swscale: use 15-bit intermediates for 9/10-bit scaling.

1 files changed, 50 insertions, 25 deletions

diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 9492303301..60787aed91 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -64,11 +64,16 @@ typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t* src[],
  * without any additional vertical scaling (or point-scaling).
  *
  * @param c       SWS scaling context
- * @param lumSrc  scaled luma (Y) source data, 15bit for 8bit output
- * @param chrUSrc scaled chroma (U) source data, 15bit for 8bit output
- * @param chrVSrc scaled chroma (V) source data, 15bit for 8bit output
- * @param alpSrc  scaled alpha (A) source data, 15bit for 8bit output
- * @param dest    pointer to the 4 output planes (Y/U/V/A)
+ * @param lumSrc  scaled luma (Y) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param alpSrc  scaled alpha (A) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param dest    pointer to the 4 output planes (Y/U/V/A). For >8bit
+ *                output, this is in uint16_t
  * @param dstW    width of dest[0], dest[3], lumSrc and alpSrc in pixels
  * @param chrDstW width of dest[1], dest[2], chrUSrc and chrVSrc
  */
@@ -82,14 +87,19 @@ typedef void (*yuv2planar1_fn) (struct SwsContext *c,
  *
  * @param c             SWS scaling context
  * @param lumFilter     vertical luma/alpha scaling coefficients, 12bit [0,4096]
- * @param lumSrc        scaled luma (Y) source data, 15bit for 8bit output
+ * @param lumSrc        scaled luma (Y) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
  * @param lumFilterSize number of vertical luma/alpha input lines to scale
  * @param chrFilter     vertical chroma scaling coefficients, 12bit [0,4096]
- * @param chrUSrc       scaled chroma (U) source data, 15bit for 8bit output
- * @param chrVSrc       scaled chroma (V) source data, 15bit for 8bit output
+ * @param chrUSrc       scaled chroma (U) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
+ * @param chrVSrc       scaled chroma (V) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
  * @param chrFilterSize number of vertical chroma input lines to scale
- * @param alpSrc        scaled alpha (A) source data, 15bit for 8bit output
- * @param dest          pointer to the 4 output planes (Y/U/V/A)
+ * @param alpSrc        scaled alpha (A) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
+ * @param dest          pointer to the 4 output planes (Y/U/V/A). For >8bit
+ *                      output, this is in uint16_t
  * @param dstW          width of dest[0], dest[3], lumSrc and alpSrc in pixels
  * @param chrDstW       width of dest[1], dest[2], chrUSrc and chrVSrc
  */
@@ -105,11 +115,16 @@ typedef void (*yuv2planarX_fn) (struct SwsContext *c, const int16_t *lumFilter,
  * that this function may do chroma scaling, see the "uvalpha" argument.
  *
  * @param c       SWS scaling context
- * @param lumSrc  scaled luma (Y) source data, 15bit for 8bit output
- * @param chrUSrc scaled chroma (U) source data, 15bit for 8bit output
- * @param chrVSrc scaled chroma (V) source data, 15bit for 8bit output
- * @param alpSrc  scaled alpha (A) source data, 15bit for 8bit output
- * @param dest    pointer to the output plane
+ * @param lumSrc  scaled luma (Y) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param alpSrc  scaled alpha (A) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param dest    pointer to the output plane. For 16bit output, this is
+ *                uint16_t
  * @param dstW    width of lumSrc and alpSrc in pixels, number of pixels
  *                to write into dest[]
  * @param uvalpha chroma scaling coefficient for the second line of chroma
@@ -132,11 +147,16 @@ typedef void (*yuv2packed1_fn) (struct SwsContext *c,  const int16_t *lumSrc,
  * output by doing bilinear scaling between two input lines.
  *
  * @param c       SWS scaling context
- * @param lumSrc  scaled luma (Y) source data, 15bit for 8bit output
- * @param chrUSrc scaled chroma (U) source data, 15bit for 8bit output
- * @param chrVSrc scaled chroma (V) source data, 15bit for 8bit output
- * @param alpSrc  scaled alpha (A) source data, 15bit for 8bit output
- * @param dest    pointer to the output plane
+ * @param lumSrc  scaled luma (Y) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param alpSrc  scaled alpha (A) source data, 15bit for 8-10bit output,
+ *                19-bit for 16bit output (in int32_t)
+ * @param dest    pointer to the output plane. For 16bit output, this is
+ *                uint16_t
  * @param dstW    width of lumSrc and alpSrc in pixels, number of pixels
  *                to write into dest[]
  * @param yalpha  luma/alpha scaling coefficients for the second input line.
@@ -160,14 +180,19 @@ typedef void (*yuv2packed2_fn) (struct SwsContext *c,  const int16_t *lumSrc[2],
  *
  * @param c             SWS scaling context
  * @param lumFilter     vertical luma/alpha scaling coefficients, 12bit [0,4096]
- * @param lumSrc        scaled luma (Y) source data, 15bit for 8bit output
+ * @param lumSrc        scaled luma (Y) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
  * @param lumFilterSize number of vertical luma/alpha input lines to scale
  * @param chrFilter     vertical chroma scaling coefficients, 12bit [0,4096]
- * @param chrUSrc       scaled chroma (U) source data, 15bit for 8bit output
- * @param chrVSrc       scaled chroma (V) source data, 15bit for 8bit output
+ * @param chrUSrc       scaled chroma (U) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
+ * @param chrVSrc       scaled chroma (V) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
  * @param chrFilterSize number of vertical chroma input lines to scale
- * @param alpSrc        scaled alpha (A) source data, 15bit for 8bit output
- * @param dest          pointer to the output plane
+ * @param alpSrc        scaled alpha (A) source data, 15bit for 8-10bit output,
+ *                      19-bit for 16bit output (in int32_t)
+ * @param dest          pointer to the output plane. For 16bit output, this is
+ *                      uint16_t
  * @param dstW          width of lumSrc and alpSrc in pixels, number of pixels
  *                      to write into dest[]
  * @param y             vertical line number for this output. This does not need