Commit Graph

152 Commits

Author SHA1 Message Date
James Zern
0ccad4d649 Revert "VPX: x86 asm version of vpx_idct32x32_34_add()"
This reverts commit 9aeaa2016e.

This causes some test vectors to fail.

Change-Id: I3659a2068404ec5a0591fba5c88b1bec0c9059a4
2015-11-11 11:12:38 -08:00
James Zern
e3efed7f4c Merge "convolve_copy_sse2: replace SSE w/SSE2 code" 2015-11-10 22:35:12 +00:00
Scott LaVarnway
f48321974b Merge "VPX: x86 asm version of vpx_idct32x32_34_add()" 2015-11-10 21:40:11 +00:00
Scott LaVarnway
9aeaa2016e VPX: x86 asm version of vpx_idct32x32_34_add()
Change-Id: I8a933c63b7fbf3c65e2c06dbdca9646cadd0b7cb
2015-11-10 11:54:56 -08:00
James Zern
40dab58941 convolve_copy_sse2: replace SSE w/SSE2 code
this should be neutral or slightly faster on modern (P4+) architectures

Change-Id: Iec4c080275941eb8c9e05a66a2daf0405d86a69b
2015-11-09 23:45:16 -08:00
Debargha Mukherjee
65dd056e41 Merge "Optimize vpx_quantize_{b,b_32x32} assembler." 2015-10-26 18:04:49 +00:00
Ronald S. Bultje
53dc9fd0a0 vp10: merge ext_ipred_bltr experiment into misc_fixes.
Change-Id: I2f2deb700748408b8278b7f5c29ee1f2e39785ec
2015-10-21 22:27:34 -04:00
Geza Lore
9cfba09ac0 Optimize vpx_quantize_{b,b_32x32} assembler.
Added optimization of the 8 bit assembly quantizer routines. This makes
these functions up to 100% faster, depending on encoding parameters.

This patch maskes the encoder faster in both the high bitdepth and 8bit
configurations. In the high bitdepth configuration, it effects profile 0
only.

Based on my profiling using 1080p input the net gain is between 1-3% for
the 8 bit config, and around 2.5-4.5% for the high bitdepth config,
depending on target bitrate. The difference between the 8 bit and high
bitdepth configurations for the same encoder run is reduced by 1% in all
cases I have profiled.

Change-Id: I86714a6b7364da20cd468cd784247009663a5140
2015-10-20 10:11:19 +01:00
Ronald S. Bultje
c7dc1d78bf vp10: add extended-intra prediction edges experiment.
This experiment allows using full above/right edges for all transform
sizes whenever available (for d45/d63), and adds bottom/left edges for
d207.

See issue 1043.

Change-Id: I5cf7f345e783e8539bb6b6d2c9972fb1d6d0a78b
2015-10-16 19:30:39 -04:00
Johann
ec623a0bb7 Upstream Mozilla fix for older Apple clang builds
Also use the _mm_broadcastsi128_si256 intrisic for
Apple clang versions 4.[012]

https://bugzilla.mozilla.org/show_bug.cgi?id=1085607
https://code.google.com/p/webm/issues/detail?id=1082

Change-Id: I6bc821d8163387194ef663e94bfed91fa7281d88
2015-10-14 07:41:23 -07:00
hui su
6f31722950 Fix compiler warnings
Change-Id: I761256a8100d83abf1b937f3739580237e3fad2a
2015-10-13 10:33:17 -07:00
Alex Converse
0c00af126d Add vpx_highbd_convolve_{copy,avg}_sse2
single-threaded:
swanky (silvermont): ~1% faster overall
peppy (celeron,haswell): ~1.5% faster overall

Change-Id: Ib74f014374c63c9eaf2d38191cbd8e2edcc52073
2015-10-09 11:50:25 -07:00
Geza Lore
cbada4a982 Remove 4 mova insts from quantize_ssse3_x86_64.asm
Change-Id: If3cb9345b44162e600e6c74873e0cb4c207fc7fb
2015-10-09 07:52:04 -07:00
Julia Robson
37c68efee2 SSSE3 optimisation for quantize in high bit depth
When configured with high bit detpth enabled, the 8bit quantize
function stopped using optimised code. This made 8bit content
decode slowly. This commit re-enables the SSSE3 optimisations.

Change-Id: I194b505dd3f4c494e5c5e53e020f5d94534b16b5
2015-10-06 13:32:02 +01:00
Scott LaVarnway
b212094839 Merge "VPX: refactor vpx_idct32x32_1_add_sse2()" 2015-10-06 11:35:15 +00:00
Julia Robson
5e6533e707 SSE2 optimisation for quantize in high bit depth
When configured with high bit detpth enabled, the 8bit quantize
function stopped using optimised code. This made 8bit content
decode slowly. This commit re-enables the SSE2 optimisation
(but not the SSSE3 optimisation).

Change-Id: Id015fe3c1c44580a4bff3f4bd985170f2806a9d9
2015-10-05 10:59:16 -07:00
Scott LaVarnway
23d1c06268 VPX: refactor vpx_idct32x32_1_add_sse2()
Change-Id: Ia1a2cac0e9dc05f3207b3433a6c1589fa7f2aee3
2015-10-05 06:33:42 -07:00
Ronald S. Bultje
3fedf4a59b Merge "vp10: reimplement d45/4x4 to match vp8 instead of vp9." 2015-10-02 17:15:59 +00:00
Debargha Mukherjee
cb5c47f20d Merge "Accelerated transform in high bit depth" 2015-10-02 06:55:55 +00:00
Ronald S. Bultje
62a1579525 vp10: reimplement d45/4x4 to match vp8 instead of vp9.
This is more a proof of concept than anything else. The problem here
isn't so much how to code it, but rather where to place the resulting
code. All intrapred DSP code lives in vpx_dsp, so do we want the vp10
specific intra pred functions to live there, or in vp10/?

See issue 1015.

Change-Id: I675f7badcc8e18fd99a9553910ecf3ddf81f0a05
2015-10-01 10:11:54 -04:00
Ronald S. Bultje
c26a9ecaa2 vp8: change build_intra4x4_predictors() to use vpx_dsp.
I've added a few new functions (d45e, d63e, he, ve) to cover the
filtered h/v 4x4 predictors that are vp8-specific, the "correct"
d45 with the correctly filtered bottom-right pixel (as opposed to
the unfiltered version in vp9), and the "broken" d63 with weirdly
filtered bottom-right pixels (which is correctly filtered in vp9).

There may be a minor performance impact on all systems because we
have to do an extra copy of the Above pixel array to incorporate
the topleft pixel in the same array (thus fitting the vpx_dsp API).
In addition, armv6 will have a more serious performance impact b/c
I removed the armv6/vp8-specific assembly. I'm not sure anyone
cares...

Change-Id: I7f9e5ebee11d8e21aca2cd517a69eefc181b2e86
2015-09-30 18:45:49 -04:00
Ronald S. Bultje
54d48955f6 vp8: change build_intra_predictors_mby_s to use vpx_dsp.
Change-Id: I2000820e0c04de2c975d370a0cf7145330289bb2
2015-09-30 18:45:40 -04:00
Julia Robson
406030d1b0 Accelerated transform in high bit depth
When configured with high bitdepth enabled, the 8bit transform
stopped using optimised code. This made 8bit content decode slowly.

Change-Id: I67d91f9b212921d5320f949fc0a0d3f32f90c0ea
2015-09-28 21:09:16 -07:00
Johann
dd4f953350 Remove vpx_filter_block1d16_v8_intrin_ssse3
This was rewritten and moved to vpx_dsp/x86/vpx_subpixel_8t_ssse3.asm
in 195883023b

Change-Id: I117ce983dae12006e302679ba7f175573dd9e874
2015-09-18 16:05:43 -07:00
James Zern
683b5a3161 vpx_subpixel_8t_ssse3: fix reg counts/access
fixes build on windows x64; previously 'heightq' i.e., the 64-bit register
was accessed when only the 32-bit value was needed. given this is from a
stack variable the upper bits were undefined.

+ bump register/xmm counts; users of SETUP_LOCAL_VARS touch xmm13 in
64-bit builds and filter_block1d16_v* uses one extra temp variable

Change-Id: I9c768c0b2047481d1d3b11c2e16b2f8de6eb0d80
2015-09-17 12:27:34 -07:00
Ronald S. Bultje
a3df343cda vp10: code sign bit before absolute value in non-arithcoded header.
For reading, this makes the operation branchless, although it still
requires two shifts. For writing, this makes the operation as fast
as writing an unsigned value, branchlessly. This is also how other
codecs typically code signed, non-arithcoded bitstream elements.

See issue 1039.

Change-Id: I6a8182cc88a16842fb431688c38f6b52d7f24ead
2015-09-16 19:35:03 -04:00
Debargha Mukherjee
1c8567ff09 Remove some trailing whitespaces
Change-Id: Icf06d35ca347713253d1eba341a894b51efa81a9
2015-09-08 01:31:04 -07:00
Scott LaVarnway
195883023b VPX: subpixel_8t_ssse3 asm using x86inc
This is based on the original patch optimized for 32bit
platforms by Tamar/Ilya and now uses the x86inc style asm.
The assembly was also modified to support 64bit platforms.

Change-Id: Ice12f249bbbc162a7427e3d23fbf0cbe4135aff2
2015-09-03 20:35:51 -07:00
Johann
c5f11912ae Include vpx_dsp_common.h when using VPXMIN/MAX
Change-Id: I2e387a06484a06301f3cd6600c4ba2f4335b61ee
2015-08-31 14:36:35 -07:00
Angie Chiang
45db71d0ac Expand the idct4_c() function in idct8_c()
Change-Id: I5afa3c351ba7c5e7deb3889f7471619ac60af255
2015-08-28 10:53:11 -07:00
Johann Koenig
5c245a46d8 Merge changes I53b5bdc5,Ib81168a7,Ie0113945
* changes:
  Only build ssse3 filter functions on 64 bit
  Clean up unused function warnings in vp8 encoder
  Clean up unused function warnings in vp8 onyx_if.c
2015-08-27 20:58:53 +00:00
Johann Koenig
18ea2a7e0c Merge "Add sse2 versions of halfpix variance" 2015-08-27 20:56:32 +00:00
Johann
a28b2c6ff0 Add sse2 versions of halfpix variance
These were lost in the great sub pixel variance move of
6a82f0d7fb

Not having these functions caused a ~10% performance regression in
some realtime vp8 encodes.

Change-Id: I50658483d9198391806b27899f2c0d309233c4b5
2015-08-27 11:58:38 -07:00
James Zern
5e16d397bd vpx_dsp_common: add VPX prefix to MIN/MAX
prevents redeclaration warnings;
vp8 has its own define which will be resolved in a future commit

Change-Id: Ic941fef3dd4262fcdce48b73075fe6b375f11c9c
2015-08-26 20:11:32 -07:00
Johann
f5507b514c Only build ssse3 filter functions on 64 bit
Avoid an unused function warning by only building the functions when
they will be used.

Change-Id: I53b5bdc5a180c79d63b34e4c8921d679bbc54009
2015-08-26 10:32:18 -07:00
Scott LaVarnway
6c0f6dd817 Merge "VPX: scaled convolve : fix windows build errors" 2015-08-21 12:06:34 +00:00
Scott LaVarnway
acf24cc1b8 VPX: scaled convolve : fix windows build errors
Change-Id: Ic81d435ea928183197040cdf64b6afd7dbaf57e4
2015-08-20 13:09:27 -07:00
Scott LaVarnway
6a21ca20cc Merge "VPX ssse3 scaled convolve" 2015-08-19 22:12:21 +00:00
Jingning Han
b1339751b9 Merge "Rename inv_txfm_sse2.asm to inv_wht_sse2.asm" 2015-08-19 18:26:30 +00:00
Jingning Han
49f6ff1103 Rename inv_txfm_sse2.asm to inv_wht_sse2.asm
Change-Id: I43bcc70680503e4c18d8f021097307778cf9ea70
2015-08-19 10:29:53 -07:00
Scott LaVarnway
2030c49cf8 VPX ssse3 scaled convolve
Change-Id: I71d5994e21813554a927d35ebcc26bf7a68984fd
2015-08-18 15:13:02 -07:00
Jingning Han
5de049b067 Turn on dspr2 loop filter functions in vpx_dsp
Add the dspr2 files to vpx_dsp.mk and enable these functions in
vpx_dsp_rtcd_defs.pl file.

Change-Id: I79feb5af24f174f4a0788dc6f3b6df7f4e1fa467
2015-08-17 16:15:24 -07:00
James Zern
1794624c18 Merge changes I2fe52bfb,I5e5084eb
* changes:
  VPX: removed filter == 128 checks from mips convolve code
  VPX: removed step checks from mips convolve code
2015-08-14 19:45:27 +00:00
James Zern
78629508f2 Merge "VPX: removed step checks from neon convolve code" 2015-08-14 19:23:46 +00:00
Yaowu Xu
94ba3939cd vpx_highbd_ssim_parms_8x8: make parameter types consistent
Change-Id: Ie1fe6603232adc22dbe4d51bd1008c856a6d40ca
2015-08-14 09:18:07 -07:00
Scott LaVarnway
89dcc13939 VPX: removed filter == 128 checks from mips convolve code
The check is handled by the predictor table.

Change-Id: I2fe52bfbbfccb2edd13ba250986e3a4b4b589459
2015-08-13 12:57:01 -07:00
Scott LaVarnway
aeea00cc4f VPX: removed step checks from mips convolve code
The check is handled by the predictor table.

Change-Id: I5e5084ebb46be8087c8c9d80b5f76e919a1cd05b
2015-08-13 11:27:04 -07:00
Scott LaVarnway
fa47212933 VPX: removed step checks from neon convolve code
The check is handled by the predictor table.

Change-Id: I42479f843e77a2d40cdcdfc9e2e6c48a05a36561
2015-08-12 16:46:53 -07:00
Scott LaVarnway
6cf95bd1e7 Merge "VPX: remove step == 16 and filter[3] != 128 checks" 2015-08-12 20:13:33 +00:00
James Zern
345b11cd73 Merge "fix build w/only mmx+sse enabled" 2015-08-12 02:26:08 +00:00