From 138fe832152d4b30528933ca681b39831dd94569 Mon Sep 17 00:00:00 2001 From: Mike Melanson Date: Tue, 22 Sep 2009 15:58:08 +0000 Subject: [PATCH] Perform the DC prediction reversal immediately after decoding all of the DC coefficients. This has a greater probability of leveraging the coefficients while they are still cached. When testing with the Big Buck Bunny 1080p video, I consistently saw improvements of 500k-600k dezicycles per run (through reverse_dc_prediction()) thanks to this move. Originally committed as revision 19966 to svn://svn.ffmpeg.org/ffmpeg/trunk --- libavcodec/vp3.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) diff --git a/libavcodec/vp3.c b/libavcodec/vp3.c index 0f4f6dafb3..e2b17062dd 100644 --- a/libavcodec/vp3.c +++ b/libavcodec/vp3.c @@ -1099,6 +1099,10 @@ static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb, return eob_run; } +static void reverse_dc_prediction(Vp3DecodeContext *s, + int first_fragment, + int fragment_width, + int fragment_height); /* * This function unpacks all of the DCT coefficient data from the * bitstream. @@ -1120,10 +1124,22 @@ static int unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb) residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_y_table], 0, s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run); + /* reverse prediction of the Y-plane DC coefficients */ + reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height); + /* unpack the C plane DC coefficients */ residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_c_table], 0, s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run); + /* reverse prediction of the C-plane DC coefficients */ + if (!(s->avctx->flags & CODEC_FLAG_GRAY)) + { + reverse_dc_prediction(s, s->fragment_start[1], + s->fragment_width / 2, s->fragment_height / 2); + reverse_dc_prediction(s, s->fragment_start[2], + s->fragment_width / 2, s->fragment_height / 2); + } + /* fetch the AC table indexes */ ac_y_table = get_bits(gb, 4); ac_c_table = get_bits(gb, 4); @@ -1996,14 +2012,6 @@ static int vp3_decode_frame(AVCodecContext *avctx, return -1; } - reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height); - if ((avctx->flags & CODEC_FLAG_GRAY) == 0) { - reverse_dc_prediction(s, s->fragment_start[1], - s->fragment_width / 2, s->fragment_height / 2); - reverse_dc_prediction(s, s->fragment_start[2], - s->fragment_width / 2, s->fragment_height / 2); - } - for (i = 0; i < s->macroblock_height; i++) render_slice(s, i); -- GitLab