Martin Storsjö
062937ac5a
Unify the indentation of the new aarch64 assembly
2014-06-17 10:01:23 +03:00
Martin Storsjö
d15534ecb8
Get rid of mixed tabs and spaces in the aarch64 assembly
2014-06-17 10:00:07 +03:00
ruil2
1111757977
Merge pull request #967 from dongzha/Deblock_AArch64
...
add arm 64 deblock code and Unit Test code
2014-06-16 17:19:25 +08:00
huili2
91cd93e5d0
Merge pull request #962 from dongzha/UseIntInRC
...
Use Int instead of Double in Rate Control and Modify anchor SHA1 value
2014-06-13 10:59:50 +08:00
dongzha
f6ce43f83b
Use Int instead of Double in Rate Control and Modify anchor SHA1 value
2014-06-12 17:30:13 +08:00
ruil2
44b048edd6
move trace related info to interface header
2014-06-11 17:05:40 +08:00
Martin Storsjö
dc91e0958b
Integrate the lone function from logging.cpp into welsCodecTrace.cpp
2014-06-11 08:08:56 +03:00
Martin Storsjö
6e5f31214a
Add a method for overriding the logging function in welsCodecTrace
2014-06-11 08:08:56 +03:00
Martin Storsjö
ce8065fe68
Don't use global variables in welsCodecTrace
...
This allows actually honoring the requested log level
properly if there are multiple codec instances within
the same process.
2014-06-11 08:08:56 +03:00
Martin Storsjö
cb5ee6c239
Remove the global log callback function
...
Now all logging should use a non-null log context, allowing to
pass the messages to the right recipient.
2014-06-11 08:08:56 +03:00
Martin Storsjö
8bac9315e6
Expose a SLogContext from welsCodecTrace
2014-06-11 08:08:29 +03:00
Martin Storsjö
4e428ab020
Add a log context to the encoder and decoder contexts
...
This will allow setting non-global logging callbacks, that
are different for each encoder or decoder instance.
2014-06-11 08:08:29 +03:00
Martin Storsjö
c8b81b4239
Only keep one single trace function pointer in welsCodecTrace
2014-06-11 08:08:29 +03:00
Martin Storsjö
20e889fadb
Change CM_WELS_TRACE to take a plain string, not a format and variadic arguments
...
The format string was always "%s" anyway.
2014-06-11 08:08:29 +03:00
Martin Storsjö
5e22d5366e
Remove the unused level parameter to welsStderrLevelTrace
2014-06-11 08:08:29 +03:00
Martin Storsjö
cfc9367610
Remove WelsStderrSetTraceLevel
...
The logging level is checked in welsCodecTrace anyway.
Previously, error logging wasn't ever shown if the trace
level was set to WELS_LOG_ERROR (as it was by default),
since welsStderrLevelTrace required the message level to
be strictly lower than the trace level.
2014-06-11 08:08:29 +03:00
Martin Storsjö
90be3d8215
Don't treat log levels as a bitmask
...
All use of log levels in the library just do a numerical
greater-than comparison between the set log level and the
level of the current message.
2014-06-11 08:08:29 +03:00
Martin Storsjö
4f1ea1c4f8
Remove some unused typedefs
2014-06-11 08:08:29 +03:00
Martin Storsjö
cc65a1d76c
Don't include a [ENCODER]: prefix in all logging
...
The same trace module is used for the decoder now as well.
2014-06-10 10:52:26 +03:00
Martin Storsjö
17fc6bd66e
Remove the unnecessary method WelsTraceModuleIsExist(), which always returned true
2014-06-10 10:52:26 +03:00
Martin Storsjö
d93488448e
Remove some commented out lines
2014-06-10 10:52:26 +03:00
Martin Storsjö
968d87045d
Remove an unnecessary local function
2014-06-10 10:52:26 +03:00
Martin Storsjö
40af75c19d
Remove the unnecessary WelsSet/GetLogLevel functions
...
Nothing actually used the variable that these functions
handled.
2014-06-10 10:52:06 +03:00
Martin Storsjö
ba1de16ac2
Make internal logging variables static
...
This avoids polluting the global namespace.
2014-06-10 09:28:45 +03:00
Martin Storsjö
ab4fe3fdf4
Remove an unused variable
2014-06-10 09:27:54 +03:00
dongzhang
0e0c8b5569
add arm 64 deblock code and Unit Test code
2014-06-10 11:23:51 +08:00
ruil2
4c12f8970c
cleanup trace module
2014-06-10 10:24:45 +08:00
Martin Storsjö
7bc3e944ad
Get rid of uneven spacing after WELS_EXTERN
2014-06-09 11:03:25 +03:00
Martin Storsjö
57f6bcc4b0
Convert all tabs to spaces in assembly sources, unify indentation
...
Previously the assembly sources had mixed indentation consisting
of both spaces and tabs, making it quite hard to read unless
the right tab size was used in the editor.
Tabs have been interpreted as 4 spaces in most cases, matching
the surrounding code.
2014-06-01 01:35:43 +03:00
Martin Storsjö
faaf62afad
Get rid of double spaces in macro declarations
2014-06-01 01:13:01 +03:00
Martin Storsjö
ac03b8b503
Avoid unnecessary tabs in macro declarations
2014-06-01 01:13:01 +03:00
Martin Storsjö
932a38abc0
Reformat the copyright header of deblocking_neon.S
...
This makes it identical to the ones in the other files.
2014-05-31 13:44:21 +03:00
ruil2
14e5d740cd
clean up expand picture.
2014-05-30 11:05:31 +08:00
dongzha
80fdf09b26
Merge pull request #903 from zhilwang/arm64-sad
...
Add Arm64 sad code
2014-05-30 09:26:04 +08:00
Sijia Chen
7413032185
using WelsRound for all the double-int32_t conversion
2014-05-20 14:06:31 +08:00
zhiliang wang
e6c9eb9824
Add Sad arm64 code
2014-05-14 17:06:48 +08:00
Martin Storsjö
3cc01c6239
Use CCASFLAGS when assembling .S sources
...
This allows overriding whether all of CFLAGS should be passed
when assembling.
2014-05-13 19:39:26 +03:00
sijchen
31a4d2aa3e
Merge pull request #829 from dongzha/FixBugforDeblocking
...
Fix a bug in deblocking for neon 32 bit arm implementation for master
2014-05-13 17:21:48 +08:00
Martin Storsjö
6b9167199f
Use the built-in define __linux__ instead of the manually set LINUX
2014-05-12 12:14:33 +03:00
dongzhang
218adc7e29
Fix a bug in deblocking for neon 32 bit arm implementation
2014-05-09 14:06:16 +08:00
Martin Storsjö
6e715ddc10
Make an endif comment match the actual condition
2014-05-08 11:14:24 +03:00
huili2
5ed24f216b
astyle all files
2014-05-05 19:30:21 -07:00
Martin Storsjö
b8eeda1740
Properly back up and restore XMM registers on win64 in WelsSampleSadFour4x4_sse2
2014-05-04 15:47:56 +03:00
Licai Guo
fe5b8d1a69
refine format
2014-05-04 14:51:05 +08:00
Licai Guo
485b2b5b43
Add IntraSad asm code.
...
Enable intraSad ASM code
Refine format
Add X86_ASM pretect for intraSad ASM code UT
remove duplicated code.
2014-05-04 12:12:38 +08:00
Martin Storsjö
23f57adaea
Do full register loads instead of single-lane loads in DeblockLumaEq4H_neon
...
Instead of loading the registers one lane at a time, load full
registers and then transpose them.
This is faster, reducing the runtime for the function from about
506 cycles to 434 cycles (tested on a Cortex A8).
This also avoids an issue which seems like a cpu bug, present
on Sony Xperia T (cpu implementer 0x51 architecture 7 variant 0x1
part 0x04d). On such a device, it seemed like the "vswp q9, q10"
could start executing before the previous
vld4.u8 {d20[x],d21[x],d22[x],d23[x]}, [r3], r1
had finished and written back their result. Changing the
"vswp q9, q10" into "vswp q10, q9", or into separate
"vswp d18, d20; vswp d19, d21" (or the other way around) seemed to
avoid the issue. This happened occasionally (a couple times per
100000 invocations or so).
2014-04-28 10:12:16 +03:00
volvet
c65e286036
Merge pull request #738 from mstorsjo/gnu-aarch64
...
Fix building the aarch64 assembly using gnu binutils
2014-04-25 09:07:43 +08:00
Martin Storsjö
66f58e8357
Add macros for the non-standard mov.16b/mov.8b/ext.16b/ext.8b
...
This fixes building with gnu binutils, which don't support this
nonstandard form of the instructions.
Once Apple's tools support the proper standard form of the
instructions, the code should be updated to use that everywhere
instead, and these macros should be removed.
2014-04-23 11:47:12 +03:00
Martin Storsjö
7cd175d097
Use the correct ext syntax in the gnu version of macros
2014-04-23 11:47:12 +03:00
Martin Storsjö
b13a399ab5
Use a plain "ret" instead of "ret lr"
...
This fixes an issue with assembling with gnu binutils.
2014-04-23 11:47:12 +03:00