Age | Commit message (Collapse) | Author |
|
* commit 'ceb214959c6af1945c48599feb0e6b8e4e42c1e0':
Fix libvpx clang build
|
|
|
|
64-bit clang builds with ninja were failing because LOCAL_CLANG_ASFLAGS
was set to always point to the intermediates directory for the primary
architecture. Ninja happens to build the secondary architecture first,
and if the translated assembly for the primary architecture hasn't been
built yet it results in:
out/target/product/flounder/obj_arm/STATIC_LIBRARIES/libvpx_intermediates/vp8/encoder/arm/neon/subtract_neon.asm.s: Assembler messages:
out/target/product/flounder/obj_arm/STATIC_LIBRARIES/libvpx_intermediates/vp8/encoder/arm/neon/subtract_neon.asm.s:21: Error: can't open vp8_asm_enc_offsets.asm for reading: No such file or directory
Modify libvpx.mk to use the primary intermediates for the primary
architecture and the secondary intermediates for the secondary architecture.
Change-Id: I3bf6e63e846ad5985b2f46f72995664e183de394
|
|
\'efc44fee\' into HEAD
* commit '9dbb32694529ff1dd285c5f2fad497ed937a5c44':
Disable bilinear_predict4x4_neon
|
|
* commit '8a9c5a1de95ee2cb19b57c811723d32d7f770c67':
Disable bilinear_predict4x4_neon
|
|
When built with newer versions of clang the alignments hints may cause
the loads to fail.
BUG=22943772
Change-Id: I7464de73f260458bc490f9178d0cfcc4eca8661b
(cherry picked from commit 939bda3fbdcbb7d56301f31010d10970c5d10612)
|
|
* commit '4a94e3b4897601af1d3ba5a0d8a57c4a82e9b05d':
Disable bilinear_predict4x4_neon
|
|
* commit '217c6ceb16a7ec83f027e1963ba9cbe542097179':
Disable bilinear_predict4x4_neon
|
|
Change-Id: I850a320e2231f641bb0164398f1fb04555874a56
|
|
When built with newer versions of clang the alignments hints may cause
the loads to fail.
BUG=22943772
Change-Id: I7464de73f260458bc490f9178d0cfcc4eca8661b
(cherry picked from commit 939bda3fbdcbb7d56301f31010d10970c5d10612)
|
|
* commit '16f9a4cb3024b600733eafaf841f76ac70ba881a':
Add MODULE_LICENSE
|
|
|
|
Change-Id: Ia6e149207680e18b69f8d2f8fcdce6d2c0051342
|
|
Cherry-picking change from https://gerrit.chromium.org/gerrit/#/c/73959/
Change-Id: I8d98543f226594b28a88d5042f43c76f1224c0bb
|
|
NOT MERGE]
* commit '9ded949dcf45f8a869662b1c10ff6f38c7f62142':
Cherry-pick "Use lrand48 on Android" [DO NOT MERGE]
|
|
optimizations [DO NOT MERGE]
* commit 'f9a5c1b2bb051b7089c0f49e06a1727a8b7d2a8a':
Turn on libvpx x86 optimizations [DO NOT MERGE]
|
|
* commit '69f5594ef6a4a13fac8b001421b302a9a0b4a009':
Cherry-pick "Use lrand48 on Android" [DO NOT MERGE]
|
|
* commit '470ef077237c82c7ab7ba2def07241a6a1f9349c':
Turn on libvpx x86 optimizations [DO NOT MERGE]
|
|
Also remove from x86/libvpx_srcs.txt
Original change: https://gerrit.chromium.org/gerrit/70494/
When building x86 assembly use lrand48 instead of the
undocumented inlined _rand function.
Android now supports rand()
https://android-review.googlesource.com/97731
but only for new versions. Original workaround:
https://gerrit.chromium.org/gerrit/15744
(cherry-picked from 614a6a21483b59b4ab557785c160c8ca4722b062)
Change-Id: I3c17cf37c40623f8353d8836fb4ac02c35a4d8b8
|
|
With yasm available we can use the assembly:
https://android-review.googlesource.com/#/c/96972/
Comparing with the VP9 video playback with C-reference software decode,
before and after these patches applied,
Performance data with 2 test clips:
Clips bit rate fps(after) fps(before)
vp9_720p_qp10to20_1510Kbps_30fps.webm 1510Kbps 30 12
vp9_720p_qp20to30_840Kbps_30fps.webm 840Kbps 30 13
(cherry-picked from afecaee6099284387688b9ce17ec3642ec4e5c1b)
Change-Id: I9c67e5ae302f923134e073f4bafb09fb1f3fe94d
|
|
* commit '03bb428a6a1414279f936c9da635899f043e156d':
allow subtitle tracks with ContentEncodings
|
|
* commit '004e13f2a5a0cf8e2abf94af22cfcee4572313ce':
allow subtitle tracks with ContentEncodings
|
|
Bug: 18999967
Change-Id: I1842b6e7c6bab4ba0f86938bbe6b110be0c5435f
|
|
* commit 'bb1f43c64d1a04850df4bcfd9ae1c5537524de45':
Suppress clang warning until upstream fix.
|
|
* commit '4d59fcfbd820e0a43367fb91c49cccaff4655ba6':
Suppress clang warning until upstream fix.
|
|
|
|
Change-Id: Ic0966bd89bca5502aa525414b31b29ca1a2f29a5
|
|
level."
* commit 'dce19c030c07256b6860a9f4ea2804be2f75f7c0':
|
|
* commit 'acfe8bb381dd96f2a1cfe44c882ecd06fbce27f1':
|
|
* commit '4c6460b45f379691b5e40859c26fbf2ef0441a30':
Add -no-integrated-as at local level.
|
|
* commit '7fb54bb13de3ee9c7d769f7599bc157aebb178ff':
Add -no-integrated-as at local level.
|
|
* commit '7fb54bb13de3ee9c7d769f7599bc157aebb178ff':
Add -no-integrated-as at local level.
|
|
|
|
Later we will enable integrated-as as default at the global level.
Also pass include path explicitly to the GCC assembler.
Unlike gcc, clang doesn't pass -I flags on to the assembler - causing
the .include statements not to find the generated asm files.
This is reported as a bug in clang at
http://llvm.org/bugs/show_bug.cgi?id=21000
BUG: 17820427
BUG: 17137906
Change-Id: I788f3c8a3ca71948476c469b21dfda14483d269d
|
|
* commit '0e7430e0f21c927ef7c81ccc3cdbb557a1a92556':
libvpx: disable accidentally enabled neon code
|
|
* commit '9921efda82d7b1f6410ad01c2cb4549b3bc1bb5c':
libvpx: disable accidentally enabled neon code
|
|
|
|
upstream's configure behavior changed and an extra flag needed to
be passed to disable all the neon code for target armv7a.
Change-Id: I773fdf29bdd678aa56cd7cf5d0942cd3f456de73
|
|
upstream's configure behavior changed and an extra flag needed to
be passed to disable all the neon code for target armv7a.
Change-Id: I773fdf29bdd678aa56cd7cf5d0942cd3f456de73
|
|
* commit 'a695ee8b5b191737fec62b386272e28a4c452d14':
libvpx: Pull from upstream
|
|
|
|
Upstream hash: d4a47a6cc0d869bea3071c15bc61da6836026d0b
Pull latest libvpx from upstream. This fixes a few vp9 encoder bugs and
includes some optimizations. Also fixes a couple of configure flags in
x86 to be consistent with the rest.
Change-Id: Ic58e0b03cce832571a35ec73eec559cdf881d1f5
|
|
* commit 'bbeabeb879e7fa51c6395ee7ad590617dfbd5299':
Cherry-pick "Use lrand48 on Android"
|
|
|
|
Also remove from x86/libvpx_srcs.txt
Original change: https://gerrit.chromium.org/gerrit/70494/
When building x86 assembly use lrand48 instead of the
undocumented inlined _rand function.
Android now supports rand()
https://android-review.googlesource.com/97731
but only for new versions. Original workaround:
https://gerrit.chromium.org/gerrit/15744
Change-Id: I3c17cf37c40623f8353d8836fb4ac02c35a4d8b8
|
|
* commit 'd64f247f64fbb814c9ecf06a56bcb0948bfca21f':
Turn on libvpx x86 optimizations
|
|
|
|
Rolling mkvparser from upstream. Primarily for fixing a bug on parsing
failures with certain Opus WebM files.
Upstream commit hash of this pull: 574045edd4ecbeb802ee3f1d214b5510269852ae
The diff is so huge because there were some style clean ups upstream.
But it was ensured that there were no breaking changes when the style
clean ups was done upstream.
Change-Id: Ib6e907175484b4b0ae1b55ab39522ea3188ad039
|
|
With yasm available we can use the assembly:
https://android-review.googlesource.com/#/c/96972/
Comparing with the VP9 video playback with C-reference software decode,
before and after these patches applied,
Performance data with 2 test clips:
Clips bit rate fps(after) fps(before)
vp9_720p_qp10to20_1510Kbps_30fps.webm 1510Kbps 30 12
vp9_720p_qp20to30_840Kbps_30fps.webm 840Kbps 30 13
Change-Id: I9c67e5ae302f923134e073f4bafb09fb1f3fe94d
|
|
\'4fb68e5dd4e93c7599dc905d861de11ac39c5585\' into HEAD
* commit 'a18cc111cea31a23de108aacbb02a888f62ab4be':
|