On 10/9/17, Michael Niedermayer <mich...@niedermayer.cc> wrote:
> On Mon, Oct 09, 2017 at 09:02:38AM -0400, Ronald S. Bultje wrote:
>> Hi,
>>
>> On Mon, Oct 9, 2017 at 6:46 AM, Ivan Kalvachev <ikalvac...@gmail.com>
>> wrote:
>>
>> > On 10/9/17, Ronald S. Bultje <rsbul...@gmail.com> wrote:
>> > > On Sun, Oct 8, 2017 at 6:52 PM, Ivan Kalvachev <ikalvac...@gmail.com>
>> > wrote:
>> > > [..]
>> > >
>> > > Indentation is off in the second hunk, can you fix that?
>> >
>> > You want it 4 spaces to the right
Done.
>>
>> Yes, please.
>>
>> BTW, I think it would be better to use "127" number.
>> >
>>
>> I don't really mind either way. The number 128 suggests it may have been
>> intended as a bitmask. Michael is probably better positioned to comment
>> on
>> this.
>
> I don't really remember but i think 128 was chosen for ABI
> compatibility with additions to it from libav. So it should no longer
> matter what values are used on additions
Then I'm using the next free number "24".
Please, commit when you think it is appropriate.
Best Regards
Ivan Kalvachev
From 8842a69091b5eb5cf9b704b3ff504d21db4aad9b Mon Sep 17 00:00:00 2001
From: Ivan Kalvachev <ikalvac...@gmail.com>
Date: Mon, 9 Oct 2017 01:25:00 +0300
Subject: [PATCH] Fix visual glitch with XvMC, caused by wrong idct
permutation.
In the past XvMC forced simple_idct since
it was using FF_IDCT_PERM_NONE.
However now we have SIMD variants of simple_idct that
are using FF_IDCT_PERM_TRANSPOSE and if they are selected
XvMC would get coefficients in the wrong order.
The patch creates new FF_IDCT_NONE that
is used only for this kind of hardware decoding
and that fallbacks to the old C only simple idct.
Signed-off-by: Ivan Kalvachev <ikalvac...@gmail.com>
---
libavcodec/avcodec.h | 1 +
libavcodec/idctdsp.c | 1 +
libavcodec/mpeg12dec.c | 2 +-
3 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
index 52cc5b0ca..18c3e3ea1 100644
--- a/libavcodec/avcodec.h
+++ b/libavcodec/avcodec.h
@@ -3146,6 +3146,7 @@ typedef struct AVCodecContext {
#if FF_API_ARCH_ALPHA
#define FF_IDCT_SIMPLEALPHA 23
#endif
+#define FF_IDCT_NONE 24 /* Used by XvMC to extract IDCT coefficients with FF_IDCT_PERM_NONE */
#define FF_IDCT_SIMPLEAUTO 128
/**
diff --git a/libavcodec/idctdsp.c b/libavcodec/idctdsp.c
index d596aed1a..0122d29ef 100644
--- a/libavcodec/idctdsp.c
+++ b/libavcodec/idctdsp.c
@@ -279,6 +279,7 @@ av_cold void ff_idctdsp_init(IDCTDSPContext *c, AVCodecContext *avctx)
c->perm_type = FF_IDCT_PERM_NONE;
#endif /* CONFIG_FAANIDCT */
} else { // accurate/default
+ /* Be sure FF_IDCT_NONE will select this one, since it uses FF_IDCT_PERM_NONE */
c->idct_put = ff_simple_idct_put_8;
c->idct_add = ff_simple_idct_add_8;
c->idct = ff_simple_idct_8;
diff --git a/libavcodec/mpeg12dec.c b/libavcodec/mpeg12dec.c
index 22c29c150..4e68be27f 100644
--- a/libavcodec/mpeg12dec.c
+++ b/libavcodec/mpeg12dec.c
@@ -1217,7 +1217,7 @@ static void setup_hwaccel_for_pixfmt(AVCodecContext *avctx)
#endif
)
if (avctx->idct_algo == FF_IDCT_AUTO)
- avctx->idct_algo = FF_IDCT_SIMPLE;
+ avctx->idct_algo = FF_IDCT_NONE;
if (avctx->hwaccel && avctx->pix_fmt == AV_PIX_FMT_XVMC) {
Mpeg1Context *s1 = avctx->priv_data;
--
2.14.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
