Correct the condition for freeing references in dav1d_parse_obus
obu: tile_group and frame OBUs do not have trailing bits
meson: fix disabling asm for arm/arm64
CI: bump the dav1d-debian-unstable image version.
Fix operator order in obu.c
dav1d_fuzzer: skip empty frames to avoid assertions after 026069693ef
Correctly flush at the end of OBUs
Fix how we read the UV quantization level
Segmentation map reference logic
Fix parsing segmentation data in parse_frame_hdr
Fix segmentation map size check
mc: use width/height of reference frame in warp_affine
mc: ensure order of evaluation of macro arguments in FILTER_BILIN
Support skip/globalmv/ref segmentation features
mc: use PXSTRIDE consistently in {put,prep}_8tap_scaled_c
implement a decoupled decode API
Add support for scaled references
Pass dimensions to cfl_ac and derive log2sz
Fix parsing of render size
Align block size to bw/bh for z2 filter extension
intra_pred: fix ubsan issues with max_width/height in 2f251bd11528
Add missing .data suffix so we don't overwrite in memset()
Fix dav1d CLI tool skip short option
Add a max_width/height argument to angular_ipred_fn
Re-add imax(v, 0) in SGR calculation
Fix type mismatch (int32_t vs. int)
Use prev_frame seg_map if temporal=0,update=0 also
Use negative return value in case of segmentation setup errors
do not try to build the windows .rc file if we're building for a static target
include errno.h in the public headers
lf mask: calculate vertical mask correctly
MSVC Compat: Move stdatomic fallback to its own directory
Add stdatomic fallback for GCC 4.7/4.8
frame-mt: do not derive_warpmv intrabc blocks
reduce size of Av1FrameHeader by 7940 bytes
frame-mt: set b->filter2d for intra bc blocks
Always free references upon sequence header refresh
fix backup of t->l.tx_lpf_uv at tile boundaries for 4:2:2 and 4:4:4
cdef: simplify sec_taps
Fix context for comp_type and intrabc
Add blend AVX2 asm
Add blend checkasm unit test
cdef: make scalar types non-sized
decode: clear the temporal tile data cache in Dav1dContext after copying to the frame context
Change stride of emu_edge buffer from 160 to 192
Change emu_edge function argument order for easier SIMD
Add emu_edge checkasm unit test
Fix crash if --muxer is not set
Add 'null' muxer in help
recon: fix bilinear entry in dav1d_filter_dir table
coef/dequant: clip coefs before and after dequantization
picture: add dav1d_picture_move_ref()
Make the public headers C++ friendly
Fix GCC aliasing warnings
threading: fix deadlock with multiple tile rows
frame threading: signal reference frame decoding errors
tile threading: pass decoding errors through progress tracking
do not use the picture after an allocation error
submit_frame: frame threading cleanup on decoding errors
submit_frame: always 'goto error' for proper cleanup
Use grouped context setting
Move emu_edge() from recon_tmpl.c to Dav1dMCDSPContext in mc_tmpl.c
checkasm: Don't benchmark all the 8tap filters
checkasm: Extend mc testing for any combination of 2-32 x 2-32, for OBMC
Add "disable_cdf_update" support
getbits: prevent undefined shift in dav1d_get_vlc()
Fix SGR AVX2 asm on Windows
x86: Add warp8x8 and warp8x8t AVX2 asm
Add checkasm unit tests for warp8x8 and warp8x8t
Reorder the mc warp filter array
Add AVX2 implementation for SGR looprestoration
Add unit test for SGR looprestoration
Simplify rate calculation in update_cdf
remove unused function av1_clamp_mv()
move rl32() to its only user tools/dav1d/input/ivf.c
Remove debug and redundant zero-inits from ref_mvs.c
loopfilter: use width aligned to 4 instead of 8
loop_filter_h_*_c: removed unnecessary mask to width
loopfilter: limit filter width to the frame edge
calculate and cache the picture dimension in 4x4 blocks
Add simpler version of update_cdf() for boolean symbols
Simplify blend() by making intermediate buffer contiguous
CI: update master in the cached test data repository during fetch
Fix typo in dav1d_cli_parse.c
dav1d: Only return 0/1 from md5_verify
meson: prevent memory sanitizer builds with asm
arm: Improve use of getauxval in quirky environments
Include <stdlib.h> for win32/thread.c
mem: use memalign as fallback for posix_memalign and _aligned_malloc
arm: use defined() for defines which are not guaranteed to be defined
obu: max_tile_height_sb is only used for non-uniform tiles
CI: run testdata tests on all branches and merge requests
arm: add /proc/cpuinfo based NEON detection for Android J-B
arm: add cpu flag detection
arm: call dav1d_get_cpu_flags_arm only when asm is enabled