.forgejo/CODEOWNERS: add myself for the AC-3 and E-AC-3 decoders

20 September 13:11 - FFmpeg - Ayoub Nabil Boubagrat

vulkan_decode, vulkan_encode: report codec profiles the device does not support

20 September 02:59 - FFmpeg - Lynne

VK_ERROR_VIDEO_PROFILE_CODEC_NOT_SUPPORTED_KHR, which drivers return for a codec profile they have no notion of, such as H.264 High 4:2:2, fell through to a silent generic failure.

avformat/matroskadec: add support for more than one BlockAdditionMapping of type ITU-T T.35

19 September 21:42 - FFmpeg - James Almer

.forgejo/CODEOWNERS: add myself for the D3D12 video filters

19 September 00:14 - FFmpeg - Steven Xiao

I am listed as maintainer for vf_*_d3d12.c.

avformat/mov, movenc: add the dav1 sample entry for AV1 Dolby Vision

18 September 23:21 - FFmpeg - Vincent Herbst

MP4RA registers 'dav1' as a variant of 'av01' that signals Dolby Vision.

[varc] Support static component axes

18 September 15:07 - FreeType - Behdad Esfahbod

The VARC draft permits static fonts without a 'fvar' table to retain 'gvar' or 'CFF2' variation data for component-internal coordinates.

avutil/base64: add aarch64 NEON implementation for base64 encode

18 September 14:20 - FFmpeg - Shreesh Adiga

Add NEON implementation using LD3 to load 48 bytes with de-interleaving using which 6 bits are extracted using USHR and SLI to get 4 registers, which is used as a index into lookup table using TBL with 4 registers to translate into the 64-byte output, which is written using ST4 to interleave.

avcodec/mfenc: guard D3D11 input support when disabled

17 September 03:10 - FFmpeg - Michael Niedermayer

The D3D11 header is conditional, but the context members and frame processing helpers using its types are not.

avfilter/vf_sr_amf: convert YUV input on the GPU for sr1-1 and point

16 September 22:05 - FFmpeg - Julius Bairaktaris

sr1-1 and point need a packed RGB surface.

avfilter/amf: make the input format lists describe what AMF accepts

16 September 22:05 - FFmpeg - Julius Bairaktaris

amf_setup_input_output_formats() builds both negotiated lists from output_pix_fmts and ignores its input_pix_fmts argument, so the input lists have never been used and have drifted from what the filters handle. None of them lists AV_PIX_FMT_D3D11 or AV_PIX_FMT_DXVA2_VLD even though amf_avframe_to_amfsurface() wraps both.

avfilter/amf: negotiate the input link from the input format list

16 September 22:05 - FFmpeg - Julius Bairaktaris

amf_setup_input_output_formats() built both negotiated lists from output_pix_fmts, so every filter's input_pix_fmts argument was dead and the input link only ever offered what the filter could output. For vpp_amf, whose input list is deliberately wider than its output list, libavfilter therefore inserted a software conversion in front of the filter for any input format missing from the output list.

avfilter/amf: pin the output format for components that cannot convert

16 September 22:05 - FFmpeg - Julius Bairaktaris

amf_init_filter_config() falls back to outlink->format for the output frames context when no format is requested.

avfilter/amf: copy D3D11 decoder textures the components can read

16 September 22:05 - FFmpeg - Julius Bairaktaris

The AMF filter components read their input with a shader and reject a texture created without D3D11_BIND_SHADER_RESOURCE, which is what a D3D11VA decoder pool provides.

avfilter/amf: flush the component at EOF with an activate callback

16 September 22:05 - FFmpeg - Julius Bairaktaris

The AMF filters ran off a filter_frame callback, which has no way to tell a component that no more input is coming.

avfilter/vf_vqe_amf: leave the output colorspace untouched

16 September 22:05 - FFmpeg - Julius Bairaktaris

The shared output path retags every frame from the filter's color_profile unless that is AMF_VIDEO_CONVERTER_COLOR_PROFILE_UNKNOWN, which is -1.

avfilter/amf: hold the wrapped D3D11 and DXVA2 frame until AMF releases it

16 September 22:05 - FFmpeg - Julius Bairaktaris

CreateSurfaceFromDX11Native() and CreateSurfaceFromDX9Native() wrap the texture without taking a reference on anything FFmpeg owns.

avcodec/hevc: fill the stream parameters when all frames are skipped

16 September 18:59 - FFmpeg - Michael Niedermayer

With skip_frame at AVDISCARD_ALL the parameter set and SEI NAL units are still decoded, and as long as no decoded frame has configured the decoder, the parameters of the SPS activated by the first slice are exported.

avutil/aes: add aarch64 NEON implementation

16 September 06:29 - FFmpeg - Shreesh Adiga

Ported the x86 AES-NI implementation to aarch64 NEON using AES extension.

drivers/char: Panic when the requested UART fails to initialise

15 September 15:13 - Xen - Michal Orzel

uart_init() cannot tell its caller that the UART the user asked for did not come up: every failure path only printks.

xen/arm: report proper GIC version via XEN_DOMCTL_getdomaininfo

15 September 15:13 - Xen - Julian Vetter

When creating a domain on ARM, and passing XEN_DOMCTL_CONFIG_GIC_NATIVE for the gic_version field in the struct xen_arch_domainconfig, arch_sanitise_domain_config() resolves this to the approrpiate GIC_V2 or GIC_V3 version the domain actually has, based on the host's gic_hw_version().

avfilter/vf_colordetectdsp: rewrite C implementations as macros

15 September 13:12 - FFmpeg - Niklas Haas

In the process, also add the range assumption and cast to the detect_alpha intermediates, which seems to help codegen: checkasm:- CPU: AMD Ryzen 9 9950X3D 16-Core Processor (00B40F40)- Timing source: x86 (rdtsc)- Bench duration: 100000 µs per function (442374016 cycles)- Random seed: 2809724650 Benchmark results: (old) name cycles (vs ref) detect_alpha_8_full_c: 7582.7 detect_alpha_8_full_off_c: 7587.4 detect_alpha_8_limited_c: 13306.7 detect_alpha_16_full_c: 3482.5 detect_alpha_16_full_off_c: 3480.4 detect_alpha_16_limited_c: 8055.1 detect_range_8_c: 1271.4 detect_range_16_c: 1213.0 Benchmark results: (new) name cycles (vs ref) detect_alpha_8_full_c: 4716.6 detect_alpha_8_full_off_c: 4715.4 detect_alpha_8_limited_c: 7943.2 detect_alpha_16_full_c: 2920.2 detect_alpha_16_full_off_c: 2918.9 detect_alpha_16_limited_c: 9952.0 detect_range_8_c: 1253.1 detect_range_16_c: 1194.8 The ~20% slowdown for detect_alpha_16_limited_c is because the previous version of the code short-circuited this condition, which is faster when the compiler is not able to vectorize the loop body and falls back to a pure scalar loop. However, I don't think optimizing around this is necessary as it's a pure compiler optimization question.

avformat/shared: add -ignore_errors option

15 September 10:04 - FFmpeg - Niklas Haas

This allows the shared: cache protocol to ignore errors in the underlying URL (e.g.

avformat/shared: add spacemap metadata for the number of cached blocks

15 September 10:04 - FFmpeg - Niklas Haas

This can only ever be a lower bound in theory, because a process could crash exactly in between committing the cached block and updating the count, though that edge case is incredibly rare. This is a backwards-compatible addition so it doesn't require a cache version update.

avformat/shared: add -cache_size_max option

15 September 10:04 - FFmpeg - Niklas Haas

This allows users to limit the amount of cached data.