Document the thread context in which Dav1dPicAllocator's functions are called
Allow redundant sequence header
itx: cancel 16 out of '(x * 2896) >> 12' to avoid integer overflows
inv_identity: avoid signed int overflow for 12-bit files
x86: Add chroma-from-luma AC 4:2:2 AVX2 asm
meson: fix installing the includes
allocation fail fuzzer: use a static library and objcopy
do not build the allocation fail fuzzer when lto is enabled
fuzzing: add memory allocation fail fuzzer
cdf: fix memleak in dav1d_cdf_thread_unref
Make static CDF tables source directly from rodata
svc: move spatial layer check out of output_picture
annexb: don't use a int64_t intermediary for leb128
picture: free pic_ctx on dav1d_ref_wrap error
output_picture: unref pictures on allocation errors
memory allocation error checks
decode_frame: use goto error after allocation errors
checkasm: replace rand() with xorshift xor128
fix mvs_ref NULL ptr dererence after allocation error
cdf: make CdfThreadContext allocation failure safe
dav1d_open: free context completely on allocation errors
CI: set empty dependecies for the tests jobs
annexb: ensure that OBU size <= frame size <= TU size
Add Length delimited bitstream (Annex B) demuxer
checkasm: Try to handle crashes more gracefully
Implement --cpumask to mask out CPU features
Fix mc.avg/w_avg/mask for x86-32
Add SSSE3 implementation for the 4x4 blocks in itx
Add SSSE3 implementation for dav1d_ipred_h
12 bits/component support
Change type of MC intermediates from coef to int16_t
Fix 12-bit support in yuv4mpeg muxer
Use adapted CDF for invisible keyframes if refresh_context=1
Add ipred_z1 AVX2 asm
Make some minor ipred_z simplifications
frame-mt: drain delayed pictures fully before returning -EAGAIN
frame-mt: output delayed frames as soon as possible
Correctly use movd instead of movq in blend_h_avx2.w4
Add SSSE3 implementation for dav1d_inv_txfm_add_dct_dct_4x4
film_grain: limit overlapped pixels to block boundaries
film_grain: copy unmodified planes before applying noise
Make per-width versions of cfl_ac
x86: Add chroma-from-luma AC 4:2:0 AVX2 asm
frame-mt: output only fully decoded frames
obu: fix setting num_ticks_per_picture in sequence headers
data: fix setting Dav1dDataProps defaults in dav1d_data_create()
common: make 0 the default for Dav1dDataProps.duration
common: make it clear that the default values in Dav1dDataProps mean unknown
common: make -1 the default for Dav1dDataProps.offset
common: make Dav1dDataProps.duration signed
backup_lpf: increment row by the correct stripe_h
tools/dav1d: abort on unused/unknown CLI arguments
Make Dav1dDataProps::timestamp and offset signed.
ivf: Use _ftelli64 instead of ftello for MSVC
ivf: Use int64_t instead of off_t
build: use join_paths() for doxygen paths
tile-mt: reset tile tasks when the number of tile rows changes
doc: extend dav1d_flush Doxygen
add a Doxygen html target
add SSSE3 avg/w_avg/mask
checkasm: add --list tests
obu: ensure that the frame header reference is writable
obu/film grain: check ref frame frame_hdr before use
clean state in dav1d_flush()
Round up film grain block count
Remove source padding in film grain
Don't use string literals in assert()
CI: use the sanitizer test setup to prevent timeouts
CI: make sure MSAN and UBSAN warnings are errors
Clear coefficient cache upon error
Use a single atomic integer for flush
ivf: store offset and timestamp in pkt.m
Add a simple API usage example to the doxy
Reset the random seed when testing each CPU type.
mc/svc: adjust frame-mt wait scanline for 8-tap filter
obu: error out on missing frame_hdrs in ref frames with frame_ref_short_signaling
dav1d/md5: avoid unaligned 32-bit reads
dav1d: add public defines for max amount of tile and frame threads
dav1d/cli: start decoding on a sequence header after '--skip'
fuzzer: use dav1d_parse_sequence_header()
fix memleaks with ref counted sequence and frame headers
memset frame_hdr to fix use of uninitilized value regressions
Add a public function to parse the Sequence Header from raw bitstream data
Put layout back in Dav1dSequenceHeader
arm64: looprestoration: NEON optimized wiener filter
Store literal bitstream values in Dav1dSequence/FrameHeader
Move Dav1dPictureParameters-related entries to top of Dav1dSeqHdr
Add some more constants from section 3
Fix some crashes (ninja test) from the previous patches
Remove entries from c->refs[] that can be retrieved from frame_hdr
Add frame_hdr/seq_hdr to Dav1dPicture
Move Av1FrameHeader and Av1SequenceHeader into public headers
Make frame_hdr a referenced object