Your request is on the way, soon we will share the performance related
details. Thanks.
Regards,
Praveen Tiwari
On Fri, Apr 6, 2018 at 9:36 PM, Vittorio Giovara wrote:
> just curious, what kind of general speed improvement does this give?
> I could have missed them
Sorry, I miss a line, resend with addition comment
At 2018-04-07 01:27:34, "chen" wrote:
At 2018-04-06 21:17:37, mythr...@multicorewareinc.com wrote:
># HG changeset patch
># User Jayashree
># Date 1517283539 28800
># Mon Jan 29 19:38:59 2018 -0800
># Node ID
At 2018-04-06 21:17:37, mythr...@multicorewareinc.com wrote:
># HG changeset patch
># User Jayashree
># Date 1517283539 28800
># Mon Jan 29 19:38:59 2018 -0800
># Node ID 3c6e5ce07dbca7f967e4b5b62fe450979da3bf81
># Parent 624c83571d1df840e1206c46e589044fbf87ff32
>x86: AVX512
just curious, what kind of general speed improvement does this give?
I could have missed them in the series, but it would be nice to have some
sort of benchmarks
thanks
Vittorio
On Sat, Apr 7, 2018 at 4:29 AM, wrote:
> This series of patches enables AVX-512 in
# HG changeset patch
# User Jayashree
# Date 1515992814 -19800
# Mon Jan 15 10:36:54 2018 +0530
# Node ID 3a310b157fdf345023ff4e96e7de316cee79b954
# Parent c1daa99a8c14edbe5e9e5a59a74a6b0936c27a82
x86 : AVX512 intra_pred_ang16 mode 5 and 31 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashree
# Date 1516361209 -19800
# Fri Jan 19 16:56:49 2018 +0530
# Node ID e82bfd58acb99cd4c2e4767b1afdd3750881a68e
# Parent f56354b2b542aaafa389a226f0fb3b41e4d33803
Fix SIMD register count for intra_pred_ang modes
diff -r f56354b2b542 -r e82bfd58acb9
# HG changeset patch
# User Jayashri Murugan
# Date 1517294626 -19800
# Tue Jan 30 12:13:46 2018 +0530
# Node ID b80e844209ecd0abc896df94306a5ef96b27b918
# Parent e82bfd58acb99cd4c2e4767b1afdd3750881a68e
X86: AVX512 intra_pred_ang16 mode 8 and 28 high bit
# HG changeset patch
# User Jayashree
# Date 1522928767 -19800
# Thu Apr 05 17:16:07 2018 +0530
# Node ID f6ad2fa637fd3c8f9e2811982b89aa28228e9f6b
# Parent 876b6e006f2080072c0684dbf75e7cfde974ba79
x86:AVX512 Set run time flag to enable/disable avx512
diff
# HG changeset patch
# User Jayashree
# Date 1516018946 -19800
# Mon Jan 15 17:52:26 2018 +0530
# Node ID a2b347ed81f90ac82f59d891deba7fa876df7f62
# Parent 1107c2def5f9dbee9947a2c9c41f50961fa31bc6
x86:AVX512 intra_pred-ang32 mode 4 and 32 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashree
# Date 1516212669 28800
# Wed Jan 17 10:11:09 2018 -0800
# Node ID ae3f7bd65b45df716f1cd56b6b15d91643772621
# Parent 3a08a957d4cd2bf0eb57524651a824513378e0a3
X86:AVX512 intra_pred_ang32 mode 7 and 29 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashree
# Date 1516212669 28800
# Wed Jan 17 10:11:09 2018 -0800
# Node ID f56354b2b542aaafa389a226f0fb3b41e4d33803
# Parent ae3f7bd65b45df716f1cd56b6b15d91643772621
X86:AVX512 intra_pred_ang16 mode 7 and 29 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashri Murugan
# Date 1517855940 28800
# Mon Feb 05 10:39:00 2018 -0800
# Node ID 876b6e006f2080072c0684dbf75e7cfde974ba79
# Parent b80e844209ecd0abc896df94306a5ef96b27b918
x86: AVX512 fix intra_pred_ang16_mode_8_and_28 kernel
# HG changeset patch
# User Jayashree
# Date 1515990226 -19800
# Mon Jan 15 09:53:46 2018 +0530
# Node ID c1daa99a8c14edbe5e9e5a59a74a6b0936c27a82
# Parent e4983d90f403d968d6760ae044f86a7a2e1865a2
x86:AVX512 intra_pred_ang32 mode 5 and 31 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashri Murugan
# Date 1515789616 28800
# Fri Jan 12 12:40:16 2018 -0800
# Node ID 624c83571d1df840e1206c46e589044fbf87ff32
# Parent b0d00ca83af0cb2053d6eda82b6d4081236a0f5f
X86: AVX512 intra_pred_ang32 mode 8 and 28 high bit depth
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515661273 -19800
# Thu Jan 11 14:31:13 2018 +0530
# Node ID d43237051962eab3cd761cf24f3971de09c07aa5
# Parent 5a90661c7fbf2fbacbd6b8afde64368147c29674
x86: AVX512 intra_pred_ang32 mode 11 and 25,
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515575222 -19800
# Wed Jan 10 14:37:02 2018 +0530
# Node ID 59e596ff83801d7c3e3e01f6d6f64d26b2e8010f
# Parent a4d60c45fdce6797486f25f5f319615b25bd86f0
x86: AVX512 intra_pred_ang32 mode 9 and 27 for high bit
# HG changeset patch
# User Jayashree
# Date 1516014532 -19800
# Mon Jan 15 16:38:52 2018 +0530
# Node ID ab1c3fd4fc64e0239e4e632f03cd616e857718ed
# Parent a2b347ed81f90ac82f59d891deba7fa876df7f62
x86 : AVX512 intra_pred_ang16 mode 4 and 32 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515745345 -19800
# Fri Jan 12 13:52:25 2018 +0530
# Node ID 1c2875198a213a5f8d84bff57fcec15727f94a4f
# Parent d43237051962eab3cd761cf24f3971de09c07aa5
x86: AVX512 pixel_satd_32xN
Size | AVX2 performance
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515750476 -19800
# Fri Jan 12 15:17:56 2018 +0530
# Node ID e4983d90f403d968d6760ae044f86a7a2e1865a2
# Parent 1c2875198a213a5f8d84bff57fcec15727f94a4f
x86: AVX512 pixel_satd_64xN
Size | AVX2 performance
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515999160 -19800
# Mon Jan 15 12:22:40 2018 +0530
# Node ID 1107c2def5f9dbee9947a2c9c41f50961fa31bc6
# Parent 3a310b157fdf345023ff4e96e7de316cee79b954
x86: AVX512 planecopy_sp_shl for input 10bit, output
# HG changeset patch
# User Jayashree
# Date 1516097338 -19800
# Tue Jan 16 15:38:58 2018 +0530
# Node ID b0d00ca83af0cb2053d6eda82b6d4081236a0f5f
# Parent ce088a0f4c0ede8ff2e5f0ed6faa005c280acd84
X86:AVX512 intra_pred_ang16 mode 6 and 30 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515642236 -19800
# Thu Jan 11 09:13:56 2018 +0530
# Node ID 5a90661c7fbf2fbacbd6b8afde64368147c29674
# Parent b31c44c9bd690a67837f857081f1a27fa5512d59
x86: AVX512 Link ssd_s_16 and blockfill_s for high bit
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515581418 -19800
# Wed Jan 10 16:20:18 2018 +0530
# Node ID b31c44c9bd690a67837f857081f1a27fa5512d59
# Parent 59e596ff83801d7c3e3e01f6d6f64d26b2e8010f
x86: AVX512 intra_pred_ang16 mode 9 and 27 for high bit
# HG changeset patch
# User Jayashree
# Date 1517285149 28800
# Mon Jan 29 20:05:49 2018 -0800
# Node ID 3a08a957d4cd2bf0eb57524651a824513378e0a3
# Parent 3c6e5ce07dbca7f967e4b5b62fe450979da3bf81
x86: AVX512 'count_nonzero_32x32' avx-512 kernel
diff -r 3c6e5ce07dbc -r 3a08a957d4cd
# HG changeset patch
# User Jayashree
# Date 1516094731 -19800
# Tue Jan 16 14:55:31 2018 +0530
# Node ID ce088a0f4c0ede8ff2e5f0ed6faa005c280acd84
# Parent ab1c3fd4fc64e0239e4e632f03cd616e857718ed
X86:AVX512 intra_pred_ang32 mode 6 and 30 high bit depth
Mode | AVX2 performance | AVX512
# HG changeset patch
# User Jayashree
# Date 1517283539 28800
# Mon Jan 29 19:38:59 2018 -0800
# Node ID 3c6e5ce07dbca7f967e4b5b62fe450979da3bf81
# Parent 624c83571d1df840e1206c46e589044fbf87ff32
x86: AVX512 'count_nonzero_16x16' avx-512 kernel, 22% speedup over avx2
count_nonzero[16x16]
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515128523 -19800
# Fri Jan 05 10:32:03 2018 +0530
# Node ID 8d466bd92e8fac3f24526c01a9532cb05ca82fa1
# Parent b192baa7583385b4715f07f481aae959d1777a74
x86: AVX512 Fix linux build error from psyRdoQuant
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515559539 -19800
# Wed Jan 10 10:15:39 2018 +0530
# Node ID a4d60c45fdce6797486f25f5f319615b25bd86f0
# Parent a5d29083237f28a944143862f980960c3f2b15ff
x86: AVX512 optimise cpy1Dto2D_shr
Work on shift in xmm
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1514282724 -19800
# Tue Dec 26 15:35:24 2017 +0530
# Node ID e883724b1af9f60e9d91be3aa6fe7b949e782684
# Parent 3d780e0d48827cd1cc4e664c3bf96dce6f515810
[x265-avx512]x86: AVX512 optimize
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1513752346 -19800
# Wed Dec 20 12:15:46 2017 +0530
# Node ID 9a2c5411769847c4283594b99c1b07a99e92ea4a
# Parent a2224f4d257cf5f5cd391f455aae3117b7fe65ab
x86: nonPsyRdoQuant primitive
This patch also adds
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515409019 -19800
# Mon Jan 08 16:26:59 2018 +0530
# Node ID a5d29083237f28a944143862f980960c3f2b15ff
# Parent 6d1c8baabf78dba44a500f2039eace3e31abf69c
x86: AVX512 cpy2Dto1D_shr_16
AVX2 performance :
# HG changeset patch
# User Jayashree
# Date 1514886668 -19800
# Tue Jan 02 15:21:08 2018 +0530
# Node ID 80775bda5ec16735e7b1de97dedeb7f7ed391c8f
# Parent 47fd272d3c7002b5a84067a818ca4ae1c61276c1
x86: AVX512 fix testbench crash for ssd_s aligned
diff -r 47fd272d3c70 -r 80775bda5ec1
# HG changeset patch
# User Jayashree
# Date 1513585304 -19800
# Mon Dec 18 13:51:44 2017 +0530
# Node ID fd28f49cb7b30aab97105a59ec841812af205cb9
# Parent 265fd2e1e49587837ebed4e7efcc38a0f6e79346
Aligned routine implementation of ssd_s primitive
diff -r 265fd2e1e495 -r fd28f49cb7b3
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515407502 -19800
# Mon Jan 08 16:01:42 2018 +0530
# Node ID 48917be3e409f917468ff2f73302b62afef492fb
# Parent c9f8c315a900c488e41bb39955a1c893e35a66d4
x86: AVX512 cpy1Dto2D_shr_32
AVX2 performance :
# HG changeset patch
# User Jayashree
# Date 1514549317 -19800
# Fri Dec 29 17:38:37 2017 +0530
# Node ID 47fd272d3c7002b5a84067a818ca4ae1c61276c1
# Parent 74965520283a92095a542ba1997798d6b3af7281
x86:AVX512 intra_pred_ang32 mode 26 for high bit depth
Primitive | AVX2 performance
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515057707 -19800
# Thu Jan 04 14:51:47 2018 +0530
# Node ID b192baa7583385b4715f07f481aae959d1777a74
# Parent c225f79142184cb78ded05ba791edc9b3a40b3da
x86: Fix primitive names displayed in testharness for
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515407134 -19800
# Mon Jan 08 15:55:34 2018 +0530
# Node ID c9f8c315a900c488e41bb39955a1c893e35a66d4
# Parent 8d466bd92e8fac3f24526c01a9532cb05ca82fa1
x86: AVX512 cpy1Dto2D_shr_16
AVX2 performance :
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1515047865 -19800
# Thu Jan 04 12:07:45 2018 +0530
# Node ID c225f79142184cb78ded05ba791edc9b3a40b3da
# Parent 4e9f2efdfd097910aa5bf704a4bbf38b0a28f2a5
x86: AVX512 ssd_ss and ssd_pp for high bit depth
Size
# HG changeset patch
# User Jayashri Murugan
# Date 1514366500 -19800
# Wed Dec 27 14:51:40 2017 +0530
# Node ID 74965520283a92095a542ba1997798d6b3af7281
# Parent 8036bbce3d26fbebd3408a7e17a76206275fbde9
x86: AVX512 intra_pred_ang32x32 mode 18 for high bit
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1514520036 -19800
# Fri Dec 29 09:30:36 2017 +0530
# Node ID ca3c04bd0a71bb263b8084283acce012f0cc397c
# Parent fd28f49cb7b30aab97105a59ec841812af205cb9
x86: AVX512 nonPsyRdoQuant optimise load and floating
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1514521347 -19800
# Fri Dec 29 09:52:27 2017 +0530
# Node ID 4e9f2efdfd097910aa5bf704a4bbf38b0a28f2a5
# Parent 80775bda5ec16735e7b1de97dedeb7f7ed391c8f
x86: psyRdoQuant primitive
This patch also adds AVX512
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1513073610 -19800
# Tue Dec 12 15:43:30 2017 +0530
# Node ID a9be28cde01fd379dff1aec4bfcf809c7c96f9d2
# Parent 9a2c5411769847c4283594b99c1b07a99e92ea4a
[x265-avx512] x86: AVX512 psyCost_pp
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1513930936 -19800
# Fri Dec 22 13:52:16 2017 +0530
# Node ID a2224f4d257cf5f5cd391f455aae3117b7fe65ab
# Parent e883724b1af9f60e9d91be3aa6fe7b949e782684
[x265-avx512]x86: AVX512 optimize
# HG changeset patch
# User Jayashri Murugan
# Date 1513231602 -19800
# Thu Dec 14 11:36:42 2017 +0530
# Node ID 0b38182facb66543adfbf9664cda085f7e641327
# Parent 1480076a7bdda5ca31776adf31c087268f232107
x86: AVX512 Clean up of luma_hps and chroma_hps
diff -r
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1513084512 -19800
# Tue Dec 12 18:45:12 2017 +0530
# Node ID 3d780e0d48827cd1cc4e664c3bf96dce6f515810
# Parent 0b38182facb66543adfbf9664cda085f7e641327
x86: AVX512 intra_pred_ang32x32 mode 2 and 34 for high
# HG changeset patch
# User Jayashri Murugan
# Date 1512992616 -19800
# Mon Dec 11 17:13:36 2017 +0530
# Node ID d6873e0a0786cd732304a94812a28914978113e3
# Parent 73488dec74e1882a16751252398207007b3a0c29
x86: AVX512 interp_8tap_horiz_ps_8xN for high bit depth
# HG changeset patch
# User Jayashri Murugan
# Date 1513228979 -19800
# Thu Dec 14 10:52:59 2017 +0530
# Node ID 1480076a7bdda5ca31776adf31c087268f232107
# Parent 458b708e6d17aafb49a5fd369b2e9540d0268726
x86: AVX512 Clean up of luma_vps and chroma_vps
diff -r
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1512721859 -19800
# Fri Dec 08 14:00:59 2017 +0530
# Node ID ab5b1becd807647d5264381c1fb74750c20fdfae
# Parent 42fe321e5cdf9ad260e4e5c7a64137a8b7601915
[x265-avx512] x86: AVX512 psyCost_pp
# HG changeset patch
# User Jayashri Murugan
# Date 1512627824 -19800
# Thu Dec 07 11:53:44 2017 +0530
# Node ID 862e490b929710720e0066335c95a6e8951b2194
# Parent c335a7ca4304001e245dea7977cde1c2e0c0a8ee
x86: AVX512 interp_4tap_vert_ps_16xN
Disabled lesser
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1513073695 -19800
# Tue Dec 12 15:44:55 2017 +0530
# Node ID b858f80e3ff03118abb1ef3e4ea56059f9ec5af4
# Parent 75d5a01d97daad790cecd35b40ff4b0e4cc34cac
x86: AVX512 pixel_satd_48x64 for high bit depth
AVX2
# HG changeset patch
# User Jayashri Murugan
# Date 1512724598 -19800
# Fri Dec 08 14:46:38 2017 +0530
# Node ID b3f877f022e33733d4bd6ec5292e4325d18ced12
# Parent 9ca6f6a66eabf5bfdecc3a8472c1137d16b1c722
x86: AVX512 interp_8tap_vert_ps_48x64
AVX2 performance
# HG changeset patch
# User Jayashri Murugan
# Date 1512989555 -19800
# Mon Dec 11 16:22:35 2017 +0530
# Node ID 73488dec74e1882a16751252398207007b3a0c29
# Parent 6cef9dfdbef43d36b4e6e4659a8db31d9f2529ba
x86: AVX512 interp_8tap_horiz_ps_24x32 for high bit
# HG changeset patch
# User Vignesh Vijayakumar>
# Date 1522981507 25200
# Thu Apr 05 19:25:07 2018 -0700
# Node ID 75d5a01d97daad790cecd35b40ff4b0e4cc34cac
# Parent ddd64f4b2ff382d05e86708750b20332ed93f3c9
x86: AVX512 optimise chroma_hps_16xN for high bit depth
# HG changeset patch
# User Jayashri Murugan
# Date 1513143310 -19800
# Wed Dec 13 11:05:10 2017 +0530
# Node ID 458b708e6d17aafb49a5fd369b2e9540d0268726
# Parent ab5b1becd807647d5264381c1fb74750c20fdfae
x86: Aligned routine implementation for cpy1Dto2D_shl
# HG changeset patch
# User Jayashri Murugan
# Date 1512627824 -19800
# Thu Dec 07 11:53:44 2017 +0530
# Node ID e2a41cef45ae06d73c0c17d99a853df6e81562ee
# Parent 862e490b929710720e0066335c95a6e8951b2194
x86: AVX512 interp_8tap_vert_ps_64xN
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1513077484 -19800
# Tue Dec 12 16:48:04 2017 +0530
# Node ID 42fe321e5cdf9ad260e4e5c7a64137a8b7601915
# Parent d6873e0a0786cd732304a94812a28914978113e3
x86: AVX512 optimise intra_pred_dc_32 for high bit
# HG changeset patch
# User Jayashri Murugan
# Date 1512637020 -19800
# Thu Dec 07 14:27:00 2017 +0530
# Node ID 6cef9dfdbef43d36b4e6e4659a8db31d9f2529ba
# Parent b3f877f022e33733d4bd6ec5292e4325d18ced12
x86: AVX512 interp_4tap_vert_ps_48x64
AVX2 performance
# HG changeset patch
# User Jayashri Murugan
# Date 1512722575 -19800
# Fri Dec 08 14:12:55 2017 +0530
# Node ID 617aa7cf2c76368cb8a3b252175c1b3d6f716915
# Parent 3d7a32f87793527d40d17024c1bb415a95cbdeae
x86: AVX512 interp_8tap_vert_ps_16xN
Disabled lesser
# HG changeset patch
# User Jayashri Murugan
# Date 1512717026 -19800
# Fri Dec 08 12:40:26 2017 +0530
# Node ID 3d7a32f87793527d40d17024c1bb415a95cbdeae
# Parent e2a41cef45ae06d73c0c17d99a853df6e81562ee
x86: AVX512 interp_8tap_vert_ps_32xN
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512715363 -19800
# Fri Dec 08 12:12:43 2017 +0530
# Node ID fa954ed4a1e7ce2741f3cac14006f78c3199191b
# Parent 86d3d34de566d7696028b5e798a79b9de3a6e62b
x86: AVX512 pixel_var_64x64
AVX2 performance:
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512723573 -19800
# Fri Dec 08 14:29:33 2017 +0530
# Node ID ddd64f4b2ff382d05e86708750b20332ed93f3c9
# Parent fa954ed4a1e7ce2741f3cac14006f78c3199191b
x86: AVX512 intra_pred_dc32 for high bit depth
AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1513072665 -19800
# Tue Dec 12 15:27:45 2017 +0530
# Node ID 9ca6f6a66eabf5bfdecc3a8472c1137d16b1c722
# Parent b858f80e3ff03118abb1ef3e4ea56059f9ec5af4
x86: AVX512 luma_hvpp
Calling interp_8tap_hv_pp_cpu C
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512377610 -19800
# Mon Dec 04 14:23:30 2017 +0530
# Node ID 3e8615bc86537e07754a1c023ade702a837042a8
# Parent 465682e66d91ecf207feae78c33e32f0eaaf45c4
x86: AVX512 optimise interp_4tap_vert_pp_16xN
i444
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1512637265 -19800
# Thu Dec 07 14:31:05 2017 +0530
# Node ID 0ffc9c56a0a7361e98e6388e3067e4a78e8cd252
# Parent 931dd781dc0c6de76bb31d0215db7a7af885f9bf
[x265-avx512]x86: AVX512 sad_x3_16xN
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512537795 -19800
# Wed Dec 06 10:53:15 2017 +0530
# Node ID 2d298099a8d6b266a32b975de4b6a369988d3887
# Parent 8b1c9d9c5bd8135dc11b6d031b990bfe47e3bcd8
x86: AVX512 interp_8tap_vert_pp_24xN and
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1512640914 -19800
# Thu Dec 07 15:31:54 2017 +0530
# Node ID 81a870948ac446b36c248325e0c7264cf8f3f09e
# Parent 0ffc9c56a0a7361e98e6388e3067e4a78e8cd252
[x265-avx512]x86: AVX512 sad_x4_16xN
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512622521 -19800
# Thu Dec 07 10:25:21 2017 +0530
# Node ID 9bd38bd06850914d1cbf617063ea0e1e60f66219
# Parent 2d298099a8d6b266a32b975de4b6a369988d3887
x86: AVX512 pixel_satd_64xN and 32xN for high bit depth
# HG changeset patch
# User Jayashri Murugan
# Date 1512455478 -19800
# Tue Dec 05 12:01:18 2017 +0530
# Node ID c335a7ca4304001e245dea7977cde1c2e0c0a8ee
# Parent 81a870948ac446b36c248325e0c7264cf8f3f09e
x86: AVX512 interp_4tap_vert_ps_32xN
i420
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512041776 -19800
# Thu Nov 30 17:06:16 2017 +0530
# Node ID e77ef4964dd04de6a8b84378f7a46219f34bf1b5
# Parent 9c652d9062d29607cdb3392567817e4e2ab7f6bb
x86: AVX512 interp_4tap_vert_sp_48x64
AVX2 performance
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1512036841 -19800
# Thu Nov 30 15:44:01 2017 +0530
# Node ID f86b11b8c629b0e4bf8342d42a0e9c475d7c3a7d
# Parent e77ef4964dd04de6a8b84378f7a46219f34bf1b5
[x265-avx512]x86: AVX512 denoise DCT
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511953084 -19800
# Wed Nov 29 16:28:04 2017 +0530
# Node ID 834a8f52a976a6c5da294267392bcd6da1aa6d6e
# Parent 9f2c4a0d09f3405f9c28cd3ebf229617c2278681
x86: AVX512 interp_8tap_vert_sp_64xN
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512465067 -19800
# Tue Dec 05 14:41:07 2017 +0530
# Node ID f92128e41ac3c1da210c1c665d97061539821aaf
# Parent ca6bb5919227672e0cf98b785acf099531c32945
x86: AVX512 interp_8tap_vert_pp_16xN and
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512389309 -19800
# Mon Dec 04 17:38:29 2017 +0530
# Node ID 283aa4d77cef296699167c041763d7115e7a88aa
# Parent ae75b2d09d10f28391d573507c13512360593386
x86: AVX512 interp_4tap_vert_pp_48x64
AVX2 performance
# HG changeset patch
# User Jayashri Murugan
# Date 1511781308 -19800
# Mon Nov 27 16:45:08 2017 +0530
# Node ID 1cd123613bbb28fd00da36a3cfe3765f8e07d00e
# Parent 283aa4d77cef296699167c041763d7115e7a88aa
x86: AVX512 interp_4tap_vert_ps_64xN
Size | AVX2
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1512625055 -19800
# Thu Dec 07 11:07:35 2017 +0530
# Node ID 931dd781dc0c6de76bb31d0215db7a7af885f9bf
# Parent 9bd38bd06850914d1cbf617063ea0e1e60f66219
[x265-avx512]x86: AVX512 sad_16x32 and
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512371012 -19800
# Mon Dec 04 12:33:32 2017 +0530
# Node ID 465682e66d91ecf207feae78c33e32f0eaaf45c4
# Parent 4f690222337dbc1757665729ea15f2380a11c329
x86: AVX512 ssd_ss_16x16
AVX2 performance : 43.55x
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512460722 -19800
# Tue Dec 05 13:28:42 2017 +0530
# Node ID ca6bb5919227672e0cf98b785acf099531c32945
# Parent c3a341391f0c777665e191a4cd172f08a5a313f9
x86: AVX512 interp_8tap_vert_pp_64xN and
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511522096 -19800
# Fri Nov 24 16:44:56 2017 +0530
# Node ID c3a341391f0c777665e191a4cd172f08a5a313f9
# Parent 1cd123613bbb28fd00da36a3cfe3765f8e07d00e
x86: AVX512 interp_8tap_vert_pp_32xN and
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512475230 -19800
# Tue Dec 05 17:30:30 2017 +0530
# Node ID 8b1c9d9c5bd8135dc11b6d031b990bfe47e3bcd8
# Parent f92128e41ac3c1da210c1c665d97061539821aaf
x86: AVX512 interp_8tap_vert_pp_48x64 and
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512021637 -19800
# Thu Nov 30 11:30:37 2017 +0530
# Node ID 6137bed68dac85ab475b8be28fdba7f5787ac551
# Parent a78e09e144582bd52c52d3475aa1922fc2ae8893
x86: AVX512 interp_4tap_vert_sp_32xN
i444
Size |
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512018320 -19800
# Thu Nov 30 10:35:20 2017 +0530
# Node ID a78e09e144582bd52c52d3475aa1922fc2ae8893
# Parent 3e14c3f607d0f9ec6dd3735d21fc2e698217fe71
x86: AVX512 interp_8tap_vert_sp_16xN
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512037814 -19800
# Thu Nov 30 16:00:14 2017 +0530
# Node ID 3e14c3f607d0f9ec6dd3735d21fc2e698217fe71
# Parent adbbeb6c2f1e3697515561179abe156bacc7cd1c
x86: AVX512 interp_8tap_vert_sp_48x64
AVX2 performance
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512380104 -19800
# Mon Dec 04 15:05:04 2017 +0530
# Node ID ae75b2d09d10f28391d573507c13512360593386
# Parent 3e8615bc86537e07754a1c023ade702a837042a8
x86: AVX512 optimise interp_4tap_vert_ss_8xN
i444 8x4
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512024031 -19800
# Thu Nov 30 12:10:31 2017 +0530
# Node ID 76cec343d9305e800a5fd74a5066c0fd919eb8d8
# Parent 6137bed68dac85ab475b8be28fdba7f5787ac551
x86: AVX512 interp_4tap_vert_sp_64xN
i444
Size |
# HG changeset patch
# User Praveen Tiwari
# Date 1512104438 -19800
# Fri Dec 01 10:30:38 2017 +0530
# Node ID 4f690222337dbc1757665729ea15f2380a11c329
# Parent f86b11b8c629b0e4bf8342d42a0e9c475d7c3a7d
x86: AVX512 - dct16 and dct32 for main10 profile
diff -r
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511948305 -19800
# Wed Nov 29 15:08:25 2017 +0530
# Node ID 9f2c4a0d09f3405f9c28cd3ebf229617c2278681
# Parent 3e2058cec6c6f4ad49d92f9df7fbc110a54f4b4b
x86: AVX512 interp_8tap_vert_sp_32xN
Size | AVX2
# HG changeset patch
# User Praveen Tiwari
# Date 1512035958 -19800
# Thu Nov 30 15:29:18 2017 +0530
# Node ID d7af8d747bffacafa5dfe8f4d513bbd09314ad63
# Parent aff686238f2a30ab42b0e2ad296be54e77179531
x86: disable similar perfoamnce kernels - need more
# HG changeset patch
# User Praveen Tiwari
# Date 1512016927 28800
# Wed Nov 29 20:42:07 2017 -0800
# Node ID aff686238f2a30ab42b0e2ad296be54e77179531
# Parent 63bedd49719fe9094ffdcbb88ac8512dccc120d2
x86: AVX512 idct32 optimize load operations
diff -r
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511864533 -19800
# Tue Nov 28 15:52:13 2017 +0530
# Node ID af867976d51969b1770e6bcffd80e0389c88b561
# Parent 651bf679ed5c7ec6b68714e81d3c24664f08ec6a
x86: AVX512 optimise scale1D128to64 code
Previous
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511870877 -19800
# Tue Nov 28 17:37:57 2017 +0530
# Node ID 240ae5a46e63d3bebd8a4db63a5662a4000d70a7
# Parent af867976d51969b1770e6bcffd80e0389c88b561
x86: AVX512 cleanup main profile sad, sad_x3 sad_x4
# HG changeset patch
# User Praveen Tiwari
# Date 1512016234 28800
# Wed Nov 29 20:30:34 2017 -0800
# Node ID 63bedd49719fe9094ffdcbb88ac8512dccc120d2
# Parent 2a79f5eb0a9897f8703dafadfa034ba68b5955a9
x86: AVX512 idct16 kernel - optimize to use align load
diff
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1512037629 -19800
# Thu Nov 30 15:57:09 2017 +0530
# Node ID adbbeb6c2f1e3697515561179abe156bacc7cd1c
# Parent 834a8f52a976a6c5da294267392bcd6da1aa6d6e
x86: Fix merge errors
diff -r 834a8f52a976 -r
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1511947290 -19800
# Wed Nov 29 14:51:30 2017 +0530
# Node ID 3e2058cec6c6f4ad49d92f9df7fbc110a54f4b4b
# Parent d7af8d747bffacafa5dfe8f4d513bbd09314ad63
[x265-avx512]x86: AVX512 nquant
AVX2
# HG changeset patch
# User Praveen Tiwari
# Date 1512003711 28800
# Wed Nov 29 17:01:51 2017 -0800
# Node ID bf62120c026fd079b8a1411eed1c655afc859671
# Parent 0c4c63916d6e5af144bad8280706da6f3a613e75
dct32 AVX512 Kernel
diff -r 0c4c63916d6e -r bf62120c026f
# HG changeset patch
# User Praveen Tiwari
# Date 1512013411 28800
# Wed Nov 29 19:43:31 2017 -0800
# Node ID 2a79f5eb0a9897f8703dafadfa034ba68b5955a9
# Parent bf62120c026fd079b8a1411eed1c655afc859671
dct16 AVX-512 kernel
diff -r bf62120c026f -r 2a79f5eb0a98
# HG changeset patch
# User Gopi Satykrishna Akisetty
# Date 1511857266 -19800
# Tue Nov 28 13:51:06 2017 +0530
# Node ID 0c4c63916d6e5af144bad8280706da6f3a613e75
# Parent 8eeff916ebe608526f167177a19c4516266ba513
[x265-avx512]x86: AVX512 Quant
AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511929966 -19800
# Wed Nov 29 10:02:46 2017 +0530
# Node ID 2ebaab8b8d9be6f8c1c89699f818f6426cc3fbea
# Parent 240ae5a46e63d3bebd8a4db63a5662a4000d70a7
x86: AVX512 interp_8tap_horiz_pp_8xN for high bit depth
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511785721 -19800
# Mon Nov 27 17:58:41 2017 +0530
# Node ID 8abe9a2bb0f1290db65c73416fd01fcf21465460
# Parent 3de532ebcd766f11661ca023e144e8db0db9cd56
x86: AVX512 interp_8tap_vert_pp_64xN
Size | AVX2
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511775875 -19800
# Mon Nov 27 15:14:35 2017 +0530
# Node ID 06cc7db3bf0d1a6afb98bb797d70d35ebea5fe32
# Parent 8e38b952769d76e17b81fdc692956d42e8d45df1
x86: AVX512 interp_8tap_vert_ss_48x64
AVX2 performance
# HG changeset patch
# User Vignesh Vijayakumar
# Date 1511847523 -19800
# Tue Nov 28 11:08:43 2017 +0530
# Node ID 6c434e27f2657e97c5ee16d71f6e0e8218ea94ad
# Parent 7ec19c5b18e76fb3ffe2c5dece61ef67d18448ff
x86: AVX512 interp_8tap_vert_pp_16xN
Size | AVX2
1 - 100 of 290 matches
Mail list logo