On 11/06/2025 03:58, averne wrote:
This fixes a few leaks reported by LeakSanitizer.
I tested this on my Intel laptop running anv, with
these patches no more leaks are detected.
Signed-off-by: averne
---
averne (3):
vulkan: fix leak in FFVkExecPool
vulkan/ffv1dec: fix leak in FFVu
If its a problem, you'll likely want to set it to 1 than more fine-grained
control, which you can already do via the API.
---
libavutil/hwcontext_vulkan.c | 21 +
1 file changed, 17 insertions(+), 4 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_v
On NVIDIA, there's a global maximum limit of approximately 112 queues,
which means it takes ONLY 7 total programs using the maximum amount of
queues to cause the driver to error out/*segfault* during initialization.
Also, each queue takes about 30ms to allocate, which quickly adds up.
This reduce
We don't use them, and on NVIDIA, each queue takes around 30ms
to allocate, and the driver has a global limit of ONLY 112 queues.
---
libavutil/hwcontext_vulkan.c | 8
1 file changed, 8 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index ce485a85a2
On NVIDIA, there's a global maximum limit of approximately 112 queues,
which means it takes ONLY 7 total programs using the maximum amount of
queues to cause the driver to error out/*segfault* during initialization.
Also, each queue takes about 30ms to allocate, which quickly adds up.
This reduce
We don't use them, and on NVIDIA, each queue takes around 30ms
to allocate, and the driver has a global limit of ONLY 112 queues.
---
libavutil/hwcontext_vulkan.c | 8
1 file changed, 8 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index aa0bfa3756
You don't get to say "oh, there were no more objections, so it was fine"
or "the use is evident" after the mess that your original patchsets were.
You're still not using the native time fields, nor the packet buffers,
which is a very hard NAK from me.
On 03/06/2025 23:20, softworkz . wrote:
---
libavutil/hwcontext_vulkan.c | 84 +++-
libavutil/vulkan_functions.h | 6 +++
libavutil/vulkan_loader.h| 1 +
3 files changed, 89 insertions(+), 2 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index ce485a85a2..b22e
On 23/05/2025 15:50, Kieran Kunhya via ffmpeg-devel wrote:
On Fri, 23 May 2025, 04:44 Lynne, wrote:
On 23/05/2025 08:42, Kieran Kunhya via ffmpeg-devel wrote:
Hello,
I wanted to put on the record that adding RaptorQ to FFmpeg isn't
maintenance of FFmpeg.
It isn't -- it's
On 23/05/2025 08:55, Devin Heitmueller wrote:
On Thu, May 22, 2025 at 7:42 PM Kieran Kunhya via ffmpeg-devel
wrote:
I wanted to put on the record that adding RaptorQ to FFmpeg isn't
maintenance of FFmpeg.
It's adding an obscure FEC protocol to FFmpeg, which is not going to be
implemented well
On 23/05/2025 08:42, Kieran Kunhya via ffmpeg-devel wrote:
Hello,
I wanted to put on the record that adding RaptorQ to FFmpeg isn't
maintenance of FFmpeg.
It isn't -- it's research.
It's adding an obscure FEC protocol to FFmpeg, which is not going to be
implemented well without an event loop
The issue is that state->cur[] is 8-bits, but a+b+1 can overflow
before being clipped to 0xF in the following line, causing an incorrect
state to be saved for the next symbol.
This solves numerous bitstream desyncs, particularly when coefficients
with magnitude greater than 127 are sent.
---
liba
On 18/05/2025 21:11, Niklas Haas wrote:
From: Niklas Haas
Carbon copy of vf_scdet.
Signed-off-by: Niklas Haas
Sponsored-by: nxtedition
---
configure | 1 +
libavfilter/Makefile | 1 +
libavfilter/allfilters.c | 1 +
libavfilter/vf_scdet_vulkan.c |
On 18/05/2025 21:11, Niklas Haas wrote:
From: Niklas Haas
Like vf_blackdetect but better, faster, stronger, harder.
Signed-off-by: Niklas Haas
Sponsored-by: nxtedition
---
configure | 1 +
doc/filters.texi| 2 +-
libavfilter/Makefile
On 18/05/2025 21:11, Niklas Haas wrote:
From: Niklas Haas
Implied internally now when needed.
---
libavfilter/vf_gblur_vulkan.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/libavfilter/vf_gblur_vulkan.c b/libavfilter/vf_gblur_vulkan.c
index 80b66de735..fb676a7fc9 100644
--- a/libavfilt
On 18/05/2025 21:11, Niklas Haas wrote:
From: Niklas Haas
We require this internally when using descriptor buffers, so it makes sense
to enable it internally, also.
---
libavutil/vulkan.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/libavutil/vulkan.c b/libavutil/vulkan.c
index 5f2
On 22/05/2025 06:21, sohzm wrote:
Fix inconsistent sample format reporting between probing and decoding.
Previously, avformat_find_stream_info reported fltp format for MP3
streams but frames were decoded as s16p.
Fixes ticket/11561
---
libavcodec/allcodecs.c | 2 +-
1 file changed, 1 insertio
LGTM
On 21/05/2025 23:12, Niklas Haas wrote:
From: Niklas Haas
ol->frame_rate is 0/0, so we need to calcalute the correct value based on
the il->frame_rate instead. Also adjust the time base, PTS and frame_duration
values accordingly. (Logic taken from vf_tinterlace.c)
---
libavfilter/vf_int
On 16/05/2025 02:39, James Almer wrote:
On 5/15/2025 9:36 PM, softworkz . wrote:
-Original Message-
From: ffmpeg-devel On Behalf Of
softworkz .
Sent: Freitag, 16. Mai 2025 02:33
To: FFmpeg development discussions and patches
Subject: Re: [FFmpeg-devel] [FFmpeg-cvslog] fftools/grap
---
libavutil/hwcontext_vulkan.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index 9f9df91e5d..4f205137eb 100644
--- a/libavutil/hwcontext_vulkan.c
+++ b/libavutil/hwcontext_vulkan.c
@@ -2643,11 +2643,12 @@ st
By pure coincidence, BUFFER and IMAGE flags were equal for those
two usage types.
---
libavutil/hwcontext_vulkan.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index eded36bc01..9f9df91e5d 100644
--- a/libavutil
Yet another thing that should've been always present.
---
libavutil/hwcontext_vulkan.c | 5 +
libavutil/vulkan_functions.h | 1 +
libavutil/vulkan_loader.h| 1 +
3 files changed, 7 insertions(+)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index 978d7e29d3..ede
---
libavcodec/vulkan/ffv1_dec.comp | 4 ++
libavcodec/vulkan/ffv1_dec_setup.comp | 4 +-
libavcodec/vulkan_decode.c| 1 +
libavcodec/vulkan_decode.h| 1 +
libavcodec/vulkan_ffv1.c | 60 +++
5 files changed, 52 insertions(+), 1
This simply makes a get_rac/put_rac_internal variant that can be
reused.
---
libavcodec/vulkan/rangecoder.comp | 57 +--
1 file changed, 17 insertions(+), 40 deletions(-)
diff --git a/libavcodec/vulkan/rangecoder.comp
b/libavcodec/vulkan/rangecoder.comp
index 9e2c5fbe
ice(inout SliceContext sc, const uint slice_idx)
{
/* Set coordinates */
uvec2 img_size = imageSize(src[0]);
@@ -37,11 +37,13 @@ void init_slice(out SliceContext sc, const uint slice_idx)
sc.slice_pos = ivec2(sxs, sys);
sc.slice_dim = ivec2(sxe - sxs, sye - sys);
-sc.sli
This gives a 35% speedup on AMD and 50% on Nvidia.
---
libavcodec/ffv1enc_vulkan.c | 6 ++-
libavcodec/vulkan/ffv1_enc.comp | 68 ++---
2 files changed, 50 insertions(+), 24 deletions(-)
diff --git a/libavcodec/ffv1enc_vulkan.c b/libavcodec/ffv1enc_vulkan.c
index
writeout_rgb requires that all subgroups are active.
---
libavcodec/vulkan/ffv1_dec.comp | 7 +--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index c74af4bf6a..e73b3f1dc0 100644
--- a/libavcodec/vulkan/ffv1_dec
We used to create our own buffer, but still used the DR1 flag,
which is not how it's supposed to work.
Instead, use ff_get_encode_buffer, and either host-map the buffer
before copying each slice via GPU transfers, or just copy each
slice manually if that fails or is unavailable.
---
libavcodec/ff
---
libavcodec/vulkan/ffv1_common.comp | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_common.comp
b/libavcodec/vulkan/ffv1_common.comp
index 1f222bdc42..3d40592739 100644
--- a/libavcodec/vulkan/ffv1_common.comp
+++ b/libavcodec/vulkan/ffv1_common
---
libavcodec/ffv1enc_vulkan.c| 379 +
libavcodec/vulkan/ffv1_common.comp | 87 +++
libavcodec/vulkan/ffv1_dec.comp| 91 +--
libavcodec/vulkan/ffv1_enc.comp| 155 ++--
libavcodec/vulkan_ffv1.c | 5 +-
5 files changed, 288 i
---
libavcodec/vulkan/rangecoder.comp | 19 ++-
1 file changed, 6 insertions(+), 13 deletions(-)
diff --git a/libavcodec/vulkan/rangecoder.comp
b/libavcodec/vulkan/rangecoder.comp
index badc65293f..9e2c5fbecf 100644
--- a/libavcodec/vulkan/rangecoder.comp
+++ b/libavcodec/vulkan/
---
libavcodec/vulkan/ffv1_dec.comp | 32 ++---
libavcodec/vulkan/ffv1_enc.comp | 85 -
2 files changed, 68 insertions(+), 49 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index fc0175c715..1c313b3168 100644
---
---
libavcodec/ffv1enc_vulkan.c | 1 +
libavcodec/vulkan/ffv1_enc.comp | 7 ---
libavcodec/vulkan/ffv1_enc_setup.comp | 10 +-
libavcodec/vulkan/rangecoder.comp | 23 +++
4 files changed, 17 insertions(+), 24 deletions(-)
diff --git a/libavcod
const uint slice_idx = gl_WorkGroupID.y*gl_NumWorkGroups.x +
gl_WorkGroupID.x;
encode_slice(slice_ctx[slice_idx], slice_idx);
+finalize_slice(slice_ctx[slice_idx], slice_idx);
}
diff --git a/libavcodec/vulkan/ffv1_enc_ac.comp
b/libavcodec/vulkan/ffv1_enc_ac.comp
deleted file mode 10
---
libavcodec/ffv1enc_vulkan.c | 21 -
libavcodec/vulkan/ffv1_enc_setup.comp | 65 +++
libavcodec/vulkan/rangecoder.comp | 28 +++-
3 files changed, 42 insertions(+), 72 deletions(-)
diff --git a/libavcodec/ffv1enc_vulkan.c b/libavcodec/ffv1e
On 08/05/2025 14:26, softworkz . wrote:
-Original Message-
From: ffmpeg-devel On Behalf Of Lynne
Sent: Donnerstag, 8. Mai 2025 14:00
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 0/5] doc/developer: Add subsection about
patch submission via FFstaging
On 08/05/2025
On 08/05/2025 13:33, softworkz . wrote:
-Original Message-
From: ffmpeg-devel On Behalf Of Lynne
Sent: Donnerstag, 8. Mai 2025 13:20
To: ffmpeg-devel@ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 0/5] doc/developer: Add subsection about
patch submission via FFstaging
On 08/05/2025
On 08/05/2025 12:22, ffmpegagent wrote:
Signed-off-by: softworkz softwo...@hotmail.com
softworkz (5):
doc/developer: Move checklist into Submitting Patches chapter
doc/developer: Move codecs/formats checklist into Development Policy
chapter
doc/developer: Reorder Submission procedu
On 06/05/2025 01:37, Andreas Rheinhardt wrote:
Patches attached.
- Andreas
Vulkan patches LGTM.
OpenPGP_0xA2FEA5F03F034464.asc
Description: OpenPGP public key
OpenPGP_signature.asc
Description: OpenPGP digital signature
___
ffmpeg-devel mailing l
On 05/05/2025 21:37, Link Mauve wrote:
In the call to vkGetPhysicalDeviceImageFormatProperties2(), we were
previously requesting the properties of the first fallback format (e.g.
VK_FORMAT_R8_UNORM for VK_FORMAT_G8_B8R8_2PLANE_420_UNORM) instead of
the actual format in use.
We don’t do anything
On 05/05/2025 16:57, Andreas Rheinhardt wrote:
Lynne:
On 05/05/2025 15:52, Andreas Rheinhardt wrote:
Lynne:
---
libavcodec/aac/aacdec_tab.c | 54 ++
libavcodec/aactab.c | 1820 +++
libavcodec/aactab.h | 20 +
3 files changed, 1894
On 05/05/2025 15:52, Andreas Rheinhardt wrote:
Lynne:
---
libavcodec/aac/aacdec_tab.c | 54 ++
libavcodec/aactab.c | 1820 +++
libavcodec/aactab.h | 20 +
3 files changed, 1894 insertions(+)
1. This should only be applied if it is used
---
libavcodec/aac/aacdec_tab.c | 54 ++
libavcodec/aactab.c | 1820 +++
libavcodec/aactab.h | 20 +
3 files changed, 1894 insertions(+)
diff --git a/libavcodec/aac/aacdec_tab.c b/libavcodec/aac/aacdec_tab.c
index 45a84a9a72..5ba20c0d8a 100644
-
On 03/05/2025 17:01, Russell Greene wrote:
vkGetPhysicalDeviceToolPropertiesEXT(dev, nb, array)
If pToolProperties is NULL, then the number of tools currently active on
physicalDevice is returned in pToolCount. Otherwise, pToolCount must
point to a variable set by the application to the number of
On 03/05/2025 04:39, Russell Greene wrote:
Is this documented anywhere? Should I maybe have a fixed sized stack
buffer and fallback to an allocation if it's just a rule of thumb that
it'll be less than some known small number. I just somehow doubt this
is in the vulkan spec or any sort of guarant
On 01/05/2025 07:05, Russell Greene wrote:
From: Russell Greene
Previously, it was assumed that `drmFormatModifierPlaneCount` was one
for all modifiers when exporting, which is not always the case, in
particular for AMD GPUs and maybe others.
Fetch the number of memory planes and fill the stru
On 30/04/2025 10:11, Dawid Kozinski wrote:
- Added APV decoder wrapper
- Changes in project configuration file and libavcodec Makefile
- Added documentation for APV decoder wrapper
Signed-off-by: Dawid Kozinski
---
configure | 1 +
doc/decoders.texi | 27 ++
libavcodec/
This adds support for NV12 and P010.
More will be added later.
---
libavcodec/ffv1enc.c| 2 ++
libavcodec/ffv1enc_vulkan.c | 8 ++--
libavcodec/vulkan/ffv1_enc.comp | 21 ++---
3 files changed, 22 insertions(+), 9 deletions(-)
diff --git a/libavcodec/ffv1enc.
Support was partially added previously in vulkan.c, but now it's fully
supported.
---
libavutil/hwcontext_vulkan.c | 1 +
libavutil/vulkan.c | 6 --
2 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index ade023
It gets parsed after SBR, even if there is no SBR.
---
libavcodec/aac/aacdec_usac.c | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/libavcodec/aac/aacdec_usac.c b/libavcodec/aac/aacdec_usac.c
index ccdf58bc8e..ef0c115aa0 100644
--- a/libavcodec/aac/aacdec_usac.c
+++ b/l
On 26/04/2025 05:11, Michael Niedermayer wrote:
On Tue, Apr 22, 2025 at 04:44:07PM -0500, Romain Beauxis wrote:
---
tests/Makefile | 4 +
tests/api/Makefile | 2 +-
tests/api/api-dump-stream-meta-test.c | 177 +
This patch adds support for accepting yuv420p10 in encoders and cuda.
---
libavcodec/nvenc.c | 2 ++
libavutil/hwcontext_cuda.c | 7 +--
2 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/libavcodec/nvenc.c b/libavcodec/nvenc.c
index 41a4dc55f4..ccf8cc6d39 100644
--- a/liba
The field represents the maximum Vulkan version that an application
supports. We do not have any code which require us to stay at 1.3.
This enables validation against 1.4, and enables all 1.4 features.
---
libavutil/hwcontext_vulkan.c | 7 ++-
1 file changed, 6 insertions(+), 1 deletion(-)
di
This allows for it to be shared.
Technically, implementations should not give drivers structs
that the drivers are not familiar with.
---
libavutil/hwcontext_vulkan.c | 63 +-
libavutil/vulkan.c | 75 ++--
libavutil/vulkan.h
Just correctness. We don't exceed this on any known hardware, but
its better to check.
If we do, we simply fall back to regular descriptors.
---
libavutil/vulkan.c | 15 ++-
libavutil/vulkan.h | 1 +
2 files changed, 11 insertions(+), 5 deletions(-)
diff --git a/libavutil/vulkan.c b/
Allows for it to be reused.
---
libavutil/hwcontext_vulkan.c | 42 ++--
libavutil/vulkan.c | 34 +
libavutil/vulkan.h | 6 ++
3 files changed, 46 insertions(+), 36 deletions(-)
diff --git a/libavutil/hwcontext_vu
On 19/04/2025 03:01, Michael Niedermayer wrote:
On Wed, Apr 16, 2025 at 04:09:13PM +0200, Lynne wrote:
This lets us test features that were broken earlier, as well as
test the hardware decoder by using the HWACCEL=vulkan option.
---
tests/Makefile| 1 +
tests/fate/ffv1
This implements support for reading and writing storage images with
no format.
The issue is that we define our images as arrays, and arrays can
only have a single type, which means that f.ex. NV12 needs two
different images, R8 and RG8.
Not using any formats fixes this.
---
libavutil/vulkan.c | 2
On 18/04/2025 01:55, IndecisiveTurtle wrote:
From: IndecisiveTurtle
Performance wise, encoding a 1080p 1-minute video is performed in about 2.5
minutes with the cpu encoder running on my Ryzen 5 4600H, while it takes about
30 seconds on my NVIDIA GTX 1650
Haar shader has a subgroup optimized
This lets us test features that were broken earlier, as well as
test the hardware decoder by using the HWACCEL=vulkan option.
---
tests/Makefile| 1 +
tests/fate/ffv1.mak | 46 +++
tests/ref/fate/ffv1-s1024-bgra| 4 +++
tests/ref/
We violated the spec, which, despite the actual command buffer pool
*not* being involved in any functions which require external synchronization
of the pool, *require* external synchronization even if only the
command buffers are used.
This also has the effect of *significantly* speeding up execut
On 16/04/2025 12:22, Andreas Rheinhardt wrote:
Lynne:
This lets us test features that were broken earlier, as well as
test the hardware decoder by using the HWACCEL=vulkan option.
Can't you just adjust some of vsynth FFV1 tests to use this many slices?
(Can one even use 1024 slices for s
This lets us test features that were broken earlier, as well as
test the hardware decoder by using the HWACCEL=vulkan option.
---
tests/Makefile | 1 +
tests/fate/ffv1.mak | 53 +
tests/ref/fate/ffv1-s1024-bgra | 4 +++
tests/
On 16/04/2025 08:12, Shiqi Zhu wrote:
h264bsd url https://github.com/oneam/h264bsd
./configure --enable-decoder="h264_bsd" --extra-cflags="-I/h264bsd/src"
--extra-ldflags="-L/h264bsd/posix/lib" --extra-libs="-lh264bsd"
Test
./ffmpeg -y -codec:v h264_bsd -i /h264bsd/test/test_640x360.h264
./test
---
libavutil/hwcontext_vulkan.c | 8
libavutil/vulkan_loader.h| 2 ++
2 files changed, 10 insertions(+)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index 8823af0175..f878dfec76 100644
--- a/libavutil/hwcontext_vulkan.c
+++ b/libavutil/hwcontext_vulkan.c
On 15/04/2025 12:04, Lynne wrote:
Most of the work was done by Paul B Mahol.
I cleaned the patch up, improved the code a bit, and added slice
threading.
---
configure | 1 +
libavcodec/Makefile| 1 +
libavcodec/allcodecs.c | 1 +
libavcodec/braw.c | 666
ndex 00..2246b2d645
--- /dev/null
+++ b/libavcodec/braw.c
@@ -0,0 +1,666 @@
+/*
+ * Blackmagic RAW codec
+ *
+ * Copyright (c) 2019 Paul B Mahol
+ * Copyright (c) 2025 Lynne
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under
This adds codec entries for the Blackmagic RAW codec.
---
libavcodec/codec_desc.c | 7 +++
libavcodec/codec_id.h | 1 +
2 files changed, 8 insertions(+)
diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c
index 9fb190e35a..90e31a8b6a 100644
--- a/libavcodec/codec_desc.c
+++ b/lib
---
libavformat/isom_tags.c | 7 +++
libavformat/mov.c | 2 +-
2 files changed, 8 insertions(+), 1 deletion(-)
diff --git a/libavformat/isom_tags.c b/libavformat/isom_tags.c
index f05762beec..7bf2bf7d5b 100644
--- a/libavformat/isom_tags.c
+++ b/libavformat/isom_tags.c
@@ -286,6 +286,13
This commit adds a Vulkan AV1 encoder, using the native acceleration API.
---
configure |1 +
libavcodec/Makefile|2 +
libavcodec/allcodecs.c |1 +
libavcodec/vulkan_encode_av1.c | 1385
libavcodec/vulkan_video.c
This commit adds support for compiler hints.
While on AMD these are not used/needed, Nvidia benefits from them, and gives
a sizeable 10% speedup on 4k.
---
libavcodec/vulkan/ffv1_dec.comp | 16
libavcodec/vulkan/rangecoder.comp | 12 ++--
libavutil/hwcontext_vulkan.c
This commit also makes it possible for the encoder to choose a different
quantization table on a per-slice basis, as well as adding this capability
to the decoder.
Also, this commit fully fixes decoding of context=1 encoded files.
---
libavcodec/ffv1_vulkan.h | 2 +-
libavcodec/ffv1
Temporary workaround. Will be replaced with a version check once a fix is
in the works and a known next version for Mesa with a fix is known.
---
libavutil/hwcontext_vulkan.c | 5 +
1 file changed, 5 insertions(+)
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index
Speeds up everything on AMD by 3x.
This uses 32 local invocations to load state into cache, as well
as to do the RCT faster.
---
libavcodec/vulkan/ffv1_dec.comp | 71 -
libavcodec/vulkan_ffv1.c| 7 +++-
2 files changed, 50 insertions(+), 28 deletions(-)
di
Slightly faster, and allows for further optimizations.
---
libavcodec/vulkan/ffv1_dec.comp | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index 4cc3b9987f..fd9b98023c 100644
--- a/libavcodec/vulkan/ffv1_dec.
This saves on some VRAM, but mainly allows for a more unified path.
---
libavcodec/vulkan/ffv1_dec_setup.comp | 55 ++-
libavcodec/vulkan/rangecoder.comp | 17 +
libavcodec/vulkan_ffv1.c | 23 +--
3 files changed, 46 insertions(+), 49 deleti
s, run_index);
+decode_line(sc, sp, w, y, p, bits,
+slice_state_off[p], run_index);
+
+writeout_rgb(sc, sp, w, y, true);
}
#endif
}
diff --git a/libavcodec/vulkan/ffv1_dec_rct.comp
b/libavcodec/vulkan/ffv1_dec_rct.comp
deleted
8% speedup on nvidia on 4k.
---
libavcodec/vulkan/ffv1_dec.comp | 3 +--
libavcodec/vulkan_ffv1.c| 6 ++
2 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index a6272d4832..4cc3b9987f 100644
--- a/libavcodec
Typo.
Not enough to fix context=1, but its a start.
---
libavcodec/vulkan/ffv1_dec.comp | 14 --
1 file changed, 8 insertions(+), 6 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index 7d3150ed63..1954c050f8 100644
--- a/libavcodec/vulkan/f
This is likely a nanooptimization, but its more correct.
---
libavcodec/vulkan_ffv1.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libavcodec/vulkan_ffv1.c
index d90db291aa..e511840a01 100644
--- a/libavcodec/vulkan_ffv1.c
+++ b/libavcodec/vul
Without a barrier upfront, the reset shader may read data fields not
yet set by the setup shader.
---
libavcodec/vulkan_ffv1.c | 36 +---
1 file changed, 17 insertions(+), 19 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libavcodec/vulkan_ffv1.c
index ccff92
The commit which added support for host mapping accidentally broke the
original, upload route.
For drivers without host-mapping (very few), fix it.
---
libavcodec/vulkan_ffv1.c | 19 +++
1 file changed, 7 insertions(+), 12 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libav
Rather than always using the maximum allowed slices, just use the number
of slices present in this frame.
---
libavcodec/vulkan_ffv1.c | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libavcodec/vulkan_ffv1.c
index b6c9320ec2..9747721f0d 100644
GPUs have cmovs as standard.
---
libavcodec/vulkan/rangecoder.comp | 9 -
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/libavcodec/vulkan/rangecoder.comp
b/libavcodec/vulkan/rangecoder.comp
index 4272b2a42f..ba8a6cfd9d 100644
--- a/libavcodec/vulkan/rangecoder.comp
+++ b/l
This was the fastest variant tested.
---
libavcodec/vulkan/ffv1_dec.comp | 12 +---
1 file changed, 5 insertions(+), 7 deletions(-)
diff --git a/libavcodec/vulkan/ffv1_dec.comp b/libavcodec/vulkan/ffv1_dec.comp
index f9ffe1cee1..7d3150ed63 100644
--- a/libavcodec/vulkan/ffv1_dec.comp
+++
Forgotten.
---
libavcodec/vulkan/rangecoder.comp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libavcodec/vulkan/rangecoder.comp
b/libavcodec/vulkan/rangecoder.comp
index ba8a6cfd9d..e332bce8a5 100644
--- a/libavcodec/vulkan/rangecoder.comp
+++ b/libavcodec/vulkan/rangec
Fixed by previous commit.
---
libavcodec/vulkan_ffv1.c | 14 --
1 file changed, 14 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libavcodec/vulkan_ffv1.c
index 17bfc943d4..1156d6749b 100644
--- a/libavcodec/vulkan_ffv1.c
+++ b/libavcodec/vulkan_ffv1.c
@@ -1142,20 +1142,6 @@ sta
Leftover debug macro.
---
libavcodec/vulkan_ffv1.c | 2 --
1 file changed, 2 deletions(-)
diff --git a/libavcodec/vulkan_ffv1.c b/libavcodec/vulkan_ffv1.c
index 1156d6749b..b6c9320ec2 100644
--- a/libavcodec/vulkan_ffv1.c
+++ b/libavcodec/vulkan_ffv1.c
@@ -41,8 +41,6 @@ const FFVulkanDecodeDescri
On 22/03/2025 18:49, Massimo Eynard wrote:
This patch adds support for decoding the fourth MLP substream
which contains the 16-channel presentation used for Atmos
audio objects.
By default only the first three substreams are decoded
unless the new extract_objects flag is enabled as the resulting
On 04/04/2025 19:20, Niklas Haas wrote:
Hi all,
My current workstation (first generation Zen) is getting rather old (8 years).
On top of being slow and power hungry, it is also not very representative
anymore of modern hardware.
Additionally, I would like to try and incorporate AVX-512 instruct
On 02/04/2025 16:23, Andreas Rheinhardt wrote:
Andreas Rheinhardt:
Patch attached.
- Andreas
Will apply this patch tomorrow unless there are objections.
LGTM
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo
---
libavcodec/vulkan_decode.c | 38 ++
libavcodec/vulkan_decode.h | 6 ++
2 files changed, 40 insertions(+), 4 deletions(-)
diff --git a/libavcodec/vulkan_decode.c b/libavcodec/vulkan_decode.c
index 058efe3037..893f8fca3d 100644
--- a/libavcodec/vulkan_de
---
libavcodec/vulkan_hevc.c | 291 ---
1 file changed, 150 insertions(+), 141 deletions(-)
diff --git a/libavcodec/vulkan_hevc.c b/libavcodec/vulkan_hevc.c
index e83f8346a0..baf8b0c3fd 100644
--- a/libavcodec/vulkan_hevc.c
+++ b/libavcodec/vulkan_hevc.c
@@ -11
This commit moves all temporary structs used for decoding out
of the common picture header and onto stack.
---
libavcodec/vulkan_decode.c | 1 +
libavcodec/vulkan_decode.h | 3 +
libavcodec/vulkan_h264.c | 322 +++--
3 files changed, 168 insertions(+), 158 del
Required for next refactor.
---
libavcodec/vulkan_av1.c| 30 +-
libavcodec/vulkan_decode.c | 20 ++--
libavcodec/vulkan_decode.h | 13 +
libavcodec/vulkan_h264.c | 32 ++--
libavcodec/vulkan_hevc.c | 28 +++
This allows for the upcoming refactor.
---
libavcodec/vulkan_av1.c| 6 +++---
libavcodec/vulkan_decode.c | 23 ++-
libavcodec/vulkan_decode.h | 5 +++--
libavcodec/vulkan_ffv1.c | 22 --
libavcodec/vulkan_h264.c | 9 +++--
libavcodec/vulkan_he
We tried to create sequence params in both start_frame and end_frame.
This was redundant.
Just always create them in end_frame.
---
libavcodec/vulkan_av1.c | 8
libavcodec/vulkan_h264.c | 9 -
libavcodec/vulkan_hevc.c | 9 -
3 files changed, 26 deletions(-)
diff --git a
On 20/03/2025 23:30, Michael Niedermayer wrote:
This performs about as good as the non LRU system for 16bit and
better than then the LRU system for 16 converted to 32. So
its basically performing best in all cases we have atm making
the LRU system unneeded.
Test on *real* 32-bit content, please
On 18/03/2025 16:25, Andreas Rheinhardt wrote:
Patch attached.
- Andreas
LGTM
Thanks
OpenPGP_0xA2FEA5F03F034464.asc
Description: OpenPGP public key
OpenPGP_signature.asc
Description: OpenPGP digital signature
___
ffmpeg-devel mailing list
ffmpeg-
On 15/03/2025 20:10, James Almer wrote:
Will prevent decoding frame data during probing.
Signed-off-by: James Almer
---
libavcodec/ffv1dec.c | 4
1 file changed, 4 insertions(+)
diff --git a/libavcodec/ffv1dec.c b/libavcodec/ffv1dec.c
index b731f11297..998b981a33 100644
--- a/libavcode
1 - 100 of 1412 matches
Mail list logo
