Merge pull request #4228 from afedchin/dxva_h264_intel

[DXVA] Fix h264 deconding with refs > 11 on recent Intel GPUs (SNB/IVB)

4 files changed, 147 insertions, 4 deletions

diff --git a/lib/ffmpeg/libavcodec/dxva2.h b/lib/ffmpeg/libavcodec/dxva2.h
index ac39e06917..2639d89d90 100644
--- a/lib/ffmpeg/libavcodec/dxva2.h
+++ b/lib/ffmpeg/libavcodec/dxva2.h
@@ -49,6 +49,7 @@
  */
 
 #define FF_DXVA2_WORKAROUND_SCALING_LIST_ZIGZAG 1 ///< Work around for DXVA2 and old UVD/UVD+ ATI video cards
+#define FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO    2 ///< Work around for DXVA2 and old Intel GPUs with ClearVideo interface
 
 /**
  * This structure is used to provides the necessary configurations and data
diff --git a/lib/ffmpeg/libavcodec/dxva2_h264.c b/lib/ffmpeg/libavcodec/dxva2_h264.c
index 6ae5cd919f..cbb05e2dd0 100644
--- a/lib/ffmpeg/libavcodec/dxva2_h264.c
+++ b/lib/ffmpeg/libavcodec/dxva2_h264.c
@@ -115,6 +115,8 @@ static void fill_picture_parameters(struct dxva_context *ctx, const H264Context
     pp->bit_depth_chroma_minus8       = h->sps.bit_depth_chroma - 8;
     if (ctx->workaround & FF_DXVA2_WORKAROUND_SCALING_LIST_ZIGZAG)
         pp->Reserved16Bits            = 0;
+    else if (ctx->workaround & FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO)
+        pp->Reserved16Bits            = 0x34c;
     else
         pp->Reserved16Bits            = 3; /* FIXME is there a way to detect the right mode ? */
     pp->StatusReportFeedbackNumber    = 1 + ctx->report_id++;
@@ -194,8 +196,18 @@ static void fill_slice_short(DXVA_Slice_H264_Short *slice,
     slice->wBadSliceChopping     = 0;
 }
 
+static int get_refpic_index(const DXVA_PicParams_H264 *pp, int surface_index)
+{
+    int i;
+    for (i = 0; i < FF_ARRAY_ELEMS(pp->RefFrameList); i++) {
+        if ((pp->RefFrameList[i].bPicEntry & 0x7f) == surface_index)
+          return i;
+    }
+    return 0x7f;
+}
+
 static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
-                            unsigned position, unsigned size)
+                            const DXVA_PicParams_H264 *pp, unsigned position, unsigned size)
 {
     const H264Context *h = avctx->priv_data;
     struct dxva_context *ctx = avctx->hwaccel_context;
@@ -228,8 +240,12 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
             if (list < h->list_count && i < h->ref_count[list]) {
                 const Picture *r = &h->ref_list[list][i];
                 unsigned plane;
-                fill_picture_entry(&slice->RefPicList[list][i],
-                                   ff_dxva2_get_surface_index(ctx, r),
+                unsigned index;
+                if (ctx->workaround & FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO)
+                    index = ff_dxva2_get_surface_index(ctx, r);
+                else
+                    index = get_refpic_index(pp, ff_dxva2_get_surface_index(ctx, r));
+                fill_picture_entry(&slice->RefPicList[list][i], index,
                                    r->f.reference == PICT_BOTTOM_FIELD);
                 for (plane = 0; plane < 3; plane++) {
                     int w, o;
@@ -414,7 +430,7 @@ static int dxva2_h264_decode_slice(AVCodecContext *avctx,
                          position, size);
     else
         fill_slice_long(avctx, &ctx_pic->slice_long[ctx_pic->slice_count],
-                        position, size);
+                        &ctx_pic->pp, position, size);
     ctx_pic->slice_count++;
 
     if (h->slice_type != AV_PICTURE_TYPE_I && h->slice_type != AV_PICTURE_TYPE_SI)
diff --git a/lib/ffmpeg/patches/0069-dxva2_h264-set-the-correct-ref-frame-index-in-the-long-slice-struct.patch b/lib/ffmpeg/patches/0069-dxva2_h264-set-the-correct-ref-frame-index-in-the-long-slice-struct.patch
new file mode 100644
index 0000000000..94f12d881e
--- /dev/null
+++ b/lib/ffmpeg/patches/0069-dxva2_h264-set-the-correct-ref-frame-index-in-the-long-slice-struct.patch
@@ -0,0 +1,66 @@
+From patchwork Sat Mar  2 11:15:29 2013
+Content-Type: text/plain; charset="utf-8"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Subject: [1/2] dxva2_h264: set the correct ref frame index in the long slice
+ struct
+Date: Sat, 02 Mar 2013 10:15:29 -0000
+From: Hendrik Leppkes <h.leppkes@gmail.com>
+X-Patchwork-Id: 35320
+Message-Id: <1362222930-4764-2-git-send-email-h.leppkes@gmail.com>
+To: libav-devel@libav.org
+
+The latest H.264 DXVA specification states that the index in this
+structure should refer to a valid entry in the RefFrameList of the picture
+parameter structure, and not to the actual surface index.
+
+Fixes H.264 DXVA2 decoding on recent Intel GPUs (tested on Sandy and Ivy)
+
+---
+libavcodec/dxva2_h264.c | 18 ++++++++++++++----
+ 1 file changed, 14 insertions(+), 4 deletions(-)
+
+diff --git a/libavcodec/dxva2_h264.c b/libavcodec/dxva2_h264.c
+index 7009d6c..9dcb005 100644
+--- a/libavcodec/dxva2_h264.c
++++ b/libavcodec/dxva2_h264.c
+@@ -194,8 +194,18 @@ static void fill_slice_short(DXVA_Slice_H264_Short *slice,
+     slice->wBadSliceChopping     = 0;
+ }
+ 
++static int get_refpic_index(const DXVA_PicParams_H264 *pp, int surface_index)
++{
++    int i;
++    for (i = 0; i < FF_ARRAY_ELEMS(pp->RefFrameList); i++) {
++        if ((pp->RefFrameList[i].bPicEntry & 0x7f) == surface_index)
++          return i;
++    }
++    return 0x7f;
++}
++
+ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
+-                            unsigned position, unsigned size)
++                            const DXVA_PicParams_H264 *pp, unsigned position, unsigned size)
+ {
+     const H264Context *h = avctx->priv_data;
+     struct dxva_context *ctx = avctx->hwaccel_context;
+@@ -228,8 +238,8 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
+             if (list < h->list_count && i < h->ref_count[list]) {
+                 const Picture *r = &h->ref_list[list][i];
+                 unsigned plane;
+-                fill_picture_entry(&slice->RefPicList[list][i],
+-                                   ff_dxva2_get_surface_index(ctx, r),
++                unsigned index = get_refpic_index(pp, ff_dxva2_get_surface_index(ctx, r));
++                fill_picture_entry(&slice->RefPicList[list][i], index,
+                                    r->f.reference == PICT_BOTTOM_FIELD);
+                 for (plane = 0; plane < 3; plane++) {
+                     int w, o;
+@@ -413,7 +423,7 @@ static int decode_slice(AVCodecContext *avctx,
+                          position, size);
+     else
+         fill_slice_long(avctx, &ctx_pic->slice_long[ctx_pic->slice_count],
+-                        position, size);
++                        &ctx_pic->pp, position, size);
+     ctx_pic->slice_count++;
+ 
+     if (h->slice_type != AV_PICTURE_TYPE_I && h->slice_type != AV_PICTURE_TYPE_SI)
diff --git a/lib/ffmpeg/patches/0070-dxva2_h264-add-a-workaround-for-old-intel-GPUs.patch b/lib/ffmpeg/patches/0070-dxva2_h264-add-a-workaround-for-old-intel-GPUs.patch
new file mode 100644
index 0000000000..d87b19f433
--- /dev/null
+++ b/lib/ffmpeg/patches/0070-dxva2_h264-add-a-workaround-for-old-intel-GPUs.patch
@@ -0,0 +1,60 @@
+From patchwork Sat Mar  2 11:15:30 2013
+Content-Type: text/plain; charset="utf-8"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Subject: [2/2] dxva2_h264: add a workaround for old intel GPUs
+Date: Sat, 02 Mar 2013 10:15:30 -0000
+From: Hendrik Leppkes <h.leppkes@gmail.com>
+X-Patchwork-Id: 35321
+Message-Id: <1362222930-4764-3-git-send-email-h.leppkes@gmail.com>
+To: libav-devel@libav.org
+
+Old Intel GPUs expect the reference frame index to the actual surface,
+instead of the index into RefFrameList as specified by the spec.
+
+This workaround should be set when using one of the "ClearVideo" decoder
+devices.
+
+---
+libavcodec/dxva2.h      | 1 +
+ libavcodec/dxva2_h264.c | 8 +++++++-
+ 2 files changed, 8 insertions(+), 1 deletion(-)
+
+diff --git a/libavcodec/dxva2.h b/libavcodec/dxva2.h
+index d161eb9..78939be 100644
+--- a/libavcodec/dxva2.h
++++ b/libavcodec/dxva2.h
+@@ -42,6 +42,7 @@
+  */
+ 
+ #define FF_DXVA2_WORKAROUND_SCALING_LIST_ZIGZAG 1 ///< Work around for DXVA2 and old UVD/UVD+ ATI video cards
++#define FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO    2 ///< Work around for DXVA2 and old Intel GPUs with ClearVideo interface
+ 
+ /**
+  * This structure is used to provides the necessary configurations and data
+diff --git a/libavcodec/dxva2_h264.c b/libavcodec/dxva2_h264.c
+index 9dcb005..4b3f83c 100644
+--- a/libavcodec/dxva2_h264.c
++++ b/libavcodec/dxva2_h264.c
+@@ -115,6 +115,8 @@ static void fill_picture_parameters(struct dxva_context *ctx, const H264Context
+     pp->bit_depth_chroma_minus8       = h->sps.bit_depth_chroma - 8;
+     if (ctx->workaround & FF_DXVA2_WORKAROUND_SCALING_LIST_ZIGZAG)
+         pp->Reserved16Bits            = 0;
++    else if (ctx->workaround & FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO)
++        pp->Reserved16Bits            = 0x34c;
+     else
+         pp->Reserved16Bits            = 3; /* FIXME is there a way to detect the right mode ? */
+     pp->StatusReportFeedbackNumber    = 1 + ctx->report_id++;
+@@ -238,7 +240,11 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice,
+             if (list < h->list_count && i < h->ref_count[list]) {
+                 const Picture *r = &h->ref_list[list][i];
+                 unsigned plane;
+-                unsigned index = get_refpic_index(pp, ff_dxva2_get_surface_index(ctx, r));
++                unsigned index;
++                if (ctx->workaround & FF_DXVA2_WORKAROUND_INTEL_CLEARVIDEO)
++                    index = ff_dxva2_get_surface_index(ctx, r);
++                else
++                    index = get_refpic_index(pp, ff_dxva2_get_surface_index(ctx, r));
+                 fill_picture_entry(&slice->RefPicList[list][i], index,
+                                    r->f.reference == PICT_BOTTOM_FIELD);
+                 for (plane = 0; plane < 3; plane++) {