libvpx dropped vpx_mem_set_functions,
only use it if an external libvpx
is used and still has it.
update update.py
add vpx_dsp_rtcd.h
rebase disable_pthread_on_mingw.patch
add vp9_filter_restore_aligment.patch
drop msvc2015.patch
vp9_thread.c has changed directories, which mach doesn't handle
properly, resulting in "No rule to make target" for the old source
file path.
Bump clobber for libvpx file renames.
From df2e0cd12753c147a3bd59f300c507a8246bcd52 Mon Sep 17 00:00:00 2001
This speeds up incremental updates from upstream, or debugging
update script issues.
Note that this changes the behaviour of the update script without
the --commit argument. Previously that would do a fresh clone and
one would get the lastest upstream default branch. Now without a
specific --commit argument, it will re-use an existing checkout
if one is in tree, so take care to update the repo's HEAD to what
you want first.
From 79bdcf3f30dd70c3fe95fc6e5adb2f8961e640fd Mon Sep 17 00:00:00 2001
This fixes a build problem with the most recent import on msvc.
In general upstream doesn't support our method of concatenating
source files to speed up compilation. Rather than hand-maintaining
a list of exceptions and risking silent mis-compiles, it's better
just to compile every module individually.
From b40321d9379faaa62fa16d6d412098fca115042d Mon Sep 17 00:00:00 2001
Result of running ./update.py --ndk ~/android/android-ndk-r9 --commit
afad1a84c15b9af8298a37c0fa449e0af40931fd
Update our in-tree libvpx code to upstream master from September 25.
This includes a work-around added to skip neon intrinsics which
android toolchains fail to compile properly. See bug 1087096.
Results of running ./update.py --ndk ~/android/android-ndk-r9 --commit
587ff646f
The assembly relies on __ANDROID__ to decide whether to call
'lrand48' or 'rand'. The later is available on desktop but
only on Android L or later. However, this macro isn't defined
by the linux vpx_config asm file we use for x86 Android builds,
so we must pass it manually when invoking yasm.
Temporarily disable AVX2 simd acceleration on linux to work around
issues with old toolchains on the integration server builds.
Linux valgrind and B2G desktop builds use an older gcc which doesn't
support the new AVX2 intrinsics. This can be resolved either by updating
the build, or splitting the target platforms into AVX and no-AVX variants.
Propagate ssse3 and sse4.1 build flags from configure.
This is required by new optimations in upstream commit
c731d6a4f19eea861ceb2ff31399420b2452eb74.
Remove upstreamed patches, rebase stdint.patch.
Update script to work with the most recent chrome pull, upstream
hasn't been doing releases, so this is the next best way to stay
current.
It's no longer possible to build vp8 and vp9 in the same unified
source file, so most of the vp9 code is moved out of unified
sources.
Propagate ssse3 and sse4.1 build flags from configure.
This is required by new optimations in upstream commit
d95585fb0ec024f6abd96f7b02e0df58019d46af.
There have been releases. Bump to the git commit used by the
most recent chrome pull: d95585fb0ec024f6abd96f7b02e0df58019d46af.
This patch contains changes to the update script necessary to
build the new snapshot and removes upstreamed patches.
It's no longer possible to build vp8 and vp9 in the same unified
source file, so most of the vp9 code is moved to standalone sources.
