generic-library/openh264
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
1,346 Commits 23 Branches 9 Tags 55 MiB
f305096f01
Commit Graph

336 Commits

Author SHA1 Message Date
Licai Guo
4bbe61a783 Merge pull request #537 from mstorsjo/rename-x86-asm 
Rename the asm subdirectories to x86
 2014-03-19 08:51:39 +08:00
Licai Guo
d897d362ab Merge pull request #532 from huili2/WELS_CLIP1 
Modify MACRO WELS_CLIP1 as inline functions
 2014-03-19 08:50:04 +08:00
Martin Storsjö
9586c59b9e Use __APPLE__ instead of APPLE_IOS in the arm assembly sources 2014-03-18 23:15:49 +02:00
Martin Storsjö
ed9c03408f Rename the asm subdirectories to x86 
This is consistent with having the arm assembly in a subdirectory
called arm.
 2014-03-18 23:09:45 +02:00
Ethan Hugg
197423f271 Merge pull request #520 from ylatuya/master 
Fix compiler warnings and remove dead code
 2014-03-18 13:28:02 -07:00
Andoni Morales Alastruey
ae60f1bee9 Fix compiler warnings and remove dead code 
Fix several -Werror=unused-variable and -Werror=unused-but-set-variable
and removed dead code found with this warnings
 2014-03-18 19:15:25 +01:00
volvet
d7b7419040 add scene change detector for further extension 2014-03-18 17:54:58 +08:00
huili2
3b270aa901 remove unncessary cast 2014-03-18 02:15:57 -07:00
huili2
090e8cc1ed modify WELS_CLIP1 to be inline functions 2014-03-18 01:54:25 -07:00
volvet
b21411ad7c Merge pull request #511 from mstorsjo/remove-unused-define 
Remove the unused FORMAT_COFF define
 2014-03-18 16:11:22 +08:00
volvet
fb1958ad13 Merge pull request #519 from mstorsjo/push-xmm-registers 
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64

Reviewed by zhiliang
 2014-03-18 15:04:54 +08:00
Licai Guo
3956a32d41 Merge pull request #524 from sijchen/me_refactor33 
Expand structure of MD and ME
 2014-03-18 12:51:05 +08:00
Licai Guo
37fa5f554e Merge pull request #513 from ruil2/encoder_interface 
Encoder interface
 2014-03-18 09:51:32 +08:00
sijchen
7f0c7daad9 expand structure of MD and ME 2014-03-18 09:47:05 +08:00
volvet
b5353c8455 Merge pull request #516 from mstorsjo/fix-yasm-64bit 
Fix building with yasm in 64 bit mode
 2014-03-18 09:29:42 +08:00
volvet
e75cd2298b Merge pull request #517 from mstorsjo/simplify-x86-asm-func-macro 
Fold ALIGN 16 and the function label into WELS_EXTERN
 2014-03-18 09:29:17 +08:00
Martin Storsjö
29a0c77acf Don't clobber q4-q7 in WelsIntra16x16Combined3Satd_neon 
This is similar to what is done in other neon functions. This
function was missed since it isn't covered by the current
set of unittests.
 2014-03-17 20:04:53 +02:00
Martin Storsjö
3cf52554f7 Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64 
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.

This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.

This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.

This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
 2014-03-17 13:44:33 +02:00
Martin Storsjö
9293f2f947 Remove commented out rodata sections and tables in assembly files 2014-03-17 13:42:18 +02:00
Martin Storsjö
eec968234d Fold ALIGN 16 and the function label into WELS_EXTERN 
This simplifies the structure for all x86 assembly functions,
reducing the amount of duplicated code structure.
 2014-03-17 13:35:00 +02:00
Martin Storsjö
04f5bcd68d Use movsxd in SIGN_EXTENSION 
This is what nasm ended up assembling movsx with 32 bit input to
anyway.

Keep using plain movsx for 16 bit input.

This fixes building with yasm in 64 bit mode.
 2014-03-17 13:26:46 +02:00
Martin Storsjö
f96918283f Remove commented out code for old, 32-bit only x86 assembly function prologues/epilogues 2014-03-17 11:20:11 +02:00
Licai Guo
fc4e0cacec Merge pull request #483 from volvet/develop_b 
use large/medium/similar to define scene change result
 2014-03-17 16:32:31 +08:00
Licai Guo
b5a4d706b9 Merge pull request #496 from mstorsjo/use-sign-extend-macro 
Use the SIGN_EXTENSION macro where possible
 2014-03-17 16:31:03 +08:00
ruil2
36d56b6638 modify EncoderFrame return type. 
commit b99a307ab94183c32a293ad5fda8b0e3323546a0
Author: ruil2 <ruil2@cisco.com>
Date:   Wed Mar 12 13:34:27 2014 +0800

    fix typo
 2014-03-17 10:46:38 +08:00
Licai Guo
2c796337ba Merge pull request #510 from huili2/remove_basemb 
remove BASE_MB related code
 2014-03-17 08:46:25 +08:00
Martin Storsjö
fc260b39e0 Remove the unused FORMAT_COFF define 
Nothing in the project currently sets FORMAT_COFF - the other generic
branch works just fine on windows.
 2014-03-16 17:54:55 +02:00
Martin Storsjö
eb238e6549 Use the SIGN_EXTENSION macro where possible 
This shortens the x86 assembly by 134 lines in total.
 2014-03-16 17:54:24 +02:00
volvet
e606bae0e9 Merge pull request #504 from mstorsjo/fix-function-name-typo 
Fix a typo, Smple -> Sample
 2014-03-16 10:18:03 +08:00
Licai Guo
6f2b98975e Merge pull request #502 from mstorsjo/fix-macro-indentation 
Fix the indentation of some nasm macros
 2014-03-15 07:16:37 +08:00
Martin Storsjö
f4fdb15397 Fix a typo, Smple -> Sample 2014-03-14 23:30:09 +02:00
Martin Storsjö
4d120781c1 Fix the indentation of some nasm macros 2014-03-14 22:26:33 +02:00
Martin Storsjö
b3d04d88a0 Check for the right function pointer 
This code checked whether one function pointer was non-null,
but the went on to call a different function pointer. Check
for the one that actually was called.
 2014-03-14 22:20:40 +02:00
volvet
6da9a9e5c8 Merge pull request #489 from sijchen/me_refactor22 
refactor ME for easier adding other search methods
 2014-03-14 17:53:10 +08:00
huili2
b1f596fd69 remove BASE_MB related code 2014-03-14 02:03:41 -07:00
Martin Storsjö
9199798f22 Fix a typo in a macro name, EXTENTION -> EXTENSION 2014-03-14 10:13:18 +02:00
sijchen
6c3d83a8ac refactor ME for easier adding other search methods 2014-03-14 15:04:35 +08:00
volvet
6714b8ae99 Merge pull request #463 from mstorsjo/dont-clobber-neon-registers 
Avoid clobbering the neon registers q4-q7

Review and verified by zhilwang
 2014-03-14 10:28:55 +08:00
volvet
fc5c48830a fix the condition of scene change flag and comments 2014-03-14 09:53:24 +08:00
volvet
c8761c08ae use large/medium/similar to define scene change result 2014-03-13 10:43:20 +08:00
volvet
8962b7c98b Merge pull request #482 from sijchen/me_refactor1 
mv range setting refactor
 2014-03-13 10:21:39 +08:00
sijchen
d809a7981b mv range setting refactor 2014-03-13 10:18:01 +08:00
volvet
8b907c18fd fix idr interval issue 2014-03-12 17:38:25 +08:00
ruil2
c7f2a0b7f6 3Author: ruil2 <ruil2@cisco.com> 
modify the parameter verification for SM_AUTO_SLICE mode -- uiSliceNum
 iis ignored
 2014-03-12 10:44:13 +08:00
ruil2
7c8ce799c0 fix SM_FIXEDSLCNUM_SLICE bug, add SM_AUTO_SLICE mode 2014-03-11 10:23:46 +08:00
Martin Storsjö
c011890764 Push clobbered neon registers on the stack 
According to the calling convention, the registers q4-q7 should be
preserved by functions. The caller (generated by the compiler) could
be using those registers anywhere for any intermediate data.

Functions that use more than 12 of the qX registers must push
the clobbered registers on the stack in order to be able to restore them
afterwards.

In functions that don't use all 16 registers, but clobber some of
the callee saved registers q4-q7, one or more of them are remapped
to reduce the number of registers that have to be saved/restored.

This incurs a very small (around 0.5%) slowdown in the decoder and
encoder.
 2014-03-10 22:07:36 +02:00
Martin Storsjö
811c647c0e Remap registers to avoid clobbering the neon registers q4-q7 
According to the calling convention, the registers q4-q7 should be
preserved by functions. The caller (generated by the compiler) could
be using those registers anywhere for any intermediate data.

Functions that use 12 or less of the qX registers can avoid
violating the calling convention by simply using other registers instead
of the callee saved registers q4-q7.

This change only remaps the registers used within functions - therefore
this does not affect performance at all. E.g. in functions using
registers q0-q7, we now use q0-q3 and q8-q11 instead.
 2014-03-10 22:07:25 +02:00
ruil2
a922155c9a Merge pull request #466 from sijchen/add_memalign_test 
Add memalign unit test
 2014-03-10 17:25:41 +08:00
sijchen
385128e403 Merge pull request #465 from ruil2/encoder_trace 
use global trace in encoder
reviewed at https://rbcommons.com/s/OpenH264/r/176/
 2014-03-10 17:22:19 +08:00
sijchen
53a570556d add memalign unit test 2014-03-10 16:28:05 +08:00
ruil2
02bafd9320 Merge pull request #445 from mstorsjo/use-thread-param 
Use the iMultipleThreadIdc field from SEncParamExt
 2014-03-10 15:28:04 +08:00
ruil2
86f37f047c Merge pull request #452 from mstorsjo/use-slice-mode-enum 
Use SliceModeEnum as data type for the slice mode fields
 2014-03-10 15:27:04 +08:00
ruil2
2539d6e447 Merge pull request #462 from mstorsjo/fix-typos 
Fix two typos in variable and macro names
 2014-03-10 15:25:20 +08:00
ruil2
ba6b2a8d62 use global trace in encoder 2014-03-10 15:22:40 +08:00
Martin Storsjö
cc7b81f3c3 Fix a typo in arm assembly, LORD -> LOAD 2014-03-09 19:19:38 +02:00
Martin Storsjö
7c435ad295 Remove a stray inline keyword in a function signature comment in x86 assembly 
Assembly functions written in external assembly files is obviously
not inlined.
 2014-03-09 19:18:03 +02:00
Martin Storsjö
8d6b368a1c Remove unnecessary stray __cdecl annotations in function signature comments in x86 assembly 2014-03-09 19:18:02 +02:00
Martin Storsjö
5df2e2a996 Use SliceModeEnum as data type for the slice mode fields 
This makes the use of the field clearer and safer by allowing
the compiler check that users actually assign proper enum
values.
 2014-03-08 00:23:58 +02:00
Martin Storsjö
ce7b00ea72 Get rid of an unnecessary cast by declaring the right pointer type 2014-03-08 00:17:30 +02:00
Ethan Hugg
fb4f677f77 Merge pull request #446 from mstorsjo/remove-unnecessary-public-param 
Move the iCountThreadsNum field to SWelsSvcCodingParam
 2014-03-07 09:18:52 -08:00
Ethan Hugg
7632510209 Merge pull request #450 from mstorsjo/publish-slice-mode-enum 
Move the slice mode enum to the public API
 2014-03-07 09:17:03 -08:00
Martin Storsjö
5f1c207845 Move the slice mode enum to the public header 
This simplifies setting the slice mode in the public API.
 2014-03-07 14:53:29 +02:00
Martin Storsjö
495a4a392e Make ParamValidationExt use the actual type instead of a void pointer 2014-03-07 14:51:34 +02:00
Martin Storsjö
656e4c5c35 Move the iCountThreadsNum field to SWelsSvcCodingParam 
There is no point in the user setting this field, it's only used
as an internal field within the encoder.
 2014-03-07 14:48:38 +02:00
Martin Storsjö
dbc324d5bb Use the iMultipleThreadIdc field from SEncParamExt 2014-03-07 14:47:43 +02:00
Martin Storsjö
5b8ee37162 Merge WelsThreadDestroy into WelsThreadJoin 
Now calling WelsThreadJoin is enough to finish and clean up
the thread on all platforms.

This unifies the thread cleanup code between windows and unix.

Now all of the threading code should use the exact same codepaths
between windows and unix.
 2014-03-07 10:51:28 +02:00
Martin Storsjö
b4aa9be7de Use WelsThreadJoin on windows as well 
This avoids using a separate event just for signalling that
a thread has finished running.
 2014-03-07 10:51:28 +02:00
Martin Storsjö
baaa38737e Use pExitEncodeEvent instead of thread cancellation on unix as well 
This works now that we've got a suitably working implementation
of WelsMultipleEventsWaitSingleBlocking.
 2014-03-07 10:49:39 +02:00
volvet
38a3fada24 Merge pull request #435 from mstorsjo/threadlib-wait-single-unix 
Make WelsMultipleEventsWaitSingleBlocking usable on unix as well
 2014-03-07 16:47:38 +08:00
Licai Guo
1b9aae8434 Merge pull request #439 from zhilwang/mc-arm-asm 
mv mc_neon.S to common,add MC arm code to encoder
 2014-03-07 16:36:48 +08:00
Licai Guo
d986c27b9d remove mc_neon.S from encoder 2014-03-07 16:11:36 +08:00
Licai Guo
71467f948a mv mc_neon.S to common,add MC arm code to encoder 2014-03-07 12:18:58 +08:00
Licai Guo
a4cecd8004 Merge pull request #426 from volvet/simplify-layer-process 
simplify-layer-process
 2014-03-07 10:58:28 +08:00
volvet
14f5518e6a Merge pull request #437 from mstorsjo/fix-arm-encoder-android 
Fix building arm encoder assembly for android
 2014-03-07 10:41:34 +08:00
Martin Storsjö
c0043f7053 Use the three-operand form of add/sub with shift 
When using unified syntax, the two operand form with a shift
isn't allowed.
 2014-03-06 16:21:54 +02:00
Martin Storsjö
f1502c26e3 Don't use WELS_ASM_FUNC_END in the middle of a function 
WELS_ASM_FUNC_END declares the end of the function, and needs
to be paired with WELS_ASM_FUNC_BEGIN.
 2014-03-06 16:21:54 +02:00
Martin Storsjö
ce4fa9e272 Correct the endif comment 
The code block is about HAVE_NEON, not X86_ASM.
 2014-03-06 15:43:04 +02:00
Martin Storsjö
636df2bebb Use WelsMultipleEventsWaitSingleBlocking within the worker thread on unix as well 
This avoids using a separate thread for handling pUpdateMbListEvent
events, and later allowing using the encode exit event on unix instead
of pthread cancellation.
 2014-03-06 15:34:35 +02:00
Martin Storsjö
801da26d1d Use WelsMultipleEventsWaitSingleBlocking with a master event for waiting on finished threads 
This allows using the same codepath for both unix and windows
for distributing new slices to code to threads.

This also improves the performance on unix - instead of waiting
for all the current threads to finish their current slice
before handing out a new slice to each of them (where the threads
that finish first will just wait instead of immediately getting
a new slice to work on), we now use the same logic as on windows.

In one setup, it improves the performance of encoding from ~920 fps
to ~950 fps, and in another setup it goes from ~390 fps to ~660 fps.
(These tests were done with the SM_ROWMB_SLICE mode, which
heavily exercises the code for distributing new slices to the
worker threads.)

The extra WelsEventSignal call on windows where it isn't strictly
necessary doesn't incur any measurable slowdown, so it is kept
without any extra ifdefs to keep the code more readable and unified.
 2014-03-06 15:33:37 +02:00
Martin Storsjö
de32455d87 Remove the timeout parameter from WelsMultipleEventsWaitSingleBlocking 
All users of the function passed the value corresponding to
"infinite", and the (currently unused) unix implementation of it
only supported infinite wait as well.
 2014-03-06 15:03:59 +02:00
volvet
8cc332dea1 Merge pull request #432 from zhilwang/arm-asm 
Arm asm
 2014-03-06 16:50:56 +08:00
volvet
73452e0993 Merge pull request #429 from mstorsjo/simplify-ifdef-with-macro 
Use a macro for conditionally logging based on ENABLE_TRACE_MT
 2014-03-06 16:01:41 +08:00
Licai Guo
7bfe801874 Remove trailing space 2014-03-06 14:55:36 +08:00
Licai Guo
67534b0fc0 arm asm code refine. 2014-03-06 14:30:16 +08:00
Martin Storsjö
fd6f8a83b3 Use a macro for conditionally logging based on ENABLE_TRACE_MT 
This avoids having an extra ifdef around every single WelsLog
call.
 2014-03-06 08:06:34 +02:00
ruil2
28a56a6752 Merge pull request #415 from volvet/remove-useless-mgs-code 
remove un-supported mgs code
 2014-03-06 14:05:04 +08:00
volvet
50fe120a3e simplify-layer-process 2014-03-06 11:19:33 +08:00
ruil2
334c5765c7 remove inter-deblock related parameters 2014-03-06 10:26:53 +08:00
Licai Guo
e7cc8c2780 Add arm asm code for processing. 2014-03-05 16:54:05 +08:00
Martin Storsjö
d4bdef2916 Use an event name that contains the process id 
This reduces the risk for namespace collisions if two processes
run the encoder simultaneously without address space layout
randomization.
 2014-03-05 09:36:46 +02:00
Martin Storsjö
5480ffafdf Use the WelsEventOpen interface with an event name on windows as well 
This unifies the event creation interface, even if the event
name itself is unused on windows, allowing use the exact same
code to initialize events regardless of the actual platform.

Some ifdefs still remain in the event initialization code, since
some events are only used on windows.
 2014-03-05 09:36:04 +02:00
volvet
e9395bbd35 remove un-supported mgs code 2014-03-05 15:17:07 +08:00
volvet
adb27ff0b1 Merge pull request #405 from mstorsjo/simplify-threads 
Adjust WELS_EVENT definitions to allow sharing more code between unix and win32 codepaths
 2014-03-05 12:31:15 +08:00
Licai Guo
ced9e41b5d Merge pull request #399 from volvet/refine-multi-layer-process 
refine-multi-layer-process
 2014-03-05 10:45:35 +08:00
Licai Guo
248f324c62 Add intra predictor arm asm code. 2014-03-05 10:25:15 +08:00
Licai Guo
efcee63692 Remove .DS_Store file. 2014-03-05 10:24:05 +08:00
Licai Guo
bb244d736b Partly add arm asm code to encoder. 2014-03-05 10:24:05 +08:00
volvet
7150adc91b Merge pull request #407 from mstorsjo/do-blocking-wait 
Do a blocking wait with WelsMultipleEventsWaitSingleBlocking
 2014-03-05 09:18:45 +08:00
Martin Storsjö
cf07d61f06 Do a blocking wait with WelsMultipleEventsWaitSingleBlocking 
There is no point in doing a timed wait here - there's no work
that we can do if the wait timed out, and sleeping for 1 ms
inbetween doesn't help, it only adds potential extra latency
to reacting to threads that need more work to do.
 2014-03-04 14:51:33 +02:00
Martin Storsjö
1eaa38b130 Simplify code by allocating the arrays of events and thread handles statically 
This avoids having to malloc a whole lot of separate arrays,
all which are all bounded by MAX_THREADS_NUM.
 2014-03-04 12:17:32 +02:00