Commit Graph

12345 Commits

Author SHA1 Message Date
James Zern
84eca07c61 tests/examples.sh: skip env. check for filtered tests
Change-Id: Ide670b00ae3ecd9e18f394905341bf793349f247
2014-12-17 19:02:00 -08:00
Yaowu Xu
09b9a59fb5 Merge "Corrected value range of --cpu-used for vp9" 2014-12-17 17:12:14 -08:00
Yaowu Xu
a16f075375 Corrected value range of --cpu-used for vp9
This commit removes undefined value options of cpu-used for VP9 and
changed vpxenc prompt to reflect the usable range of [-8,8]

Change-Id: Ib80fef3dbb6ec9aabac45ed13e8ab6fbaf94f55e
2014-12-17 15:18:01 -08:00
JackyChen
9bc7974552 Merge "Add rectangle block support for MFQE." 2014-12-17 15:10:02 -08:00
Deb Mukherjee
4dd5b88535 Merge "Enhance the end to end psnr tests" 2014-12-17 13:49:13 -08:00
Deb Mukherjee
b62064f338 Enhance the end to end psnr tests
Includes more speed settings and also real-time mode now.

Change-Id: I71c77c4a2471d715c61cc30db092aa053cf012e1
2014-12-17 12:00:20 -08:00
Yunqing Wang
28c9bed70c Merge "Improve the libvpx encoder test driver" 2014-12-17 10:06:33 -08:00
JackyChen
021e244a51 Merge "Use bit_depth in VP9Common as the flag of highbit." 2014-12-17 09:30:32 -08:00
Jingning Han
00d2211929 Merge "Remove reset mode_info array per frame" 2014-12-17 09:24:44 -08:00
Jingning Han
cc8a11d8a1 Merge "Set second ref frame to be NONE in key frame coding" 2014-12-17 09:24:39 -08:00
Yunqing Wang
3666478195 Improve the libvpx encoder test driver
The encoder initialization is called in EncodeFrame(). Therefore,
in the unit tests, the set control is done when video->frame() is 1.
This didn't cause problem since current tests mainly test lag_frame
> 0 case, or no encoding option that needs to allocate memory before
1st frame is used. If use lag_frame = 0 and encoding multiple tiles,
the unit tests crash. The issue is fixed by doing the initialization
before encoding frames.

Change-Id: I43102048f88448bcf27e9c60e0ec06c176b02e5c
2014-12-17 08:49:46 -08:00
JackyChen
b363cedcd1 Use bit_depth in VP9Common as the flag of highbit.
Change-Id: I881aefbe68f9c10bb4629a2a5ee1e42a225d5ab7
2014-12-16 21:45:01 -08:00
James Yu
aeeaa67987 VP9 common for ARMv8 by using NEON intrinsics 15
Re-write
- vp9_lpf_horizontal_4_dual_neon
in vp9_loopfilter_16_neon.c

Change-Id: Ie14f63d352f9564ad01db3939a61d91cf6d21a31
Signed-off-by: James Yu <james.yu@linaro.org>
2014-12-16 20:00:26 -08:00
Johann
ebc1951c7c Merge "Use defines for inline and __builtin_prefetch" 2014-12-16 18:04:04 -08:00
Jingning Han
200d93545e Merge "Fix intra mode update process in vp9_pick_inter_mode" 2014-12-16 17:04:04 -08:00
JackyChen
9931070094 Add rectangle block support for MFQE.
Only for the rectangle blocks larger than 16X16, SAD and Variance are
still based on the internal square blocks.

Change-Id: I3754da1b0254147313f86a0140dbf4f980f06a5a
2014-12-16 16:35:54 -08:00
Johann
4f7060a431 Merge "VP9 common for ARMv8 by using NEON intrinsics 16" 2014-12-16 16:15:48 -08:00
Johann
9022ba3cb9 Merge "Allow building multi-res encoder with --shared" 2014-12-16 15:54:36 -08:00
Johann
5eb9bfa4e3 Merge "Silence -Werror=unused-parameter" 2014-12-16 15:54:29 -08:00
Jingning Han
ccdc448b70 Remove reset mode_info array per frame
The mode_info array was unnecessarily reset to zero every frame
when error resilient mode turned on, given that the mode info
values per block will be assigned during mode search stage.

This commit removes this reset operation. It reduces the runtime
cost on memset operation to 1/3. The overall speed -6 runtime is
reduced by 2%.

Change-Id: I32ecb73338d8995cc0c5147de09357364f13d45b
2014-12-16 15:54:24 -08:00
Jingning Han
01613aa753 Set second ref frame to be NONE in key frame coding
This commit explicitly set the second reference frame type to be
NONE in key frame coding mode. This fixes a subtle dependency of
reference motion vector used by next inter frame on mode_info
reset before key frame coding.

Change-Id: I5ff0359753fdc9992b0bfe889490f7a32d7d5f6a
2014-12-16 15:49:58 -08:00
Johann
2fdbf70d40 Use defines for inline and __builtin_prefetch
These were established for compatibility. Make sure to use them.

Most frequently they manifest as issues on Visual Studio builds.

Change-Id: I39d764d2eb341b999d7a6132cb44b2acfc511160
2014-12-16 15:21:19 -08:00
Frank Galligan
5fdd0f1fe0 Merge "Revert "Revert "Add support for setting byte alignment.""" 2014-12-16 15:14:17 -08:00
James Yu
aa8dd897c1 VP9 common for ARMv8 by using NEON intrinsics 16
Add vp9_reconintra_neon.c
- vp9_v_predictor_4x4_neon
- vp9_v_predictor_8x8_neon
- vp9_v_predictor_16x16_neon
- vp9_v_predictor_32x32_neon
- vp9_h_predictor_4x4_neon
- vp9_h_predictor_8x8_neon
- vp9_h_predictor_16x16_neon
- vp9_h_predictor_32x32_neon
- vp9_tm_predictor_4x4_neon
- vp9_tm_predictor_8x8_neon
- vp9_tm_predictor_16x16_neon
- vp9_tm_predictor_32x32_neon

Change-Id: Ib5d54a4766a1b5127169045659974f33aa98376d
Signed-off-by: James Yu <james.yu@linaro.org>
2014-12-16 12:57:52 -08:00
Johann
73261850fe Allow building multi-res encoder with --shared
Export vpx_codec_enc_init_multi_ver so the vp8 multi res encoder example
can see it when building shared.

Change-Id: Ic5222b1b6d949f39c7e50c3bc58fb76bece2a3f1
2014-12-16 12:47:53 -08:00
Johann
80b344dec5 Silence -Werror=unused-parameter
Cast away remaining issues so that new ones don't get lost in the noise.

Change-Id: Iacd6999b0686ce80f9835730d68db6382690fa92
2014-12-16 12:47:08 -08:00
James Zern
2a64ed9b70 Merge "rtcd: handle armv7s architecture" 2014-12-16 12:22:55 -08:00
James Yu
ba05a4c640 VP9 common for ARMv8 by using NEON intrinsics 19
Delete vp9_dc_only_idct_add_neon.c

The function was merged with vp9_short_idct4x4_1_add (later
vp9_idct4x4_1_add) in d2de1ca and should have been deleted then.

Change-Id: Ie58ba3dd9dc7330a8f1238dd7dd71c9ed4639b94
Signed-off-by: James Yu <james.yu@linaro.org>
2014-12-16 11:14:12 -08:00
JackyChen
603cdcfce5 Merge "Fixed MFQE crash issue for highbit depth." 2014-12-16 11:12:03 -08:00
JackyChen
e7bad92689 Fixed MFQE crash issue for highbit depth.
Check the flags, no MFQE for highbit now. Will add highbit support
latter.

Change-Id: I548c27593e0f47ab7f4c92b45f14fb037dc86591
2014-12-16 10:07:38 -08:00
Jingning Han
581c8dbd33 Merge "Initialize best_tx_size with invalid value" 2014-12-16 10:01:03 -08:00
Yaowu Xu
b60ae45f36 Merge "Prevent decoder from using uninitialized entropy context." 2014-12-16 09:30:24 -08:00
Jingning Han
b47f9c5802 Merge "Use right shift to replace division in vp9_pick_inter_mode" 2014-12-16 09:26:51 -08:00
Debargha Mukherjee
4ebdb4a1b9 Merge "Fix for crash in highbitdepth rt mode" 2014-12-16 06:41:54 -08:00
Jim Bankoski
abc5a66770 Merge "Fix the comments." 2014-12-16 06:25:01 -08:00
Peter de Rivaz
e3d19bfc63 Fix for crash in highbitdepth rt mode
Change 72141 introduced a new use of vp9_avg_4x4.
This call needs to switch to using vp9_highbd_avg_4x4
when performing high bitdepth encodes.

Change-Id: I6a8ba4b62f8a75d0a917b365a55245e2f0438ea1
2014-12-16 10:55:49 +00:00
James Zern
a2fed22b6c Merge "iosbuild.sh: allow for out of tree builds" 2014-12-15 22:04:46 -08:00
James Zern
0e94f3cf06 iosbuild.sh: allow for out of tree builds
Change-Id: Id89ed4a4153dd6e02380c4b8ae8ad6c1f8d21944
2014-12-15 18:41:46 -08:00
James Zern
6ea881cc12 rtcd: handle armv7s architecture
quiets build warnings about implicit (armv6) function declarations

Change-Id: I896a5ef3e367e5bb92777a60d34003eb3a040997
2014-12-15 18:39:51 -08:00
Jingning Han
df3e3ab6ff Fix intra mode update process in vp9_pick_inter_mode
When multiple intra modes are tested, the previous mode info
update process may overwrite the selected best intra mode and make
the final selection use an inter mode. This commit fixes this
issue by moving the mode_info reset outside the intra mode search
loop.

Change-Id: I15ed4288a6b3cb0832104a5e6d5d9a25cd1a5b2b
2014-12-15 17:52:09 -08:00
Johann
1d059fa23e Merge "VP9 common for ARMv8 by using NEON intrinsics 06" 2014-12-15 14:49:33 -08:00
Johann
37ea1e1218 Merge "VP9 common for ARMv8 by using NEON intrinsics 05" 2014-12-15 14:48:53 -08:00
Jingning Han
5c93dca3d3 Merge "Simplify rate-distortion modeling function" 2014-12-15 14:37:19 -08:00
Jingning Han
c2c7596fc7 Initialize best_tx_size with invalid value
If vp9_pick_inter_mode works properly, it should at least check
one coding mode and hence get best_tx_size assigned a valid value.
There is no need to initialize best_tx_size with a legitimate
value before starting the mode search.

Change-Id: Ic0496cd89672ea9c2c512a9bd1da952190af9cba
2014-12-15 12:58:34 -08:00
Jingning Han
83e2c62aba Use right shift to replace division in vp9_pick_inter_mode
Make the variable reduction_fac log2 based and explicitly use
right shift when computing intra_cost_penalty.

Change-Id: I208f1fb879a02debb3b3fc64f9fd06260dcf1c86
2014-12-15 12:48:07 -08:00
Frank Galligan
c4f7079ad4 Revert "Revert "Add support for setting byte alignment.""
This reverts commit 91471d6aad.

Fixes the compile issues if post_proc is enabled.

Change-Id: Ib40a15ce2c194f9b5adfa65a17ab01ddf60f5a59
2014-12-15 12:20:37 -08:00
James Yu
4f856cd7fa VP9 common for ARMv8 by using NEON intrinsics 06
Add vp9_iht8x8_add_neon.c
- vp9_iht8x8_64_add_neon

The assembly did not previously implement tx_type 0
BUG=716

Change-Id: Icfc99dd24f3d59047f9184a7d0c761ba7e3de934
Signed-off-by: James Yu <james.yu@linaro.org>
2014-12-15 12:18:06 -08:00
James Yu
6b71013277 VP9 common for ARMv8 by using NEON intrinsics 05
Add vp9_iht4x4_add_neon.c
- vp9_iht4x4_16_add_neon

The assembly did not previously implement tx_type 0
BUG=715

Change-Id: I60034d1568de034edba45c5cdd13f3d87dbc73b6
Signed-off-by: James Yu <james.yu@linaro.org>
2014-12-15 12:16:19 -08:00
James Zern
8d558f2ca5 Merge "vp9/MACROBLOCKD: reorder struct members" 2014-12-15 11:54:51 -08:00
Jingning Han
eefe869291 Simplify rate-distortion modeling function
Use left shift to replace one multiplication. The computation
outcome remains identical.

Change-Id: I1e1737af0a245de0d2a2bde10f0c171477199fc1
2014-12-15 11:51:16 -08:00