Re: [FFmpeg-devel] [RFC] What to do with 15k euro from STF?

On Mon, Nov 11, 2024 at 12:53 AM James Almer wrote: > > On 11/10/2024 8:44 PM, Michael Niedermayer wrote: > > Hi all > > > > as we will likely have 15k available from STF "2024" for another > > task/project > > what should it be used for ? > > > > I think the obvious options are: > > > > 1. gitla

[FFmpeg-devel] SIMD Motion Compensation Optimizations

Dear ffmpeg developers, I recently read a news article about impressive AVX-512 optimizations of Motion Compensation functions and I'd like to learn more. Unfortunately, the article merely linked the post on X, which included a screenshot and little else. https://x.com/FFmpeg/status/1852542388851

[FFmpeg-devel] [PATCH v2 1/6] hwcontext_vulkan: explicitly wait when uploading

--- libavutil/hwcontext_vulkan.c | 11 +-- 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c index 0b52ad5112..0c9047f4c6 100644 --- a/libavutil/hwcontext_vulkan.c +++ b/libavutil/hwcontext_vulkan.c @@ -4200,13 +4200,1

[FFmpeg-devel] [PATCH v2 2/6] .gitignore: add exclusions for shader .c files

--- .gitignore | 2 ++ 1 file changed, 2 insertions(+) diff --git a/.gitignore b/.gitignore index e810d11107..9cfc78b414 100644 --- a/.gitignore +++ b/.gitignore @@ -41,3 +41,5 @@ /src /mapfile /tools/python/__pycache__/ +/libavcodec/vulkan/*.c +/libavfilter/vulkan/*.c -- 2.45.2.753.g447d99e1

[FFmpeg-devel] [PATCH v2 6/6] ffv1enc: add a Vulkan encoder

This commit implements a standard, compliant, version 3 and version 4 FFv1 encoder, entirely in Vulkan. The encoder is written in standard GLSL and requires a Vulkan 1.3 supporting GPU with the BDA extension. The encoder can use any amount of slices, but nominally, should use 32x32 slices (1024 in

[FFmpeg-devel] [PATCH v2 3/6] ffv1enc: split off encoder initialization into a separate function

--- libavcodec/ffv1enc.c | 354 +++ libavcodec/ffv1enc.h | 30 2 files changed, 217 insertions(+), 167 deletions(-) create mode 100644 libavcodec/ffv1enc.h diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c index 7a6c718b41..ca2c9f32e2 100644 -

[FFmpeg-devel] [PATCH v2 4/6] ffv1enc: move plane info init into a separate function

--- libavcodec/ffv1enc.c | 40 libavcodec/ffv1enc.h | 2 ++ 2 files changed, 26 insertions(+), 16 deletions(-) diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c index ca2c9f32e2..a106e2130a 100644 --- a/libavcodec/ffv1enc.c +++ b/libavcodec/ffv1enc

[FFmpeg-devel] [PATCH v2 5/6] ffv1enc: move slice allocation out of generic encode init

--- libavcodec/ffv1enc.c | 36 ++-- 1 file changed, 18 insertions(+), 18 deletions(-) diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c index a106e2130a..7572594f3e 100644 --- a/libavcodec/ffv1enc.c +++ b/libavcodec/ffv1enc.c @@ -710,24 +710,6 @@ av_cold int

Re: [FFmpeg-devel] [PATCH] lavc/libx265: flag as experimental

Hi all, We would like to update that we have fixed the memory leaks and other issues reported in x265 and will be releasing v4.1 this week. *Thanks,* *Kirithika* On Fri, Oct 18, 2024 at 3:08 PM Kirithika Kalirathnam < kirith...@multicorewareinc.com> wrote: > Hi all , > > We understand your con

[FFmpeg-devel] [PATCH] avcodec/dxva2: add support for HEVC RExt DXVA profiles

Microsoft has formally standardized DXVA GUIDs for HEVC Range Extension profiles in the Windows 11 24H2 SDK. They are supported by Intel GPUs starting with Tiger Lake. Like VDPAU and VAAPI, DXVA has separate GUIDs for each RExt profile, so we must parse the SPS like those hwaccels do to figure out

Re: [FFmpeg-devel] [RFC] What to do with 15k euro from STF?

On Mon, 11 Nov 2024, Michael Niedermayer wrote: Hi all as we will likely have 15k available from STF "2024" for another task/project what should it be used for ? I think the obvious options are: 1. gitlab / forgeyo setup + hw + maintaince on our infrastructure (this was often requested

[FFmpeg-devel] [PATCH v5 07/12] swscale/internal: expose sws_init_single_context() internally

From: Niklas Haas Used by the graph API swscale wrapper, for now. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libswscale/swscale_internal.h | 3 +++ libswscale/utils.c| 4 ++-- 2 files changed, 5 insertions(+), 2 deletions(-) diff --git a/libswscale/swscale_in

[FFmpeg-devel] [PATCH v5 05/12] swscale: eliminate redundant SwsInternal accesses

From: Niklas Haas This is a purely cosmetic commit aimed at replacing accesses to SwsInternal.opts by direct access to SwsContext wherever convenient. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libswscale/options.c | 2 +- libswscale/swscale.c

[FFmpeg-devel] [PATCH v5 12/12] avfilter/vf_scale: switch to new swscale API

From: Niklas Haas Most logic from this filter has been co-opted into swscale itself, allowing the resulting filter to be substantially simpler as it no longer has to worry about context initialization, interlacing, etc. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libavfilt

[FFmpeg-devel] [PATCH v5 00/12] swscale: introduce new, dynamic scaling API

Changes since v4: - Rebased on top of current master - Fixed a segfault with cascaded scaling - Fixed garbled output with some threaded special converters - Added pixfmt to SwsImg to track cascaded context format changes ___ ffmpeg-devel mailing list ffm

[FFmpeg-devel] [PATCH v5 01/12] swscale/options: cosmetic changes

From: Niklas Haas Reorganize the list, fix whitespace, make indentation consistent, and rename some descriptions for clarity, consistency or informativeness. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libswscale/options.c | 86 ++--

[FFmpeg-devel] [PATCH v5 02/12] swscale/internal: use static_assert for enforcing offsets

From: Niklas Haas Instead of sprinkling av_assert0 into random init functions. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libswscale/swscale_internal.h | 11 +++ libswscale/utils.c| 2 -- libswscale/x86/swscale.c | 4 3 files changed, 11 in

[FFmpeg-devel] [PATCH v5 06/12] swscale: organize and better document flags

From: Niklas Haas Group them into an enum rather than random #defines, and document their behavior a bit more obviously. Of particular note, I discovered that SWS_DIRECT_BGR is not referenced anywhere else in the code base. As such, I have moved it to the deprecated section, alongside SWS_ERROR_

[FFmpeg-devel] [PATCH v5 08/12] swscale/graph: add new high-level scaler dispatch mechanism

From: Niklas Haas This interface has been designed from the ground up to serve as a new framework for dispatching various scaling operations at a high level. This will eventually replace the old ad-hoc system of using cascaded contexts, as well as allowing us to plug in more dynamic scaling passe

[FFmpeg-devel] [PATCH v5 04/12] swscale: expose SwsContext publicly

From: Niklas Haas Following in the footsteps of the work in the previous commit, it's now relatively straightforward to expose the options struct publicly as SwsContext. This is a step towards making this more user friendly, as well as following API conventions established elsewhere. Sponsored-b

[FFmpeg-devel] [PATCH v5 11/12] tests/swscale: add a benchmarking mode

From: Niklas Haas With the ability to set the thread count as well. This benchmark includes the constant overhead of context initialization. Sponsored-by: Sovereign Tech Fund Signed-off-by: Niklas Haas --- libswscale/tests/swscale.c | 93 -- 1 file changed,

Re: [FFmpeg-devel] GSoC Mentor Summit Reimbursement Request

On Wed, Nov 6, 2024 at 2:16 PM Thilo Borgmann via ffmpeg-devel wrote: > > Hi, > > I also request reimbursement for this years GSoC summit. > > My costs included: > > 903,91 EUR Flight (BER to SFO) > 248,71 EUR Hotel > - > 1152,62 EUR Total > ==

[FFmpeg-devel] [PATCH 1/3] lavc/vvc_mc: R-V V put_pixels

From: sunyuechi k230 banana_f3 put_chroma_pixels_8_4x4_c: 63.5 ( 1.00x)59.2 ( 1.00x) put_chroma_pixels_8_4x4_rvv_i32:26.5 ( 2.39x)28.0 ( 2.12x) put_chroma_pixels_8_8

[FFmpeg-devel] [PATCH 2/3] lavc/hevc: R-V V put_pixels(pow2)

From: sunyuechi k230 banana_f3 put_hevc_pel_pixels4_8_c: 61.6 ( 1.00x)69.5 ( 1.00x) put_hevc_pel_pixels4_8_rvv_i32: 24.6 ( 2.50x)28.0 ( 2.48x) put_hevc_pel_pixels8_8_c: 209.8 ( 1.00x)215.5 ( 1.00x)

Re: [FFmpeg-devel] [PATCH 3/5] lavc/vvc_mc: R-V V put_uni_pixels

> Is this going to be reused anywhere? it seems the macro is only used once atm. The next patch will use ([PATCH 4/5] lavc/hevc: R-V V pel_uni(pow2)) > Also is there a reason to use RVV here instead of just unaligned RVI? Yes, RVI is enough; I deleted it and resent it. Rémi Denis-Courmont 于202

[FFmpeg-devel] [PATCH 3/3] lavc/vvc_mc R-V V sad

From: sunyuechi k230 banana_f3 sad_8x16_c: 385.9 ( 1.00x)403.1 ( 1.00x) sad_8x16_rvv_i32:108.1 ( 3.57x)100.8 ( 4.00x) sad_16x8_c: 376.6 ( 1.00x)392.6 ( 1.00x) sad_16x8_rvv_i32: 89.3 ( 4

[FFmpeg-devel] [PATCH 4/4] fate: revert previous frequency adjustments of the sine filter

With more precise frequency support in the sine filter, several fate tests will change. Signed-off-by: Marton Balint --- tests/fate/ffmpeg.mak | 8 +- tests/fate/filter-video.mak | 2 +- tests/fate/libswresample.mak| 2 +- tests/filtergraphs/concat |

[FFmpeg-devel] [PATCH 1/4] fate: adjust frequencies of the sine filter

The filter currently use inaccurate frequencies, this is inpreparation for fixing that. Signed-off-by: Marton Balint --- tests/fate/ffmpeg.mak| 8 +- tests/fate/filter-video.mak | 2 +- tests/fate/libswresample.mak | 2 +- tests/filtergraphs/concat| 6 +- tes

[FFmpeg-devel] [PATCH 2/4] avfilter/asrc_sine: factorize sampling to a separate context

Signed-off-by: Marton Balint --- libavfilter/asrc_sine.c | 34 +++--- 1 file changed, 23 insertions(+), 11 deletions(-) diff --git a/libavfilter/asrc_sine.c b/libavfilter/asrc_sine.c index 96b603dd26..854fba6e37 100644 --- a/libavfilter/asrc_sine.c +++ b/libavfilter/a

[FFmpeg-devel] [PATCH 3/4] avfilter/asrc_sine: increase frequency accuracy

Previously the delta phase was fixed point fractional with 2^32 fractions, which caused inaccuracies in the output frequency, unless the input frequency*2^32 was divisable by the sample rate. This patch improves frequency accuracy by tracking subfractions of the delta phase fractions. For this we

Re: [FFmpeg-devel] [RFC] dormant git accounts

On 11/9/2024 6:04 PM, Rémi Denis-Courmont wrote: > What most people are concerned about right now is the incomplete > documentation > of any and all credentials - not just git write access - and more generally > the lack of transparency. Once that is sorted out, we can start arguing about > wha

Re: [FFmpeg-devel] [RFC] libpostproc splitout

On 11/7/2024 11:33 PM, Michael Niedermayer wrote: > And as a sidenote I and Derek where much friendlier towards each > other from what i remember, than in the discussion today. I hope we will be > friendlier towards each other in the future again. On the contrary, I remember being a much, much big

Re: [FFmpeg-devel] [PATCH] avformat/mov: use an array of pointers for heif_item

On 11/8/2024 8:45 PM, James Almer wrote: Pointers to specific entries in the array are stored in other structs, so in the scenario where heif_item was reallocated when parsing an iloc box after and iinf one, the pointers may end up referencing freed memory. Fixes use-after-free with such samples

[FFmpeg-devel] [PATCH 1/3] avutil/opt: do no rely on av_d2q returning higher num/den than allowed when parsing rational opts

Signed-off-by: Marton Balint --- libavutil/opt.c | 14 +++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/libavutil/opt.c b/libavutil/opt.c index ee451ae023..44b507dbd4 100644 --- a/libavutil/opt.c +++ b/libavutil/opt.c @@ -231,6 +231,14 @@ static int opt_set_init(void

[FFmpeg-devel] [PATCH 2/3] avutil/parseutils: do no rely on av_d2q returning higher num/den than allowed in av_parse_video_rate

--- libavutil/parseutils.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/libavutil/parseutils.c b/libavutil/parseutils.c index 94e88e0a79..9f57b16ec2 100644 --- a/libavutil/parseutils.c +++ b/libavutil/parseutils.c @@ -193,6 +193,9 @@ int av_parse_video_rate(AVRational *rate, const char *

[FFmpeg-devel] [PATCH 3/3] avutil/rational: never return greater num/den than the maximum allowed in av_d2q

This reverts 887d74c47efa70d7d1513e9492d41cf7f88dee0b, because it ignores the maximum allowed numerator/denominator. Even if the result was rounded to zero or infinity, it should not be cause to ignore the maximum allowed number, this "feature" is unintuitive and undocumented. Signed-off-by: Marto

Re: [FFmpeg-devel] [PATCH] fate: Add a target for listing failed tests

On Sun, 10 Nov 2024, Michael Niedermayer wrote: On Thu, Nov 07, 2024 at 11:36:24AM +0200, Martin Storsjö wrote: If running tests with "make -j fate", the execution will stop after the first failing test. To get an overview of the whole test suite, one rather would run "make -k -j fate", which t

[FFmpeg-devel] [PATCH] avcodec/x86/vvc: add prototypes for OF functions

And wrap the implemtation in preprocessor checks. Should fix ticket #11282. Signed-off-by: James Almer --- libavcodec/x86/vvc/vvcdsp_init.c | 20 +++- 1 file changed, 15 insertions(+), 5 deletions(-) diff --git a/libavcodec/x86/vvc/vvcdsp_init.c b/libavcodec/x86/vvc/vvcdsp_init.

[FFmpeg-devel] [PATCH 2/2] avcodec/rv60: simplify fill_mv_skip_cand

--- libavcodec/rv60dec.c | 54 ++-- 1 file changed, 17 insertions(+), 37 deletions(-) diff --git a/libavcodec/rv60dec.c b/libavcodec/rv60dec.c index 19cd0f2861..a54980f131 100644 --- a/libavcodec/rv60dec.c +++ b/libavcodec/rv60dec.c @@ -862,50 +862,30 @@ st

[FFmpeg-devel] [PATCH 1/2] avcodec/rv60: loosen fill_mv_skip_cand top right and bottom left criteria

Fixes ticket #11293. --- libavcodec/rv60dec.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libavcodec/rv60dec.c b/libavcodec/rv60dec.c index 091018cc56..19cd0f2861 100644 --- a/libavcodec/rv60dec.c +++ b/libavcodec/rv60dec.c @@ -881,12 +881,12 @@ static void fill_mv_skip

Re: [FFmpeg-devel] [RFC] dormant git accounts

Hi all On Sat, Nov 09, 2024 at 05:18:08PM +0100, Michael Niedermayer wrote: > Hi all > > Should we disable git accounts for developers who have not been active since > a long time (like 10 years) ? > > (if these developers come back, the account would then be enabled again) > but disabling such

[FFmpeg-devel] [PATCH] lavc/x86/videodsp: Fix clobbered FPU state on x86-32

These assembly optimisations can use MMX. They failed to reset the floating-point state when they are done, hence subsequent floating-point operations return nonsense values. This fixes the FATE failure for vvc-output-ref on x86-32, e.g. https://fate.ffmpeg.org/report.cgi?slot=x86_32-uubuntu-ming

Re: [FFmpeg-devel] [PATCH 2/2] avcodec/x86/diracdsp: Migrate last remaining MMX function to SSE2

Hi, On Sun, Nov 10, 2024 at 9:28 AM Kyosuke Kawakami wrote: > On Sat, Nov 9, 2024 at 9:07 PM Michael Niedermayer > wrote: > > > > this seems to break fate: > > Ouch, I didn't know fate needs samples to be downloaded separately. > I'll work on patch v2 tomorrow. Thanks for the review. > Maybe w

[FFmpeg-devel] [PATCH] lavc/vvc: Remove floating point logic

This was the only floating point logic in the native VVC decoder. Signed-off-by: Frank Plowman --- libavcodec/vvc/intra_utils.c | 10 +- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/libavcodec/vvc/intra_utils.c b/libavcodec/vvc/intra_utils.c index 8c40eb1b16..7229222b95

Re: [FFmpeg-devel] [PATCH] avcodec/cbs_h266: Fix regression in DVB clip introduced by 93281630a71c06642adfebebb0d4b105a4e02e91

On 10/11/2024 12:13, Nuo Mi wrote: > This commit introduced a regression to > VVC_HDR_UHDTV1_OpenGOP_3840x2160_50fps_HLG10_mosaic.ts. > > Root Cause: > The AV_CEIL_RSHIFT(a, b) macro uses bit tricks that work only when -a is a > negative value. > However, due to integer promotion rules, this

Re: [FFmpeg-devel] [PATCH 1/5] avformat/rpl: check channels

On Sun, Nov 03, 2024 at 09:12:12PM +0100, Michael Niedermayer wrote: > Fixes: > 42537199/clusterfuzz-testcase-minimized-fuzzer_loadfile_direct-5447162658357248 > Fixes: runtime error: signed integer overflow: -3330498059201358222 * 4 > cannot be represented in type 'int64_t' (aka 'long') > > Fou

Re: [FFmpeg-devel] [PATCH 2/2] avcodec/x86/diracdsp: Migrate last remaining MMX function to SSE2

On Sat, Nov 9, 2024 at 9:07 PM Michael Niedermayer wrote: > > this seems to break fate: Ouch, I didn't know fate needs samples to be downloaded separately. I'll work on patch v2 tomorrow. Thanks for the review. Kyosuke ___ ffmpeg-devel mailing list ffm

Re: [FFmpeg-devel] [RFC] dormant git accounts

On Sun, Nov 10, 2024 at 02:42:18PM +, Derek Buitenhuis wrote: > On 11/9/2024 6:04 PM, Rémi Denis-Courmont wrote: > > What most people are concerned about right now is the incomplete > > documentation > > of any and all credentials - not just git write access - and more generally > > the lack

[FFmpeg-devel] [PATCH] avcodec/cbs_h266: Fix regression in DVB clip introduced by 93281630a71c06642adfebebb0d4b105a4e02e91

This commit introduced a regression to VVC_HDR_UHDTV1_OpenGOP_3840x2160_50fps_HLG10_mosaic.ts. Root Cause: The AV_CEIL_RSHIFT(a, b) macro uses bit tricks that work only when -a is a negative value. However, due to integer promotion rules, this behavior does not extend to the unsigned int type.

Re: [FFmpeg-devel] [PATCH] lavu/common: Fix AV_CEIL_RSHIFT for unsigned LHS

On Mon, Oct 21, 2024 at 2:30 AM Michael Niedermayer wrote: > On Fri, Oct 18, 2024 at 06:48:40PM +0100, Frank Plowman wrote: > > On 15/10/2024 21:23, Frank Plowman wrote: > > > On 14/10/2024 23:26, Michael Niedermayer wrote: > > >> On Sat, Oct 05, 2024 at 03:38:05PM -0700, Frank Plowman wrote: > >

Re: [FFmpeg-devel] [RFC] libpostproc splitout

Hi Remi On Sat, Nov 09, 2024 at 08:21:24PM +0200, Rémi Denis-Courmont wrote: > Le lauantaina 9. marraskuuta 2024, 18.11.07 EET Michael Niedermayer a écrit : > > > And sorry but while I am all for splitting postproc to a separate > > > repository, it is at best a few hundreds euros worth of consult

[FFmpeg-devel] [RFC] What to do with 15k euro from STF?

Hi all as we will likely have 15k available from STF "2024" for another task/project what should it be used for ? I think the obvious options are: 1. gitlab / forgeyo setup + hw + maintaince on our infrastructure (this was often requested by people in the last years and vittorio also as

Re: [FFmpeg-devel] [RFC] What to do with 15k euro from STF?

On 11/10/2024 8:44 PM, Michael Niedermayer wrote: Hi all as we will likely have 15k available from STF "2024" for another task/project what should it be used for ? I think the obvious options are: 1. gitlab / forgeyo setup + hw + maintaince on our infrastructure (this was often requested

[FFmpeg-devel] [RFC] github assembly course ?

Hi I have been given recordings of VDD by the Reconnaissance General Bureau. (didnt listen to all yet) IIUC there was a talk where it was suggested (by kieran?) to have some assembly langauge course on FFmpeg github. IMO thats a great idea. Who intends to work on that? What do i need to setup on

Re: [FFmpeg-devel] [PATCH] lavc/x86/videodsp: Fix clobbered FPU state on x86-32

On 11/10/2024 3:38 PM, Frank Plowman wrote: These assembly optimisations can use MMX. They failed to reset the floating-point state when they are done, hence subsequent floating-point operations return nonsense values. This fixes the FATE failure for vvc-output-ref on x86-32, e.g. https://fate.