i+7] * lumFilter[j] vector signed int vf2 = vec_mergel(i1, i2); // lumSrc[j][i] * lumFilter[j] ... lumSrc[j][i+7] * lumFilter[j]
vector signed int vo1 = vec_add(v1, vf1); vector signed int vo1 = vec_add(v1, vf1);
vector signed int vo2 = vec_add(v2, vf2); vector signed int vo2 = vec_add(v2, vf2);
vec_st(vo1, offset, val); vec_st(vo1, offset, val);
vec_st(vo2, offset + 16, val); vec_st(vo2, offset + 16, val);
l1 = l2; l1 = l2;
} }
for ( ; i < dstW; i++) { for ( ; i < dstW; i++) {
val[i] += lumSrc[j][i] * lumFilter[j]; val[i] += lumSrc[j][i] * lumFilter[j];
} }
} }
altivec_packIntArrayToCharArray(val,dest,dstW); altivec_packIntArrayToCharArray(val,dest,dstW);
}
if (uDest != 0) {
int __attribute__ ((aligned (16))) u[chrDstW];
int __attribute__ ((aligned (16))) v[chrDstW];
for (i = 0; i < (chrDstW -7); i+=4) {
vec_st(vini, i << 2, u);
vec_st(vini, i << 2, v);
}
for (; i < chrDstW; i++) {
u[i] = (1 << 18);
v[i] = (1 << 18);
} }
if (uDest != 0) {
int __attribute__ ((aligned (16))) u[chrDstW];
int __attribute__ ((aligned (16))) v[chrDstW];
for (i = 0; i < (chrDstW -7); i+=4) {
vec_st(vini, i << 2, u);
vec_st(vini, i << 2, v);
}
for (; i < chrDstW; i++) {
u[i] = (1 << 18);
v[i] = (1 << 18);
}
for (j = 0; j < chrFilterSize; j++) { for (j = 0; j < chrFilterSize; j++) {
vector signed short l1, l1_V, vChrFilter = vec_ld(j << 1, chrFilter); vector signed short l1, l1_V, vChrFilter = vec_ld(j << 1, chrFilter);
vector unsigned char perm, perm0 = vec_lvsl(j << 1, chrFilter); vector unsigned char perm, perm0 = vec_lvsl(j << 1, chrFilter);
vChrFilter = vec_perm(vChrFilter, vChrFilter, perm0); vChrFilter = vec_perm(vChrFilter, vChrFilter, perm0);
vChrFilter = vec_splat(vChrFilter, 0); // chrFilter[j] is loaded 8 times in vChrFilter vChrFilter = vec_splat(vChrFilter, 0); // chrFilter[j] is loaded 8 times in vChrFilter
perm = vec_lvsl(0, chrSrc[j]); perm = vec_lvsl(0, chrSrc[j]);
l1 = vec_ld(0, chrSrc[j]); l1 = vec_ld(0, chrSrc[j]);
l1_V = vec_ld(2048 << 1, chrSrc[j]); l1_V = vec_ld(2048 << 1, chrSrc[j]);
for (i = 0; i < (chrDstW - 7); i+=8) { for (i = 0; i < (chrDstW - 7); i+=8) {
int offset = i << 2; int offset = i << 2;
vector signed short l2 = vec_ld((i << 1) + 16, chrSrc[j]); vector signed short l2 = vec_ld((i << 1) + 16, chrSrc[j]);
vector signed short l2_V = vec_ld(((i + 2048) << 1) + 16, chrSrc[j]); vector signed short l2_V = vec_ld(((i + 2048) << 1) + 16, chrSrc[j]);
vector signed int v1 = vec_ld(offset, u); vector signed int v1 = vec_ld(offset, u);
vector signed int v2 = vec_ld(offset + 16, u); vector signed int v2 = vec_ld(offset + 16, u);
vector signed int v1_V = vec_ld(offset, v); vector signed int v1_V = vec_ld(offset, v);
vector signed int v2_V = vec_ld(offset + 16, v); vector signed int v2_V = vec_ld(offset + 16, v);
vector signed short ls = vec_perm(l1, l2, perm); // chrSrc[j][i] ... chrSrc[j][i+7] vector signed short ls = vec_perm(l1, l2, perm); // chrSrc[j][i] ... chrSrc[j][i+7]
vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrSrc[j][i+2048] ... chrSrc[j][i+2055] vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrSrc[j][i+2048] ... chrSrc[j][i+2055]
vector signed int i1 = vec_mule(vChrFilter, ls); vector signed int i1 = vec_mule(vChrFilter, ls);
vector signed int i2 = vec_mulo(vChrFilter, ls); vector signed int i2 = vec_mulo(vChrFilter, ls);
vector signed int i1_V = vec_mule(vChrFilter, ls_V); vector signed int i1_V = vec_mule(vChrFilter, ls_V);
vector signed int i2_V = vec_mulo(vChrFilter, ls_V); vector signed int i2_V = vec_mulo(vChrFilter, ls_V);
vector signed int vf1 = vec_mergeh(i1, i2); vector signed int vf1 = vec_mergeh(i1, i2);
vector signed int vf2 = vec_mergel(i1, i2); // chrSrc[j][i] * chrFilter[j] ... chrSrc[j][i+7] * chrFilter[j] vector signed int vf2 = vec_mergel(i1, i2); // chrSrc[j][i] * chrFilter[j] ... chrSrc[j][i+7] * chrFilter[j]
vector signed int vf1_V = vec_mergeh(i1_V, i2_V); vector signed int vf1_V = vec_mergeh(i1_V, i2_V);
vector signed int vf2_V = vec_mergel(i1_V, i2_V); // chrSrc[j][i] * chrFilter[j] ... chrSrc[j][i+7] * chrFilter[j] vector signed int vf2_V = vec_mergel(i1_V, i2_V); // chrSrc[j][i] * chrFilter[j] ... chrSrc[j][i+7] * chrFilter[j]
vector signed int vo1 = vec_add(v1, vf1); vector signed int vo1 = vec_add(v1, vf1);
vector signed int vo2 = vec_add(v2, vf2); vector signed int vo2 = vec_add(v2, vf2);
vector signed int vo1_V = vec_add(v1_V, vf1_V); vector signed int vo1_V = vec_add(v1_V, vf1_V);
vector signed int vo2_V = vec_add(v2_V, vf2_V); vector signed int vo2_V = vec_add(v2_V, vf2_V);
vec_st(vo1, offset, u); vec_st(vo1, offset, u);
vec_st(vo2, offset + 16, u); vec_st(vo2, offset + 16, u);
vec_st(vo1_V, offset, v); vec_st(vo1_V, offset, v);
vec_st(vo2_V, offset + 16, v); vec_st(vo2_V, offset + 16, v);
l1 = l2; l1 = l2;
l1_V = l2_V; l1_V = l2_V;
} }
for ( ; i < chrDstW; i++) { for ( ; i < chrDstW; i++) {
u[i] += chrSrc[j][i] * chrFilter[j]; u[i] += chrSrc[j][i] * chrFilter[j];
v[i] += chrSrc[j][i + 2048] * chrFilter[j]; v[i] += chrSrc[j][i + 2048] * chrFilter[j];
} }
}
altivec_packIntArrayToCharArray(u,uDest,chrDstW);
altivec_packIntArrayToCharArray(v,vDest,chrDstW);
} }
altivec_packIntArrayToCharArray(u,uDest,chrDstW);
altivec_packIntArrayToCharArray(v,vDest,chrDstW);
}
} }
static inline void hScale_altivec_real(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, int16_t *filter, int16_t *filterPos, int filterSize) { static inline void hScale_altivec_real(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc, int16_t *filter, int16_t *filterPos, int filterSize) {
register int i; register int i;
int __attribute__ ((aligned (16))) tempo[4]; int __attribute__ ((aligned (16))) tempo[4];
if (filterSize % 4) { if (filterSize % 4) {
for(i=0; i<dstW; i++) { for (i=0; i<dstW; i++) {
register int j; register int j;
register int srcPos = filterPos[i]; register int srcPos = filterPos[i];
register int val = 0; register int val = 0;
for(j=0; j<filterSize; j++) { for (j=0; j<filterSize; j++) {
val += ((int)src[srcPos + j])*filter[filterSize*i + j]; val += ((int)src[srcPos + j])*filter[filterSize*i + j];
} }
dst[i] = av_clip(val>>7, 0, (1<<15)-1); dst[i] = av_clip(val>>7, 0, (1<<15)-1);
}
} }
} else
else switch (filterSize) {
switch (filterSize) { case 4:
case 4:
{ {
for(i=0; i<dstW; i++) { for (i=0; i<dstW; i++) {
register int srcPos = filterPos[i]; register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src); vector unsigned char src_v0 = vec_ld(srcPos, src);
vector unsigned char src_v1, src_vF; vector unsigned char src_v1, src_vF;
vector signed short src_v, filter_v; vector signed short src_v, filter_v;
vector signed int val_vEven, val_s; vector signed int val_vEven, val_s;
if ((((int)src + srcPos)% 16) > 12) { if ((((int)src + srcPos)% 16) > 12) {
src_v1 = vec_ld(srcPos + 16, src); src_v1 = vec_ld(srcPos + 16, src);
} }
src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src)); src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
src_v = // vec_unpackh sign-extends... src_v = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
// now put our elements in the even slots // now put our elements in the even slots
src_v = vec_mergeh(src_v, (vector signed short)vzero); src_v = vec_mergeh(src_v, (vector signed short)vzero);
filter_v = vec_ld(i << 3, filter); filter_v = vec_ld(i << 3, filter);
// the 3 above is 2 (filterSize == 4) + 1 (sizeof(short) == 2) // the 3 above is 2 (filterSize == 4) + 1 (sizeof(short) == 2)
// the neat trick : we only care for half the elements, // the neat trick : we only care for half the elements,
// high or low depending on (i<<3)%16 (it's 0 or 8 here), // high or low depending on (i<<3)%16 (it's 0 or 8 here),
// and we're going to use vec_mule, so we chose // and we're going to use vec_mule, so we chose
// carefully how to "unpack" the elements into the even slots // carefully how to "unpack" the elements into the even slots
if ((i << 3) % 16) if ((i << 3) % 16)
filter_v = vec_mergel(filter_v,(vector signed short)vzero); filter_v = vec_mergel(filter_v,(vector signed short)vzero);
else else
filter_v = vec_mergeh(filter_v,(vector signed short)vzero); filter_v = vec_mergeh(filter_v,(vector signed short)vzero);
val_vEven = vec_mule(src_v, filter_v); val_vEven = vec_mule(src_v, filter_v);
val_s = vec_sums(val_vEven, vzero); val_s = vec_sums(val_vEven, vzero);
vec_st(val_s, 0, tempo); vec_st(val_s, 0, tempo);
dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1); dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);
} }
} }
break; break;
case 8: case 8:
{ {
for(i=0; i<dstW; i++) { for (i=0; i<dstW; i++) {
register int srcPos = filterPos[i]; register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src); vector unsigned char src_v0 = vec_ld(srcPos, src);
vector unsigned char src_v1, src_vF; vector unsigned char src_v1, src_vF;
vector signed short src_v, filter_v; vector signed short src_v, filter_v;
vector signed int val_v, val_s; vector signed int val_v, val_s;
if ((((int)src + srcPos)% 16) > 8) { if ((((int)src + srcPos)% 16) > 8) {
src_v1 = vec_ld(srcPos + 16, src); src_v1 = vec_ld(srcPos + 16, src);
} }
src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src)); src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
src_v = // vec_unpackh sign-extends... src_v = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
filter_v = vec_ld(i << 4, filter); filter_v = vec_ld(i << 4, filter);
// the 4 above is 3 (filterSize == 8) + 1 (sizeof(short) == 2) // the 4 above is 3 (filterSize == 8) + 1 (sizeof(short) == 2)
val_v = vec_msums(src_v, filter_v, (vector signed int)vzero); val_v = vec_msums(src_v, filter_v, (vector signed int)vzero);
val_s = vec_sums(val_v, vzero); val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo); vec_st(val_s, 0, tempo);
dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1); dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);
} }
} }
break; break;
case 16: case 16:
{ {
for(i=0; i<dstW; i++) { for (i=0; i<dstW; i++) {
register int srcPos = filterPos[i]; register int srcPos = filterPos[i];
vector unsigned char src_v0 = vec_ld(srcPos, src); vector unsigned char src_v0 = vec_ld(srcPos, src);
vector unsigned char src_v1 = vec_ld(srcPos + 16, src); vector unsigned char src_v1 = vec_ld(srcPos + 16, src);
vector unsigned char src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src)); vector unsigned char src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
vector signed short src_vA = // vec_unpackh sign-extends... vector signed short src_vA = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
vector signed short src_vB = // vec_unpackh sign-extends... vector signed short src_vB = // vec_unpackh sign-extends...
(vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
vector signed short filter_v0 = vec_ld(i << 5, filter); vector signed short filter_v0 = vec_ld(i << 5, filter);
vector signed short filter_v1 = vec_ld((i << 5) + 16, filter); vector signed short filter_v1 = vec_ld((i << 5) + 16, filter);
// the 5 above are 4 (filterSize == 16) + 1 (sizeof(short) == 2) // the 5 above are 4 (filterSize == 16) + 1 (sizeof(short) == 2)
vector signed int val_acc = vec_msums(src_vA, filter_v0, (vector signed int)vzero); vector signed int val_acc = vec_msums(src_vA, filter_v0, (vector signed int)vzero);
vector signed int val_v = vec_msums(src_vB, filter_v1, val_acc); vector signed int val_v = vec_msums(src_vB, filter_v1, val_acc);
vector signed int val_s = vec_sums(val_v, vzero); vector signed int val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo); vec_st(val_s, 0, tempo);
dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1); dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);
} }
} }
break; break;
default: default:
{ {
for(i=0; i<dstW; i++) { for (i=0; i<dstW; i++) {
register int j; register int j;
register int srcPos = filterPos[i]; register int srcPos = filterPos[i];
vector signed int val_s, val_v = (vector signed int)vzero; vector signed int val_s, val_v = (vector signed int)vzero;
vector signed short filter_v0R = vec_ld(i * 2 * filterSize, filter); vector signed short filter_v0R = vec_ld(i * 2 * filterSize, filter);
vector unsigned char permF = vec_lvsl((i * 2 * filterSize), filter); vector unsigned char permF = vec_lvsl((i * 2 * filterSize), filter);
vector unsigned char src_v0 = vec_ld(srcPos, src); vector unsigned char src_v0 = vec_ld(srcPos, src);
vector unsigned char permS = vec_lvsl(srcPos, src); vector unsigned char permS = vec_lvsl(srcPos, src);
for (j = 0 ; j < filterSize - 15; j += 16) { for (j = 0 ; j < filterSize - 15; j += 16) {
vector unsigned char src_v1 = vec_ld(srcPos + j + 16, src); vector unsigned char src_v1 = vec_ld(srcPos + j + 16, src);
vector unsigned char src_vF = vec_perm(src_v0, src_v1, permS); vector unsigned char src_vF = vec_perm(src_v0, src_v1, permS);
vector signed short src_vA = // vec_unpackh sign-extends... vector signed short src_vA = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
vector signed short src_vB = // vec_unpackh sign-extends... vector signed short src_vB = // vec_unpackh sign-extends...
(vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
vector signed short filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter); vector signed short filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
vector signed short filter_v2R = vec_ld((i * 2 * filterSize) + (j * 2) + 32, filter); vector signed short filter_v2R = vec_ld((i * 2 * filterSize) + (j * 2) + 32, filter);
vector signed short filter_v0 = vec_perm(filter_v0R, filter_v1R, permF); vector signed short filter_v0 = vec_perm(filter_v0R, filter_v1R, permF);
vector signed short filter_v1 = vec_perm(filter_v1R, filter_v2R, permF); vector signed short filter_v1 = vec_perm(filter_v1R, filter_v2R, permF);
vector signed int val_acc = vec_msums(src_vA, filter_v0, val_v); vector signed int val_acc = vec_msums(src_vA, filter_v0, val_v);
val_v = vec_msums(src_vB, filter_v1, val_acc); val_v = vec_msums(src_vB, filter_v1, val_acc);
filter_v0R = filter_v2R; filter_v0R = filter_v2R;
src_v0 = src_v1; src_v0 = src_v1;
} }
if (j < (filterSize-7)) { if (j < (filterSize-7)) {
// loading src_v0 is useless, it's already done above // loading src_v0 is useless, it's already done above
//vector unsigned char src_v0 = vec_ld(srcPos + j, src); //vector unsigned char src_v0 = vec_ld(srcPos + j, src);
vector unsigned char src_v1, src_vF; vector unsigned char src_v1, src_vF;
vector signed short src_v, filter_v1R, filter_v; vector signed short src_v, filter_v1R, filter_v;
if ((((int)src + srcPos)% 16) > 8) { if ((((int)src + srcPos)% 16) > 8) {
src_v1 = vec_ld(srcPos + j + 16, src); src_v1 = vec_ld(srcPos + j + 16, src);
} }
src_vF = vec_perm(src_v0, src_v1, permS); src_vF = vec_perm(src_v0, src_v1, permS);
src_v = // vec_unpackh sign-extends... src_v = // vec_unpackh sign-extends...
(vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF)); (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
// loading filter_v0R is useless, it's already done above // loading filter_v0R is useless, it's already done above
//vector signed short filter_v0R = vec_ld((i * 2 * filterSize) + j, filter); //vector signed short filter_v0R = vec_ld((i * 2 * filterSize) + j, filter);
filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter); filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
filter_v = vec_perm(filter_v0R, filter_v1R, permF); filter_v = vec_perm(filter_v0R, filter_v1R, permF);
val_v = vec_msums(src_v, filter_v, val_v); val_v = vec_msums(src_v, filter_v, val_v);
} }
val_s = vec_sums(val_v, vzero); val_s = vec_sums(val_v, vzero);
vec_st(val_s, 0, tempo); vec_st(val_s, 0, tempo);
dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1); dst[i] = av_clip(tempo[3]>>7, 0, (1<<15)-1);
} }
} }
} }
} }
static inline int yv12toyuy2_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, static inline int yv12toyuy2_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) { int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY; uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
// yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); // yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
uint8_t *ysrc = src[0]; uint8_t *ysrc = src[0];
uint8_t *usrc = src[1]; uint8_t *usrc = src[1];
uint8_t *vsrc = src[2]; uint8_t *vsrc = src[2];
const int width = c->srcW; const int width = c->srcW;
const int height = srcSliceH; const int height = srcSliceH;
const int lumStride = srcStride[0]; const int lumStride = srcStride[0];
const int chromStride = srcStride[1]; const int chromStride = srcStride[1];
const int dstStride = dstStride_a[0]; const int dstStride = dstStride_a[0];
const vector unsigned char yperm = vec_lvsl(0, ysrc); const vector unsigned char yperm = vec_lvsl(0, ysrc);
const int vertLumPerChroma = 2; const int vertLumPerChroma = 2;
register unsigned int y; register unsigned int y;
if(width&15){ if (width&15) {
yv12toyuy2( ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride); yv12toyuy2( ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride);
return srcSliceH; return srcSliceH;
} }
/* this code assume:
1) dst is 16 bytes-aligned
2) dstStride is a multiple of 16
3) width is a multiple of 16
4) lum&chrom stride are multiple of 8
*/
for(y=0; y<height; y++) /* this code assume:
{
int i; 1) dst is 16 bytes-aligned
for (i = 0; i < width - 31; i+= 32) { 2) dstStride is a multiple of 16
const unsigned int j = i >> 1; 3) width is a multiple of 16
vector unsigned char v_yA = vec_ld(i, ysrc); 4) lum&chrom stride are multiple of 8
vector unsigned char v_yB = vec_ld(i + 16, ysrc); */
vector unsigned char v_yC = vec_ld(i + 32, ysrc);
vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm); for (y=0; y<height; y++) {
vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm); int i;
vector unsigned char v_uA = vec_ld(j, usrc); for (i = 0; i < width - 31; i+= 32) {
vector unsigned char v_uB = vec_ld(j + 16, usrc); const unsigned int j = i >> 1;
vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc)); vector unsigned char v_yA = vec_ld(i, ysrc);
vector unsigned char v_vA = vec_ld(j, vsrc); vector unsigned char v_yB = vec_ld(i + 16, ysrc);
vector unsigned char v_vB = vec_ld(j + 16, vsrc); vector unsigned char v_yC = vec_ld(i + 32, ysrc);
vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc)); vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm);
vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm);
vector unsigned char v_uv_b = vec_mergel(v_u, v_v); vector unsigned char v_uA = vec_ld(j, usrc);
vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a); vector unsigned char v_uB = vec_ld(j + 16, usrc);
vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a); vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc));
vector unsigned char v_yuy2_2 = vec_mergeh(v_y2, v_uv_b); vector unsigned char v_vA = vec_ld(j, vsrc);
vector unsigned char v_yuy2_3 = vec_mergel(v_y2, v_uv_b); vector unsigned char v_vB = vec_ld(j + 16, vsrc);
vec_st(v_yuy2_0, (i << 1), dst); vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc));
vec_st(v_yuy2_1, (i << 1) + 16, dst); vector unsigned char v_uv_a = vec_mergeh(v_u, v_v);
vec_st(v_yuy2_2, (i << 1) + 32, dst); vector unsigned char v_uv_b = vec_mergel(v_u, v_v);
vec_st(v_yuy2_3, (i << 1) + 48, dst); vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a);
} vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a);
if (i < width) { vector unsigned char v_yuy2_2 = vec_mergeh(v_y2, v_uv_b);
const unsigned int j = i >> 1; vector unsigned char v_yuy2_3 = vec_mergel(v_y2, v_uv_b);
vector unsigned char v_y1 = vec_ld(i, ysrc); vec_st(v_yuy2_0, (i << 1), dst);
vector unsigned char v_u = vec_ld(j, usrc); vec_st(v_yuy2_1, (i << 1) + 16, dst);
vector unsigned char v_v = vec_ld(j, vsrc); vec_st(v_yuy2_2, (i << 1) + 32, dst);
vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); vec_st(v_yuy2_3, (i << 1) + 48, dst);
vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a); }
vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a); if (i < width) {
vec_st(v_yuy2_0, (i << 1), dst); const unsigned int j = i >> 1;
vec_st(v_yuy2_1, (i << 1) + 16, dst); vector unsigned char v_y1 = vec_ld(i, ysrc);
} vector unsigned char v_u = vec_ld(j, usrc);
if((y&(vertLumPerChroma-1))==(vertLumPerChroma-1) ) vector unsigned char v_v = vec_ld(j, vsrc);
{ vector unsigned char v_uv_a = vec_mergeh(v_u, v_v);
usrc += chromStride; vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a);
vsrc += chromStride; vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a);
} vec_st(v_yuy2_0, (i << 1), dst);
ysrc += lumStride; vec_st(v_yuy2_1, (i << 1) + 16, dst);
dst += dstStride; }
if ( (y&(vertLumPerChroma-1))==(vertLumPerChroma-1) ) {
usrc += chromStride;
vsrc += chromStride;
}
ysrc += lumStride;
dst += dstStride;
} }
return srcSliceH; return srcSliceH;
} }
static inline int yv12touyvy_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, static inline int yv12touyvy_unscaled_altivec(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) { int srcSliceH, uint8_t* dstParam[], int dstStride_a[]) {
uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY; uint8_t *dst=dstParam[0] + dstStride_a[0]*srcSliceY;
// yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); // yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
uint8_t *ysrc = src[0]; uint8_t *ysrc = src[0];
uint8_t *usrc = src[1]; uint8_t *usrc = src[1];
uint8_t *vsrc = src[2]; uint8_t *vsrc = src[2];
const int width = c->srcW; const int width = c->srcW;
const int height = srcSliceH; const int height = srcSliceH;
const int lumStride = srcStride[0]; const int lumStride = srcStride[0];
const int chromStride = srcStride[1]; const int chromStride = srcStride[1];
const int dstStride = dstStride_a[0]; const int dstStride = dstStride_a[0];
const int vertLumPerChroma = 2; const int vertLumPerChroma = 2;
const vector unsigned char yperm = vec_lvsl(0, ysrc); const vector unsigned char yperm = vec_lvsl(0, ysrc);
register unsigned int y; register unsigned int y;
if(width&15){ if (width&15) {
yv12touyvy( ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride); yv12touyvy( ysrc, usrc, vsrc, dst,c->srcW,srcSliceH, lumStride, chromStride, dstStride);
return srcSliceH; return srcSliceH;
} }
/* this code assume:
1) dst is 16 bytes-aligned
2) dstStride is a multiple of 16
3) width is a multiple of 16
4) lum&chrom stride are multiple of 8
*/
for(y=0; y<height; y++) /* this code assume:
{
int i; 1) dst is 16 bytes-aligned
for (i = 0; i < width - 31; i+= 32) { 2) dstStride is a multiple of 16
const unsigned int j = i >> 1; 3) width is a multiple of 16
vector unsigned char v_yA = vec_ld(i, ysrc); 4) lum&chrom stride are multiple of 8
vector unsigned char v_yB = vec_ld(i + 16, ysrc); */
vector unsigned char v_yC = vec_ld(i + 32, ysrc);
vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm); for (y=0; y<height; y++) {
vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm); int i;
vector unsigned char v_uA = vec_ld(j, usrc); for (i = 0; i < width - 31; i+= 32) {
vector unsigned char v_uB = vec_ld(j + 16, usrc); const unsigned int j = i >> 1;
vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc)); vector unsigned char v_yA = vec_ld(i, ysrc);
vector unsigned char v_vA = vec_ld(j, vsrc); vector unsigned char v_yB = vec_ld(i + 16, ysrc);
vector unsigned char v_vB = vec_ld(j + 16, vsrc); vector unsigned char v_yC = vec_ld(i + 32, ysrc);
vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc)); vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm);
vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm);
vector unsigned char v_uv_b = vec_mergel(v_u, v_v); vector unsigned char v_uA = vec_ld(j, usrc);
vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1); vector unsigned char v_uB = vec_ld(j + 16, usrc);
vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1); vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc));
vector unsigned char v_uyvy_2 = vec_mergeh(v_uv_b, v_y2); vector unsigned char v_vA = vec_ld(j, vsrc);
vector unsigned char v_uyvy_3 = vec_mergel(v_uv_b, v_y2); vector unsigned char v_vB = vec_ld(j + 16, vsrc);
vec_st(v_uyvy_0, (i << 1), dst); vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc));
vec_st(v_uyvy_1, (i << 1) + 16, dst); vector unsigned char v_uv_a = vec_mergeh(v_u, v_v);
vec_st(v_uyvy_2, (i << 1) + 32, dst); vector unsigned char v_uv_b = vec_mergel(v_u, v_v);
vec_st(v_uyvy_3, (i << 1) + 48, dst); vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1);
} vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1);
if (i < width) { vector unsigned char v_uyvy_2 = vec_mergeh(v_uv_b, v_y2);
const unsigned int j = i >> 1; vector unsigned char v_uyvy_3 = vec_mergel(v_uv_b, v_y2);
vector unsigned char v_y1 = vec_ld(i, ysrc); vec_st(v_uyvy_0, (i << 1), dst);
vector unsigned char v_u = vec_ld(j, usrc); vec_st(v_uyvy_1, (i << 1) + 16, dst);
vector unsigned char v_v = vec_ld(j, vsrc); vec_st(v_uyvy_2, (i << 1) + 32, dst);
vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); vec_st(v_uyvy_3, (i << 1) + 48, dst);
vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1); }
vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1); if (i < width) {
vec_st(v_uyvy_0, (i << 1), dst); const unsigned int j = i >> 1;
vec_st(v_uyvy_1, (i << 1) + 16, dst); vector unsigned char v_y1 = vec_ld(i, ysrc);
} vector unsigned char v_u = vec_ld(j, usrc);
if((y&(vertLumPerChroma-1))==(vertLumPerChroma-1) ) vector unsigned char v_v = vec_ld(j, vsrc);
{ vector unsigned char v_uv_a = vec_mergeh(v_u, v_v);
usrc += chromStride; vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1);
vsrc += chromStride; vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1);
} vec_st(v_uyvy_0, (i << 1), dst);
ysrc += lumStride; vec_st(v_uyvy_1, (i << 1) + 16, dst);
dst += dstStride; }
if ( (y&(vertLumPerChroma-1))==(vertLumPerChroma-1) ) {
usrc += chromStride;
vsrc += chromStride;
}
ysrc += lumStride;
dst += dstStride;
} }
return srcSliceH; return srcSliceH;
} }

Write Preview
Loading…
Cancel
Save