Re: [FFmpeg-devel] [PATCH] lavc/pngdec: actually use SIMD version for bpp<=2

[Michael Niedermayer]

On Wed, May 12, 2021 at 07:07:27AM +0200, Nicolas Frattaroli wrote:
> The SIMD version was only used by the bpp > 2 case. This patch
> uses the SIMD version for both code paths.
> 
> Change has been tested to still work with a 1-bpp image that uses
> paeth, to generate one:

change breaks fate
--- ./tests/ref/fate/lscr   2021-05-07 19:20:39.256076206 +0200
+++ tests/data/fate/lscr2021-05-12 15:43:28.128017958 +0200
@@ -4,18 +4,18 @@
 #dimensions 0: 320x240
 #sar 0: 0/1
 0,  0,  0,1,   230400, 0x07c93bd9
-0,  1,  1,1,   230400, 0xdcacc274
-0,  2,  2,1,   230400, 0x6115dc34
-0,  3,  3,1,   230400, 0x7317f35c
-0,  4,  4,1,   230400, 0xca18d9a6
-0,  5,  5,1,   230400, 0x0c104d49
-0,  6,  6,1,   230400, 0xc4c415cf
-0,  7,  7,1,   230400, 0xd1e4149f
-0,  8,  8,1,   230400, 0x5973ae25
-0,  9,  9,1,   230400, 0xd77893e6
-0, 10, 10,1,   230400, 0x7f9cac97
-0, 11, 11,1,   230400, 0xc8f81940
-0, 12, 12,1,   230400, 0xa2b6f5d4
-0, 13, 13,1,   230400, 0x620c3222
-0, 14, 14,1,   230400, 0x60d097a0
-0, 15, 15,1,   230400, 0x47201c65
+0,  1,  1,1,   230400, 0x5af5c26e
+0,  2,  2,1,   230400, 0x115fdbbf
+0,  3,  3,1,   230400, 0x5f290ac6
+0,  4,  4,1,   230400, 0x93030c28
+0,  5,  5,1,   230400, 0x37a8c4f0
+0,  6,  6,1,   230400, 0xbaff9eb5
+0,  7,  7,1,   230400, 0x89c4c6e9
+0,  8,  8,1,   230400, 0x48dd646a
+0,  9,  9,1,   230400, 0x31c944ac
+0, 10, 10,1,   230400, 0x88ad9fd3
+0, 11, 11,1,   230400, 0x0b1fb7b1
+0, 12, 12,1,   230400, 0xe6abfaf6
+0, 13, 13,1,   230400, 0xea3c3e53
+0, 14, 14,1,   230400, 0xe8e443d3
+0, 15, 15,1,   230400, 0x4c069712
Test lscr failed. Look at tests/data/fate/lscr.err for details.
tests/Makefile:254: recipe for target 'fate-lscr' failed
make: *** [fate-lscr] Error 1


[...]
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Everything should be made as simple as possible, but not simpler.
-- Albert Einstein


signature.asc
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

[FFmpeg-devel] [PATCH] lavc/pngdec: actually use SIMD version for bpp<=2

[Nicolas Frattaroli]

The SIMD version was only used by the bpp > 2 case. This patch
uses the SIMD version for both code paths.

Change has been tested to still work with a 1-bpp image that uses
paeth, to generate one:

pngnq-s9 input.png
optipng -f4 -force input-nq8.png
---
 libavcodec/pngdec.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libavcodec/pngdec.c b/libavcodec/pngdec.c
index f2e6f689b0..2cb7946b49 100644
--- a/libavcodec/pngdec.c
+++ b/libavcodec/pngdec.c
@@ -311,7 +311,7 @@ void ff_png_filter_row(PNGDSPContext *dsp, uint8_t *dst, 
int filter_type,
 i = w;
 }
 }
-ff_add_png_paeth_prediction(dst + i, src + i, last + i, size - i, bpp);
+dsp->add_paeth_prediction(dst + i, src + i, last + i, size - i, bpp);
 break;
 }
 }
-- 
2.31.1

___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".