platform/upstream/libvpx.git
13 years agoUse correct 32 bit comparisons for SAD breakout.
James Berry [Thu, 7 Apr 2011 19:03:49 +0000 (15:03 -0400)]
Use correct 32 bit comparisons for SAD breakout.

Rax updated to eax to avoid uninitialized memory
usage.

Change-Id: Iedb953f104329ede2a786fc648a47f1be2f3798a

13 years agoMerge "use asm_offsets with vp8_fast_quantize_b_sse2"
Johann [Wed, 6 Apr 2011 17:53:55 +0000 (10:53 -0700)]
Merge "use asm_offsets with vp8_fast_quantize_b_sse2"

13 years agoMerge "Minor modification"
Yunqing Wang [Wed, 6 Apr 2011 13:12:13 +0000 (06:12 -0700)]
Merge "Minor modification"

13 years agoMinor modification
Yunqing Wang [Wed, 6 Apr 2011 13:08:47 +0000 (09:08 -0400)]
Minor modification

A small change.

Change-Id: I2e7726e58370a95d0319361f4f6ad231138d1328

13 years agouse asm_offsets with vp8_fast_quantize_b_sse2
Johann [Thu, 24 Mar 2011 17:31:10 +0000 (13:31 -0400)]
use asm_offsets with vp8_fast_quantize_b_sse2

on the same order as the regular quantize change: ~2%

Change-Id: I5c9eec18e89ae7345dd96945cb740e6f349cee86

13 years agoFixed unused variable warnings for firstpass.c
Scott LaVarnway [Mon, 4 Apr 2011 18:18:31 +0000 (14:18 -0400)]
Fixed unused variable warnings for firstpass.c

Change-Id: I8378a9a541ade2f098359a7b20fa08e6c1596d80

13 years agoMerge "Slightly simplify vp8_decode_mb_tokens."
John Koleszar [Mon, 4 Apr 2011 15:58:25 +0000 (08:58 -0700)]
Merge "Slightly simplify vp8_decode_mb_tokens."

13 years agoMerge "tweak vp8_regular_quantize_b_sse2"
Johann [Mon, 4 Apr 2011 15:56:25 +0000 (08:56 -0700)]
Merge "tweak vp8_regular_quantize_b_sse2"

13 years agoSlightly simplify vp8_decode_mb_tokens.
Gaute Strokkenes [Mon, 4 Apr 2011 15:47:22 +0000 (16:47 +0100)]
Slightly simplify vp8_decode_mb_tokens.

Change-Id: I0058ba7dcfc50a3374b712197639ac337f8726be

13 years agoMerge "Use full-pixel MV in mvsadcost calculation"
Yunqing Wang [Mon, 4 Apr 2011 15:40:51 +0000 (08:40 -0700)]
Merge "Use full-pixel MV in mvsadcost calculation"

13 years agoMerge "support obj_int_extract on cygwin"
John Koleszar [Mon, 4 Apr 2011 15:29:50 +0000 (08:29 -0700)]
Merge "support obj_int_extract on cygwin"

13 years agoUse full-pixel MV in mvsadcost calculation
Yunqing Wang [Fri, 1 Apr 2011 20:41:58 +0000 (16:41 -0400)]
Use full-pixel MV in mvsadcost calculation

MV sad cost error is only used in full-pixel motion search,
which only need full-pixel resolution instead of quarter-pixel
resolution. This change reduced mvsadcost table size, and
removed unneccessary pamameter passing since this table is
constant once it is generated.

Change-Id: I9f931e55f6abc3c99011321f1dfb2f3562e6f6b0

13 years agosupport obj_int_extract on cygwin
Johann [Thu, 31 Mar 2011 20:35:22 +0000 (16:35 -0400)]
support obj_int_extract on cygwin

cygwin doesn't support _sopen. drop down to the lowest common
denominator and merge main for all platforms. this also opens the door
for supporting multiple object formats with a single binary.

Change-Id: I7cd45091639d447434e6d5db2e19cfc9988f8630

13 years agoMerge "vpxenc: die on realloc failures"
John Koleszar [Fri, 1 Apr 2011 14:55:55 +0000 (07:55 -0700)]
Merge "vpxenc: die on realloc failures"

13 years agotweak vp8_regular_quantize_b_sse2
Johann [Fri, 1 Apr 2011 13:58:23 +0000 (09:58 -0400)]
tweak vp8_regular_quantize_b_sse2

rather than look up rc in the zig zag table, embed it in the macro. this
also allows us to shuffle some values in the macro and keep *d in rsi

gains of about the same order as the obj_int_extract implementation: ~2%

Change-Id: Ib7252dd10eee66e0af8b0e567426122781dc053d

13 years agoMerge "Wrapper function removed from vp8_subtract_b_neon function call"
Johann [Fri, 1 Apr 2011 12:47:21 +0000 (05:47 -0700)]
Merge "Wrapper function removed from vp8_subtract_b_neon function call"

13 years agoWrapper function removed from vp8_subtract_b_neon function call
Tero Rintaluoma [Wed, 30 Mar 2011 10:45:59 +0000 (13:45 +0300)]
Wrapper function removed from vp8_subtract_b_neon function call

Address calculations moved from encodemb_arm.c file to neon
optimized assembly function to save cycles in function calls.
 - vp8_subtract_b_neon_func replaced with vp8_subtract_b_neon
   that contains all needed address calculations
 - unnecessary file encodemb_arm.c removed
 - consistent with ARMv6 optimized version

Change-Id: I6cbc1a2670b56c2077f59995fcf8f70786b4990b

13 years agoMerge "ARMv6 optimized subtract functions"
Johann [Thu, 31 Mar 2011 15:40:10 +0000 (08:40 -0700)]
Merge "ARMv6 optimized subtract functions"

13 years agoMerge changes I4e32a8fb,Ic6a9d4c5
John Koleszar [Thu, 31 Mar 2011 13:21:11 +0000 (06:21 -0700)]
Merge changes I4e32a8fb,Ic6a9d4c5

* changes:
  Generate a vpx.pc file for pkg-config.
  Export the version string as a makefile variable.

13 years agoGenerate a vpx.pc file for pkg-config.
Ralph Giles [Mon, 28 Mar 2011 19:04:51 +0000 (12:04 -0700)]
Generate a vpx.pc file for pkg-config.

Rules are added to libs.mk to generate a vpx.pc, which is
installed as pkgconfig/vpx.pc under the target library directory.
This also requires the install path prefix be exported directly
in config.mk.

Some systems use a tool called pkg-config to query information
about intalled libraries or other resources, based on database
files provided by the packages themselves at install time.

Providing such a file for libvpx simplifies integration with
other build systems, and provides an easy avenue for developers
to test against their own builds of the library.

Change-Id: I4e32a8fbb53fc331aa95eb207c63dd70a76d18ed

13 years agoExport the version string as a makefile variable.
Ralph Giles [Mon, 28 Mar 2011 18:36:53 +0000 (11:36 -0700)]
Export the version string as a makefile variable.

The configure script exports the major/minor/patch version
numbers, but didn't make the full version string available
to Makefile recipes and rules, the way it is available to
C code from vpx_version.h.

Change-Id: Ic6a9d4c574a6ea66a50c928f4eedeb91d7668eb5

13 years agoFix: lpf semaphore was signaled in single threaded run
Attila Nagy [Wed, 30 Mar 2011 12:50:17 +0000 (15:50 +0300)]
Fix: lpf semaphore was signaled in single threaded run

After picking filter level, post the loopfilter semaphore
just when multiple threads are in use.

Change-Id: If7bfb64601d906adef703f454dafc25e978b93c6

13 years agovpxenc: die on realloc failures
John Koleszar [Wed, 30 Mar 2011 01:44:19 +0000 (21:44 -0400)]
vpxenc: die on realloc failures

Identified as a possible cause of issue #308, the code was silently
ignoring realloc failures, which would lead to corruption, memory
leaks, and likely a crash. The best we can do in this case is die
gracefully.

Change-Id: Ie5f6a853d367015be5b9712bd742778f3baeefd9

13 years agoMerge "Half pixel variance further optimized for ARMv6"
Johann [Tue, 29 Mar 2011 19:14:54 +0000 (12:14 -0700)]
Merge "Half pixel variance further optimized for ARMv6"

13 years agoMerge "Fix a crash while enabling shared (--enable-shared)"
Yunqing Wang [Tue, 29 Mar 2011 16:04:22 +0000 (09:04 -0700)]
Merge "Fix a crash while enabling shared (--enable-shared)"

13 years agoFix a crash while enabling shared (--enable-shared)
Yunqing Wang [Tue, 29 Mar 2011 15:31:06 +0000 (11:31 -0400)]
Fix a crash while enabling shared (--enable-shared)

Fixed a bug in SSSE3 sub-pixel filter functions.

Change-Id: I2e2126652970eb78307ffcefcace1efd5966fb0a

13 years agouse GLOBAL correctly on 32bit shared libraries
Johann [Tue, 29 Mar 2011 14:28:30 +0000 (10:28 -0400)]
use GLOBAL correctly on 32bit shared libraries

http://code.google.com/p/webm/issues/detail?id=309

Change-Id: I6fce9e2f74bc09a9f258df7f91ab599812324e8c

13 years agoMerge "configure: enable unused variable warnings"
John Koleszar [Tue, 29 Mar 2011 14:38:04 +0000 (07:38 -0700)]
Merge "configure: enable unused variable warnings"

13 years agoARMv6 optimized subtract functions
Tero Rintaluoma [Tue, 29 Mar 2011 13:52:00 +0000 (16:52 +0300)]
ARMv6 optimized subtract functions

Adds following ARMv6 optimized functions to encoder:
  - vp8_subtract_b_armv6
  - vp8_subtract_mby_armv6
  - vp8_subtract_mbuv_armv6

Gives 1-5% speed-up depending on input sequence and encoding
parameters. Functions have one stall cycle inside the loop body
on Cortex pipeline.

Change-Id: I19cca5408b9861b96f378e818eefeb3855238639

13 years agoadd asm_enc_offsets.c for all targets
Johann [Thu, 24 Mar 2011 22:12:21 +0000 (18:12 -0400)]
add asm_enc_offsets.c for all targets

now that we need asm_enc_offsets.c for x86 and arm and it is
harmless to build it for other targets, add it unconditionally

Change-Id: I320c5220afd94fee2b98bda9ff4e5e34c67062f3

13 years agoHalf pixel variance further optimized for ARMv6
Tero Rintaluoma [Mon, 28 Mar 2011 06:51:51 +0000 (09:51 +0300)]
Half pixel variance further optimized for ARMv6

Half pixel interpolations optimized in variance calculations. Separate
function calls to vp8_filter_block2d_bil_x_pass_armv6 are avoided.On
average, performance improvement is 6-7% for VGA@30fps sequences.

Change-Id: Idb5f118a9d51548e824719d2cfe5be0fa6996628

13 years agoMerge "use asm_offsets with vp8_regular_quantize_b_sse2"
Johann [Thu, 24 Mar 2011 18:06:36 +0000 (11:06 -0700)]
Merge "use asm_offsets with vp8_regular_quantize_b_sse2"

13 years agouse asm_offsets with vp8_regular_quantize_b_sse2
Johann [Thu, 10 Feb 2011 19:57:43 +0000 (14:57 -0500)]
use asm_offsets with vp8_regular_quantize_b_sse2

remove helper function and avoid shadowing all the arguments to the
stack on 64bit systems

when running with --good --cpu-used=0:
~2% on linux x86 and x86_64
~2% on win32 x86 msys and visual studio
more on darwin10 x86_64
significantly more on
x86_64-win64-vs9

Change-Id: Ib7be12edf511fbf2922f191afd5b33b19a0c4ae6

13 years agoMerge "ARMv6 optimized fdct4x4"
Johann [Wed, 23 Mar 2011 14:52:51 +0000 (07:52 -0700)]
Merge "ARMv6 optimized fdct4x4"

13 years agoMerge "Allow specifying --end-usage by enum name"
John Koleszar [Mon, 21 Mar 2011 19:29:11 +0000 (12:29 -0700)]
Merge "Allow specifying --end-usage by enum name"

13 years agoMerge "vpx_codec_dec_init: check that the iface is a decoder"
John Koleszar [Mon, 21 Mar 2011 16:12:58 +0000 (09:12 -0700)]
Merge "vpx_codec_dec_init: check that the iface is a decoder"

13 years agovpx_codec_dec_init: check that the iface is a decoder
John Koleszar [Mon, 21 Mar 2011 15:46:15 +0000 (11:46 -0400)]
vpx_codec_dec_init: check that the iface is a decoder

Make sure the given interface is actually a decoder interface before
initializing it.

Change-Id: Ie48d737f2956cc2f0891666de5ea87251e96bc49

13 years agoMerge "Fix multithreaded encoding for 1 MB wide frame"
Yunqing Wang [Mon, 21 Mar 2011 14:41:31 +0000 (07:41 -0700)]
Merge "Fix multithreaded encoding for 1 MB wide frame"

13 years agoRemove unused vp8_get4x4sse_cs_mmx declaration
John Koleszar [Fri, 11 Mar 2011 18:01:02 +0000 (13:01 -0500)]
Remove unused vp8_get4x4sse_cs_mmx declaration

This declaration did not match the prototype_sad() prototype, but was
unused in this translation unit, so it is removed instead. Fixes
issue 290.

Change-Id: I168854f88a85f73ca9aaf61d1e5dc0f43fc3fdb3

13 years agoMerge "Increase static linkage, remove unused functions"
John Koleszar [Mon, 21 Mar 2011 11:51:51 +0000 (04:51 -0700)]
Merge "Increase static linkage, remove unused functions"

13 years agoAllow specifying --end-usage by enum name
John Koleszar [Mon, 21 Mar 2011 11:50:42 +0000 (07:50 -0400)]
Allow specifying --end-usage by enum name

Map an enum to the --end-usage values, so you can specify
--end-usage=cq instead of --end-usage=2. The numerical values still
work for historical scripts, etc, but this is more user friendly.

Change-Id: I445ecd9638f801f5924a71eabf449bee293cdd34

13 years agoARMv6 optimized fdct4x4
Tero Rintaluoma [Mon, 21 Mar 2011 11:33:45 +0000 (13:33 +0200)]
ARMv6 optimized fdct4x4

Optimized fdct4x4 (8x4) for ARMv6 instruction set.
  - No interlocks in Cortex-A8 pipeline
  - One interlock cycle in ARM11 pipeline
  - About 2.16 times faster than current C-code compiled with -O3

Change-Id: I60484ecd144365da45bb68a960d30196b59952b8

13 years agoFix multithreaded encoding for 1 MB wide frame
Attila Nagy [Fri, 18 Mar 2011 08:44:08 +0000 (10:44 +0200)]
Fix multithreaded encoding for 1 MB wide frame

Thread synchronization was not correct when frame width was 1 MB.
Number of allocated encoding threads is limited by the sync_range.
There is no point having more because each thread lags sync_range MBs
behind the thread processing the row above.

http://code.google.com/p/webm/issues/detail?id=302

Change-Id: Icaf67a883beecc5ebf2f11e9be47b6997fdf6f26

13 years agoIncrease static linkage, remove unused functions
John Koleszar [Thu, 17 Mar 2011 21:07:59 +0000 (17:07 -0400)]
Increase static linkage, remove unused functions

A large number of functions were defined with external linkage, even
though they were only used from within one file. This patch changes
their linkage to static and removes the vp8_ prefix from their names,
which should make it more obvious to the reader that the function is
contained within the current translation unit. Functions that were
not referenced were removed.

These symbols were identified by:

  $ nm -A libvpx.a | sort -k3 | uniq -c -f2 | grep ' [A-Z] ' \
    | sort | grep '^ *1 '

Change-Id: I59609f58ab65312012c047036ae1e0634f795779

13 years agoSet bounds from the array when iterating mmaps.
Ralph Giles [Thu, 17 Mar 2011 07:23:36 +0000 (00:23 -0700)]
Set bounds from the array when iterating mmaps.

The mmap allocation code in vp8_dx_iface.c was inconsistent.
The static array vp8_mem_req_segs defines two descriptors,
but only the first is real. The second is a sentinel and
isn't actually allocated, so vpx_codec_alg_priv is declared
with mmaps[NELEMENTS(vp8_mem_req_segs)-1]. Some functions
use this reduced upper bound when iterating though the mmap
array, but these two functions did not.

Instead, this commit calls NELEMENTS(...->mmaps) to directly
query the bounds of the dereferenced array.

This fixes an array-bounds warning from gcc 4.6 on
vp8_xma_set_mmap.

Change-Id: I918e2721b401d134c1a9764c978912bdb3188be1

13 years agoRemove commented-out VP6 code from vp8_finalize_mmaps
Ralph Giles [Thu, 17 Mar 2011 01:30:22 +0000 (18:30 -0700)]
Remove commented-out VP6 code from vp8_finalize_mmaps

Change-Id: I48642c380353043bed96026f56de5908fcee270a

13 years agoMerge "Fix "used uninitialized" warning in vp8_pack_bitstream()"
John Koleszar [Thu, 17 Mar 2011 21:25:04 +0000 (14:25 -0700)]
Merge "Fix "used uninitialized" warning in vp8_pack_bitstream()"

13 years agoapple: include proper mach primatives
John Koleszar [Thu, 29 Jul 2010 21:04:39 +0000 (17:04 -0400)]
apple: include proper mach primatives

Fixes implicit declaration warning for 'mach_task_self'. This change
is an update to Change I9991dedd1ccfddc092eca86705ecbc3b764b799d,
which fixed this issue for the decoder but not the encoder.

Change-Id: I9df033e81f9520c4f975b7a7cf6c643d12e87c96

13 years agoRemove echoing in obj_int_extract rule
Attila Nagy [Wed, 16 Mar 2011 10:56:52 +0000 (12:56 +0200)]
Remove echoing in obj_int_extract rule

Change-Id: I9965170b40e2f32e9d84895c33a529b0d7dacdc1

13 years agoAdd vp8_variance8x8_armv6 and vp8_sub_pixel_variance8x8_armv6 functions
Attila Nagy [Wed, 9 Mar 2011 12:26:24 +0000 (14:26 +0200)]
Add vp8_variance8x8_armv6 and vp8_sub_pixel_variance8x8_armv6 functions

Change-Id: I08edaffc62514907fa5e90e1689269e467c857f5

13 years agoAvoid misspelling "dependent".
Gaute Strokkenes [Tue, 15 Mar 2011 12:20:54 +0000 (12:20 +0000)]
Avoid misspelling "dependent".

Change-Id: Ib0c280e1fcfd977e11e4390807b2c8077a87500c

13 years agoMerge "Fix an unused variable warning."
John Koleszar [Mon, 14 Mar 2011 21:13:53 +0000 (14:13 -0700)]
Merge "Fix an unused variable warning."

13 years agoImprove grammar in a comment.
Ralph Giles [Mon, 14 Mar 2011 17:50:19 +0000 (10:50 -0700)]
Improve grammar in a comment.

Change-Id: I18bfda6d420626f2718e096e338c1d0bf0ba029d

13 years agoobj_int_extract: win64 does not prefix symbols
Johann [Mon, 14 Mar 2011 15:10:24 +0000 (11:10 -0400)]
obj_int_extract: win64 does not prefix symbols

obj_int_extract was unconditionally skipping the first character in the
symbol. make sure it's actually an '_' first

Change-Id: Icfe527eb8a0028faeabaa1dcedf8cd8f51c92754

13 years agoMerge "Add vp8_mse16x16_armv6 function"
Johann [Mon, 14 Mar 2011 19:47:42 +0000 (12:47 -0700)]
Merge "Add vp8_mse16x16_armv6 function"

13 years agoAdd vp8_mse16x16_armv6 function
Attila Nagy [Tue, 8 Mar 2011 12:48:20 +0000 (14:48 +0200)]
Add vp8_mse16x16_armv6 function

Change-Id: I77e9f2f521a71089228f96e2db72524189364ffb

13 years agoFix build with xcode4 and simplify GLOBAL.
Rafael Ávila de Espíndola [Sat, 5 Mar 2011 20:36:40 +0000 (15:36 -0500)]
Fix build with xcode4 and simplify GLOBAL.

Without this change I get link errors in firefox's libxul. It looks
like the linker expect a particular pattern for getting the GOT. This
patch changes webm to use the same pattern used by the compiler.

Change-Id: Iea8c2e134ad45c1dc7d221ff885a8429bfa4e057

13 years agoMerge "Move build_intra_predictors_mby to RTCD framework"
Johann [Fri, 11 Mar 2011 18:23:48 +0000 (10:23 -0800)]
Merge "Move build_intra_predictors_mby to RTCD framework"

13 years agoMove build_intra_predictors_mby to RTCD framework
John Koleszar [Fri, 11 Mar 2011 16:35:38 +0000 (11:35 -0500)]
Move build_intra_predictors_mby to RTCD framework

The vp8_build_intra_predictors_mby and vp8_build_intra_predictors_mby_s
functions had global function pointers rather than using the RTCD
framework. This can show up as a potential data race with tools such as
helgrind. See https://bugzilla.mozilla.org/show_bug.cgi?id=640935
for an example.

Change-Id: I29c407f828ac2bddfc039f852f138de5de888534

13 years agoMerge "ARMv6 optimized quantization"
Johann [Fri, 11 Mar 2011 16:29:00 +0000 (08:29 -0800)]
Merge "ARMv6 optimized quantization"

13 years agoMerge "Only enable ssim_opt.asm on X86_64"
John Koleszar [Fri, 11 Mar 2011 16:28:05 +0000 (08:28 -0800)]
Merge "Only enable ssim_opt.asm on X86_64"

13 years agoOnly enable ssim_opt.asm on X86_64
John Koleszar [Fri, 11 Mar 2011 16:27:08 +0000 (11:27 -0500)]
Only enable ssim_opt.asm on X86_64

Fix compiling on 32 bit x86.

Change-Id: I6210573e1d9287ac49acbe3d7e5181e309316107

13 years agoClean up of vp8_init_config()
Paul Wilkins [Fri, 11 Mar 2011 14:51:40 +0000 (14:51 +0000)]
Clean up of vp8_init_config()

Clean up vp8_init_config() a bit and remove null pointer case,
as this code can't be called any more and is not an adequate
trap anyway, as a null pointer would cause exceptions before
hitting the test.

Change-Id: I937c00167cc039b3aa3f645f29c319d58ae8d3ee

13 years agoMerge "1 Pass CQ and VBR bug fixes"
John Koleszar [Fri, 11 Mar 2011 16:06:09 +0000 (08:06 -0800)]
Merge "1 Pass CQ and VBR bug fixes"

13 years ago1 Pass CQ and VBR bug fixes
Paul Wilkins [Thu, 10 Mar 2011 16:11:39 +0000 (16:11 +0000)]
1 Pass CQ and VBR bug fixes

Issue 291 highlighted  the fact that CQ mode was not working
as expected in 1 pass mode,

This commit fixes that specific problem but in so doing I also
uncovered an overflow issue in the VBR code for 1 pass and
some data values not being correctly initialized.

For some clips (particularly short clips), the resulting
improvement is dramatic.

Change-Id: Ieefd6c6e4776eb8f1b0550dbfdfb72f86b33c960

13 years agoMerge "Fix incorrect macroblock counts in twopass rate control"
John Koleszar [Fri, 11 Mar 2011 14:06:04 +0000 (06:06 -0800)]
Merge "Fix incorrect macroblock counts in twopass rate control"

13 years agoMerge "Align SAD output array to be 16-byte aligned"
Yunqing Wang [Fri, 11 Mar 2011 13:56:02 +0000 (05:56 -0800)]
Merge "Align SAD output array to be 16-byte aligned"

13 years agoMerge "vp8cx - psnr converted to call assemblerized sse"
John Koleszar [Fri, 11 Mar 2011 13:54:00 +0000 (05:54 -0800)]
Merge "vp8cx - psnr converted to call assemblerized sse"

13 years agoMerge "vp8cx- alternate ssim function with optimizations"
John Koleszar [Fri, 11 Mar 2011 13:53:41 +0000 (05:53 -0800)]
Merge "vp8cx- alternate ssim function with optimizations"

13 years agovp8cx - psnr converted to call assemblerized sse
Jim Bankoski [Tue, 8 Mar 2011 20:23:40 +0000 (15:23 -0500)]
vp8cx - psnr converted to call assemblerized sse

Change-Id: Ie388d4618c44b131f96b9fe526618b457f020dfa

13 years agovp8cx- alternate ssim function with optimizations
Jim Bankoski [Tue, 8 Mar 2011 14:05:18 +0000 (09:05 -0500)]
vp8cx- alternate ssim function with optimizations

Change-Id: I91921b0a90dbaddc7010380b038955be347964b3

13 years agoAlign SAD output array to be 16-byte aligned
Yunqing Wang [Fri, 11 Mar 2011 13:24:23 +0000 (08:24 -0500)]
Align SAD output array to be 16-byte aligned

Use aligned store.

Change-Id: Icab4c0c53da811d0c52bb7e8134927f249ba2499

13 years agoMerge "Encoder loopfilter running in its own thread"
Yunqing Wang [Fri, 11 Mar 2011 12:55:05 +0000 (04:55 -0800)]
Merge "Encoder loopfilter running in its own thread"

13 years agoFix "used uninitialized" warning in vp8_pack_bitstream()
Attila Nagy [Fri, 11 Mar 2011 10:34:57 +0000 (12:34 +0200)]
Fix "used uninitialized" warning in vp8_pack_bitstream()

Change-Id: Iadcbdba717439f47a2c24e65fd69a3a1464174b5

13 years agoEncoder loopfilter running in its own thread
Attila Nagy [Fri, 25 Feb 2011 11:42:05 +0000 (13:42 +0200)]
Encoder loopfilter running in its own thread

In multithreaded mode the loopfilter is running in its own thread (filter level
calculation and frame filtering). Filtering is mostly done in parallel with the
bitstream packing. Before starting the packing the loopfilter level has
to be calculated. Also any needed reference frame copying is done in the
filter thread.

Currently the encoder will create n+1 threads, where n > 1 is the number of
threads specified by application  and 1 is the extra filter thread. With n = 1
the encoder runs in single thread mode. There will never be more than n threads
running concurrently.

Change-Id: I4fb29b559a40275d6d3babb8727245c40fba931b

13 years agoARMv6 optimized quantization
Tero Rintaluoma [Mon, 7 Mar 2011 09:12:56 +0000 (11:12 +0200)]
ARMv6 optimized quantization

Adds new ARMv6 optimized function vp8_fast_quantize_b_armv6
to the encoder.

Change-Id: I40277ec8f82e8a6cbc453cf295a0cc9b2504b21e

13 years agoobj_int_extract for Visual Studio
Johann [Wed, 2 Mar 2011 14:44:39 +0000 (09:44 -0500)]
obj_int_extract for Visual Studio

Enable extraction of assembly offsets from compiled examples in MSVS.
This will allow us to remove some stub functions from x86 assembly since
we will be able to reliably determine structure offsets at compile time.

see ARM code for examples:
vp8/encoder/arm/armv5te/
vpx_scale/arm/neon/

Change-Id: I1852dc6b56ede0bf1dddb5552196222a7c6a902f

13 years agoAdded missing format specifier in print statement
Adrian Grange [Thu, 10 Mar 2011 20:43:49 +0000 (12:43 -0800)]
Added missing format specifier in print statement

Printout of firstpass stats for frame had one fewer
format specifiers than arguments.

Change-Id: I5a42c85aa79c471e1a70afd75e24a91546b7a1cd

13 years agoRemoved firstpass motion map
Adrian Grange [Thu, 10 Mar 2011 19:32:48 +0000 (11:32 -0800)]
Removed firstpass motion map

The firstpass motion map consists of an 8-bit flag for
each MB indicating how strongly the firstpass code
believes it should be filtered during the second pass
ARNR filtering.

For long or large format material the motion map can
become extremely large and hamper the operation of
the encoding process.

This change removes the motion map altogether, leaving
the second pass to rely on the magnitude of the motion
compensated error to determine the filter weight to
use for the MB during ARNR filtering.

Tests on the derf set indicate that the effect of this
change is neutral, with some small wins and losses. The
motion map has therefore been removed based on
a cost/benefit evaluation.

Change-Id: I53e07d236f5ce09a6f0c54e7c4ffbb490fb870f6

13 years agoFix incorrect macroblock counts in twopass rate control
James Berry [Thu, 10 Mar 2011 16:13:44 +0000 (11:13 -0500)]
Fix incorrect macroblock counts in twopass rate control

The previous calculation of macroblock count (w*h)/256
is not correct when the width/height are not multiples of
16. Use the precalculated macroblock count from
cpi->common instead. This manifested itself as a divide
by zero when the number of pixels was less than 256.
num_mbs updated in estimate_max_q, estimate_q,
 estimate_kf_group_q, and estimate_cq

Change-Id: I92ff98587864c801b1ee5485cfead964673a9973

13 years agoMerge "Add vp8_sub_pixel_variance16x8_ssse3 function"
Yunqing Wang [Wed, 9 Mar 2011 20:26:37 +0000 (12:26 -0800)]
Merge "Add vp8_sub_pixel_variance16x8_ssse3 function"

13 years agoMerge branch 'bali'
John Koleszar [Wed, 9 Mar 2011 19:11:28 +0000 (14:11 -0500)]
Merge branch 'bali'

Change-Id: Icf18b4981afb12ef255fca431d4ba45860dd22c9

13 years agoAdd missing filter.h to build system
John Koleszar [Wed, 9 Mar 2011 18:43:31 +0000 (13:43 -0500)]
Add missing filter.h to build system

Missing file causes 'make dist' to not include a complete copy of the
source.

Change-Id: I3f55aeb5a86d0e81234e4e4588cb8086ba4cfc4a

13 years agoMerge "fix obj_int_extract for MinGW"
Johann [Wed, 9 Mar 2011 17:44:48 +0000 (09:44 -0800)]
Merge "fix obj_int_extract for MinGW"

13 years agoAdd vp8_sub_pixel_variance16x8_ssse3 function
Yunqing Wang [Wed, 9 Mar 2011 16:16:30 +0000 (11:16 -0500)]
Add vp8_sub_pixel_variance16x8_ssse3 function

Added SSSE3 function

Change-Id: I8c304c92458618d93fda3a2f62bd09ccb63e75ad

13 years agoRemove unused functions
Yunqing Wang [Wed, 9 Mar 2011 15:45:03 +0000 (10:45 -0500)]
Remove unused functions

Removed some unused functions

Change-Id: Ifdfc27453e53cfc75997b38492901d193a16b245

13 years agoMerge "Improve SSE2 half-pixel filter funtions"
Yunqing Wang [Wed, 9 Mar 2011 15:23:06 +0000 (07:23 -0800)]
Merge "Improve SSE2 half-pixel filter funtions"

13 years agoMerge "Configuration updates:Making a clear distinction between Init and Change"
John Koleszar [Wed, 9 Mar 2011 13:07:11 +0000 (05:07 -0800)]
Merge "Configuration updates:Making a clear distinction between Init and Change"

13 years agoFix an unused variable warning.
Ralph Giles [Tue, 8 Mar 2011 22:51:23 +0000 (14:51 -0800)]
Fix an unused variable warning.

Move the update of the loopfilter info to the same block where it
is used. GCC 4.5 is not able trace the initialization of the local
filter_info across the other calls between the two conditionals on
pbi->common and issues an uninitialized variable warning.

Change-Id: Ie4487b3714a096b3fb21608f6b0c74e745e3c6fc

13 years agofix obj_int_extract for MinGW
Johann [Tue, 8 Mar 2011 22:41:45 +0000 (17:41 -0500)]
fix obj_int_extract for MinGW

failed to find headers in the source directory

output to stdout instead of a hardcoded file

MinGW doesn't support _sopen_s

_fstat catches non-existant files

Change-Id: I24e0aacc6f6f26e6bcfc25f9ee7821aa3c8cc7e7

13 years agoImprove SSE2 half-pixel filter funtions
Yunqing Wang [Tue, 8 Mar 2011 21:25:06 +0000 (16:25 -0500)]
Improve SSE2 half-pixel filter funtions

Rewrote these functions to process 16 pixels once instead of 8.

Change-Id: Ic67e80124467a446a3df4cfecfb76a4248602adb

13 years agoMerge "64bit mach-o support"
Johann [Tue, 8 Mar 2011 20:29:32 +0000 (12:29 -0800)]
Merge "64bit mach-o support"

13 years agoMerge "Add zero offset checking in SSE2 sub-pixel filter function"
Yunqing Wang [Tue, 8 Mar 2011 20:26:58 +0000 (12:26 -0800)]
Merge "Add zero offset checking in SSE2 sub-pixel filter function"

13 years agoAdd zero offset checking in SSE2 sub-pixel filter function
Yunqing Wang [Tue, 8 Mar 2011 20:22:07 +0000 (15:22 -0500)]
Add zero offset checking in SSE2 sub-pixel filter function

Skip filter at zero offset.

Change-Id: I95fc7e211869bc0ab5bcfb7ab2e3259d1c0ccf38

13 years agoMerge "Write SSSE3 sub-pixel filter function"
Yunqing Wang [Tue, 8 Mar 2011 18:58:30 +0000 (10:58 -0800)]
Merge "Write SSSE3 sub-pixel filter function"

13 years agoWrite SSSE3 sub-pixel filter function
Yunqing Wang [Fri, 4 Mar 2011 00:02:45 +0000 (19:02 -0500)]
Write SSSE3 sub-pixel filter function

1. Process 16 pixels at one time instead of 8.
2. Add check for both xoffset =0 and yoffset=0, which happens
   during motion search.
This change gave encoder 1%~3% performance gain.

Change-Id: Idaa39506b48f4f8b2fbbeb45aae8226fa32afb3e

13 years ago64bit mach-o support
Johann [Wed, 23 Feb 2011 21:08:10 +0000 (16:08 -0500)]
64bit mach-o support

enable parsing 64bit mach-o files (OS X)

also fixes --enable-debug issue!

Change-Id: I250ee69745cd2365e3e63264f9365cd58fbb6678

13 years ago64bit elf support
Johann [Wed, 23 Feb 2011 22:07:35 +0000 (17:07 -0500)]
64bit elf support

enable parsing 64bit elf files

Change-Id: I7981f4769cf1b822f288fe2e32166254e4394bab

13 years agoFix a multi-line format-string warning.
Ralph Giles [Tue, 8 Mar 2011 15:14:12 +0000 (07:14 -0800)]
Fix a multi-line format-string warning.

GCC 4.5 and 4.6 both issue a warning about the multi-line format
string introduced in bc9c30a0, which also changed the whitespace
in the associated stt file by line-wrapping the long format string.

Instead, use multiple string constants, which the compiler will
concatenate. This maintains the original formatting, but remains
legible within the standard line length.

Change-Id: I27c9f92d46be82d408105a3a4091f145f677e00e

13 years agoCorrected minor typos.
Paul Wilkins [Tue, 8 Mar 2011 14:46:22 +0000 (14:46 +0000)]
Corrected minor typos.

Change-Id: Icc9f12bd1e1bdaf51256dc8a90d08aa9be89ef34