- 20 Feb, 2019 4 commits
-
-
Martin Storsjö authored
Check all commits, not just the latest one, as a branch may consist of multiple commits. Exclude commits from upstream master (which may contain issues which we want to error out for on new commits).
-
Ronald S. Bultje authored
Allows simplified SIMD function implementations that don't exactly respect picture boundaries when reading picture data. Fixes #251 and #250.
-
Ronald S. Bultje authored
Allows aligned accesses to this array in SIMD. Fixes #246.
-
Ronald S. Bultje authored
Fixes #252.
-
- 19 Feb, 2019 6 commits
-
-
James Almer authored
The references in the Dav1dContext may not necessarely apply to the picture being copied.
-
Henrik Gramner authored
Eliminates segfault with clusterfuzz-testcase-minimized-dav1d_fuzzer-5697181166600192. Credit to OSS-Fuzz.
-
Martin Storsjö authored
This test builds with clang, as GCC defaults to 16 byte stack alignment on windows, while clang defaults to 4.
-
Martin Storsjö authored
This version supports running the built executables with wine.
-
Martin Storsjö authored
This is intended only for use in CI testing.
-
Xuefeng Jiang authored
cfl_pred_cfl_128_w4_8bpc_c: 291.3 cfl_pred_cfl_128_w4_8bpc_ssse3: 34.2 cfl_pred_cfl_128_w8_8bpc_c: 838.8 cfl_pred_cfl_128_w8_8bpc_ssse3: 55.4 cfl_pred_cfl_128_w16_8bpc_c: 1640.6 cfl_pred_cfl_128_w16_8bpc_ssse3: 91.8 cfl_pred_cfl_128_w32_8bpc_c: 4073.5 cfl_pred_cfl_128_w32_8bpc_ssse3: 206.6 cfl_pred_cfl_left_w4_8bpc_c: 321.6 cfl_pred_cfl_left_w4_8bpc_ssse3: 40.0 cfl_pred_cfl_left_w8_8bpc_c: 887.8 cfl_pred_cfl_left_w8_8bpc_ssse3: 62.9 cfl_pred_cfl_left_w16_8bpc_c: 1932.6 cfl_pred_cfl_left_w16_8bpc_ssse3: 97.4 cfl_pred_cfl_left_w32_8bpc_c: 4795.4 cfl_pred_cfl_left_w32_8bpc_ssse3: 213.8 cfl_pred_cfl_top_w4_8bpc_c: 359.7 cfl_pred_cfl_top_w4_8bpc_ssse3: 38.2 cfl_pred_cfl_top_w8_8bpc_c: 1007.1 cfl_pred_cfl_top_w8_8bpc_ssse3: 58.5 cfl_pred_cfl_top_w16_8bpc_c: 1873.4 cfl_pred_cfl_top_w16_8bpc_ssse3: 97.9 cfl_pred_cfl_top_w32_8bpc_c: 4523.7 cfl_pred_cfl_top_w32_8bpc_ssse3: 216.9 cfl_pred_cfl_w4_8bpc_c: 537.4 cfl_pred_cfl_w4_8bpc_ssse3: 44.9 cfl_pred_cfl_w8_8bpc_c: 1031.1 cfl_pred_cfl_w8_8bpc_ssse3: 66.7 cfl_pred_cfl_w16_8bpc_c: 1751.5 cfl_pred_cfl_w16_8bpc_ssse3: 99.9 cfl_pred_cfl_w32_8bpc_c: 4814.5 cfl_pred_cfl_w32_8bpc_ssse3: 219.8
-
- 18 Feb, 2019 2 commits
-
-
Michael Bradshaw authored
All the other headers are being included relatively.
-
Janne Grunau authored
Listing 'include/dav1d' before 'include' prefers the generated version.h in include/dav1d/ over the previous one in include/. This allows the use of '#include "version.h"' in dav1d.h. See !592.
-
- 17 Feb, 2019 7 commits
-
-
Martin Storsjö authored
-
Martin Storsjö authored
Even if they target the win32/win64 environments, they are cross compilation jobs expected to run in a debian/amd64 environment.
-
Martin Storsjö authored
These tests can now be run in wine. Setting up a persistent wineserver for this run probably isn't necessary as there will only be two wine invocations (one within meson to check that executing binaries works, and one when running checkasm), but it serves as an example of a good practice when running things with wine.
-
Martin Storsjö authored
This requires bumping the cache key, as the old image stored cache data as the user root, while the new image runs as the user videolan, which fails to update cache data owned by the user root.
-
Luc Trudeau authored
It was found by Christopher "Monty" Montgomery that when a value between 224 and 96 is clipped between -32 and 95, the resulting value will always be 95.
-
Luc Trudeau authored
Removed arrays
-
James Almer authored
-
- 16 Feb, 2019 2 commits
-
-
Henrik Gramner authored
-
Henrik Gramner authored
Warn when the following are used without the appropriate cpuflag: * 256-bit integer instructions * 'pextrw' with a memory operand * GPR instruction set extensions
-
- 15 Feb, 2019 3 commits
-
-
Henrik Gramner authored
-
Henrik Gramner authored
It's only used by msac_decode_symbol_adapt() in the same file.
-
James Almer authored
They only apply to the old one.
-
- 14 Feb, 2019 6 commits
-
-
François Cartegnie authored
-
François Cartegnie authored
-
Mark Shuttleworth authored
-
James Almer authored
They were meant to be freed by the call to dav1d_flush() at the beginning, but before the previous commit that wasn't the case when decoding with one frame thread. This ensures they are explicity freed here and not depend on other functions behaving correctly.
-
James Almer authored
-
Xuefeng Jiang authored
pal_pred_w4_8bpc_c: 141.0 pal_pred_w4_8bpc_ssse3: 23.4 pal_pred_w8_8bpc_c: 374.5 pal_pred_w8_8bpc_ssse3: 29.0 pal_pred_w16_8bpc_c: 946.3 pal_pred_w16_8bpc_ssse3: 45.6 pal_pred_w32_8bpc_c: 1946.1 pal_pred_w32_8bpc_ssse3: 92.3 pal_pred_w64_8bpc_c: 4925.9 pal_pred_w64_8bpc_ssse3: 180.1
-
- 13 Feb, 2019 10 commits
-
-
Janne Grunau authored
-
Martin Storsjö authored
Relative speedup vs C code: Cortex A53 A72 A73 warp_8x8_8bpc_neon: 3.19 2.60 3.66 warp_8x8t_8bpc_neon: 3.09 2.50 3.58
-
Martin Storsjö authored
This effectively reverts a0692eb8 for other architectures. The order that is beneficial for x86 SIMD is not beneficial for other architectures. For a NEON implementation of the warp filter, reordering the filter coefficients back in the right order took 1/4 of the filter runtime.
-
Martin Storsjö authored
Speedup vs C code: Cortex A53 A72 A73 cdef_dir_8bpc_neon: 4.43 3.51 4.39
-
Janne Grunau authored
Renames the current version function and define to dav1d_version_vcs() and DAV1D_VERSION_VCS. Fixes #241. The generated version.h is included as "dav1d/version.h" to avoid errors with a stale include/version.h in a build dir.
-
Vittorio Giovara authored
-
Vittorio Giovara authored
-
Martin Storsjö authored
A symbol starting with two leading underscores is reserved for the compiler/standard library implementation. Also remove the trailing two double underscores for consistency and symmetry.
-
Martin Storsjö authored
-
Martin Storsjö authored
-