The merge list only needs to be derived up to the merge index.
---
libavcodec/hevc_mvs.c | 35 ++-
1 file changed, 26 insertions(+), 9 deletions(-)
diff --git a/libavcodec/hevc_mvs.c b/libavcodec/hevc_mvs.c
index 2b017dd..0302c88 100644
--- a/libavcodec/hevc_mvs.c
When the candidate has been found, no need to derive others.
---
libavcodec/hevc_mvs.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/libavcodec/hevc_mvs.c b/libavcodec/hevc_mvs.c
index 0302c88..952a431 100644
--- a/libavcodec/hevc_mvs.c
+++ b/libavcodec/hevc_mvs.c
@@ -737,6 +737,11 @@
---
libavcodec/hevc.c | 4 ++--
libavcodec/hevc_cabac.c | 4 ++--
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index 2b37f53..4d3513a 100644
--- a/libavcodec/hevc.c
+++ b/libavcodec/hevc.c
@@ -100,7 +100,7 @@ static int
Save only 2 CTB lines worth of data (i.e. at worst 16 lines).
---
libavcodec/hevc.c | 4 ++--
libavcodec/hevc.h | 2 +-
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index f8a495f..bed2f5f 100644
--- a/libavcodec/hevc.c
+++
When dealing with MVs, both components may be processed at a time.
---
libavcodec/hevc.c | 3 +--
libavcodec/hevc_mvs.c | 23 +--
2 files changed, 10 insertions(+), 16 deletions(-)
diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index 4d3513a..f2bbe91 100644
---
Note: this is also used for skip blocks, so the speed improvement of
this patch should be around 1 or 2%. (amvp probably much less)
2014-07-10 8:47 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
The merge list only needs to be derived up to the merge index.
---
libavcodec
Hi,
I had missed some potential consequences of these macros, so here's
something to document these consequences.
I'm still a bit unsatisfied, as I would think AV_(ZERO|COPY|SWAP)64
are not that much affected by unaligned addresses or, if they are
aligned to 4, compared to 2 AV_*32 operations.
a patch where I align various things to 4 to allow using
AV_ZERO32/AV_RN32A.
--
Christophe
From 86fcce0a680e799eff3df86eb28ee77a88736811 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sat, 12 Jul 2014 16:51:09 +0200
Subject: [PATCH 2/4] hevc: use intreadwrite
Hi,
2014-07-19 14:17 GMT+02:00 Michael Niedermayer michae...@gmx.at:
did you intend to attach a patch ?
Here you are.
--
Christophe
From 30f5f833280c0753a000ab05fd60b9d0a9685dcb Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Thu, 10 Jul 2014 20:49:06
Hi,
2014-07-15 10:10 GMT+02:00 Clément Bœsch u...@pkh.me:
The new website can be previewed @ http://db0.galo.pe/ffmpeg-web/
Some comments:
- The FFmpeg project is proud a brand new version - lacks a to launch/... ?
- The whole section before news does not fit in my screen height and
is even
: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sun, 20 Jul 2014 09:52:50 +0200
Subject: [PATCH 2/2] hevc: use different thread wait/report for tmvp
The tmvp only needs decoded MVs, not the whole filtered pixels.
Therefore abuse field 1 to report parsing process, and use it when
checking progress
Hi,
2014-07-20 13:35 GMT+02:00 Ronald S. Bultje rsbul...@gmail.com:
Do you get a speedup? I tinkered with some related problem in vp9 once and
I never got a significant speedup out of it...
Whatever your definition of significant, I don't think the figure I have is.
It's around 1% on a dual
Use new API without introducing changes to synchronization points,
i.e. same behaviour.
---
libavcodec/hevc.c| 19 ++-
libavcodec/hevc_filter.c | 10 +-
libavcodec/hevc_mvs.c| 4 ++--
libavcodec/hevc_refs.c | 2 +-
4 files changed, 18 insertions(+), 17
In that case, there's an offset of 4 between the CTB line and the
raster line.
---
libavcodec/hevc_filter.c | 8 ++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/libavcodec/hevc_filter.c b/libavcodec/hevc_filter.c
index abf8b24..8b5ae6b 100644
--- a/libavcodec/hevc_filter.c
slower in ffmpeg or if there's something else
occurring.
--
Christophe
From 36dbc9b67269579e23345ec225ffa270d472b94e Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Thu, 24 Jul 2014 17:23:47 +0200
Subject: [PATCH 10/10] x86: hevc_mc: replace simple leas by adds
Hi,
I missed that occurrence in the previous patch.
--
Christophe
From 673b18b88e80650bfc7f898f2ac000147d2d71eb Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Fri, 25 Jul 2014 19:07:27 +0200
Subject: [PATCH 11/13] x86: hevc_mc: replace one lea by add
2014-07-27 10:20 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
This is already used for qpel/luma filters.
And with patch.
From 4c9ef00c805b5ea81d96955bdfbc57ab7b383934 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Fri, 25 Jul 2014 15:08:49
The stride is sometimes way bigger than actually needed. ptrdiff_t is
the actual type used. Aligning to 8 is needed as the lines must have
aligned addresses for SIMD.
--
Christophe
From 126adf820bc54c2d00f794629595ad6310fbfc37 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq
One may interleave even more vertical and horizontal passes, but this
is less straightforward (more code changes than just moves).
--
Christophe
From b48828b956d3a183d358ef9ad8a12f354c87da7d Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sat, 26 Jul 2014 21
Hi,
2014-07-27 19:18 GMT+02:00 Michael Niedermayer michae...@gmx.at:
It is unconditionally set to 64, which is quite higher than the
actual block size.
is this faster?
No, which surprises me, but so be it.
On the other hand, if this stride is PB_MAX_SIZE throughout the code,
then maybe its
---
libavcodec/hevc.h| 1 -
libavcodec/hevcdsp.h | 2 ++
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/libavcodec/hevc.h b/libavcodec/hevc.h
index 8420f38..0369d8f 100644
--- a/libavcodec/hevc.h
+++ b/libavcodec/hevc.h
@@ -56,7 +56,6 @@
#define MAX_TRANSFORM_DEPTH 5
---
libavcodec/hevc.c | 8 +++
libavcodec/hevcdsp.h | 8 +++
libavcodec/hevcdsp_template.c | 56 +--
libavcodec/x86/hevc_mc.asm| 18 +++---
libavcodec/x86/hevcdsp.h | 6 ++---
libavcodec/x86/hevcdsp_init.c |
The stride for some buffers is known.
---
libavcodec/x86/hevc_mc.asm | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavcodec/x86/hevc_mc.asm b/libavcodec/x86/hevc_mc.asm
index ea4447b..2c34927 100644
--- a/libavcodec/x86/hevc_mc.asm
+++ b/libavcodec/x86/hevc_mc.asm
@@
2014-07-23 21:20 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Didn't refresh the patch with update documentation. Here it is.
It didn't contain dummy functions in libavcodec/utils.c.
The attached patch should fix failures to build with pthread disabled.
--
Christophe
From
With patch.
From 0b5d56a563df7efdfcf206013680510c62e2569c Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Mon, 28 Jul 2014 11:48:21 +0200
Subject: [PATCH 10/15] x86: hevc_mc: fix register count usage
A macro was using a fixed register, causing too many GPRs
Hi,
note: I'm using step throughout the patch because of the step
function and what the causal part most often looks like. I have no
idea for another less confusing wording.
2014-07-28 23:15 GMT+02:00 Michael Niedermayer michae...@gmx.at:
maybe i misunderstand but the progress[1] y check looks
Hi,
2014-08-02 14:48 GMT+02:00 Michael Niedermayer michae...@gmx.at:
is this for apply/push or just RFC/WIP ?
in-between. I had expected Mickael Raulet to comment if he was seeing
something not compatible with this. I think the bipred code is a bit
more mature since Ronald comments (iirc), so
for the others, so I didn't bother
resending them/starting another thread.
--
Christophe
From 8b13e4350c6662ca4bd2bcab443a1e62f7751b30 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Mon, 28 Jul 2014 08:55:26 +0200
Subject: [PATCH 1/5] x86: hevc_mc: assume 2nd
, and I can't really
spend more time on it.
--
Christophe
From 57819727586c186bfea733a8f06eead22ac6a1f2 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Wed, 23 Jul 2014 23:21:20 +0200
Subject: [PATCH 08/13] x86: hevc_deblock: remove unnecessary masking
2014-07-28 19:17 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
---
libavcodec/hevc.c | 8 +++
libavcodec/hevcdsp.h | 8 +++
libavcodec/hevcdsp_template.c | 56
+--
libavcodec/x86/hevc_mc.asm| 18
Hi,
this patch is mostly cosmetical. I don't like seeing arrays passed to
dsp functions being of a type whose length may not be fixed, though
it's a small matter here.
--
Christophe
From 25a3d9fa51860b8df108f31a818e6f74ee1761da Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq
---
libavcodec/x86/constants.c | 2 --
libavcodec/x86/constants.h | 3 +--
libavcodec/x86/dwt_yasm.asm | 3 ++-
libavcodec/x86/inline_asm.h | 2 +-
4 files changed, 4 insertions(+), 6 deletions(-)
diff --git a/libavcodec/x86/constants.c b/libavcodec/x86/constants.c
index ee3f1a5..bfb0ff3
---
libavcodec/x86/constants.c | 5 +
libavcodec/x86/constants.h | 5 +
libavcodec/x86/h264_intrapred_10bit.asm | 2 +-
libavcodec/x86/hevc_deblock.asm | 2 +-
libavcodec/x86/hpeldsp.asm | 3 ++-
libavcodec/x86/vp8dsp.asm | 2 +-
---
libavcodec/x86/constants.c | 2 ++
libavcodec/x86/constants.h | 2 ++
libavcodec/x86/fft.asm | 7 ---
libavcodec/x86/sbrdsp.asm | 2 +-
4 files changed, 9 insertions(+), 4 deletions(-)
diff --git a/libavcodec/x86/constants.c b/libavcodec/x86/constants.c
index bfb0ff3..5d2c237 100644
---
libavcodec/x86/diracdsp_yasm.asm | 11 ++-
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/libavcodec/x86/diracdsp_yasm.asm b/libavcodec/x86/diracdsp_yasm.asm
index 3e9765b..d3cf9f1 100644
--- a/libavcodec/x86/diracdsp_yasm.asm
+++ b/libavcodec/x86/diracdsp_yasm.asm
@@
---
libavcodec/x86/constants.h | 1 +
libavcodec/x86/dwt_yasm.asm | 6 +++---
2 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/libavcodec/x86/constants.h b/libavcodec/x86/constants.h
index c85a54e..f73a9f2 100644
--- a/libavcodec/x86/constants.h
+++ b/libavcodec/x86/constants.h
@@
This fixes ticket #3839.
--
Christophe
From 3745a3b611159f6a373785b67cbc92d4b36af44b Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sun, 10 Aug 2014 11:43:12 +0200
Subject: [PATCH] hevc: fix incorrect sao buffer size
It previously used the output, cropped
:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sat, 9 Aug 2014 20:11:20 +0200
Subject: [PATCH] hevc_ps: verify P/T/L information
This makes the SPS parsing a little, but barely, safer.
---
libavcodec/hevc_ps.c | 43 ++-
1 file changed
Hi,
2014-08-10 11:59 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
This fixes ticket #3839.
By the way, not completely sure, but that is probably exploitable (I
am not a security expert):
- indicate large cropping in the header; this will cause an overrun of
probably (max_ctb_size
Hi,
2014-08-10 14:42 GMT+02:00 Ronald S. Bultje rsbul...@gmail.com:
Are we using the checked bitstream reader? If we are, we're fine already...
I think we are. On the other hand, it seems the top caller,
ff_hevc_decode_nal_vps, is never checking if we have read past the
bitstream end. Shouldn't
The patch fixes overreads (and crashes) introduced in 3ad04608.
--
Christophe
From adfac6472b015e162988e2e2208e95cc87248be2 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sun, 10 Aug 2014 15:02:36 +0200
Subject: [PATCH] hevc_mvs: set candidate availabilities
Hi,
the patch tries to validate these high-level syntax elements.
Unfortunately, it causes fuzzing to be less efficient, eg with the
sequence from ticket #3840 where no frame are decoded.
--
Christophe
From 6b60cf2968099fa4395e1e3120ab66d95d4c8709 Mon Sep 17 00:00:00 2001
From: Christophe
If the allocated size, despite best efforts, is too small, exit
with the appropriate error.
---
libavcodec/proresenc_kostya.c | 13 ++---
1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/libavcodec/proresenc_kostya.c b/libavcodec/proresenc_kostya.c
index 24cb333..a70ae3c
The buffer allocation may be incorrect (e.g. with an alpha plane),
and currently causes the buffer to be set to NULL by init_put_bits,
later on causing crashing.
So, detect that situation, and if detected, reallocate the buffer
and ask a sample if it happens.
Fixes ticket #2760
---
The packet buffer allocation considered as dct-coded, while it is
actually run-coded and thus requires a larger buffer.
---
libavcodec/proresenc_kostya.c | 13 ++---
1 file changed, 10 insertions(+), 3 deletions(-)
diff --git a/libavcodec/proresenc_kostya.c
---
libavcodec/proresenc_kostya.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/libavcodec/proresenc_kostya.c b/libavcodec/proresenc_kostya.c
index 93bcde7..24cb333 100644
--- a/libavcodec/proresenc_kostya.c
+++ b/libavcodec/proresenc_kostya.c
@@ -471,7 +471,6 @@ static
Hi,
2014-08-12 2:34 GMT+02:00 Michael Niedermayer michae...@gmx.at:
+if (pkt_size = buf - orig_buf) {
this isnt sufficient, there could be 1 byte space left, then the
reallocate wouldnt run and encode_slice() would run into the extra
padding and fail
Yeah, you're right. I
Hi,
2014-08-12 1:56 GMT+02:00 Michael Niedermayer michae...@gmx.at:
- (2 + 2 * ctx-num_planes +
+ (2 + 2 * FFMIN(3, ctx-num_planes) +
slice_hdr_size is 2 + 2 * (ctx-num_planes - 1);
is this difference intended ?
I don't
2014-08-12 7:56 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Yeah, you're right. I have no idea how big a slice can be, as that
seems the extra size check here. How about FF_MIN_BUFFER_SIZE ?
Then the growth would be FFMAX(FF_MIN_BUFFER_SIZE, buf - orig_buf) ?
Here's a patch
Hi,
2014-08-12 8:10 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Yeah, I'll drop it.
Done. Kostya suggested another rounding (the slice bitstreams are
byte-aligned), integrated in this patch.
--
Christophe
From 4e72ed146848b1d7b1804080f43b5d45484e3d94 Mon Sep 17 00:00:00 2001
Hi,
2014-08-12 10:19 GMT+02:00 Michael Niedermayer michae...@gmx.at:
the serious undersizing check already depends on the assumtation
that FF_MIN_BUFFER_SIZE is larger than a slice,
Yes, and here lies the issue: if we haven't been able to guess it
correctly previously, how likely are we to
Hi,
2014-08-12 12:26 GMT+02:00 Michael Niedermayer michae...@gmx.at:
On Tue, Aug 12, 2014 at 11:56:21AM +0200, Christophe Gisquet wrote:
Hi,
2014-08-12 10:19 GMT+02:00 Michael Niedermayer michae...@gmx.at:
the serious undersizing check already depends on the assumtation
Hi,
2014-08-12 12:38 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
I chose 2 times in the attached patch. I have no strong opinion on
which solution is best, though reallocating was a specific request I
got.
Forgot a parameter to the call to avpriv_request_sample, will be added
Hi,
the attached patch improves the decoding of ticket #3692 by following
S268M-2003.
Unfortunately, that probably means ffmpeg's dpx encoder is not
compliant to these specifications.
--
Christophe
From 4ba2ec15c9a111fb4e20523d5a8fa337bedee92d Mon Sep 17 00:00:00 2001
From: Christophe Gisquet
2014-08-13 9:48 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Unfortunately, that probably means ffmpeg's dpx encoder is not
compliant to these specifications.
And indeed the decoder is then no longer able to decode a file
produced by the encoder when number of bytes per line
SMPTE 268M-2003 specifies that each line starts at a 4-bytes boundary.
Therefore, modify correspondingly the input buffer strides and size.
Partially fixes ticket #3692: DLAD_8b_3c_big.dpx still has inverted
colors, which might be related to endianness.
---
libavcodec/dpx.c | 34
S268M-2003 specifies that each line start is aligned on a 4-byte boundary.
---
libavcodec/dpxenc.c | 44 +---
1 file changed, 37 insertions(+), 7 deletions(-)
diff --git a/libavcodec/dpxenc.c b/libavcodec/dpxenc.c
index 0eb1297..059d8c6 100644
---
Fixes DLAD_8b_3c_big.dpx from ticket #3692
---
libavcodec/dpx.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavcodec/dpx.c b/libavcodec/dpx.c
index 2ad7527..d4d6833 100644
--- a/libavcodec/dpx.c
+++ b/libavcodec/dpx.c
@@ -232,6 +232,8 @@ static int decode_frame(AVCodecContext *avctx,
---
libavcodec/dpx.c | 9 +
1 file changed, 9 insertions(+)
diff --git a/libavcodec/dpx.c b/libavcodec/dpx.c
index 8cd7d73..2ad7527 100644
--- a/libavcodec/dpx.c
+++ b/libavcodec/dpx.c
@@ -108,6 +108,15 @@ static int decode_frame(AVCodecContext *avctx,
av_log(avctx,
: Christophe Gisquet christophe.gisq...@gmail.com
Date: Wed, 13 Aug 2014 01:44:40 +0200
Subject: [PATCH 3/4] dpx: warn if encrypted
---
libavcodec/dpx.c | 10 ++
1 file changed, 10 insertions(+)
diff --git a/libavcodec/dpx.c b/libavcodec/dpx.c
index 8cd7d73..6c4951b 100644
--- a/libavcodec/dpx.c
Hi,
2014-08-14 4:30 GMT+02:00 Michael Niedermayer michae...@gmx.at:
probably ok
applied
Now that I think of it, this change causes older versions of ffmpeg to
be unable to decode the output of the fixed encoder. So:
- Should we bump eg a minor version number to let any decoder have the
Hi,
2014-08-14 4:17 GMT+02:00 Michael Niedermayer michae...@gmx.at:
see http://samples.ffmpeg.org/image-samples/dpx_samples.zip
Note that the date metadata in that file does not have a valid format either.
--
Christophe
___
ffmpeg-devel mailing list
Hi,
2014-08-14 5:01 GMT+02:00 Michael Niedermayer michae...@gmx.at:
On Wed, Aug 13, 2014 at 10:21:54AM +, Christophe Gisquet wrote:
case 50081:
+avctx-pix_fmt = AV_PIX_FMT_BGR24;
+break;
this possibly breaks decoding
Hi,
2014-08-12 12:44 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Forgot a parameter to the call to avpriv_request_sample, will be added later.
If there's no further comment on this option for the reallocation,
here's an updated patch.
--
Christophe
From
Hi,
2014-08-14 17:14 GMT+02:00 James Darnley james.darn...@gmail.com:
+ * This limit is used:
+ * - when CONFIG_SMALL is 0 to unroll a loop in the C template.
+ * - when SSE4 (or newer) is available on x86 to unroll a copy loop.
Does the implementer of a new dsp version really has
Hi,
2014-08-14 17:14 GMT+02:00 James Darnley james.darn...@gmail.com:
This should help to clarify the API.
OK.
--
Christophe
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Hi,
2014-08-14 17:15 GMT+02:00 James Darnley james.darn...@gmail.com:
The LPC encoder is not used at compression levels below 3 but the basic acodec
test uses -compression_level 2. This adds a test using -compression_level 5,
which is the current default.
What order is used in that case? It
Hi,
2014-08-14 18:08 GMT+02:00 James Darnley james.darn...@gmail.com:
Minimum 1, maximum 8. Should I find out what order(s) are actually chosen?
The encoder algorithm might change and make that information useless,
but that could be useful. I fear that only checking which orders it
*tests* is
Hi,
2014-08-14 11:48 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Hi,
2014-08-14 5:01 GMT+02:00 Michael Niedermayer michae...@gmx.at:
On Wed, Aug 13, 2014 at 10:21:54AM +, Christophe Gisquet wrote:
case 50081:
+avctx-pix_fmt = AV_PIX_FMT_BGR24;
+break
59ea1d72b27272d2a28c680fc87b353f1a88eb36 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Sun, 17 Aug 2014 09:47:46 +0200
Subject: [PATCH] gifdec: use truncated width for image manipulation
Some files seem to have an off-by-one error. In most cases, it appears to
be on the image
Le 13 août 2014 12:22, Christophe Gisquet christophe.gisq...@gmail.com
a écrit :
Fixes DLAD_8b_3c_big.dpx from ticket #3692
---
libavcodec/dpx.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavcodec/dpx.c b/libavcodec/dpx.c
index 2ad7527..d4d6833 100644
--- a/libavcodec/dpx.c
This change is almost cosmetical only, and reduces the changes needed to
fix the 24bps case.
---
libavcodec/alacenc.c | 19 ++-
1 file changed, 10 insertions(+), 9 deletions(-)
diff --git a/libavcodec/alacenc.c b/libavcodec/alacenc.c
index bc68a06..6345253 100644
---
The raw coded bits are extracted prior to decorrelation, as is correctly
performed by the decoder, and not after.
---
libavcodec/alacenc.c | 16 ++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/libavcodec/alacenc.c b/libavcodec/alacenc.c
index 6345253..b9ad899
the raw MSBs.
It would possible to detect bitstreams before this change and achieve
correct decoding if need be.
Christophe Gisquet (2):
alacenc: increase predictor buffer
alacenc: fix extra bits extraction
libavcodec/alacenc.c | 35 ---
1 file changed, 24
2014-08-17 19:09 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
It would possible to detect bitstreams before this change and achieve
correct decoding if need be.
fate tests still pass with the changes, which means the test sample
doesn't contain the corner case involved
Hi,
2014-08-17 20:39 GMT+02:00 Michael Niedermayer michae...@gmx.at:
+if (width s-screen_width) {
+av_log(s-avctx, AV_LOG_ERROR, Invalid image width.\n);
+return AVERROR_INVALIDDATA;
+}
+if (left + width s-screen_width) {
+/* width must be kept around
Hi,
2014-08-17 19:09 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
The raw coded bits are extracted prior to decorrelation, as is correctly
performed by the decoder, and not after.
Forgot to mention it fixes ticket #2768 (haven't checked 2497 yet), so
commit message amended
for missing the obvious, indeed in those cases pw/height would
end negative...
So I have added the checks you propose and decided to shuffle them
with the others.
Updated patch attached.
--
Christophe
From a8a329b19d72405b6cc7de173e3f160be02e212d Mon Sep 17 00:00:00 2001
From: Christophe
Hi,
2014-08-17 19:09 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
It would possible to detect bitstreams before this change and achieve
correct decoding if need be.
So after further exploration, you can't really detect this:
- there's no metadata for the alac bitstream
.
--
Christophe
From d9679c80df991e98b1f00fd15f9c1c3c689ec426 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Mon, 18 Aug 2014 11:27:50 +0200
Subject: [PATCH] proresenc_kostya: warn/reject on incorrect profile
This fixes the following situations:
- Reject
Hi,
2014-06-19 6:23 GMT+02:00 Michael Niedermayer michae...@gmx.at:
patch applied
Am I missing something or was it not applied?
--
Christophe
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Hi,
this is a proposal to allow decoding incorrect files. There is no
obvious and systematic way to detect them however.
--
Christophe
From feaacb09d8660b7e8784ad4a550dda455f172e24 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Mon, 18 Aug 2014 09:53:20
2014-08-18 15:59 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
this is a proposal to allow decoding incorrect files. There is no
obvious and systematic way to detect them however.
Ignore the version.h stray change in your review, I thought I had removed it.
--
Christophe
2014-08-18 19:26 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
Hi,
And with all patches.
--
Christophe
From 89b54fbd698eb9c68f0bf30105bcaf72c5575d27 Mon Sep 17 00:00:00 2001
From: Christophe Gisquet christophe.gisq...@gmail.com
Date: Mon, 18 Aug 2014 11:27:50 +0200
Subject: [PATCH
Hi,
2014-08-18 13:40 GMT+02:00 Carl Eugen Hoyos ceho...@ag.or.at:
Christophe Gisquet christophe.gisquet at gmail.com writes:
If a profile was set, the automatic setting should
probably not overwrite it.
(If this is possible.)
But I consider the usecase where a user wants to encode
alpha
Hi,
2014-08-18 19:22 GMT+02:00 Ronald S. Bultje rsbul...@gmail.com:
You mean the centerpoint chroma location (top-left, top-middle,
middle-middle, etc.) respective to the set of luma pixels?
I don't think it's only that, but it's probably only slightly longer
to implement.
For interlaced
Hi,
2014-08-18 21:49 GMT+02:00 Michael Niedermayer michae...@gmx.at:
+int max_slice_size = (ctx-frame_size_upper_bound - 200) /
(ctx-pictures_per_frame * ctx-slices_per_picture + 1);
Regarding the reallocation check:
pkt_size = buf - orig_buf + 2 * max_slice_size
For last slice, pkt_size
Hi,
2014-08-19 12:31 GMT+02:00 Michael Niedermayer michae...@gmx.at:
suggested commit message:
avcodec/proresenc_kostya: set initial max_slice_size based on
frame_size_upper_bound
If the initial max_slice_size is 0 then reallocation is disabled for the
first
slice.
OK.
--
bytestream2_* will not cause buffer overflow, but on the other hand,
it should be checked whether overflows have been prevented.
---
libavcodec/wavpackenc.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/libavcodec/wavpackenc.c b/libavcodec/wavpackenc.c
index 5b8973c..46c69a3 100644
---
The decoder is reporting CRC errors instead of the prime cause of
end-of-packet. This happens with current encoder when there are a few
samples and the output buffer is undersized because its size didn't
account for headers.
Christophe Gisquet (3):
wavpack: report if there is no bits left
The allocation didn't account for headers, that can be easily 79 bytes.
As a result, buffers allocated for a few samples (e.g. 5 in the original
bug) could be undersized.
Fixed ticket #2881.
---
libavcodec/wavpackenc.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git
Hi,
2014-08-19 16:20 GMT+02:00 Nicolas George geo...@nsup.org:
IMHO, the correct error depends on how sure you are that a buffer too small
SHOULD not happen.
If you are very sure, then av_assert0().
That would be it: I'm sure that, if the condition occurs and the
packet is written anyway,
Hi,
2014-08-20 4:55 GMT+02:00 James Almer jamr...@gmail.com:
~15% faster than sse2
[...]
@@ -509,7 +509,11 @@ void ff_hevc_dsp_init_x86(HEVCDSPContext *c, const int
bit_depth)
if (ARCH_X86_64) {
c-hevc_v_loop_filter_luma =
ff_hevc_v_loop_filter_luma_8_avx;
and bits_per_raw_sample==12 is e.g. not rescaled
to 16bits to match its rgb48 colorspace.
One may argue that a proper solution is to reduce as often as possible
references to bits_per_raw_sample and, instead, always use the
colorspace information and to introduce new colorspaces as needed.
Christophe
When input had 12 bits, it was invariably treated as packed RGB
with 12 bits per component.
---
libavcodec/dpxenc.c | 37 -
1 file changed, 36 insertions(+), 1 deletion(-)
diff --git a/libavcodec/dpxenc.c b/libavcodec/dpxenc.c
index aca745b..d02684b 100644
---
pnmdec currently respects the consign of rgb48 content. Instead, do not rescale
and make the bits_per_raw_sample valid and useful.
---
libavcodec/pnmdec.c | 6 --
1 file changed, 6 deletions(-)
diff --git a/libavcodec/pnmdec.c b/libavcodec/pnmdec.c
index c84b6eb..36bdfe7 100644
---
This allows writing actual bitdepth in RGB48 when it isn't actually 16.
---
libavcodec/pnmenc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/libavcodec/pnmenc.c b/libavcodec/pnmenc.c
index e6c3635..9198ddb 100644
--- a/libavcodec/pnmenc.c
+++ b/libavcodec/pnmenc.c
@@ -84,6 +84,8 @@
Thanks to av_get_pix_fmt_loss, we can determine more precisely if a
conversion will incur some kind of loss. If this loss doesn't modify
in particular bitdepth, the input bitdepth can be reused for the
output.
---
ffmpeg.c | 9 +
1 file changed, 9 insertions(+)
diff --git a/ffmpeg.c
Hi,
2014-08-20 10:10 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
+if (!(av_get_pix_fmt_loss(enc_ctx-pix_fmt,
dec_ctx-pix_fmt, 0)
+ (FF_LOSS_DEPTH|FF_LOSS_COLORSPACE|FF_LOSS_CHROMA))) {
If it was ever useful, this is probably wrong
Hi,
2014-08-20 10:10 GMT+02:00 Christophe Gisquet christophe.gisq...@gmail.com:
One may argue that a proper solution is to reduce as often as possible
references to bits_per_raw_sample and, instead, always use the
colorspace information and to introduce new colorspaces as needed.
I'd prefer
1 - 100 of 560 matches
Mail list logo