Commit Graph

3441 Commits

Author SHA1 Message Date
Yaowu Xu
1c61e1960d Move vp9_extend.{h,c} from common to encoder
Since they used in encoder only. This commit also re-order includes
for the files that include vp9_extend.h

Change-Id: I929fc113f2135d3198cd1fc6a17434e5a2f8a459
2013-11-18 12:43:36 -08:00
Jingning Han
46ce6ddec4 Merge "Constrain encoder motion search range" 2013-11-18 12:35:34 -08:00
Jingning Han
bbe68fbd2a Constrain encoder motion search range
Explicitly constrain the upper limit of motion search range (in the
unit of full pixel) to be [-1023, +1023]. It is intended to control
the effective motion search range for 4K sequences.

Change-Id: I645539c70885eec0f155781f439d97d333336e88
2013-11-18 11:19:45 -08:00
Yunqing Wang
e3168b0c54 Merge "Do horizontal loopfiltering in parallel" 2013-11-18 10:03:41 -08:00
Jim Bankoski
83eb1975df partition context update speedup
This removes a lot of operations in setting partition context...

Change-Id: I365e6f5607ece85190cb21443988816dfa510ce3
2013-11-17 06:58:08 -08:00
Yunqing Wang
64f728caef Do horizontal loopfiltering in parallel
This patch followed "Rewrite filter_selectively_horiz for parallel
loopfiltering" commit, and added x86 SSE2 optimization to do
16-pixel filtering in parallel. Also, corrected the declaration
of aligned arrays. For 8-pixel-in-parallel case, improved the
calculation of the masks and filters. Updated the threshold loading
since the thresholds were already duplicated. Updated neon C functions
to call neon loopfilters twice.

Using tulip clip, tests showed it gave a ~1.5% decoder speed gain.

Change-Id: Id02638626ac27a4b0e0b09d71792a24c0499bd35
2013-11-15 16:18:43 -08:00
hkuang
7fb5e73897 Merge "Let the idct vp9_idct32x32_34_add = vp9_idct32x32_1024_add on arm until we implenment real vp9_idct32x32_34_add_neon." 2013-11-15 15:45:49 -08:00
Jingning Han
bdc4371174 Take out assertion from inverse transforms
Separate the rounding and right shift operations of forward transform
from those of inverse transform. Take out the assertion check from
inverse transforms. If the transform coefficients were constructed to
cause intermediate steps of inverse transform overflow, the codec will
just let it overflow without breaking the decoding flow.

Change-Id: I73cfc3706c4e840fc543a77cbc4cdb0b05d07730
2013-11-15 15:30:47 -08:00
Yaowu Xu
dc90541563 Merge "Renamed two files" 2013-11-15 15:20:23 -08:00
hkuang
7424492a0b Let the idct vp9_idct32x32_34_add = vp9_idct32x32_1024_add
on arm until we implenment real vp9_idct32x32_34_add_neon.

This issue is due to commit 47665452f0
Merge "Add 32x32 idct function for eob<=34 case".

Change-Id: I56b5f0abc20e7dd1bba521f78a995e85d65ea296
2013-11-15 14:59:16 -08:00
Yaowu Xu
49cbe4580d Renamed two files
from vp9_decodframe.{c,h} to vp9_decodeframe.{c,h}

Change-Id: I21ac4b14fc90246e3f16bd90c52c12d126d791f8
2013-11-15 12:48:43 -08:00
Dmitry Kovalev
5380739a87 Removing vp9_encodeintra.{h, c} files.
There was only one function in *.c file, so moving it to vp9_encodemb.c.

Change-Id: I728859d08b3d6c05c33c1c5b21f0ea1d0e0f83af
2013-11-15 12:17:16 -08:00
Guillaume Martres
17084657e6 vpxenc: add --aq-mode flag to control adaptive quantization
Change-Id: I57e1ad4bed3487df12893ced77c49093f8755706
2013-11-15 19:42:20 +01:00
Dmitry Kovalev
8d7bd4d126 Merge "Cleaning up vp9_loopfilter.c file." 2013-11-15 10:10:59 -08:00
Jingning Han
a9b9f22bcd Merge "Fix coding format in vp9_idct" 2013-11-15 08:59:14 -08:00
Jim Bankoski
e1b6c42eed partition plane context speed up
Removes silly operations inside loop.

Change-Id: I9eeab1e914e715a887f86cf1089de508e2364165
2013-11-15 08:00:43 -08:00
Jim Bankoski
ffb17e2c09 Merge "loop filter assert cleanout" 2013-11-15 07:48:36 -08:00
Dmitry Kovalev
38e6cb8c7b Merge "Cleaning up vp9_tile_common.{h, c} files." 2013-11-14 20:55:01 -08:00
Jingning Han
7637387cf1 Fix coding format in vp9_idct
Change-Id: If97ae16a4478717933345b6b9d5bc1b417b8dd84
2013-11-14 16:05:22 -08:00
Adrian Grange
38144ed8b2 fix scalling bug by buffer auto-reallocation
Change-Id: Ib748eb287520c794631697204da6ebe19523ce95
2013-11-14 15:53:09 -08:00
Dmitry Kovalev
3f9fc6f6f8 Cleaning up vp9_loopfilter.c file.
Change-Id: Ic6770072f80dfb54d2725ed96370d4f243a9f474
2013-11-14 15:04:14 -08:00
Dmitry Kovalev
49fbbf72fa Finally removing txfrm_block_to_raster_block() function.
We only use txfrm_block_to_raster_xy() now.

Change-Id: I4242cd592da99e761041acf9fef1bac3d55a48e1
2013-11-14 13:45:51 -08:00
Dmitry Kovalev
f91ac9b436 Cleaning up vp9_tile_common.{h, c} files.
Change-Id: I9d18f351abe7614107f34f47eeb38a234a9937c9
2013-11-14 13:40:56 -08:00
Dmitry Kovalev
e6b72d0119 Removing unused coefband_trans_8x8plus array from VP9Decompressor.
Change-Id: Ic1367d767705377402ebfec0705f9f553a834400
2013-11-14 13:37:18 -08:00
Jim Bankoski
ef99b7b884 loop filter assert cleanout
Change-Id: I4e2ad4b7342681e6ac236356ef3a4927a54f105b
2013-11-14 12:25:32 -08:00
Dmitry Kovalev
58f754374d Merge "Eliminating usage of txfrm_block_to_raster_block() from encode_block()." 2013-11-14 10:12:54 -08:00
Marco Paniconi
b6ca9d917d Merge "For CBR, keep rate-correction damping factor to 2." 2013-11-14 08:11:43 -08:00
Deb Mukherjee
cfcd5c4f61 Simplifies band-getting with a static array
Simplifies the code by implementing band mapping with static arrays.
A lot of the code complexity introduced in a previous patch
disappears.

Change-Id: Ia3fac36e594fb5ad2d55ae141c58bba4c55c2d28
2013-11-13 22:15:16 -08:00
Dmitry Kovalev
7bfc20ac7a Eliminating usage of txfrm_block_to_raster_block() from encode_block().
Change-Id: I7d11f1b6075a1115cdc2dcd605225b9c9c9b39c7
2013-11-13 19:33:12 -08:00
Dmitry Kovalev
8282c1a68d Merge "Cleaning up decode_coefs() function." 2013-11-13 18:39:14 -08:00
Dmitry Kovalev
11631fec16 Cleaning up decode_coefs() function.
Removing vp9_read_and_apply_sign macro which was used only once.

Change-Id: I6a1625b720d89fc1291c99deccd6638b705f9b06
2013-11-13 16:46:21 -08:00
Marco Paniconi
9977332615 For CBR, keep rate-correction damping factor to 2.
The switch to the rate-correction damping factor
in https://gerrit.chromium.org/gerrit/#/c/67536/ was not conditioned on CBR mode.

Change-Id: I2326704e8ac030a4f7b592dd3fedb94c7dd0644d
2013-11-13 16:14:31 -08:00
Jingning Han
697846d76e Merge "Dual buffer encoding for intra modes" 2013-11-13 15:43:00 -08:00
Jingning Han
fabc783695 Fix an overflow issue in SSE2 forward ADST
The step that sums three input samples could potentially cause the
intermediate result go beyond 16 bit limit, when operating as the
second 1-D transform. This commit fixes the issue.

Change-Id: Iaf512449ac2d25ddd8a806d760afab362c62a516
2013-11-13 15:15:59 -08:00
Dmitry Kovalev
b3c75a2d6c Merge "Replacing raster_block with block in the encoder." 2013-11-13 14:14:27 -08:00
Dmitry Kovalev
26a1ad604f Merge "Removing function pointers from inter prediction." 2013-11-13 13:54:15 -08:00
Jingning Han
b6b9143218 Dual buffer encoding for intra modes
Overall change (using dual buffer scheme for superblocks of both inter
and intra modes) reduces speed 2 runtime:
bluesky_1080p at 6000kbps:   263553ms -> 257441ms
riverbed_1080p at 8000kbps:  233230ms -> 225308ms.

Change-Id: Idf8d70f768a4b0d97b2a8506372c57b7b4022119
2013-11-13 12:57:03 -08:00
Dmitry Kovalev
d1899557eb Merge "Syncing write_modes_{b, sb} implementation with decode_modes_{b, sb}." 2013-11-13 10:47:46 -08:00
Dmitry Kovalev
60d1a52995 Merge "Optimizing set_contexts() function." 2013-11-13 10:01:05 -08:00
Yunqing Wang
8ce0967df8 Merge "Use 1D array to store super block filter levels" 2013-11-13 09:40:14 -08:00
Johann
4da2a8b718 Merge "mips dsp-ase r2 vp9 decoder intra module optimizations (rebase)" 2013-11-13 09:00:09 -08:00
Parag Salasakar
1530a6b77f mips dsp-ase r2 vp9 decoder intra module optimizations (rebase)
Change-Id: Ib27fc4f3dbe01fe8adfa04a61aaba21b3480e75c
2013-11-13 11:17:14 +05:30
Parag Salasakar
248cf6f69f mips dsp-ase r2 vp9 decoder loopfilter module optimizations (rebase)
Change-Id: Ia7f640ca395e8deaac5986f19d11ab18d85eec2d
2013-11-13 10:53:16 +05:30
Dmitry Kovalev
3f3d14e1d3 Moving q_index from MACROBLOCKD to MACROBLOCK.
Moving because q_index is used only by encoder.

Change-Id: I0b96175614ed4fd3d76ee56a0ba36258e1e896f6
2013-11-12 18:13:19 -08:00
Jingning Han
e69461593d Merge "Enable dual buffer rd search and encoding scheme" 2013-11-12 18:11:41 -08:00
Dmitry Kovalev
919eeef5c8 Merge "Calculating transform block offsets (x and y) only once." 2013-11-12 16:57:30 -08:00
Dmitry Kovalev
73a5cbeba4 Merge "Using max_tx_size instead of bsize when possible." 2013-11-12 16:54:30 -08:00
Dmitry Kovalev
3a2ea76469 Merge "Moving {sb, mb, b, ab}_index from MACROBLOCKD to MACROBLOCK." 2013-11-12 15:59:28 -08:00
Dmitry Kovalev
58b004ff64 Merge "Adding const to tree pointer inside vp9_extra_bit struct." 2013-11-12 15:48:07 -08:00
Johann
8dd3905163 Merge "Added optimized vp9_idct32x32_34_add_dspr2" 2013-11-12 15:30:00 -08:00
Dmitry Kovalev
20f34ff0db Adding const to tree pointer inside vp9_extra_bit struct.
Change-Id: I60e02fa3de930ff1f969687ab5af93dee40d86ad
2013-11-12 14:21:15 -08:00
Dmitry Kovalev
ed5a993a97 Calculating transform block offsets (x and y) only once.
Change-Id: I4b5106bdc08fd4551339b968c13428a8f43122e2
2013-11-12 13:55:13 -08:00
Yunqing Wang
ce89309b45 Use 1D array to store super block filter levels
As Jim suggested, 1D array was used to store filter levels instead
of 2D array. This used shift_y in setup_mask directly, and saved
few cycles.

Change-Id: If61ab298784861f1806b1cd396d4e4e2e0f097b9
2013-11-12 12:07:57 -08:00
Deb Mukherjee
a33a84b11a Merge "Removes conditional statements from band getting" 2013-11-12 11:22:21 -08:00
Johann
e72d49a97a Use lowercase 'b' to branch
iOS doesn't recognize B:
bad instruction `B idct32_pass_loop'

Change-Id: I3cf6aede4639f1d9efa97f7962fa287ba6feaaef
2013-11-12 10:41:06 -08:00
Yunqing Wang
17322275dd Merge "Rewrite filter_selectively_horiz for parallel loopfiltering" 2013-11-12 10:20:49 -08:00
Yunqing Wang
7989768766 Merge "Improve loopfilter function" 2013-11-12 10:19:56 -08:00
Deb Mukherjee
5ade423774 Removes conditional statements from band getting
Implements scan order to band map with arrays in both the encoder
and decoder to remove conditional statements.

Encoding seems to be about 1% faster at speed 0, tested on football.
Decoding seems to be about 0.5-1% faster on a set of 25 videos.

Change-Id: Idb233ca0b9e0efd790e30880642e8717e1c5c8dd
2013-11-12 10:13:27 -08:00
Dmitry Kovalev
e5ed605f01 Merge "Removing redundant assignment." 2013-11-11 18:37:22 -08:00
Dmitry Kovalev
50f97cf7fb Removing function pointers from inter prediction.
Removing foreach_predicted_block_visitor and calling build_inter_predictors
directly.

Change-Id: I11bb3c872b99b47c2680b01b0dbcc01c558c4a2b
2013-11-11 18:37:00 -08:00
Jingning Han
34b6abefa2 Enable dual buffer rd search and encoding scheme
This commit enables the dual buffer rate-distortion optimization
and encoding scheme. It stacks the original transform coefficients,
quantized levels, and reconstructed coefficients, in the rate-
distortion optimization search process, hence eliminates the need
to re-run residual generation, forward transform, and quantization
in the encoding stage.

Change-Id: I011bfad3a59a380a869ee552e91dae0394ec492e
2013-11-11 18:32:55 -08:00
Jingning Han
e5741c56d1 Merge "Allocate dual buffer sets for encoding" 2013-11-11 18:00:57 -08:00
Dmitry Kovalev
42b1f62085 Removing redundant assignment.
xd->mi_8x8 is assigned inside set_offsets() for each prediction block.

Change-Id: I20e5974a9eaf105e5a04fc7f99b7a93bd50e3d0a
2013-11-11 17:39:43 -08:00
Dmitry Kovalev
3740d67d76 Syncing write_modes_{b, sb} implementation with decode_modes_{b, sb}.
Change-Id: Iaee740ec3bfb2b5328c24f4641c285e5a4a046dc
2013-11-11 17:29:31 -08:00
Yunqing Wang
b45438181c Rewrite filter_selectively_horiz for parallel loopfiltering
Added loop filter mask checking, and made the caller function
ready for implementation of parallel loopfiltering in horizontal
direction.

Next, we need to go through the loopfilter functions (both c and
optimized versions), and provide 16-byte wide loopfiltering for
each filter type.

Change-Id: Ifef47e7ef9086ebc2fd6ca7ede8f27c9bbf79e66
2013-11-11 17:06:01 -08:00
Dmitry Kovalev
4e39d530f0 Merge "Cleaning up joint_motion_search function." 2013-11-11 16:34:39 -08:00
Jingning Han
3b3aea6834 Allocate dual buffer sets for encoding
Allocate memory space of dual buffer sets that store the coeff, qcoeff,
dqcoeff, and eobs. Connect the pointers of macroblock_plane and
macroblockd_plane to the actual buffer in use accordingly.

Change-Id: I2f0b5f482ca879fae39095013eaf8901db20a5a4
2013-11-11 16:24:39 -08:00
Dmitry Kovalev
14f2cf1757 Cleaning up joint_motion_search function.
Change-Id: I70a0878b23bda0ac3ff8733b4c96d5c636bc551c
2013-11-11 16:04:02 -08:00
Dmitry Kovalev
3551e25099 Moving {sb, mb, b, ab}_index from MACROBLOCKD to MACROBLOCK.
We use {sb, mb, b, ab}_index only inside encoder, so moving them into
appropriate data structure.

Change-Id: Ib5c1036716354d9d321e11a60c1634c1cb8f9716
2013-11-11 15:58:57 -08:00
Jingning Han
d8b4c79270 Decouple macroblockd_plane buffer usage
Make the macroblockd_plane contain dynamic buffer pointers instead
static pointers to the memory space allocated therein. The decoder
uses the buffer allocated in pbi, while encoder will use a dual
buffer approach for rate-distortion optimization search.

Change-Id: Ie6f24be2dcda35df7c15b4014e5ccf236fb3f76c
2013-11-11 15:26:10 -08:00
Dmitry Kovalev
94d4add1f7 Replacing raster_block with block in the encoder.
We only used "ib" to call get_scan() function, which in turn calls
get_tx_type_4x4() function. The latter one only needs block index if
bsize < BLOCK_8X8 -- under that condition raster_block == block.

Change-Id: I697306a0c3cf937acdd4f5e623d4367c5acc0b2f
2013-11-11 15:18:48 -08:00
hkuang
c689a126ed Fix a bug in the assembly code.
Change-Id: Ic416e3f8a11e82ee298e6f709b2119a9ddf1e2f8
2013-11-11 12:49:12 -08:00
Dmitry Kovalev
ec8128e27f Merge "Replacing (raster_block >> tx_size) with (block >> (tx_size << 1))." 2013-11-11 12:27:42 -08:00
Dmitry Kovalev
c53a9c70fb Merge "Localizing NEARESTMV special cases in the code." 2013-11-11 11:12:06 -08:00
Dmitry Kovalev
f6baa62cd8 Merge "Cleaning up vp9_quantize_b_c() function." 2013-11-11 11:02:32 -08:00
Dmitry Kovalev
3aa4b42a35 Merge "Cleaning up read_mv_probs() function." 2013-11-11 11:01:35 -08:00
Dmitry Kovalev
974a27131e Merge "Adding read_reference_mode() function." 2013-11-11 11:00:51 -08:00
Yaowu Xu
af2559c0d6 Merge "[BITSTREAM]Fix row tile mode_info pointer setup" 2013-11-09 13:59:19 -08:00
Yaowu Xu
cae7e0741a [BITSTREAM]Fix row tile mode_info pointer setup
This commit fixes the assignment of mode_info pointer per tile. It
makes recognition of tiles in both row and column formats and properly
arrange the use of mode_info.

The bug was first introduced in
I6226456dd11f275fa991e4a7a930549da6675915
https://gerrit.chromium.org/gerrit/#/c/67492/

Change-Id: Ie12cd209f53241513728c461ee3d7b9599ddb860
2013-11-08 22:06:53 -08:00
Yaowu Xu
ee1e4e645a Merge "Correct a couple of typos" 2013-11-08 16:17:38 -08:00
Dmitry Kovalev
22a001988b Optimizing set_contexts() function.
Inlining set_contexts_on_border() into set_contexts(). The only difference
is the additional check that "has_eob != 0" in addition to
"xd->mb_to_right_edge < 0" and "xd->mb_to_right_edge < 0". If has_eob == 0
then memset does the right thing and works faster.

Change-Id: I5206f767d729f758b14c667592b7034df4837d0e
2013-11-08 12:44:56 -08:00
Yaowu Xu
a596975eb6 Correct a couple of typos
Change-Id: Ic470c6c9ce27b615c9645b9cb0d67526417bc374
2013-11-08 12:43:51 -08:00
Yunqing Wang
e731b2ba2c Merge "Improve vp9_idct4x4_1_add_sse2" 2013-11-08 12:00:36 -08:00
Yunqing Wang
49cf335e7f Improve loopfilter function
This patch continued the work done in "Rewrite loop_filter_info_n
struct"(commit:00dbd369c70270428d56da6d15ea5486fc821c52) to further
improve loopfilter function.

1. Instead of storing pointers to thresholds, store loopfilter
levels within 64x64 SB;
2. Since loopfilter levels are already calculated in setup_mask,
we don't need call build_lfi to look up them again. Just save
loopfilter levels in setup_mask.
3. Reorganized and simplified filter_block_plane().

Tests showed a ~0.8% decoder speedup.

Change-Id: I723c7779738bbc2afcb9afa2c6f78580ee6c3af7
2013-11-08 11:48:31 -08:00
Yaowu Xu
814112d0f6 Merge "Disable zeroblock forcing for lossless coding mode" 2013-11-08 11:11:39 -08:00
Dmitry Kovalev
614effc0f6 Merge "Unifying tile decoding for both direct and inverse tile order." 2013-11-08 10:59:02 -08:00
Yaowu Xu
a4a5a210cb Disable zeroblock forcing for lossless coding mode
This to make sure that prediction residue always get coded in lossless
mode.

This commit also fixed lossless unit test

Change-Id: I537726ee55328d4e4cf0a0196393a67e12bfcde1
2013-11-08 10:32:44 -08:00
Yunqing Wang
283427c053 Merge "Remove TEXTREL from 32bit encoder" 2013-11-08 08:26:30 -08:00
Paul Wilkins
0ed606fd40 Merge "Removed unused rate parameter." 2013-11-08 04:24:51 -08:00
Dmitry Kovalev
d28f30ef4e Replacing (raster_block >> tx_size) with (block >> (tx_size << 1)).
The new expression is much more logical than previous one. Surprisingly
both expressions give exactly the same set of dependent values
-- have_top, have_left, have_right -- in vp9_predict_intra_block.

Change-Id: I63eb1b592b8c37883b3a0dbb1f3daa271e446109
2013-11-07 15:26:57 -08:00
hkuang
a6462990e6 Merge "Add back vp9_short_idct32x32_1_add_neon which is deleted in cleanup I63df79a13cf62aa2c9360a7a26933c100f9ebda3." 2013-11-07 14:42:29 -08:00
Yunqing Wang
d7289658fb Remove TEXTREL from 32bit encoder
This patch fixed the issue reported in "Issue 655: remove textrel's
from 32-bit vp9 encoder". The set of vp9_subpel_variance functions
that used x86inc.asm ABI didn't build correctly for 32bit PIC. The
fix was carefully done under the situation that there was not
enough registers.

After the change, we got
$ eu-findtextrel libvpx.so
eu-findtextrel: no text relocations reported in 'libvpx.so'

Change-Id: I1b176311dedaf48eaee0a1e777588043c97cea82
2013-11-07 13:39:40 -08:00
Jingning Han
abdefeaa89 Merge "Fix the variable naming in encode_block" 2013-11-07 11:39:04 -08:00
Deb Mukherjee
6cf3b98ac2 Merge "Miscelleneous changes in detokenize for speed" 2013-11-07 09:40:07 -08:00
Jingning Han
e91d770511 Fix the variable naming in encode_block
The term x represents macroblock pointer across encode_block. Change
the two local variable names to avoid confusion.

Change-Id: Ic732e73023525d673c0a678ed2708ac1edf5a3f9
2013-11-07 08:59:14 -08:00
Paul Wilkins
84b3b03705 Removed unused rate parameter.
Change-Id: I6e4a266fdbad1d222eb45d45b67bbb82d091821a
2013-11-07 09:59:45 +00:00
Dmitry Kovalev
672ba3ddf5 Unifying tile decoding for both direct and inverse tile order.
Now tile decoding consists of two stages:
1. Find tile buffer start and its size, put this info into tile_buffers.
2. Decode each tile based on information from tile_buffers.

It seems that stage 1 can also be reused by multithreaded tile decoder.

Change-Id: If0cdaefdd6d10bb41c63561346c9ae4cfac081dd
2013-11-06 18:15:33 -08:00
Ivan Maltz
741c14fcf0 Merge "Move SVC per-frame loop from sample app into libvpx proper" 2013-11-06 17:24:05 -08:00
Dmitry Kovalev
af36c1f2e7 Merge "Using pd->dqcoeff instead of pd->qcoeff in the decoder." 2013-11-06 16:59:57 -08:00
Dmitry Kovalev
a1dc97beb1 Using pd->dqcoeff instead of pd->qcoeff in the decoder.
It is more logical to use dqcoeff buffer to put there *dequantized*
transform coefficients (inside inverse_transform_block and
decode_coefs functions). Dequantization happens inside WRITE_COEF_CONTINUE
macro.

qcoeff buffer should be only used in the encoder for *quantized*
transform coefficients.

Change-Id: Ifd54bef272bbf5311ced6669c4f1079f998af5d7
2013-11-06 16:14:45 -08:00