From 6a583261ea631e6da5593c4477e36fde9bc731ab Mon Sep 17 00:00:00 2001
From: Martin Vignali <martin.vignali@gmail.com>
Date: Sat, 17 Nov 2018 23:37:50 +0100
Subject: [PATCH] avcodec/proresdec : add 12b prores idct

based on patch by Kieran Kunhya
---
 libavcodec/proresdsp.c   | 31 ++++++++++++++++++++++++++++---
 libavcodec/simple_idct.c | 16 ++++++++++++++++
 libavcodec/simple_idct.h |  1 +
 3 files changed, 45 insertions(+), 3 deletions(-)

diff --git a/libavcodec/proresdsp.c b/libavcodec/proresdsp.c
index 7be7e68287..a3c618cdd1 100644
--- a/libavcodec/proresdsp.c
+++ b/libavcodec/proresdsp.c
@@ -29,36 +29,61 @@
 
 #define CLIP_MIN (1 << 2)                     ///< minimum value for clipping resulting pixels
 #define CLIP_MAX_10 (1 << 10) - CLIP_MIN - 1  ///< maximum value for clipping resulting pixels
+#define CLIP_MAX_12 (1 << 12) - CLIP_MIN - 1  ///< maximum value for clipping resulting pixels
 
 #define CLIP_10(x) (av_clip((x), CLIP_MIN, CLIP_MAX_10))
+#define CLIP_12(x) (av_clip((x), CLIP_MIN, CLIP_MAX_12))
 
 /**
  * Add bias value, clamp and output pixels of a slice
  */
-static void put_pixels_10(uint16_t *dst, ptrdiff_t linesize, const int16_t *in)
-{
+
+static inline void put_pixel(uint16_t *dst, ptrdiff_t linesize, const int16_t *in, int bits_per_raw_sample) {
     int x, y, src_offset, dst_offset;
 
     for (y = 0, dst_offset = 0; y < 8; y++, dst_offset += linesize) {
         for (x = 0; x < 8; x++) {
             src_offset = (y << 3) + x;
 
-            dst[dst_offset + x] = CLIP_10(in[src_offset]);
+            if (bits_per_raw_sample == 10) {
+                dst[dst_offset + x] = CLIP_10(in[src_offset]);
+            } else {//12b
+                dst[dst_offset + x] = CLIP_12(in[src_offset]);
+            }
         }
     }
 }
 
+static void put_pixels_10(uint16_t *dst, ptrdiff_t linesize, const int16_t *in)
+{
+    put_pixel(dst, linesize, in, 10);
+}
+
+static void put_pixels_12(uint16_t *dst, ptrdiff_t linesize, const int16_t *in)
+{
+    put_pixel(dst, linesize, in, 12);
+}
+
 static void prores_idct_put_10_c(uint16_t *out, ptrdiff_t linesize, int16_t *block, const int16_t *qmat)
 {
     ff_prores_idct_10(block, qmat);
     put_pixels_10(out, linesize >> 1, block);
 }
 
+static void prores_idct_put_12_c(uint16_t *out, ptrdiff_t linesize, int16_t *block, const int16_t *qmat)
+{
+    ff_prores_idct_12(block, qmat);
+    put_pixels_12(out, linesize >> 1, block);
+}
+
 av_cold int ff_proresdsp_init(ProresDSPContext *dsp, AVCodecContext *avctx)
 {
     if (avctx->bits_per_raw_sample == 10) {
         dsp->idct_put = prores_idct_put_10_c;
         dsp->idct_permutation_type = FF_IDCT_PERM_NONE;
+    } else if (avctx->bits_per_raw_sample == 12) {
+        dsp->idct_put = prores_idct_put_12_c;
+        dsp->idct_permutation_type = FF_IDCT_PERM_NONE;
     } else {
         return AVERROR_BUG;
     }
diff --git a/libavcodec/simple_idct.c b/libavcodec/simple_idct.c
index 2171d71d06..3b2e736538 100644
--- a/libavcodec/simple_idct.c
+++ b/libavcodec/simple_idct.c
@@ -251,3 +251,19 @@ void ff_prores_idct_10(int16_t *block, const int16_t *qmat)
         idctSparseCol_extrashift_10(block + i);
     }
 }
+
+void ff_prores_idct_12(int16_t *block, const int16_t *qmat)
+{
+    int i;
+
+    for (i = 0; i < 64; i++)
+        block[i] *= qmat[i];
+
+    for (i = 0; i < 8; i++)
+        idctRowCondDC_int16_12bit(block + i*8, 0);
+
+    for (i = 0; i < 8; i++) {
+        block[i] += 8192;
+        idctSparseCol_int16_12bit(block + i);
+    }
+}
diff --git a/libavcodec/simple_idct.h b/libavcodec/simple_idct.h
index 634a78e59e..20578b3347 100644
--- a/libavcodec/simple_idct.h
+++ b/libavcodec/simple_idct.h
@@ -53,6 +53,7 @@ void ff_simple_idct_int16_12bit(int16_t *block);
  * for larger scale of input coefficients.
  */
 void ff_prores_idct_10(int16_t *block, const int16_t *qmat);
+void ff_prores_idct_12(int16_t *block, const int16_t *qmat);
 
 void ff_simple_idct248_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block);