[FFmpeg-cvslog] swscale: un-special-case yuv2yuvX16_c().
Ronald S. Bultje
git at videolan.org
Wed Jun 8 05:31:32 CEST 2011
ffmpeg | branch: master | Ronald S. Bultje <rsbultje at gmail.com> | Sun Jun 5 21:15:44 2011 -0400| [bda9b20fa49975c6b9c39308818e7773eb78b411] | committer: Ronald S. Bultje
swscale: un-special-case yuv2yuvX16_c().
Make yuv2yuvX16_c a function pointer for yuv2yuvX(), so that the
function pointer becomes bitdepth-independent.
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=bda9b20fa49975c6b9c39308818e7773eb78b411
---
libswscale/ppc/swscale_altivec.c | 6 +++-
libswscale/swscale.c | 63 ++++++++++++++++++++----------------
libswscale/x86/swscale_template.c | 5 ++-
3 files changed, 44 insertions(+), 30 deletions(-)
diff --git a/libswscale/ppc/swscale_altivec.c b/libswscale/ppc/swscale_altivec.c
index 7f4dfcd..423297a 100644
--- a/libswscale/ppc/swscale_altivec.c
+++ b/libswscale/ppc/swscale_altivec.c
@@ -403,11 +403,15 @@ static void hScale_altivec_real(int16_t *dst, int dstW,
void ff_sws_init_swScale_altivec(SwsContext *c)
{
+ enum PixelFormat dstFormat = c->dstFormat;
+
if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
return;
c->hScale = hScale_altivec_real;
- c->yuv2yuvX = yuv2yuvX_altivec_real;
+ if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat)) {
+ c->yuv2yuvX = yuv2yuvX_altivec_real;
+ }
/* The following list of supported dstFormat values should
* match what's found in the body of ff_yuv2packedX_altivec() */
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index fd64b81..35b0233 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -258,17 +258,20 @@ yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
output_pixel(&aDest[i], val);
}
}
+#undef output_pixel
}
#define yuv2NBPS(bits, BE_LE, is_be) \
-static void yuv2yuvX ## bits ## BE_LE ## _c(const int16_t *lumFilter, \
+static void yuv2yuvX ## bits ## BE_LE ## _c(SwsContext *c, const int16_t *lumFilter, \
const int16_t **lumSrc, int lumFilterSize, \
const int16_t *chrFilter, const int16_t **chrUSrc, \
const int16_t **chrVSrc, \
int chrFilterSize, const int16_t **alpSrc, \
- uint16_t *dest, uint16_t *uDest, uint16_t *vDest, \
- uint16_t *aDest, int dstW, int chrDstW) \
+ uint8_t *_dest, uint8_t *_uDest, uint8_t *_vDest, \
+ uint8_t *_aDest, int dstW, int chrDstW) \
{ \
+ uint16_t *dest = (uint16_t *) _dest, *uDest = (uint16_t *) _uDest, \
+ *vDest = (uint16_t *) _vDest, *aDest = (uint16_t *) _aDest; \
yuv2yuvX16_c_template(lumFilter, lumSrc, lumFilterSize, \
chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
alpSrc, \
@@ -282,24 +285,24 @@ yuv2NBPS(10, LE, 0);
yuv2NBPS(16, BE, 1);
yuv2NBPS(16, LE, 0);
-static inline void yuv2yuvX16_c(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
+static inline void yuv2yuvX16_c(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
const int16_t *chrFilter, const int16_t **chrUSrc, const int16_t **chrVSrc, int chrFilterSize,
- const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
+ const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW,
enum PixelFormat dstFormat)
{
#define conv16(bits) \
if (isBE(dstFormat)) { \
- yuv2yuvX ## bits ## BE_c(lumFilter, lumSrc, lumFilterSize, \
- chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
- alpSrc, \
- dest, uDest, vDest, aDest, \
- dstW, chrDstW); \
+ yuv2yuvX ## bits ## BE_c(c, lumFilter, lumSrc, lumFilterSize, \
+ chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
+ alpSrc, \
+ dest, uDest, vDest, aDest, \
+ dstW, chrDstW); \
} else { \
- yuv2yuvX ## bits ## LE_c(lumFilter, lumSrc, lumFilterSize, \
- chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
- alpSrc, \
- dest, uDest, vDest, aDest, \
- dstW, chrDstW); \
+ yuv2yuvX ## bits ## LE_c(c, lumFilter, lumSrc, lumFilterSize, \
+ chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
+ alpSrc, \
+ dest, uDest, vDest, aDest, \
+ dstW, chrDstW); \
}
if (is16BPS(dstFormat)) {
conv16(16);
@@ -1783,14 +1786,7 @@ static int swScale(SwsContext *c, const uint8_t* src[],
} else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12 like
const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
- if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
- yuv2yuvX16_c(vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize,
- vChrFilter+chrDstY*vChrFilterSize, chrUSrcPtr,
- chrVSrcPtr, vChrFilterSize,
- alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest,
- (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW,
- dstFormat);
- } else if (vLumFilterSize == 1 && vChrFilterSize == 1) { // unscaled YV12
+ if (c->yuv2yuv1 && vLumFilterSize == 1 && vChrFilterSize == 1) { // unscaled YV12
const int16_t *lumBuf = lumSrcPtr[0];
const int16_t *chrUBuf= chrUSrcPtr[0];
const int16_t *chrVBuf= chrVSrcPtr[0];
@@ -1870,9 +1866,9 @@ static int swScale(SwsContext *c, const uint8_t* src[],
const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
- yuv2yuvX16_c(vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize,
+ yuv2yuvX16_c(c, vLumFilter+dstY*vLumFilterSize , lumSrcPtr, vLumFilterSize,
vChrFilter+chrDstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
- alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW,
+ alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW,
dstFormat);
} else {
yuv2yuvX_c(c, vLumFilter+dstY*vLumFilterSize,
@@ -1921,11 +1917,22 @@ static int swScale(SwsContext *c, const uint8_t* src[],
static void sws_init_swScale_c(SwsContext *c)
{
- enum PixelFormat srcFormat = c->srcFormat;
+ enum PixelFormat srcFormat = c->srcFormat,
+ dstFormat = c->dstFormat;
c->yuv2nv12X = yuv2nv12X_c;
- c->yuv2yuv1 = yuv2yuv1_c;
- c->yuv2yuvX = yuv2yuvX_c;
+ if (is16BPS(dstFormat)) {
+ c->yuv2yuvX = isBE(dstFormat) ? yuv2yuvX16BE_c : yuv2yuvX16LE_c;
+ } else if (is9_OR_10BPS(dstFormat)) {
+ if (dstFormat == PIX_FMT_YUV420P9BE || dstFormat == PIX_FMT_YUV420P9LE) {
+ c->yuv2yuvX = isBE(dstFormat) ? yuv2yuvX9BE_c : yuv2yuvX9LE_c;
+ } else {
+ c->yuv2yuvX = isBE(dstFormat) ? yuv2yuvX10BE_c : yuv2yuvX10LE_c;
+ }
+ } else {
+ c->yuv2yuv1 = yuv2yuv1_c;
+ c->yuv2yuvX = yuv2yuvX_c;
+ }
c->yuv2packed1 = yuv2packed1_c;
c->yuv2packed2 = yuv2packed2_c;
c->yuv2packedX = yuv2packedX_c;
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index efaadaa..174c3fb 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -2185,8 +2185,10 @@ static inline void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *d
static void RENAME(sws_init_swScale)(SwsContext *c)
{
- enum PixelFormat srcFormat = c->srcFormat;
+ enum PixelFormat srcFormat = c->srcFormat,
+ dstFormat = c->dstFormat;
+ if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat)) {
if (!(c->flags & SWS_BITEXACT)) {
if (c->flags & SWS_ACCURATE_RND) {
c->yuv2yuv1 = RENAME(yuv2yuv1_ar );
@@ -2236,6 +2238,7 @@ static void RENAME(sws_init_swScale)(SwsContext *c)
break;
}
}
+ }
#if !COMPILE_TEMPLATE_MMX2
c->hScale = RENAME(hScale );
More information about the ffmpeg-cvslog
mailing list