build: Enable SSE2 by default on x86-32
x86: Enable msac asm on x86-32
Update THANKS.md
Hard wrap contribs. Add self for logo
ci: Add full testdata tests on aarch64
checkasm: Update the mc test to check all valid heights
arm: mc: Fix 8tap_v w8 with OBMC 3/4 heights
Optimize obmc blend
Remove one multiply in Z2 filter top left
Reduce branching in intra angle to mode
small code cleaning in intra_edge init_mode
Fix unused function warning on parse_proc_cpuinfo() for Android
Use size_t for the msac window size
arm64: msac: Add handwritten versions of msac_decode_bool functions
arm64: msac: Fix a typo in a comment
x86-64: Add msac_decode_bool and msac_decode_bool_adapt asm
Add dav1d logo
x86-64: Add msac_decode_bool_equi asm
Add a hard upper frame size limit on 32-bit systems
Add an option to limit the maximum frame size
obu: add missing break to the default case of a switch statement
obu: don't abort on unknown OBUs
Update NEWS and version for 0.3.1
tools: Add a cast to silence an MSVC warning
build: Group asm files per instruction set
Add __attribute__((cold)) to rarely used functions
Add fps counter and --realtime, --frametimes and --realtimecache options
build: Use -mcmodel=small on 64-bit Windows
Increase thread stack size
arm: Simplify includes for util.S
fuzzer: fix oss-fuzz undefined behavior sanitizer build
Add a DAV1D_ERR define to negate errno values when needed
Fix buffer overflow in 64x16 ssse3 idct
arm64: remove invalid macro argument delimiter
Add SSSE3 implementation for ipred_filter
ci: Check for unprefixed global symbols
Fix all remaining symbols without a dav1d prefix
ci: Ignore binary files in style check
Add missing dav1d prefixes to picture allocation functions
Control the stack size of spawned threads
arm64: msac: Implement NEON msac_decode_symbol_adapt
itx_tmpl: Fix the assert in inv_txfm_add_c
ci: Add a test for x86-64 with 16-byte stack alignment
Update NEWS for 0.3.0 - Sailfish
Fix crash in SSSE3 inverse transform
Update NEWS for 0.2.2
Add SSSE3 implementation for the {16, 32, 64}x64 and 64 x{16, 32} blocks in itx
Over-allocate level array by 3-bytes
arm64: loopfilter: Implement NEON loop filters
arm64: looprestoration: Add a NEON implementation of SGR
msac: Add a cast to indicate intended narrowing from size_t to unsigned
x86-64: Add msac_decode_symbol_adapt SSE2 asm
Add SSSE3 implementation for ipred_paeth
arm: Add a _neon suffix to all internal functions
arm: Fix typos in comments
arm: Consistently use 8/24 columns indentation for assembly
Add SSSE3 implementation for ipred_cfl_ac_444
CI: Check for newline at end of file
x86: cdef_dir: optimize best cost finding for SSE
x86: cdef_filter: use 8-bit arithmetic for SSE
x86: cdef_filter: use a better constant for SSE4
x86: cdef_filter: fix macro case (lower to upper)
Add SSSE3 implementation for the 16x32,32x16 and 32x32 blocks in itx
build: Split x86 asm files per bitdepth
Only define DAV1D_API to dllexport when building dav1d itself
Simplify C for inverse transforms
x86: Add minor CDEF AVX2 optimizations
Add SSSE3 implementation for the 8x32 and 32x8 blocks in itx
Add SSSE3 implementation for ipred_cfl_ac_420 and ipred_cfl_ac_422
decode: add a frame tile data buffer size check
decode: don't realloc the tile data buffer when it needs to be enlarged
tools/dav1d/md5: bswap big endian high bit depth pixel data
tools/dav1d: make the md5 muxer endian-aware
On the road to 0.2.2
build: use the project version as product version in the windows resource file
x86: remove unused macro arg in SSSE3 cdef_dir
x86: fix number of xmm used in sgr_calc_ab1_ssse3
x86: optimize SSSE3 sgr_calc_ab{1,2}
x86: optimize AVX2 cdef_dir
build: add "api" to api/soname related versions
Update NEWS for 0.2.1
let dav1d_version() return the project version
x86: add SSSE3 cdef dir implementation
arm: looprestoration: Simplify a few padding cases in wiener_filter_h_neon
Use some 8 bit arithmetic in AVX2 CDEF filter
arm: looprestoration: Remove unnecessary includes
arm64: cdef: Clarify a slightly confusing comment
arm64: cdef: Use a smarter padding constant
arm64: cdef: Do saturating subtractions to avoid max operations with 0
Utilize a better CDEF constant for avx2
On the road to 0.2.1
Remove unused data from x86/cdef.asm
x86: add SSSE3 mc prep_8tap implementation
Speed up finding the best cost in avx2 cdef
windres: fix the version number string
x86: Check for BMI1 and BMI2 flags in addition to AVX2
picture: fix default_picture_allocator() return value on failure
picture: make mastering display and content light metadata accessible when allocating pictures