generic-library/openh264
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
4,219 Commits 23 Branches 9 Tags 55 MiB
96b2a87030
Commit Graph

456 Commits

Author SHA1 Message Date
ruil2
7d65687284 Merge pull request #2441 from saamas/encoder-add-avx2-4x4-quantization-routines 
[Encoder] Add AVX2 4x4 quantization routines
 2016-04-28 09:08:31 +08:00
Karina
dd340b7fe7 modify neon comment 2016-04-14 14:49:11 +08:00
Karina
d34e209266 fix 32-bit parameters issue on arm64 assembly function 2016-04-13 19:30:08 +08:00
Sindre Aamås
bb49e23719 [Encoder] Add AVX2 4x4 quantization routines 
WelsQuantFour4x4Max_avx2 (~2.06x speedup over SSE2)
WelsQuantFour4x4_avx2    (~2.32x speedup over SSE2)
WelsQuant4x4Dc_avx2      (~1.49x speedup over SSE2)
WelsQuant4x4_avx2        (~1.42x speedup over SSE2)
 2016-04-13 11:56:47 +02:00
Karina
7943764869 add missing sign extension for arm64 2016-04-12 16:27:58 +08:00
Martin Storsjö
a4e71d6662 Add missing sign extension for x86_64 in mb_copy.asm 
This fixes running the code built for x86_64 OS X with Xcode 7.3.
 2016-03-24 10:20:42 +02:00
Sindre Aamås
b6c4a5447c [Decoder/x86] IDCT one block at a time with SSE2 
At lower bitrates, it is overall faster to conditionally do one block
at a time with SSE2 on Haswell and likely other common architectures.
At higher bitrates, it is faster to use the wider routine that IDCTs
four blocks at a time. To avoid potential performance regressions
as compared to MMX, stick with single-block IDCTs with SSE2. There
is still a performance advantage as compared to MMX because the
single-block SSE2 routine is faster than the corresponding MMX
routine.

Stick with four blocks at a time with AVX2 for which that appears
to be consistently faster on Haswell.
 2016-03-16 19:55:11 +01:00
Sindre Aamås
98042f1600 [Decoder] Use encoder x86 IDCT routines 
Move asm routines to common. Delete obsolete decoder routines.

Use wider routines where applicable.

~1.07x overall faster decode on a quick 720p30 4Mbps test on Haswell.
 2016-03-09 10:41:42 +01:00
Sindre Aamås
48a520915a [Encoder/x86] Add AVX2 SATD routines 
WelsSampleSatd16x16_avx2 (~2.31x speedup over SSE4.1 on Haswell).
WelsSampleSatd16x8_avx2  (~2.19x speedup over SSE4.1 on Haswell).
WelsSampleSatd8x16_avx2  (~1.68x speedup over SSE4.1 on Haswell).
WelsSampleSatd8x8_avx2   (~1.53x speedup over SSE4.1 on Haswell).
 2016-03-08 11:31:17 +01:00
volvet
d4c68527b1 Merge pull request #2389 from saamas/common-x86-deblock-chroma-horizontal-ssse3-optimizations 
[Common/x86] Deblock chroma horizontal ssse3 optimizations
 2016-03-08 17:09:08 +08:00
sijchen
4db9c32976 remove sink in WelsThreadPool and hide the construtor to finish the singleTon 2016-03-02 17:08:09 -08:00
sijchen
d4f09d9048 put CWelsThreadPool to singleTon for future usage (including add sink for IWelsTask) 2016-02-29 11:40:25 -08:00
Sindre Aamås
a009153741 [Common/x86] DeblockChromaEq4H_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

~5.80x speedup on Haswell (x86-64).
~1.69x speedup on Haswell (x86 32-bit).
 2016-02-26 10:58:16 +01:00
Sindre Aamås
9909c306f1 [Common/x86] DeblockChromaLt4H_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

~5.72x speedup on Haswell (x86-64).
~1.85x speedup on Haswell (x86 32-bit).
 2016-02-26 10:58:16 +01:00
ruil2
2754129064 Merge pull request #2360 from saamas/common-x86-deblock-optimizations 
[Common/x86] Deblocking optimizations
 2016-02-19 09:52:39 +08:00
sijchen
e07ee9c096 use WELS_DELETE_OP for deleting 2016-02-17 10:07:33 -08:00
sijchen
74955c877f set pointers to null and call uninit 2016-02-17 10:07:33 -08:00
sijchen
cc675f9fd1 add error handling in memory allocation failed case 2016-02-17 10:07:33 -08:00
sijchen
71aa533038 move the printing of MEMORY_CHECK part to more reasonable 2016-02-15 10:12:34 -08:00
Sindre Aamås
e96a7b5c92 [Common/x86] DeblockChromaEq4V_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

Avoid spills.

~2.07x speedup on Haswell (x86-64).
~2.12x speedup on Haswell (x86 32-bit).
 2016-02-15 02:08:03 +01:00
Sindre Aamås
fc16010583 [Common/x86] DeblockChromaLt4V_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

Avoid spills.

~2.68x speedup on Haswell (x86-64).
~2.38x speedup on Haswell (x86 32-bit).
 2016-02-15 02:07:25 +01:00
Sindre Aamås
62fb37d096 [Common/x86] DeblockLumaEq4_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

Minimize spills.

~2.31x speedup on Haswell (x86-64).
~2.40x speedup on Haswell (x86 32-bit).
 2016-02-15 02:06:39 +01:00
Sindre Aamås
732e1c5f78 [Common/x86] DeblockLumaLt4_ssse3 optimizations 
Use packed 8-bit operations rather than unpack to 16-bit.

Avoid spills.

~1.97x speedup on Haswell (x86-64).
~3.09x speedup on Haswell (x86 32-bit).
 2016-02-15 02:06:18 +01:00
sijchen
e5e7013b73 Merge pull request #2350 from sijchen/th00 
[Common] Add sink to IWelsTask
 2016-02-08 14:59:38 -08:00
unknown
3873addc3d fix frame size constraints for width and height 2016-02-01 15:55:53 +08:00
Sindre Aamås
3088d96978 [Encoder] Add an AVX2 4x4 IDCT implementation 
~2.03x faster on Haswell as compared to the SSE2 version.
 2016-01-19 13:12:28 +01:00
sijchen
5eb18b101e change the output way of debug trace 2016-01-13 22:13:43 -08:00
sijchen
cce1c29844 add sink to IWelsTask (for further enhancements) 2016-01-13 16:24:54 -08:00
Sijia Chen
3e0ee69812 remove unneeded codes and add some logs 2015-11-02 23:15:29 -08:00
Sijia Chen
054a297ca7 adjust encoder tasks, add ut and enable new thread pool under some slice modes 2015-10-28 09:39:26 -07:00
HaiboZhu
e0cee02d77 Merge pull request #2177 from sijchen/thp21 
[Encoder] add encoder tasks and task-management class
 2015-10-23 13:21:42 +08:00
Martin Storsjö
80c8b7b1cc Add a missing include of stdlib.h 
This is required for malloc in this header.

This fixes building for Windows Phone.
 2015-10-20 08:59:41 +03:00
Sijia Chen
819f6f5d93 [Encoder] add encoder tasks and task-management class 
https://rbcommons.com/s/OpenH264/r/1334/
 2015-10-19 22:48:28 -07:00
Martin Storsjö
dac26cf923 Remove unused STL includes 
This fixes building for Android, where libopenh264.so is intended
not to link to any particular STL implementation.
 2015-10-19 11:21:29 +03:00
Sijia Chen
b29760ee31 remove unneeded parts 2015-10-15 11:31:34 -07:00
Sijia Chen
ade32f5c48 implementation for WelsSleep on WP8.0 
https://rbcommons.com/s/OpenH264/r/1315/
 2015-10-15 11:27:43 -07:00
Sijia Chen
a3f606e58a replacement of std::list for m_cBusyThreads 
https://rbcommons.com/s/OpenH264/r/1320/
 2015-10-15 11:17:29 -07:00
Sijia Chen
bc566f0923 put m_cIdleThreads to CWelsCircleQueue rather than std::map 
https://rbcommons.com/s/OpenH264/r/1313/
 2015-10-15 10:24:48 -07:00
Sijia Chen
eb00d5cb9e change std::list to internal implementation and add the new ut file for CWelsCircleQueue 
https://rbcommons.com/s/OpenH264/r/1310/
 2015-10-15 10:11:29 -07:00
Sijia Chen
757a596e97 add basic threadpool functions 
https://rbcommons.com/s/OpenH264/r/1294/
 2015-10-15 10:04:00 -07:00
Haibo Zhu
03d16bb4d1 Remove UBSAN warnings about negative left shift 2015-10-14 19:43:19 -07:00
HaiboZhu
3067d127aa Merge pull request #2153 from mstorsjo/fix-warnings 
Fix warnings when building for iOS with xcode
 2015-10-13 18:26:56 +08:00
Martin Storsjö
8363d43588 Fix warnings when building for iOS with xcode 2015-10-13 12:27:11 +03:00
Martin Storsjö
837599becc Revert an accidental change that broke MSVC compilation 
This reverts an unrelated part of e7e3b4f37f.

Since the function still is declared as taking an int32_t parameter
in the header, changing the function implementation makes it end
up as a different function.
 2015-10-13 12:15:01 +03:00
Haibo Zhu
e7e3b4f37f Init the string value and add protection for WelsStrcat() 2015-10-10 08:45:48 -07:00
HaiboZhu
d0d7ad57c2 Merge pull request #2116 from mstorsjo/remove-tab-indentation 
Fix indentation to consistently use spaces instead of tabs
 2015-09-16 09:12:07 +08:00
Martin Storsjö
c31e4e23f2 Fix indentation to consistently use spaces instead of tabs 
Also get rid of other stray tabs in scripts.
 2015-09-15 08:41:19 +03:00
fstd
4d063b84cc Build successfully on OpenBSD (which lacks sysctlbyname(3)) 2015-09-12 21:31:39 +02:00
Nathan Kidd
fdabca4cc9 Only use CPU_COUNT if available 
Fixes build error on Linux hosts with GLIBC < 2.6.

Resolves issue #2089
 2015-09-02 18:24:24 -04:00
Guangwei Wang
1f8ef8f0a3 Add new x86 assembly functions to support sub8x8 mode 2015-07-10 09:00:05 +08:00