platform/upstream/libvpx.git
17 months agoMove TPL to a new file
Yunqing Wang [Mon, 6 Feb 2023 22:48:34 +0000 (14:48 -0800)]
Move TPL to a new file

This is a refactoring CL.

Change-Id: Ic8c1575601d27f14ecd1b1bf0a038e447eaae458

17 months agoMerge "Fix uninitialized mesh feature for BEST mode" into main
Yunqing Wang [Fri, 3 Feb 2023 23:22:58 +0000 (23:22 +0000)]
Merge "Fix uninitialized mesh feature for BEST mode" into main

17 months agoSet _img->bit_depth in y4m_input_fetch_frame()
Wan-Teh Chang [Fri, 3 Feb 2023 22:07:09 +0000 (14:07 -0800)]
Set _img->bit_depth in y4m_input_fetch_frame()

This is a port of
https://aomedia-review.googlesource.com/c/aom/+/169961.

Change-Id: I2aa0d12cafde0c73448bf8c57eab0cd92e846468

17 months agoFix uninitialized mesh feature for BEST mode
Yunqing Wang [Fri, 3 Feb 2023 00:30:09 +0000 (16:30 -0800)]
Fix uninitialized mesh feature for BEST mode

At BEST encoding mode, the mesh search range wasn't initialized for
non FC_GRAPHICS_ANIMATION content type, which actually/mistakenly
used speed 0's setting. Fixed it by adding the initialization.

There were 2 ways to fix this. Patchset 1 set to use speed 0's setting
for non FC_GRAPHICS_ANIMATION type. This didn't change BEST mode's
encoding results much, and only a couple of clips' results were changed.

Borg result for BEST mode:
         avg_psnr:  ovr_psnr:  ssim:  encoding_spdup:
lowres2:  -0.004     -0.003   -0.000    0.030
midres2:  -0.006     -0.009   -0.012    0.033
hdres2:    0.002      0.002    0.004    0.015

Patchset 2 set to use BEST's setting for non FC_GRAPHICS_ANIMATION type.
However, the majority of test clips' BDrate got changed up to
~0.5% (gain or loss), and overall it didn't give better performance
than patchset 1. So, we chose to use patchset 1.

Change-Id: Ibbf578dad04420e6ba22cb9a3ddec137a7e4deef

17 months agovp9_diamond_search_sad_neon: use DECLARE_ALIGNED
James Zern [Wed, 1 Feb 2023 21:27:06 +0000 (13:27 -0800)]
vp9_diamond_search_sad_neon: use DECLARE_ALIGNED

rather than the gcc specific __attribute__((aligned())); fixes build
targeting ARM64 windows.

Bug: webm:1788
Change-Id: I2210fc215f44d90c1ce9dee9b54888eb1b78c99e

17 months agoMerge "Use load_unaligned mem_neon.h helpers in SAD and SAD4D" into main
James Zern [Tue, 31 Jan 2023 21:20:16 +0000 (21:20 +0000)]
Merge "Use load_unaligned mem_neon.h helpers in SAD and SAD4D" into main

17 months agoUse load_unaligned mem_neon.h helpers in SAD and SAD4D
Jonathan Wright [Tue, 31 Jan 2023 13:32:33 +0000 (13:32 +0000)]
Use load_unaligned mem_neon.h helpers in SAD and SAD4D

Use the load_unaligned helper functions in mem_neon.h to load strided
sequences of 4 bytes where alignment is not guaranteed in the Neon
SAD and SAD4D paths.

Change-Id: I941d226ef94fd7a633b09fc92165a00ba68a1501

17 months agoFix unsigned integer overflow in sse computation
Cheng Chen [Mon, 30 Jan 2023 19:51:58 +0000 (11:51 -0800)]
Fix unsigned integer overflow in sse computation

Basically port the fix from libaom:
https://aomedia-review.googlesource.com/c/aom/+/169361

Change-Id: Id06a5db91372037832399200ded75d514e096726

17 months agoMerge "Refactor 8x8 16-bit Neon transpose functions" into main
James Zern [Mon, 30 Jan 2023 19:30:45 +0000 (19:30 +0000)]
Merge "Refactor 8x8 16-bit Neon transpose functions" into main

17 months agoRefactor Neon implementation of SAD4D functions
Salome Thirot [Fri, 27 Jan 2023 16:16:16 +0000 (16:16 +0000)]
Refactor Neon implementation of SAD4D functions

Refactor and optimize the Neon implementation of SAD4D functions -
effectively backporting these libaom changes[1,2].

[1] https://aomedia-review.googlesource.com/c/aom/+/162181
[2] https://aomedia-review.googlesource.com/c/aom/+/162183

Change-Id: Icb04bd841d86f2d0e2596aa7ba86b74f8d2d360b

17 months agoMerge "Add encoder component timing information" into main
Yunqing Wang [Sat, 28 Jan 2023 00:27:57 +0000 (00:27 +0000)]
Merge "Add encoder component timing information" into main

17 months agoAdd encoder component timing information
Yunqing Wang [Fri, 27 Jan 2023 01:20:54 +0000 (17:20 -0800)]
Add encoder component timing information

Change-Id: Iaa5b73a9593ecfd74b6426ed47d2b529ec7ae2b5

17 months agoRefactor 8x8 16-bit Neon transpose functions
Gerda Zsejke More [Thu, 26 Jan 2023 15:12:55 +0000 (16:12 +0100)]
Refactor 8x8 16-bit Neon transpose functions

Refactor the Neon implementation of transpose_s16_8x8(q) and
transpose_u16_8x8 so that the final step compiles to 8 ZIP1/ZIP2
instructions as opposed to 8 EXT, MOV pairs. This change removes 8
instructions per call to transpose_s16_8x8(q), transpose_u16_8x8
where the result stays in registers for further processing - rather
than being stored to memory - like in vpx_hadamard_8x8_neon, for
example.

This is a backport of this libaom patch[1].
[1] https://aomedia-review.googlesource.com/c/aom/+/169426

Change-Id: Icef3e51d40efeca7008e1c4fc701bf39bd319c88

17 months agoMerge "Fix per frame qp for temporal layers" into main
Jerome Jiang [Thu, 26 Jan 2023 21:31:14 +0000 (21:31 +0000)]
Merge "Fix per frame qp for temporal layers" into main

17 months agoFix per frame qp for temporal layers
Jerome Jiang [Thu, 26 Jan 2023 00:25:12 +0000 (19:25 -0500)]
Fix per frame qp for temporal layers

Also add tests with fixed temporal layering mode.

Change-Id: If516fe94e3fb7f5a745821d1788bfe6cf90edaac

17 months agoMerge "Refactor Neon implementation of SAD functions" into main
James Zern [Thu, 26 Jan 2023 03:26:38 +0000 (03:26 +0000)]
Merge "Refactor Neon implementation of SAD functions" into main

17 months agoMerge "[NEON] Add Highbd FHT 8x8/16x16 functions" into main
James Zern [Thu, 26 Jan 2023 03:23:31 +0000 (03:23 +0000)]
Merge "[NEON] Add Highbd FHT 8x8/16x16 functions" into main

17 months agoRefactor Neon implementation of SAD functions
Salome Thirot [Tue, 24 Jan 2023 14:27:14 +0000 (14:27 +0000)]
Refactor Neon implementation of SAD functions

Refactor and optimize the Neon implementation of SAD functions -
effectively backporting these libaom changes[1,2,3].

[1] https://aomedia-review.googlesource.com/c/aom/+/161921
[2] https://aomedia-review.googlesource.com/c/aom/+/161923
[3] https://aomedia-review.googlesource.com/c/aom/+/166963

Change-Id: I2d72fd0f27d61a3e31a78acd33172e2afb044cb8

17 months ago[NEON] Add Highbd FHT 8x8/16x16 functions
Konstantinos Margaritis [Tue, 24 Jan 2023 20:48:06 +0000 (20:48 +0000)]
[NEON] Add Highbd FHT 8x8/16x16 functions

In total this gives about 9% extra performance for both rt/best
profiles.
Furthermore, add transpose_s32 16x16 function

Change-Id: Ib6f368bbb9af7f03c9ce0deba1664cef77632fe2

17 months agoSkip calculating internal stats when frame dropped
Jerome Jiang [Tue, 24 Jan 2023 19:08:17 +0000 (14:08 -0500)]
Skip calculating internal stats when frame dropped

Bug: webm:1771
Change-Id: I30cd5b7ec0945b521a1cc03999d39ec6a25f1696

17 months agoSpecialize Neon averaging subpel variance by filter value
Salome Thirot [Fri, 20 Jan 2023 11:42:06 +0000 (11:42 +0000)]
Specialize Neon averaging subpel variance by filter value

Use the same specialization for averaging subpel variance functions
as used for the non-averaging variants. The rationale for the
specialization is as follows:

The optimal implementation of the bilinear interpolation depends on
the filter values being used. For both horizontal and vertical
interpolation this can simplify to just taking the source values, or
averaging the source and reference values - which can be computed
more easily than a bilinear interpolation with arbitrary filter
values.

This patch introduces tests to find the most optimal bilinear
interpolation implementation based on the filter values being used.
This new specialization is only used for larger block sizes

This is a backport of this libaom change[1].

After this change, the only differences between the code in libvpx and
libaom are due to libvpx being compiled with ISO C90, which forbids
mixing declarations and code [-Wdeclaration-after-statement].

[1] https://aomedia-review.googlesource.com/c/aom/+/166962

Change-Id: I7860c852db94a7c9c3d72ae4411316685f3800a4

17 months agoRefactor Neon averaging subpel variance functions
Salome Thirot [Fri, 20 Jan 2023 11:21:02 +0000 (11:21 +0000)]
Refactor Neon averaging subpel variance functions

Merge the computation of vpx_comp_avg_pred into the second pass of the
bilinear filter - avoiding the overhead of loading and storing the
entire block again.

This is a backport of this libaom change[1].

[1] https://aomedia-review.googlesource.com/c/aom/+/166961

Change-Id: I9327ff7382a46d50c42a5213a11379b957146372

17 months agoSpecialize Neon subpel variance by filter value for large blocks
Salome Thirot [Fri, 20 Jan 2023 10:35:34 +0000 (10:35 +0000)]
Specialize Neon subpel variance by filter value for large blocks

The optimal implementation of the bilinear interpolation depends on
the filter values being used. For both horizontal and vertical
interpolation this can simplify to just taking the source values, or
averaging the source and reference values - which can be computed
more easily than a bilinear interpolation with arbitrary filter
values.

This patch introduces tests to find the most optimal bilinear
interpolation implementation based on the filter values being used.
This new specialization is only used for larger block sizes
(>= 16x16) as we need to be doing enough work to make the cost of
finding the optimal implementation worth it.

This is a backport of this libaom change[1].

After this change, the only differences between the code in libvpx and
libaom are due to libvpx being compiled with ISO C90, which forbids
mixing declarations and code [-Wdeclaration-after-statement].

[1] https://aomedia-review.googlesource.com/c/aom/+/162463

Change-Id: Ia818e148f6fd126656e8411d59c184b55dd43094

17 months agoRefactor Neon subpel variance functions
Salome Thirot [Thu, 19 Jan 2023 18:02:52 +0000 (18:02 +0000)]
Refactor Neon subpel variance functions

Refactor the Neon implementation of the sub-pixel variance bilinear
filter helper functions - effectively backporting this libaom patch[1].

[1] https://aomedia-review.googlesource.com/c/aom/+/162462

Change-Id: I3dee32e8125250bbeffeb63d1fef5da559bacbf1

18 months agoMerge "Add codec control to set per frame QP" into main
Jerome Jiang [Fri, 20 Jan 2023 17:14:04 +0000 (17:14 +0000)]
Merge "Add codec control to set per frame QP" into main

18 months agoAdd codec control to set per frame QP
Jerome Jiang [Thu, 12 Jan 2023 20:58:00 +0000 (15:58 -0500)]
Add codec control to set per frame QP

Use case is for 1 pass encoding.
Forces max_quantizer = min_quantizer and aq-mode = 0.
Applicalble to spatial layers, where user may set
the QP per spatial layer.

Change-Id: Idfcb7daefde94c475ed1bc0eb8af47c9f309110b

18 months agoMerge "Refactor Neon implementation of variance functions" into main
James Zern [Thu, 19 Jan 2023 19:44:43 +0000 (19:44 +0000)]
Merge "Refactor Neon implementation of variance functions" into main

18 months ago*/Android.mk: add a check for NDK_ROOT
James Zern [Thu, 19 Jan 2023 03:19:01 +0000 (19:19 -0800)]
*/Android.mk: add a check for NDK_ROOT

This simplifies integration with the Android platform and avoids the
files from being used when a non-NDK build is performed. In that case
Android.bp is preferred.

Change-Id: I803912146dac788b7f0af27199c7613cabbc9fa0

18 months agoRefactor Neon implementation of variance functions
Salome Thirot [Mon, 16 Jan 2023 16:44:04 +0000 (16:44 +0000)]
Refactor Neon implementation of variance functions

Refactor and optimize the Neon implementation of variance functions -
effectively backporting these libaom changes[1,2].

After this change, the only differences between the code in libvpx and
libaom are due to libvpx being compiled with ISO C90, which forbids
mixing declarations and code [-Wdeclaration-after-statement].

[1] https://aomedia-review.googlesource.com/c/aom/+/162241
[2] https://aomedia-review.googlesource.com/c/aom/+/162262

Change-Id: Ia4e8fff4d53297511d1a1e43bca8053bf811e551

18 months agoMerge "Fix to segfault for external resize test in vp9" into main
Marco Paniconi [Wed, 18 Jan 2023 02:04:18 +0000 (02:04 +0000)]
Merge "Fix to segfault for external resize test in vp9" into main

18 months agoFix to segfault for external resize test in vp9
Marco Paniconi [Sat, 14 Jan 2023 03:46:10 +0000 (19:46 -0800)]
Fix to segfault for external resize test in vp9

Failure occurs for 1 pass non-realtime mode at speed 0.
Due to speed feautre rd_ml_partition.var_pruning, which
doesn't check for scaled reference in simple_motion_search().

Bug: webm:1768

Change-Id: Iddcb56033bac042faebb5196eed788317590b23f

18 months agovariance_test.cc: Enable HBDMse speed test.
Scott LaVarnway [Fri, 13 Jan 2023 15:30:07 +0000 (07:30 -0800)]
variance_test.cc: Enable HBDMse speed test.

Change-Id: If0226307a6efd704f8a35cb986f570304d698b95

18 months agoMerge "variance_test.cc: Enable VpxHBDMseTest for C and SSE2." into main
Scott LaVarnway [Fri, 13 Jan 2023 13:36:15 +0000 (13:36 +0000)]
Merge "variance_test.cc: Enable VpxHBDMseTest for C and SSE2." into main

18 months agovariance_test.cc: Enable VpxHBDMseTest for C and SSE2.
Scott LaVarnway [Thu, 12 Jan 2023 19:03:28 +0000 (11:03 -0800)]
variance_test.cc: Enable VpxHBDMseTest for C and SSE2.

Change-Id: I66c0db6c605876d6757684fd715614881ca261e7

18 months agoMerge changes Ifbf46768,If19f5872 into main
James Zern [Thu, 12 Jan 2023 18:41:27 +0000 (18:41 +0000)]
Merge changes Ifbf46768,If19f5872 into main

* changes:
  Implement vertical convolutions using Neon USDOT instruction
  Implement horizontal convolutions using Neon USDOT instruction

18 months agoImplement vertical convolutions using Neon USDOT instruction
Jonathan Wright [Wed, 18 May 2022 15:58:50 +0000 (16:58 +0100)]
Implement vertical convolutions using Neon USDOT instruction

Add additional AArch64 paths for vpx_convolve8_vert_neon and
vpx_convolve8_avg_vert_neon that use the Armv8.6-A USDOT (mixed-sign
dot-product) instruction. The USDOT instruction takes an 8-bit
unsigned operand vector and a signed 8-bit operand vector to produce
a signed 32-bit result. This is helpful because convolution filters
often have both positive and negative values, while the 8-bit pixel
channel data being filtered is all unsigned. As a result, the USDOT
convolution paths added here do not have to do the "transform the
pixel channel data to [-128, 128) and correct for it later" dance
that we have to do with the SDOT paths.

The USDOT instruction is optional from Armv8.2 to Armv8.5 but
mandatory from Armv8.6 onwards. The availability of the USDOT
instruction is indicated by the feature macro
__ARM_FEATURE_MATMUL_INT8. The SDOT paths are retained for use on
target CPUs that do not implement the USDOT instructions.

Change-Id: Ifbf467681dd53bb1d26e22359885e6edde3c5c72

18 months agoImplement horizontal convolutions using Neon USDOT instruction
Jonathan Wright [Wed, 18 May 2022 13:14:56 +0000 (14:14 +0100)]
Implement horizontal convolutions using Neon USDOT instruction

Add additional AArch64 paths for vpx_convolve8_horiz_neon and
vpx_convolve8_avg_horiz_neon that use the Armv8.6-A USDOT (mixed-sign
dot-product) instruction. The USDOT instruction takes an 8-bit
unsigned operand vector and a signed 8-bit operand vector to produce
a signed 32-bit result. This is helpful because convolution filters
often have both positive and negative values, while the 8-bit pixel
channel data being filtered is all unsigned. As a result, the USDOT
convolution paths added here do not have to do the "transform the
pixel channel data to [-128, 128) and correct for it later" dance
that we have to do with the SDOT paths.

The USDOT instruction is optional from Armv8.2 to Armv8.5 but
mandatory from Armv8.6 onwards. The availability of the USDOT
instruction is indicated by the feature macro
__ARM_FEATURE_MATMUL_INT8. The SDOT paths are retained for use on
target CPUs that do not implement the USDOT instructions.

Change-Id: If19f5872c3453458a8cfb7c7d2be82a2c0eab46a

18 months agobuild: replace egrep with grep -E
James Zern [Tue, 10 Jan 2023 21:49:15 +0000 (13:49 -0800)]
build: replace egrep with grep -E

avoids a warning on some platforms:
egrep: warning: egrep is obsolescent; using grep -E

Bug: webm:1786
Change-Id: Ia434297731303aacb0b02cf3dcbfd8e03936485d
Fixed: webm:1786

18 months agoUse Neon load/store helper functions consistently
Jonathan Wright [Thu, 5 Jan 2023 15:04:53 +0000 (15:04 +0000)]
Use Neon load/store helper functions consistently

Define all Neon load/store helper functions in mem_neon.h and use
them consistently in Neon convolution functions.

Change-Id: I57905bc0a3574c77999cf4f4a73442c3420fa2be

18 months agoUse lane-referencing intrinsics in Neon convolution kernels
Jonathan Wright [Thu, 5 Jan 2023 12:20:03 +0000 (12:20 +0000)]
Use lane-referencing intrinsics in Neon convolution kernels

The Neon convolution helper functions take a pointer to a filter and
load the 8 values into a single Neon register. For some reason,
filter values 3 and 4 are then duplicated into their own separate
registers.

This patch modifies these helper functions so that they access filter
values 3 and 4 via the lane-referencing versions of the various Neon
multiply instructions. This reduces register pressure and tidies up
the source code quite a bit.

Change-Id: Ia4aeee8b46fe218658fb8577dc07ff04a9324b3e

19 months agoRemove references to deprecated NumPy type aliases
Jerome Jiang [Wed, 21 Dec 2022 16:13:40 +0000 (11:13 -0500)]
Remove references to deprecated NumPy type aliases

This change replaces references to a number of deprecated NumPy type
aliases (np.bool, np.int, np.float, np.complex, np.object, np.str)
with their recommended replacement
(bool, int, float, complex, object, str).

NumPy 1.24 drops the deprecated aliases
so we must remove uses before updating NumPy.

Change-Id: I9f5dfcbb11fe6534fce358054f210c7653f278c3

19 months ago[x86]: Add vpx_highbd_comp_avg_pred_sse2().
Scott LaVarnway [Tue, 20 Dec 2022 23:43:44 +0000 (15:43 -0800)]
[x86]: Add vpx_highbd_comp_avg_pred_sse2().

C vs SSE2

4x4: 3.38x
8x8: 3.45x
16x16: 2.06x
32x32: 2.19x
64x64: 1.39x

Change-Id: I46638fe187b49a78fee554114fac51c485d74474

19 months agoAdd vpx_highbd_comp_avg_pred_c() test.
Scott LaVarnway [Fri, 16 Dec 2022 18:21:00 +0000 (10:21 -0800)]
Add vpx_highbd_comp_avg_pred_c() test.

Change-Id: I6b2c3379c49a62e56e5ac56fd4782a50b3c4e12a

19 months agoMerge "rc-svc: Add tests for dynamic svc in external RC" into main
Marco Paniconi [Wed, 14 Dec 2022 17:08:21 +0000 (17:08 +0000)]
Merge "rc-svc: Add tests for dynamic svc in external RC" into main

19 months agorc-svc: Add tests for dynamic svc in external RC
Marco Paniconi [Wed, 7 Dec 2022 08:17:22 +0000 (00:17 -0800)]
rc-svc: Add tests for dynamic svc in external RC

Test to verify RC for going down and back up in
spatial layers. Going back up has an issue so added
a TODO.

Make the test more flexible to handle dynamic layers.
Test for dyanmic change in temporal layers to follow.

Change-Id: Ic5542f7b274135277429e116f56ba54e682e96a0

19 months agoAdd additional ARM targets for Visual Studio.
Anton Venema [Tue, 13 Dec 2022 18:27:37 +0000 (10:27 -0800)]
Add additional ARM targets for Visual Studio.

configure: Add an armv7-win32-vs16 target
configure: Add an armv7-win32-vs17 target
configure: Add an arm64-win64-vs16 target
configure: Add an arm64-win64-vs17 target

Change-Id: I11d6cd6e51f7703939d6fd3fc6a7469591e3b09d

19 months agoMerge "L2E: Add a new interface to control rdmult" into main
Cheng Chen [Tue, 13 Dec 2022 01:24:00 +0000 (01:24 +0000)]
Merge "L2E: Add a new interface to control rdmult" into main

19 months ago[x86]: Add vpx_highbd_subtract_block_avx2().
Scott LaVarnway [Tue, 6 Dec 2022 21:13:30 +0000 (13:13 -0800)]
[x86]: Add vpx_highbd_subtract_block_avx2().

Up to 4x faster than "sse2 vectorized C".

Change-Id: Ie9b3c12a437c5cddf92c4d5349c4f659ca6b82ea

19 months agoAdd vpx highbd subtract test.
Scott LaVarnway [Tue, 6 Dec 2022 22:18:03 +0000 (14:18 -0800)]
Add vpx highbd subtract test.

Change-Id: I069ae0fe22bfc82ad5083df85a7fdf9058a285eb

19 months agoL2E: Add a new interface to control rdmult
Cheng Chen [Sat, 3 Dec 2022 02:04:32 +0000 (18:04 -0800)]
L2E: Add a new interface to control rdmult

Allow external model to control frame rdmult.

A function is called per frame to get the value of rdmult from
the external model.

The external rdmult will overwrite libvpx's default rdmult unless
a reserved value is selected.

A unit test is added to test when the default rdmult value is set.

Change-Id: I2f17a036c188de66dc00709beef4bf2ed86a919a

19 months agorc-rtc: Test for periodic key in SVC external RC
Marco Paniconi [Mon, 5 Dec 2022 22:30:40 +0000 (14:30 -0800)]
rc-rtc: Test for periodic key in SVC external RC

This test catches the fix merged in here:
https://chromium-review.googlesource.com/c/webm/libvpx/+/4022904

Change-Id: Ib68fbcba694b5d465a9faf3ca7d6880bfe8eabb3

19 months agorc-rtc: Remove frame_flags_ change in svc ratectril rtc test
Marco Paniconi [Mon, 5 Dec 2022 19:54:33 +0000 (11:54 -0800)]
rc-rtc: Remove frame_flags_ change in svc ratectril rtc test

SVC test is only in CBR and the frame_flags are
set by the SVC pattern, so we shouldn't undo them
for svc mode.

Change-Id: I5ffa65dd58a7b47f287d124d9e71ba1dc7c5a549

20 months agoMerge "vp9/rate_ctrl_rtc: Improve get cyclic refresh data" into main
Marco Paniconi [Fri, 18 Nov 2022 04:16:26 +0000 (04:16 +0000)]
Merge "vp9/rate_ctrl_rtc: Improve get cyclic refresh data" into main

20 months agovp9/rate_ctrl_rtc: Improve get cyclic refresh data
Hirokazu Honda [Thu, 17 Nov 2022 07:05:28 +0000 (16:05 +0900)]
vp9/rate_ctrl_rtc: Improve get cyclic refresh data

A client of the vp9 rate controller needs to know whether the
segmentation is enabled and the size of delta_q. It is also nicer to
know the size of map. This CL changes the interface to achieve these.

Bug: b:259487065
Test: Build

Change-Id: If05854530f97e1430a7b97788910f277ab673a87

20 months agoMerge "vp9-svc: Fixes to make SVC work with VBR" into main
Marco Paniconi [Tue, 15 Nov 2022 21:45:07 +0000 (21:45 +0000)]
Merge "vp9-svc: Fixes to make SVC work with VBR" into main

20 months agovp9-svc: Fixes to make SVC work with VBR
Marco Paniconi [Tue, 15 Nov 2022 06:11:19 +0000 (22:11 -0800)]
vp9-svc: Fixes to make SVC work with VBR

Prior to this CL SVC with VBR mode was broken.
Fixes made here to make VBR rate control work for SVC.
Rename is_one_pass_cbr_svc() --> is_one_pass_svc(),
as it can be used now for both CBR and VBR.

Added rate targetting unittest for (2SL, 3TL).

Bug: chromium:1375111
Change-Id: I5a62ffe7fbea29dc5949c88a284768386b1907a9

20 months agoMerge "[NEON] Optimize FHT functions, add highbd FHT 4x4" into main
James Zern [Tue, 15 Nov 2022 19:19:43 +0000 (19:19 +0000)]
Merge "[NEON] Optimize FHT functions, add highbd FHT 4x4" into main

20 months agoquantize: remove vp9_regular_quantize_b_4x4
Johann [Mon, 14 Nov 2022 08:59:45 +0000 (17:59 +0900)]
quantize: remove vp9_regular_quantize_b_4x4

This was just a helper function which called vpx_quantize_b or
vpx_highbd_quantize_b. It also checked for skip_block, which was
necessary when webm:1439 was filed but does not appear to be
necessary now.

Removes a quantize variant and makes subsequent cleanups easier.

Change-Id: Ibe545eccd19370f07ff26c8e151f290c642efd2a

20 months ago[NEON] Optimize FHT functions, add highbd FHT 4x4
Konstantinos Margaritis [Wed, 9 Nov 2022 09:30:58 +0000 (09:30 +0000)]
[NEON] Optimize FHT functions, add highbd FHT 4x4

Refactor & optimize FHT functions further, use new butterfly functions
4x4 5% faster, 8x8 & 16x16 10% faster than previous versions.
Highbd 4x4 FHT version 2.27x faster than C version for --rt.

Change-Id: I3ebcd26010f6c5c067026aa9353cde46669c5d94

20 months agovp9-rc: Fix key frame setting in external RC
Marco Paniconi [Fri, 11 Nov 2022 02:50:19 +0000 (18:50 -0800)]
vp9-rc: Fix key frame setting in external RC

Bug: b/257368998

Change-Id: I03e35915ac99b50cb6bdf7bce8b8f9ec5aef75b7

20 months agoMerge "Add Neon implementation of vpx_hadamard_32x32" into main
James Zern [Mon, 7 Nov 2022 21:48:50 +0000 (21:48 +0000)]
Merge "Add Neon implementation of vpx_hadamard_32x32" into main

20 months agobuild: fix -Wimplicit-int (Clang 16)
Sam James [Sun, 6 Nov 2022 04:11:59 +0000 (04:11 +0000)]
build: fix -Wimplicit-int (Clang 16)

Clang 16 will make -Wimplicit-int error by default which can, in addition to
other things, lead to some configure tests silently failing/returning the wrong result.

Fixes this error:
```
+/var/tmp/portage/media-libs/libvpx-1.12.0/temp/vpx-conf-1802-30624.c:1:15: error: type specifier missing, defaults to 'int'; ISO C99 and later do not support implicit int [-Wimplicit-int]
```

For more information, see LWN.net [0] or LLVM's Discourse [1], gentoo-dev@ [2],
or the (new) c-std-porting mailing list [3].

[0] https://lwn.net/Articles/913505/
[1] https://discourse.llvm.org/t/configure-script-breakage-with-the-new-werror-implicit-function-declaration/65213
[2] https://archives.gentoo.org/gentoo-dev/message/dd9f2d3082b8b6f8dfbccb0639e6e240
[3] hosted at lists.linux.dev.

Bug: https://bugs.gentoo.org/879705
Change-Id: Id73a98944ab3c99a368b9da7a5e902ddff9d937f
Signed-off-by: Sam James <sam@gentoo.org>
20 months agoAdd Neon implementation of vpx_hadamard_32x32
Andrew Salkeld [Thu, 13 Oct 2022 15:28:41 +0000 (16:28 +0100)]
Add Neon implementation of vpx_hadamard_32x32

Add an Arm Neon implementation of vpx_hadamard_32x32 and use it
instead of the scalar C implementation.

Also add test coverage for the new Neon implementation.

Change-Id: Iccc018eec4dbbe629fb0c6f8ad6ea8554e7a0b13

20 months ago[NEON] Optimize highbd 32x32 DCT
Konstantinos Margaritis [Wed, 26 Oct 2022 22:09:32 +0000 (22:09 +0000)]
[NEON] Optimize highbd 32x32 DCT

For --best quality, resulting function
vpx_highbd_fdct32x32_rd_neon takes 0.27% of cpu time in
profiling, vs 6.27% for the sum of scalar functions:
vpx_fdct32, vpx_fdct32.constprop.0, vpx_fdct32x32_rd_c for rd.
For --rt quality, the function takes 0.19% vs 4.57% for the scalar
version.
Overall, this improves encoding time by ~6% compared for highbd
for --best and ~9% for --rt.

Change-Id: I1ce4bbef6e364bbadc76264056aa3f86b1a8edc5

20 months agoMerge "[NEON] Optimize and homogenize Butterfly DCT functions" into main
James Zern [Wed, 2 Nov 2022 02:21:18 +0000 (02:21 +0000)]
Merge "[NEON] Optimize and homogenize Butterfly DCT functions" into main

20 months ago[NEON] Optimize and homogenize Butterfly DCT functions
Konstantinos Margaritis [Wed, 26 Oct 2022 21:37:31 +0000 (21:37 +0000)]
[NEON] Optimize and homogenize Butterfly DCT functions

Provide a set of commonly used Butterfly DCT functions for use in
DCT 4x4, 8x8, 16x16, 32x32 functions. These are provided in various
forms, using vqrdmulh_s16/vqrdmulh_s32 for _fast variants, which
unfortunately are only usable in pass1 of most DCTs, as they do not
provide the necessary precision in pass2.
This gave a performance gain ranging from 5% to 15% in 16x16 case.
Also, for 32x32, the loads were rearranged, along with the butterfly
optimizations, this gave 10% gain in 32x32_rd function.
This refactoring was necessary to allow easier porting of highbd
32x32 functions -follows this patchset.

Change-Id: I6282e640b95a95938faff76c3b2bace3dc298bc3

20 months agoMerge "MacOS 13 is darwin22" into main
Johann Koenig [Thu, 27 Oct 2022 08:38:48 +0000 (08:38 +0000)]
Merge "MacOS 13 is darwin22" into main

20 months agoMerge "rtcd: allow disabling neon on armv8" into main
Johann Koenig [Thu, 27 Oct 2022 08:38:18 +0000 (08:38 +0000)]
Merge "rtcd: allow disabling neon on armv8" into main

20 months agoMacOS 13 is darwin22
Johann [Thu, 27 Oct 2022 02:40:19 +0000 (11:40 +0900)]
MacOS 13 is darwin22

Bug: webm:1783
Change-Id: I97d94ab8c8aebe13aedb58e280dc37474814ad5d

20 months agortcd: allow disabling neon on armv8
Johann [Wed, 26 Oct 2022 23:49:37 +0000 (08:49 +0900)]
rtcd: allow disabling neon on armv8

Change-Id: Idef943775456eb95b46be5c92c114c1d215f38d7

20 months agomailmap: add johann@duck.com
Johann [Wed, 26 Oct 2022 08:14:21 +0000 (17:14 +0900)]
mailmap: add johann@duck.com

Change-Id: I3b48951e69ba1f4a9fafdbb81fac48f79587a342

20 months agoMerge changes I36545ff4,Id1aa29da into main
James Zern [Tue, 25 Oct 2022 19:16:46 +0000 (19:16 +0000)]
Merge changes I36545ff4,Id1aa29da into main

* changes:
  vp9_highbd_quantize_fp*_neon: normalize fn param name
  highbd_sad_avx2: normalize function param names

20 months agoMerge "SAD*Test: mark virtual Run() as overridden" into main
James Zern [Tue, 25 Oct 2022 19:16:08 +0000 (19:16 +0000)]
Merge "SAD*Test: mark virtual Run() as overridden" into main

20 months agoMerge "quantize: consolidate sse2 conditionals" into main
Johann Koenig [Tue, 25 Oct 2022 13:26:37 +0000 (13:26 +0000)]
Merge "quantize: consolidate sse2 conditionals" into main

20 months agoMerge "vp9 quantize: rewrite ssse3 in intrinsics" into main
Johann Koenig [Tue, 25 Oct 2022 13:26:22 +0000 (13:26 +0000)]
Merge "vp9 quantize: rewrite ssse3 in intrinsics" into main

20 months agoSAD*Test: mark virtual Run() as overridden
James Zern [Mon, 24 Oct 2022 22:37:26 +0000 (15:37 -0700)]
SAD*Test: mark virtual Run() as overridden

this comes from AbstractBench

Change-Id: Ie0b5a26a68bfbffd80f132125d15a1bdfc990c22

20 months agovp9_highbd_quantize_fp*_neon: normalize fn param name
James Zern [Mon, 24 Oct 2022 22:28:47 +0000 (15:28 -0700)]
vp9_highbd_quantize_fp*_neon: normalize fn param name

count -> n_coeffs. aligns the name with the rtcd header; clears a
clang-tidy warning

Change-Id: I36545ff479df92b117c95e494f16002e6990f433

20 months agohighbd_sad_avx2: normalize function param names
James Zern [Mon, 24 Oct 2022 22:24:51 +0000 (15:24 -0700)]
highbd_sad_avx2: normalize function param names

(src|ref)8_ptr -> (src|ref)_ptr. aligns the names with the rtcd header;
clears some clang-tidy warnings

Change-Id: Id1aa29da8c0fa5860b46ac902f5b2620c0d3ff54

21 months agoFix to VP8 external RC for buffer levels
Marco Paniconi [Tue, 18 Oct 2022 05:36:25 +0000 (22:36 -0700)]
Fix to VP8 external RC for buffer levels

On a dynamic change of temporal layers:
starting/maimum/optimal were being set twice,
causing incorrect large values.

Bug: b/253927937
Change-Id: I204e885cff92530336a9ed9a4363c486c5bf80ae

21 months agoquantize: consolidate sse2 conditionals
Johann [Mon, 17 Oct 2022 07:22:23 +0000 (16:22 +0900)]
quantize: consolidate sse2 conditionals

Change-Id: I43de579e30f2967b97064063e29676e0af1a498f

21 months agovp9 quantize: rewrite ssse3 in intrinsics
Johann [Sat, 1 Oct 2022 02:47:05 +0000 (11:47 +0900)]
vp9 quantize: rewrite ssse3 in intrinsics

Change-Id: I3177251a5935453a23a23c39ea5f6fd41254775e

21 months agoMerge "Fix to VP8 external RC for dynamic update of layers" into main
Marco Paniconi [Sat, 15 Oct 2022 01:56:46 +0000 (01:56 +0000)]
Merge "Fix to VP8 external RC for dynamic update of layers" into main

21 months agoFix to VP8 external RC for dynamic update of layers
Marco Paniconi [Wed, 12 Oct 2022 07:10:47 +0000 (00:10 -0700)]
Fix to VP8 external RC for dynamic update of layers

On change/update of rc_cfg: when number of temporal
layers change call vp8_reset_temporal_layer_change(),
which in turn will call vp8_init_temporal_layer_context()
only for the new layers.

Bug:b/249644737

Change-Id: Ib20d746c7eacd10b78806ca6a5362c750d9ca0b3

21 months ago[NEON] fix clang compile warnings
Konstantinos Margaritis [Thu, 13 Oct 2022 15:19:46 +0000 (15:19 +0000)]
[NEON] fix clang compile warnings

Change-Id: Ib7ce7a774ec89ba51169ea64d24c878109ef07d1

21 months agoMerge "Add vpx_highbd_sad64x{64,32}_avg_avx2." into main
Scott LaVarnway [Thu, 13 Oct 2022 11:31:51 +0000 (11:31 +0000)]
Merge "Add vpx_highbd_sad64x{64,32}_avg_avx2." into main

21 months ago[NEON] Add highbd FDCT 16x16 function
Konstantinos Margaritis [Fri, 7 Oct 2022 15:13:29 +0000 (15:13 +0000)]
[NEON] Add highbd FDCT 16x16 function

90-95% faster than C version in best/rt profiles

Change-Id: I41d5e9acdc348b57153637ec736498a25ed84c25

21 months agoMerge "[NEON] Add highbd FDCT 8x8 function" into main
James Zern [Wed, 12 Oct 2022 20:07:51 +0000 (20:07 +0000)]
Merge "[NEON] Add highbd FDCT 8x8 function" into main

21 months agoMerge "Add vpx_highbd_sad32x{64,32,16}_avg_avx2." into main
Scott LaVarnway [Wed, 12 Oct 2022 19:50:55 +0000 (19:50 +0000)]
Merge "Add vpx_highbd_sad32x{64,32,16}_avg_avx2." into main

21 months agoMerge "Add vpx_highbd_sad16x{32,16,8}_avg_avx2." into main
Scott LaVarnway [Wed, 12 Oct 2022 19:44:44 +0000 (19:44 +0000)]
Merge "Add vpx_highbd_sad16x{32,16,8}_avg_avx2." into main

21 months ago[NEON] Add highbd FDCT 8x8 function
Konstantinos Margaritis [Thu, 6 Oct 2022 16:00:43 +0000 (16:00 +0000)]
[NEON] Add highbd FDCT 8x8 function

50% faster than C version in best/rt profiles

Change-Id: I0f9504ed52b5d5f7722407e91108ed4056d66bc2

21 months agoAdd vpx_highbd_sad64x{64,32}_avg_avx2.
Scott LaVarnway [Wed, 12 Oct 2022 17:26:43 +0000 (10:26 -0700)]
Add vpx_highbd_sad64x{64,32}_avg_avx2.

~2.8x faster than the sse2 version.

Bug: b/245917257

Change-Id: Ib727ba8a8c8fa4df450bafdde30ed99fd283f06d

21 months ago[NEON] Add highbd FDCT 4x4 function
Konstantinos Margaritis [Thu, 6 Oct 2022 14:53:56 +0000 (14:53 +0000)]
[NEON] Add highbd FDCT 4x4 function

~80% faster than C version for both best/rt profiles.

Change-Id: Ibb3c8e1862131d2a020922420d53c66b31d5c2c3

21 months agoAdd vpx_highbd_sad32x{64,32,16}_avg_avx2.
Scott LaVarnway [Wed, 12 Oct 2022 13:05:46 +0000 (06:05 -0700)]
Add vpx_highbd_sad32x{64,32,16}_avg_avx2.

2.1x to 2.8x faster than the sse2 version.

Bug: b/245917257

Change-Id: I1aaffa4a1debbe5559784e854b8fc6fba07e5000

21 months agoAdd vpx_highbd_sad16x{32,16,8}_avg_avx2.
Scott LaVarnway [Mon, 10 Oct 2022 15:38:44 +0000 (08:38 -0700)]
Add vpx_highbd_sad16x{32,16,8}_avg_avx2.

1.6x to 2.1x faster than the sse2 version.

Bug: b/245917257

Change-Id: I56c467a850297ae3abcca4b4843302bb8d5d0ac1

21 months ago[NEON] Move helper functions for reuse
Konstantinos Margaritis [Thu, 6 Oct 2022 13:05:01 +0000 (13:05 +0000)]
[NEON] Move helper functions for reuse

Move all butterfly functions to fdct_neon.h
Slightly optimize load/scale/cross functions
in fdct 16x16.
These will be reused in highbd variants.

Change-Id: I28b6e0cc240304bab6b94d9c3f33cca77b8cb073

21 months agoMerge "SADavgTest: Add speed test." into main
Scott LaVarnway [Mon, 10 Oct 2022 20:34:02 +0000 (20:34 +0000)]
Merge "SADavgTest: Add speed test." into main

21 months agoSADavgTest: Add speed test.
Scott LaVarnway [Mon, 10 Oct 2022 19:20:37 +0000 (12:20 -0700)]
SADavgTest: Add speed test.

Change-Id: Ie14c0f6d15f410adf749f7ab74cf9f2bf35f3d5f

21 months ago[NEON] move transpose_8x8 to reuse
Konstantinos Margaritis [Thu, 6 Oct 2022 10:58:27 +0000 (10:58 +0000)]
[NEON] move transpose_8x8 to reuse

Change-Id: I3915b6c9971aedaac9c23f21fdb88bc271216208

21 months agoMerge "[NEON] highbd partial DCT functions" into main
James Zern [Mon, 10 Oct 2022 18:37:05 +0000 (18:37 +0000)]
Merge "[NEON] highbd partial DCT functions" into main

21 months ago[NEON] highbd partial DCT functions
Konstantinos Margaritis [Thu, 6 Oct 2022 10:26:05 +0000 (10:26 +0000)]
[NEON] highbd partial DCT functions

Change-Id: I7dd4e698469562f5b1f948cc36f8403b490dcb6a