[FFmpeg-cvslog] avcodec/nvdec_mpeg4: fix order of quant matrix coefficients

Hendrik Leppkes git at videolan.org
Mon May 15 19:49:23 EEST 2023


ffmpeg | branch: release/4.3 | Hendrik Leppkes <h.leppkes at gmail.com> | Mon May 15 13:33:18 2023 +0200| [410ccbbe8c6171b0c751d8d05f1edf999916b700] | committer: Hendrik Leppkes

avcodec/nvdec_mpeg4: fix order of quant matrix coefficients

The matrix coefficients are stored permutated for the IDCT,
rather then in plain raster order, and need to be un-permutated
for the hardware.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=410ccbbe8c6171b0c751d8d05f1edf999916b700
---

 libavcodec/nvdec_mpeg4.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/libavcodec/nvdec_mpeg4.c b/libavcodec/nvdec_mpeg4.c
index 739b049933..5404e4e330 100644
--- a/libavcodec/nvdec_mpeg4.c
+++ b/libavcodec/nvdec_mpeg4.c
@@ -86,8 +86,9 @@ static int nvdec_mpeg4_start_frame(AVCodecContext *avctx, const uint8_t *buffer,
     };
 
     for (i = 0; i < 64; ++i) {
-        ppc->QuantMatrixIntra[i] = s->intra_matrix[i];
-        ppc->QuantMatrixInter[i] = s->inter_matrix[i];
+        int n = s->idsp.idct_permutation[i];
+        ppc->QuantMatrixIntra[i] = s->intra_matrix[n];
+        ppc->QuantMatrixInter[i] = s->inter_matrix[n];
     }
 
     // We need to pass the full frame buffer and not just the slice



More information about the ffmpeg-cvslog mailing list