#include "libavutil/mathematics.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
+#include "libavutil/ppc/cpu.h"
#include "libavutil/x86/asm.h"
#include "libavutil/x86/cpu.h"
#include "rgb2rgb.h"
#define RET 0xC3 // near return opcode for x86
typedef struct FormatEntry {
- int is_supported_in, is_supported_out;
+ uint8_t is_supported_in :1;
+ uint8_t is_supported_out :1;
+ uint8_t is_supported_endianness :1;
} FormatEntry;
static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
[AV_PIX_FMT_YUVJ420P] = { 1, 1 },
[AV_PIX_FMT_YUVJ422P] = { 1, 1 },
[AV_PIX_FMT_YUVJ444P] = { 1, 1 },
+ [AV_PIX_FMT_YVYU422] = { 1, 1 },
[AV_PIX_FMT_UYVY422] = { 1, 1 },
[AV_PIX_FMT_UYYVYY411] = { 0, 0 },
[AV_PIX_FMT_BGR8] = { 1, 1 },
[AV_PIX_FMT_RGBA] = { 1, 1 },
[AV_PIX_FMT_ABGR] = { 1, 1 },
[AV_PIX_FMT_BGRA] = { 1, 1 },
+ [AV_PIX_FMT_GRAY12BE] = { 1, 1 },
+ [AV_PIX_FMT_GRAY12LE] = { 1, 1 },
[AV_PIX_FMT_GRAY16BE] = { 1, 1 },
[AV_PIX_FMT_GRAY16LE] = { 1, 1 },
[AV_PIX_FMT_YUV440P] = { 1, 1 },
[AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
[AV_PIX_FMT_RGB48BE] = { 1, 1 },
[AV_PIX_FMT_RGB48LE] = { 1, 1 },
+ [AV_PIX_FMT_RGBA64BE] = { 0, 0, 1 },
+ [AV_PIX_FMT_RGBA64LE] = { 0, 0, 1 },
[AV_PIX_FMT_RGB565BE] = { 1, 1 },
[AV_PIX_FMT_RGB565LE] = { 1, 1 },
[AV_PIX_FMT_RGB555BE] = { 1, 1 },
[AV_PIX_FMT_RGB444BE] = { 1, 1 },
[AV_PIX_FMT_BGR444LE] = { 1, 1 },
[AV_PIX_FMT_BGR444BE] = { 1, 1 },
- [AV_PIX_FMT_Y400A] = { 1, 0 },
+ [AV_PIX_FMT_YA8] = { 1, 0 },
+ [AV_PIX_FMT_YA16BE] = { 1, 0 },
+ [AV_PIX_FMT_YA16LE] = { 1, 0 },
[AV_PIX_FMT_BGR48BE] = { 1, 1 },
[AV_PIX_FMT_BGR48LE] = { 1, 1 },
+ [AV_PIX_FMT_BGRA64BE] = { 0, 0, 1 },
+ [AV_PIX_FMT_BGRA64LE] = { 0, 0, 1 },
[AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
+ [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
+ [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
[AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
+ [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
+ [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
[AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
[AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
[AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
+ [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
+ [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
[AV_PIX_FMT_GBRP] = { 1, 1 },
[AV_PIX_FMT_GBRP9LE] = { 1, 1 },
[AV_PIX_FMT_GBRP9BE] = { 1, 1 },
[AV_PIX_FMT_GBRP10LE] = { 1, 1 },
[AV_PIX_FMT_GBRP10BE] = { 1, 1 },
+ [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
+ [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
[AV_PIX_FMT_GBRP16LE] = { 1, 0 },
[AV_PIX_FMT_GBRP16BE] = { 1, 0 },
+ [AV_PIX_FMT_GBRAP] = { 1, 1 },
+ [AV_PIX_FMT_GBRAP12LE] = { 1, 1 },
+ [AV_PIX_FMT_GBRAP12BE] = { 1, 1 },
+ [AV_PIX_FMT_GBRAP16LE] = { 1, 0 },
+ [AV_PIX_FMT_GBRAP16BE] = { 1, 0 },
+ [AV_PIX_FMT_XYZ12BE] = { 0, 0, 1 },
+ [AV_PIX_FMT_XYZ12LE] = { 0, 0, 1 },
+ [AV_PIX_FMT_P010LE] = { 1, 0 },
+ [AV_PIX_FMT_P010BE] = { 1, 0 },
};
int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
format_entries[pix_fmt].is_supported_out : 0;
}
-extern const int32_t ff_yuv2rgb_coeffs[8][4];
+int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
+{
+ return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
+ format_entries[pix_fmt].is_supported_endianness : 0;
+}
const char *sws_format_name(enum AVPixelFormat format)
{
dist - 1.0);
}
-static int initFilter(int16_t **outFilter, int32_t **filterPos,
- int *outFilterSize, int xInc, int srcW, int dstW,
- int filterAlign, int one, int flags, int cpu_flags,
- SwsVector *srcFilter, SwsVector *dstFilter,
- double param[2], int is_horizontal)
+static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
+ int *outFilterSize, int xInc, int srcW,
+ int dstW, int filterAlign, int one,
+ int flags, int cpu_flags,
+ SwsVector *srcFilter, SwsVector *dstFilter,
+ double param[2], int is_horizontal)
{
int i;
int filterSize;
xDstInSrc = xInc - 0x10000;
for (i = 0; i < dstW; i++) {
- int xx = (xDstInSrc - ((filterSize - 2) << 16)) / (1 << 17);
+ int xx = (xDstInSrc - ((int64_t)(filterSize - 2) << 16)) / (1 << 17);
int j;
(*filterPos)[i] = xx;
for (j = 0; j < filterSize; j++) {
}
coeff *= fone >> (30 + 24);
}
-#if 0
- else if (flags & SWS_X) {
- double p = param ? param * 0.01 : 0.3;
- coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
- coeff *= pow(2.0, -p * d * d);
- }
-#endif
else if (flags & SWS_X) {
double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
double c;
minFilterSize = min;
}
- if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) {
+ if (PPC_ALTIVEC(cpu_flags)) {
// we can handle the special case 4, so we don't want to go the full 8
if (minFilterSize < 5)
filterAlign = 4;
}
#if HAVE_MMXEXT_INLINE
-static int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
- int16_t *filter, int32_t *filterPos,
- int numSplits)
+static av_cold int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
+ int16_t *filter, int32_t *filterPos,
+ int numSplits)
{
uint8_t *fragmentA;
x86_reg imm8OfPShufW1A;
"jmp 9f \n\t"
// Begin
"0: \n\t"
- "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
- "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
- "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
+ "movq (%%"FF_REG_d", %%"FF_REG_a"), %%mm3 \n\t"
+ "movd (%%"FF_REG_c", %%"FF_REG_S"), %%mm0 \n\t"
+ "movd 1(%%"FF_REG_c", %%"FF_REG_S"), %%mm1 \n\t"
"punpcklbw %%mm7, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm1, %%mm1 \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
- "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
+ "movl 8(%%"FF_REG_b", %%"FF_REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
- "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
+ "movq %%mm0, (%%"FF_REG_D", %%"FF_REG_a") \n\t"
- "add $8, %%"REG_a" \n\t"
+ "add $8, %%"FF_REG_a" \n\t"
// End
"9: \n\t"
// "int $3 \n\t"
"jmp 9f \n\t"
// Begin
"0: \n\t"
- "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
- "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
+ "movq (%%"FF_REG_d", %%"FF_REG_a"), %%mm3 \n\t"
+ "movd (%%"FF_REG_c", %%"FF_REG_S"), %%mm0 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm0, %%mm1 \n\t"
"1: \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
- "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
+ "movl 8(%%"FF_REG_b", %%"FF_REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
- "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
+ "movq %%mm0, (%%"FF_REG_D", %%"FF_REG_a") \n\t"
- "add $8, %%"REG_a" \n\t"
+ "add $8, %%"FF_REG_a" \n\t"
// End
"9: \n\t"
// "int $3 \n\t"
contrast, saturation);
// FIXME factorize
- if (HAVE_ALTIVEC && av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)
- ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness,
- contrast, saturation);
+ if (ARCH_PPC)
+ ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
+ contrast, saturation);
return 0;
}
SwsContext *c = av_mallocz(sizeof(SwsContext));
if (c) {
- c->av_class = &sws_context_class;
+ c->av_class = &ff_sws_context_class;
av_opt_set_defaults(c);
}
flags = c->flags;
emms_c();
if (!rgb15to16)
- sws_rgb2rgb_init();
+ ff_rgb2rgb_init();
unscaled = (srcW == dstW && srcH == dstH);
+ if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
+ av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
if (!sws_isSupportedInput(srcFormat)) {
av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
sws_format_name(srcFormat));
sws_format_name(dstFormat));
return AVERROR(EINVAL);
}
+ }
i = flags & (SWS_POINT |
SWS_AREA |
SWS_SINC |
SWS_SPLINE |
SWS_BICUBLIN);
- if (!i || (i & (i - 1))) {
+
+ /* provide a default scaler if not set by caller */
+ if (!i) {
+ if (dstW < srcW && dstH < srcH)
+ flags |= SWS_GAUSS;
+ else if (dstW > srcW && dstH > srcH)
+ flags |= SWS_SINC;
+ else
+ flags |= SWS_LANCZOS;
+ c->flags = flags;
+ } else if (i & (i - 1)) {
av_log(c, AV_LOG_ERROR,
"Exactly one scaler algorithm must be chosen\n");
return AVERROR(EINVAL);
srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
+ srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
(flags & SWS_FAST_BILINEAR)))
c->chrSrcHSubSample = 1;
- // Note the -((-x)>>y) is so that we always round toward +inf.
- c->chrSrcW = -((-srcW) >> c->chrSrcHSubSample);
- c->chrSrcH = -((-srcH) >> c->chrSrcVSubSample);
- c->chrDstW = -((-dstW) >> c->chrDstHSubSample);
- c->chrDstH = -((-dstH) >> c->chrDstVSubSample);
+ // Note the AV_CEIL_RSHIFT is so that we always round toward +inf.
+ c->chrSrcW = AV_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
+ c->chrSrcH = AV_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
+ c->chrDstW = AV_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
+ c->chrDstH = AV_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
/* unscaled special cases */
if (unscaled && !usesHFilter && !usesVFilter &&
(c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
ff_get_unscaled_swscale(c);
- if (c->swScale) {
+ if (c->swscale) {
if (flags & SWS_PRINT_INFO)
av_log(c, AV_LOG_INFO,
"using unscaled %s -> %s special converter\n",
}
}
- c->srcBpc = 1 + desc_src->comp[0].depth_minus1;
+ c->srcBpc = desc_src->comp[0].depth;
if (c->srcBpc < 8)
c->srcBpc = 8;
- c->dstBpc = 1 + desc_dst->comp[0].depth_minus1;
+ c->dstBpc = desc_dst->comp[0].depth;
if (c->dstBpc < 8)
c->dstBpc = 8;
if (c->dstBpc == 16)
FF_ALLOC_OR_GOTO(c, c->formatConvBuffer,
(FFALIGN(srcW, 16) * 2 * FFALIGN(c->srcBpc, 8) >> 3) + 16,
fail);
- if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 10) {
+ if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 12) {
c->canMMXEXTBeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
(srcW & 15) == 0) ? 1 : 0;
if (!c->canMMXEXTBeUsed && dstW >= srcW && (srcW & 15) == 0
} else
#endif /* HAVE_MMXEXT_INLINE */
{
- const int filterAlign =
- (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
- (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
- 1;
+ const int filterAlign = X86_MMX(cpu_flags) ? 4 :
+ PPC_ALTIVEC(cpu_flags) ? 8 : 1;
if (initFilter(&c->hLumFilter, &c->hLumFilterPos,
&c->hLumFilterSize, c->lumXInc,
/* precalculate vertical scaler filter coefficients */
{
- const int filterAlign =
- (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
- (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
- 1;
+ const int filterAlign = X86_MMX(cpu_flags) ? 2 :
+ PPC_ALTIVEC(cpu_flags) ? 8 : 1;
if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
av_log(c, AV_LOG_INFO, "using 3DNOW\n");
else if (INLINE_MMX(cpu_flags))
av_log(c, AV_LOG_INFO, "using MMX\n");
- else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC)
+ else if (PPC_ALTIVEC(cpu_flags))
av_log(c, AV_LOG_INFO, "using AltiVec\n");
else
av_log(c, AV_LOG_INFO, "using C\n");
c->chrXInc, c->chrYInc);
}
- c->swScale = ff_getSwsFunc(c);
+ c->swscale = ff_getSwsFunc(c);
return 0;
fail: // FIXME replace things by appropriate error codes
return -1;
}
-#if FF_API_SWS_GETCONTEXT
SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
int dstW, int dstH, enum AVPixelFormat dstFormat,
int flags, SwsFilter *srcFilter,
return c;
}
-#endif
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
float lumaSharpen, float chromaSharpen,
filter->chrV = sws_getIdentityVec();
}
+ if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
+ goto fail;
+
if (chromaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec();
+ if (!id)
+ goto fail;
sws_scaleVec(filter->chrH, -chromaSharpen);
sws_scaleVec(filter->chrV, -chromaSharpen);
sws_addVec(filter->chrH, id);
if (lumaSharpen != 0.0) {
SwsVector *id = sws_getIdentityVec();
+ if (!id)
+ goto fail;
sws_scaleVec(filter->lumH, -lumaSharpen);
sws_scaleVec(filter->lumV, -lumaSharpen);
sws_addVec(filter->lumH, id);
sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
return filter;
+
+fail:
+ sws_freeVec(filter->lumH);
+ sws_freeVec(filter->lumV);
+ sws_freeVec(filter->chrH);
+ sws_freeVec(filter->chrV);
+ av_freep(&filter);
+ return NULL;
}
SwsVector *sws_allocVec(int length)