sws: support 12&14 bit planar colorspaces

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

7 files changed, 126 insertions, 16 deletions

diff --git a/libswscale/input.c b/libswscale/input.c
index c9c91d0bca..9d03c681ac 100644
--- a/libswscale/input.c
+++ b/libswscale/input.c
@@ -727,6 +727,26 @@ static void planar_rgb10be_to_y(uint8_t *dst, const uint8_t *src[4], int w)
     planar_rgb16_to_y(dst, src, w, 10, 1);
 }
 
+static void planar_rgb12le_to_y(uint8_t *dst, const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_y(dst, src, w, 12, 0);
+}
+
+static void planar_rgb12be_to_y(uint8_t *dst, const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_y(dst, src, w, 12, 1);
+}
+
+static void planar_rgb14le_to_y(uint8_t *dst, const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_y(dst, src, w, 14, 0);
+}
+
+static void planar_rgb14be_to_y(uint8_t *dst, const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_y(dst, src, w, 14, 1);
+}
+
 static void planar_rgb16le_to_y(uint8_t *dst, const uint8_t *src[4], int w)
 {
     planar_rgb16_to_y(dst, src, w, 16, 0);
@@ -780,6 +800,30 @@ static void planar_rgb10be_to_uv(uint8_t *dstU, uint8_t *dstV,
     planar_rgb16_to_uv(dstU, dstV, src, w, 10, 1);
 }
 
+static void planar_rgb12le_to_uv(uint8_t *dstU, uint8_t *dstV,
+                                 const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_uv(dstU, dstV, src, w, 12, 0);
+}
+
+static void planar_rgb12be_to_uv(uint8_t *dstU, uint8_t *dstV,
+                                 const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_uv(dstU, dstV, src, w, 12, 1);
+}
+
+static void planar_rgb14le_to_uv(uint8_t *dstU, uint8_t *dstV,
+                                 const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_uv(dstU, dstV, src, w, 14, 0);
+}
+
+static void planar_rgb14be_to_uv(uint8_t *dstU, uint8_t *dstV,
+                                 const uint8_t *src[4], int w)
+{
+    planar_rgb16_to_uv(dstU, dstV, src, w, 14, 1);
+}
+
 static void planar_rgb16le_to_uv(uint8_t *dstU, uint8_t *dstV,
                                  const uint8_t *src[4], int w)
 {
@@ -823,6 +867,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_GBRP10LE:
         c->readChrPlanar = planar_rgb10le_to_uv;
         break;
+    case PIX_FMT_GBRP12LE:
+        c->readChrPlanar = planar_rgb12le_to_uv;
+        break;
+    case PIX_FMT_GBRP14LE:
+        c->readChrPlanar = planar_rgb14le_to_uv;
+        break;
     case PIX_FMT_GBRP16LE:
         c->readChrPlanar = planar_rgb16le_to_uv;
         break;
@@ -832,6 +882,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_GBRP10BE:
         c->readChrPlanar = planar_rgb10be_to_uv;
         break;
+    case PIX_FMT_GBRP12BE:
+        c->readChrPlanar = planar_rgb12be_to_uv;
+        break;
+    case PIX_FMT_GBRP14BE:
+        c->readChrPlanar = planar_rgb14be_to_uv;
+        break;
     case PIX_FMT_GBRP16BE:
         c->readChrPlanar = planar_rgb16be_to_uv;
         break;
@@ -845,6 +901,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_YUV422P10LE:
     case PIX_FMT_YUV444P10LE:
     case PIX_FMT_YUV420P10LE:
+    case PIX_FMT_YUV422P12LE:
+    case PIX_FMT_YUV444P12LE:
+    case PIX_FMT_YUV420P12LE:
+    case PIX_FMT_YUV422P14LE:
+    case PIX_FMT_YUV444P14LE:
+    case PIX_FMT_YUV420P14LE:
     case PIX_FMT_YUV420P16LE:
     case PIX_FMT_YUV422P16LE:
     case PIX_FMT_YUV444P16LE:
@@ -857,6 +919,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_YUV444P10BE:
     case PIX_FMT_YUV422P10BE:
     case PIX_FMT_YUV420P10BE:
+    case PIX_FMT_YUV444P12BE:
+    case PIX_FMT_YUV422P12BE:
+    case PIX_FMT_YUV420P12BE:
+    case PIX_FMT_YUV444P14BE:
+    case PIX_FMT_YUV422P14BE:
+    case PIX_FMT_YUV420P14BE:
     case PIX_FMT_YUV420P16BE:
     case PIX_FMT_YUV422P16BE:
     case PIX_FMT_YUV444P16BE:
@@ -1028,6 +1096,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_GBRP10LE:
         c->readLumPlanar = planar_rgb10le_to_y;
         break;
+    case PIX_FMT_GBRP12LE:
+        c->readLumPlanar = planar_rgb12le_to_y;
+        break;
+    case PIX_FMT_GBRP14LE:
+        c->readLumPlanar = planar_rgb14le_to_y;
+        break;
     case PIX_FMT_GBRP16LE:
         c->readLumPlanar = planar_rgb16le_to_y;
         break;
@@ -1037,6 +1111,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_GBRP10BE:
         c->readLumPlanar = planar_rgb10be_to_y;
         break;
+    case PIX_FMT_GBRP12BE:
+        c->readLumPlanar = planar_rgb12be_to_y;
+        break;
+    case PIX_FMT_GBRP14BE:
+        c->readLumPlanar = planar_rgb14be_to_y;
+        break;
     case PIX_FMT_GBRP16BE:
         c->readLumPlanar = planar_rgb16be_to_y;
         break;
@@ -1050,6 +1130,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_YUV444P10LE:
     case PIX_FMT_YUV422P10LE:
     case PIX_FMT_YUV420P10LE:
+    case PIX_FMT_YUV444P12LE:
+    case PIX_FMT_YUV422P12LE:
+    case PIX_FMT_YUV420P12LE:
+    case PIX_FMT_YUV444P14LE:
+    case PIX_FMT_YUV422P14LE:
+    case PIX_FMT_YUV420P14LE:
     case PIX_FMT_YUV420P16LE:
     case PIX_FMT_YUV422P16LE:
     case PIX_FMT_YUV444P16LE:
@@ -1063,6 +1149,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c)
     case PIX_FMT_YUV444P10BE:
     case PIX_FMT_YUV422P10BE:
     case PIX_FMT_YUV420P10BE:
+    case PIX_FMT_YUV444P12BE:
+    case PIX_FMT_YUV422P12BE:
+    case PIX_FMT_YUV420P12BE:
+    case PIX_FMT_YUV444P14BE:
+    case PIX_FMT_YUV422P14BE:
+    case PIX_FMT_YUV420P14BE:
     case PIX_FMT_YUV420P16BE:
     case PIX_FMT_YUV422P16BE:
     case PIX_FMT_YUV444P16BE:
diff --git a/libswscale/ppc/swscale_altivec.c b/libswscale/ppc/swscale_altivec.c
index df8afb0ba2..694c8d8b9a 100644
--- a/libswscale/ppc/swscale_altivec.c
+++ b/libswscale/ppc/swscale_altivec.c
@@ -317,7 +317,7 @@ void ff_sws_init_swScale_altivec(SwsContext *c)
     if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
         return;
 
-    if (c->srcBpc == 8 && c->dstBpc <= 10) {
+    if (c->srcBpc == 8 && c->dstBpc <= 14) {
         c->hyScale = c->hcScale = hScale_altivec_real;
     }
     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index 5dd4124eba..1f807852d5 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -691,7 +691,7 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
 
 
     if (c->srcBpc == 8) {
-        if (c->dstBpc <= 10) {
+        if (c->dstBpc <= 14) {
             c->hyScale = c->hcScale = hScale8To15_c;
             if (c->flags & SWS_FAST_BILINEAR) {
                 c->hyscale_fast = hyscale_fast_c;
@@ -701,12 +701,12 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
             c->hyScale = c->hcScale = hScale8To19_c;
         }
     } else {
-        c->hyScale = c->hcScale = c->dstBpc > 10 ? hScale16To19_c
+        c->hyScale = c->hcScale = c->dstBpc > 14 ? hScale16To19_c
                                                  : hScale16To15_c;
     }
 
     if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
-        if (c->dstBpc <= 10) {
+        if (c->dstBpc <= 14) {
             if (c->srcRange) {
                 c->lumConvertRange = lumRangeFromJpeg_c;
                 c->chrConvertRange = chrRangeFromJpeg_c;
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index ae53b88e7f..5a584f04ca 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -558,8 +558,8 @@ const char *sws_format_name(enum PixelFormat format);
     (av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 15)
 
 #define is9_OR_10BPS(x) \
-    (av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 8 || \
-     av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 9)
+    (av_pix_fmt_descriptors[x].comp[0].depth_minus1 >= 8 && \
+     av_pix_fmt_descriptors[x].comp[0].depth_minus1 <= 13)
 
 #define isNBPS(x) is9_OR_10BPS(x)
 
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 9508413664..4047fb57ee 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -145,19 +145,35 @@ static const FormatEntry format_entries[PIX_FMT_NB] = {
     [PIX_FMT_YUV420P9LE]  = { 1, 1 },
     [PIX_FMT_YUV420P10BE] = { 1, 1 },
     [PIX_FMT_YUV420P10LE] = { 1, 1 },
+    [PIX_FMT_YUV420P12BE] = { 1, 1 },
+    [PIX_FMT_YUV420P12LE] = { 1, 1 },
+    [PIX_FMT_YUV420P14BE] = { 1, 1 },
+    [PIX_FMT_YUV420P14LE] = { 1, 1 },
     [PIX_FMT_YUV422P9BE]  = { 1, 1 },
     [PIX_FMT_YUV422P9LE]  = { 1, 1 },
     [PIX_FMT_YUV422P10BE] = { 1, 1 },
     [PIX_FMT_YUV422P10LE] = { 1, 1 },
+    [PIX_FMT_YUV422P12BE] = { 1, 1 },
+    [PIX_FMT_YUV422P12LE] = { 1, 1 },
+    [PIX_FMT_YUV422P14BE] = { 1, 1 },
+    [PIX_FMT_YUV422P14LE] = { 1, 1 },
     [PIX_FMT_YUV444P9BE]  = { 1, 1 },
     [PIX_FMT_YUV444P9LE]  = { 1, 1 },
     [PIX_FMT_YUV444P10BE] = { 1, 1 },
     [PIX_FMT_YUV444P10LE] = { 1, 1 },
+    [PIX_FMT_YUV444P12BE] = { 1, 1 },
+    [PIX_FMT_YUV444P12LE] = { 1, 1 },
+    [PIX_FMT_YUV444P14BE] = { 1, 1 },
+    [PIX_FMT_YUV444P14LE] = { 1, 1 },
     [PIX_FMT_GBRP]        = { 1, 0 },
     [PIX_FMT_GBRP9LE]     = { 1, 0 },
     [PIX_FMT_GBRP9BE]     = { 1, 0 },
     [PIX_FMT_GBRP10LE]    = { 1, 0 },
     [PIX_FMT_GBRP10BE]    = { 1, 0 },
+    [PIX_FMT_GBRP12LE]    = { 1, 0 },
+    [PIX_FMT_GBRP12BE]    = { 1, 0 },
+    [PIX_FMT_GBRP14LE]    = { 1, 0 },
+    [PIX_FMT_GBRP14BE]    = { 1, 0 },
     [PIX_FMT_GBRP16LE]    = { 1, 0 },
     [PIX_FMT_GBRP16BE]    = { 1, 0 },
 };
@@ -1006,7 +1022,7 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
         dst_stride <<= 1;
     FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
     if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 &&
-        c->srcBpc == 8 && c->dstBpc <= 10) {
+        c->srcBpc == 8 && c->dstBpc <= 14) {
         c->canMMX2BeUsed = (dstW >= srcW && (dstW & 31) == 0 &&
                             (srcW & 15) == 0) ? 1 : 0;
         if (!c->canMMX2BeUsed && dstW >= srcW && (srcW & 15) == 0
@@ -1036,7 +1052,7 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
             c->chrXInc += 20;
         }
         // we don't use the x86 asm scaler if MMX is available
-        else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX && c->dstBpc <= 10) {
+        else if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX && c->dstBpc <= 14) {
             c->lumXInc = ((int64_t)(srcW       - 2) << 16) / (dstW       - 2) - 20;
             c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
         }
@@ -1207,7 +1223,7 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
     // try to avoid drawing green stuff between the right end and the stride end
     for (i = 0; i < c->vChrBufSize; i++)
         if(av_pix_fmt_descriptors[c->dstFormat].comp[0].depth_minus1 == 15){
-            av_assert0(c->dstBpc > 10);
+            av_assert0(c->dstBpc > 14);
             for(j=0; j<dst_stride/2+1; j++)
                 ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
         } else
diff --git a/libswscale/x86/swscale_mmx.c b/libswscale/x86/swscale_mmx.c
index 9f7aa80e2b..bb88d3478a 100644
--- a/libswscale/x86/swscale_mmx.c
+++ b/libswscale/x86/swscale_mmx.c
@@ -22,6 +22,7 @@
 #include "config.h"
 #include "libswscale/swscale.h"
 #include "libswscale/swscale_internal.h"
+#include "libavutil/avassert.h"
 #include "libavutil/intreadwrite.h"
 #include "libavutil/x86_cpu.h"
 #include "libavutil/cpu.h"
@@ -378,19 +379,19 @@ void ff_sws_init_swScale_mmx(SwsContext *c)
 #if HAVE_YASM
 #define ASSIGN_SCALE_FUNC2(hscalefn, filtersize, opt1, opt2) do { \
     if (c->srcBpc == 8) { \
-        hscalefn = c->dstBpc <= 10 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
+        hscalefn = c->dstBpc <= 14 ? ff_hscale8to15_ ## filtersize ## _ ## opt2 : \
                                      ff_hscale8to19_ ## filtersize ## _ ## opt1; \
     } else if (c->srcBpc == 9) { \
-        hscalefn = c->dstBpc <= 10 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
+        hscalefn = c->dstBpc <= 14 ? ff_hscale9to15_ ## filtersize ## _ ## opt2 : \
                                      ff_hscale9to19_ ## filtersize ## _ ## opt1; \
     } else if (c->srcBpc == 10) { \
-        hscalefn = c->dstBpc <= 10 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
+        hscalefn = c->dstBpc <= 14 ? ff_hscale10to15_ ## filtersize ## _ ## opt2 : \
                                      ff_hscale10to19_ ## filtersize ## _ ## opt1; \
     } else if (c->srcBpc == 14 || ((c->srcFormat==PIX_FMT_PAL8||isAnyRGB(c->srcFormat)) && av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1<15)) { \
-        hscalefn = c->dstBpc <= 10 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \
+        hscalefn = c->dstBpc <= 14 ? ff_hscale14to15_ ## filtersize ## _ ## opt2 : \
                                      ff_hscale14to19_ ## filtersize ## _ ## opt1; \
     } else { /* c->srcBpc == 16 */ \
-        hscalefn = c->dstBpc <= 10 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
+        hscalefn = c->dstBpc <= 14 ? ff_hscale16to15_ ## filtersize ## _ ## opt2 : \
                                      ff_hscale16to19_ ## filtersize ## _ ## opt1; \
     } \
 } while (0)
@@ -412,7 +413,8 @@ switch(c->dstBpc){ \
     case 16: if (!isBE(c->dstFormat))            vscalefn = ff_yuv2plane1_16_ ## opt1; break; \
     case 10: if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_10_ ## opt2; break; \
     case 9:  if (!isBE(c->dstFormat) && opt2chk) vscalefn = ff_yuv2plane1_9_  ## opt2;  break; \
-    default:                                     vscalefn = ff_yuv2plane1_8_  ## opt1;  break; \
+    case 8:                                      vscalefn = ff_yuv2plane1_8_  ## opt1;  break; \
+    default: av_assert0(c->dstBpc>8); \
     }
 #define case_rgb(x, X, opt) \
         case PIX_FMT_ ## X: \
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index cf8e802229..245bfdeadd 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -1689,7 +1689,7 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
         }
     }
 
-    if (c->srcBpc == 8 && c->dstBpc <= 10) {
+    if (c->srcBpc == 8 && c->dstBpc <= 14) {
     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).
 #if COMPILE_TEMPLATE_MMX2
     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)