Main Menu

libx264 GIT builds

Started by LoRd_MuldeR, January 30, 2009, 12:39:23 AM

Previous topic - Next topic


libx264 SVN-r2200 library pack:
commit 999b753ff0f4dc872077f4fa90d465e948cbe656 r2200
Author: Jason Garrett-Glaser <>
Date:   Tue May 8 15:42:56 2012 -0700

    Threaded lookahead
    Split each lookahead frame analysis call into multiple threads.  Has a small
    impact on quality, but does not seem to be consistently any worse.
    This helps alleviate bottlenecks with many cores and frame threads. In many
    case, this massively increases performance on many-core systems.  For example,
    over 100% faster 1080p encoding with --preset veryfast on a 12-core i7 system.
    Realtime 1080p30 at --preset slow should now be feasible on real systems.
    For sliced-threads, this patch should be faster regardless of settings (~10%).
    By default, lookahead threads are 1/6 of regular threads.  This isn't exacting,
    but it seems to work well for all presets on real systems.  With sliced-threads,
    it's the same as the number of encoding threads.

commit ecfbf9d8025e39783bc4262dc1972ca742d8a993 r2199
Author: Anton Mitrofanov <>
Date:   Fri May 4 17:18:12 2012 +0400

    Add support for RGB formats in bit-depth conversion filter

commit 1c97f3570fba02f768fbf649b9f7d48beb720048 r2198
Author: Anton Mitrofanov <>
Date:   Sat May 12 13:57:49 2012 +0400

    Fix some bugs in mb_info code

commit 69a0443e7d8ab032a7f3c3468a42177d5e64daa2 r2197
Author: Jason Garrett-Glaser <>
Date:   Thu Mar 29 14:14:07 2012 -0700

    Add mb_info API for signalling constant macroblocks
    Some use-cases of x264 involve encoding video with large constant areas of the frame.
    Sometimes, the caller knows which areas these are, and can tell x264.
    This API lets the caller do this and adds internal tracking of modifications to macroblocks to avoid problems.
    This is really only suitable without B-frames.
    An example use-case would be using x264 for VNC.

commit df6252cfed7c23fbe883456f4e0607a7f8e91ad8 r2196
Author: Henrik Gramner <>
Date:   Sat Apr 7 00:40:09 2012 +0200

    Faster chroma weight cost calculation
    New assembly function with SSE2, SSSE3 and XOP implementations for calculating absolute sum of differences.

commit cce88ebc9e517b0fa8735b81ac30b4e6a79c8154 r2195
Author: Lucien <>
Date:   Sat Mar 31 13:42:49 2012 +0100

    Add Level 5.2 support

commit ee30c84e38b30896ffa6ddc417f3b4c281a86d1a r2194
Author: Henrik Gramner <>
Date:   Thu Apr 12 19:14:43 2012 +0200

    Eradicate all mention of Extended Profile
    x264 never supported it and never will because nobody uses it.

commit 8ca49cc5c40813d8b98544989eb684e167b06aa0 r2193
Author: Anton Mitrofanov <>
Date:   Tue Apr 3 21:46:52 2012 +0400

    Fix disabling of mbtree when using 2pass encoding and zones

commit 3691332c0b33a68f9d6f519edaa2b848ed34a38c r2192
Author: Alexander Prikhodko <>
Date:   Sat Mar 31 12:06:21 2012 +0300

    configure: force select -mXX gcc option for i386/x86-64
    Makes multilib compilation more convenient.

commit e1ccbf9bb3abdd25d3f0c76682926ec49f3f8001 r2191
Author: Rafaël Carré <>
Date:   Sun Apr 15 21:20:14 2012 -0400

    Update config.guess and config.sub
    Adds support for a bunch of targets, including:
    aarch64 (armv8)

commit f87619768dba73c1effbcfb08875d096575e079e r2190
Author: Alexander Prikhodko <>
Date:   Sat Mar 31 11:33:41 2012 +0300

    configure: correct use of RC variable and add --extra-rcflags

commit 35cf912671fddcb3e701bf667a75f77dd8b28264 r2189
Author: Steven Walters <>
Date:   Wed Mar 28 21:15:04 2012 -0400

    ICL/MSVS: Fix shared library generation and usage
    MSVS requires exported variables to be declared with the DATA keyword, and requires that imported variables be declared with dllimport.
    This does not fix x264 cli being unable to use a shared library built by ICL however.

commit 259a6e57ae25c71acc1669e0aefde7ffe7e235ec r2188
Author: Kieran Kunhya <>
Date:   Tue Mar 27 17:38:56 2012 +0100

    Fix intra-refresh + hrd

commit e0351cdfeb45bf7f891eeb1dc475292154bb9d82 r2187
Author: Anton Mitrofanov <>
Date:   Sun Mar 25 17:34:24 2012 +0400

    Fix frame input colorspace check

commit 7392c8c31f791e9b4c10e4959f8715c8a8233d25 r2186
Author: Jason Garrett-Glaser <>
Date:   Thu Mar 22 13:56:50 2012 -0700

    Fix comment in deblock.c
    The code does, in fact, handle CAVLC+8x8dct correctly already.

commit 6979713216d792e44e3cbaeeba74b455e0a07c62 r2185
Author: Jason Garrett-Glaser <>
Date:   Tue Mar 13 14:37:26 2012 -0700

    Fix sliced-threads ratecontrol bug
    Was using qp instead of qscale; could cause NANs (not to mention less accurate results).

commit 5c85e0a2b7992fcaab09418e3fcefc613cffc743 r2184
Author: Anton Mitrofanov <>
Date:   Sun Mar 11 23:08:18 2012 -0700

    Fix clobbering of mutex/cvs
    Regression in r2183.
    Bizarrely seemed to work on many platforms, but crashed on win64 and may have been slower.
    Only affected sliced threads during encoding, but could cause crashes on x264 encoder close even without sliced threads.

commit c522ad1fed167d0e985e4f9dcdee042473cf74db r2183
Author: Jason Garrett-Glaser <>
Date:   Fri Feb 24 13:34:39 2012 -0800

    Sliced-threads: do hpel and deblock after returning
    Lowers encoding latency around 14% in sliced threads mode with preset superfast.
    Additionally, even if there is no waiting time between frames, this improves parallelism, because hpel+deblock are done during
    the (singlethreaded) lookahead.
    For ease of debugging, dump-yuv forces all of the threads to wait and finish instead of setting b_full_recon.

commit 6a27a481d4c3508ce778a61a139a4734bb8126f7 r2182
Author: Jason Garrett-Glaser <>
Date:   Fri Feb 24 13:16:52 2012 -0800

    Add full-recon API option
    Fully reconstruct frames even without dump-yuv.

commit e856755d2a67f45249c24cb51aa38fc4fa192321 r2181
Author: Jason Garrett-Glaser <>
Date:   Wed Feb 22 13:33:36 2012 -0800

    x86inc: switch to amdnops
    Recent AMD CPUs' instruction decoders choke horribly on extremely long nops (i.e. with 4 prefixes).
    Won't affect much, since we don't use ALIGN much.

commit 5a242c5862baaa4bd5829bd1b43dc11cf5c86344 r2180
Author: Jason Garrett-Glaser <>
Date:   Tue Feb 14 16:54:03 2012 -0800

    BMI1 decimate functions
    Intel was nice enough to make tzcnt equal to "rep bsf", which is backwards-compatible.
    This means we don't actually have to add new functions to make it work.

commit ac31c59a98c6c690894670b9c9af2612f799d85b r2179
Author: Jason Garrett-Glaser <>
Date:   Tue Feb 14 15:07:10 2012 -0800

    Minor asm changes

commit 83561e55dde06f3247aa9b99fa62ead38d7a406e r2178
Author: Jason Garrett-Glaser <>
Date:   Thu Feb 9 14:23:52 2012 -0800

    Add row-reencoding support to VBV for improved accuracy
    Extremely accurate, possibly 100% so (I can't get it to fail even with difficult VBVs).
    Does not yet support rows split on slice boundaries (occurs often with slice-max-size/mbs).
    Still inaccurate with sliced threads, but better than before.

commit 5a69f8e105663497794d4bb4e58cf7baa5cd29cb r2177
Author: Jason Garrett-Glaser <>
Date:   Thu Feb 9 12:38:44 2012 -0800

    Abstract bitstream backup/restore functions
    Required for row re-encoding.

commit 037d123cf62c4af2dc13742b8606882b6d0d3d9e r2176
Author: Anton Mitrofanov <>
Date:   Thu Feb 9 15:27:53 2012 -0800

    Add an small per-MB cost penalty for lowres
    Helps avoid VBV predictors going nuts with very low-cost MBs.
    One particular case this fixes is zero-cost MBs: adaptive quantization decreases the QP a lot, but (before this patch),
    no cost penalty gets factored in for this, because anything times zero is zero.

commit de5a0adca1a7d08b1233b317ec092dbf19263d2f r2175
Author: Jason Garrett-Glaser <>
Date:   Mon Feb 13 18:31:51 2012 -0800

    Remove explicit run calculation from coeff_level_run
    Not necessary with the CAVLC lookup table for zero run codes.

commit 9f1ac3b36eb2666e9d2ec4b859f3b63f60827bf0 r2174
Author: Jason Garrett-Glaser <>
Date:   Mon Feb 13 13:20:06 2012 -0800

    Export PSNR/SSIM in x264 API

commit 7e85ec036df4290697239f5dc9f4a793313ceebc r2173
Author: Ronald S. Bultje <>
Date:   Wed Feb 8 13:10:31 2012 -0800

    x86inc: support yasm -f win64
    Not necessary for x264, as -m amd64 already does the right thing, but used by external users of x86inc.

commit 02c3d5ec58d6bcbc5e22715ae80d53d8556f3c8f r2172
Author: Henrik Gramner <>
Date:   Wed Feb 1 23:52:48 2012 +0100

    Fix incorrect zero-extension assumptions in x86_64 asm
    Some x264 asm assumed that the high 32 bits of registers containing "int" values would be zero.
    This is almost always the case, and it seems to work with gcc, but it is *not* guaranteed by the ABI.
    As a result, it breaks with some other compilers, like Clang, that take advantage of this in optimizations.
    Accordingly, fix all x86 code by using intptr_t instead of int or using movsxd where neccessary.
    Also add checkasm hack to detect when assembly functions incorrectly assumes that 32-bit integers are
    zero-extended to 64-bit.

commit 01f7a333e6c6a6d91a7fe977b491a448ddf4c117 r2171
Author: Jason Garrett-Glaser <>
Date:   Thu Feb 23 09:11:23 2012 -0800

    Fix possible alignment crash when linking from MSVC
    x264_cavlc_init needs to be stack-aligned now.

commit b17c247178a24c218843639c3f46bcfde0edab0a r2170
Author: Anton Mitrofanov <>
Date:   Tue Feb 21 12:58:22 2012 -0800

    Fix rare overflow in 10-bit intra_satd_x3_16x16 asm

commit 1446fe7c47cf660d764b4cbf53694bc3df9b04de r2169
Author: Steven Walters <>
Date:   Sat Feb 11 22:56:43 2012 -0500

    ICL: fix out of tree building and resource file usage on Windows

commit d3efb00abbedd2bbb70156bd989beefe06468116 r2168
Author: Oka Motofumi <>
Date:   Mon Feb 6 06:07:34 2012 +0900

    Add error handling for out-of-tree build

commit ec41b19edc67ee4eca09c0e3b37e6290844c5e1f r2167
Author: Anton Mitrofanov <>
Date:   Tue Mar 6 17:34:02 2012 +0400

    Fix RGB colorspace input
    BGR/BGRA input was correct.

commit 39a4c6fecaaa0d6cde8d89d31ef6cd1d25ab802b r2166
Author: Jason Garrett-Glaser <>
Date:   Mon Feb 13 16:40:32 2012 -0800

    Fix interlaced + extremal slice-max-size
    Broke if the first macroblock in the slice exceeded the set slice-max-size.

commit 3f72c99a15a07511b758d9e94217223480865124 r2165
Author: Henrik Gramner <>
Date:   Sun Feb 5 20:43:09 2012 +0100

    Fix regression in r2141
    Broke register preservation in x264_cpu_cpuid and x264_cpu_xgetbv.
    Did not cause any problems.

commit da19765d723b06a1fa189478e9da61a1c18490f8 r2164
Author: Jason Garrett-Glaser <>
Date:   Thu Jan 19 14:56:54 2012 -0800

    TBM, AVX2, FMA3, BMI1, and BMI2 CPU detection support
    TBM and BMI1 are supported by Trinity/Piledriver.
    The others (and BMI1) will probably appear in Intel's upcoming Haswell.
    Also update x86inc with AVX2 stuff.

commit efef20090a06a38f9d95755588d7830fb92a2a02 r2163
Author: Loren Merritt <>
Date:   Fri Feb 3 06:27:18 2012 +0000

    x86inc: add TAIL_CALL macro to abstract a common asm idiom

commit a7e6e1793b4d2b49c9449d767320c71daa855cb6 r2162
Author: Jason Garrett-Glaser <>
Date:   Wed Jan 25 16:44:38 2012 -0800

    Minor asm optimizations/cleanup

commit 56ba096141d16ffcbabd805e2d27014f62f0d722 r2161
Author: Jason Garrett-Glaser <>
Date:   Tue Jan 24 19:03:58 2012 -0800

    Clean up and optimize weightp, plus enable SSSE3 weight on SB/BDZ
    Also remove unused AVX cruft.

commit 961a278e0123eb662b46a6f136a48a43f6a2d427 r2160
Author: Jason Garrett-Glaser <>
Date:   Mon Jan 23 18:57:58 2012 -0800

    XOP frame_init_lowres
    Covers both 8-bit and 16-bit, ~5-10% faster on Bulldozer.

commit c5809994990df6c63b4250546844dc77181fee0f r2159
Author: Jason Garrett-Glaser <>
Date:   Tue Jan 17 15:25:10 2012 -0800

    XOP 8x8 zigzags
    Field: 35(mmx) ->16(xop) cycles
    Frame: 32(ssse3)->20(xop) cycles

commit 14dc11f7c52fa29576e0003c8c16857a78bf5fbf r2158
Author: Jason Garrett-Glaser <>
Date:   Mon Jan 23 15:09:38 2012 -0800

    AVX 32-bit hpel_filter_h
    Faster on Sandy Bridge.
    Also add details on unsuccessful optimizations in these functions.

commit 2fcd0446b5d91ae52e143682c30000a49441e4a1 r2157
Author: Jason Garrett-Glaser <>
Date:   Fri Jan 27 16:29:30 2012 -0800

    x86inc: add high halfword register support
    Might be useful in a few cases.

commit 5c4b8484ea9aaabfb70523ba1f9c4d8343ad3221 r2156
Author: Ronald S. Bultje <>
Date:   Wed Jan 25 13:53:59 2012 +0800

    Change %ifdef directives to %if directives in *.asm files
    This allows combining multiple conditionals in a single statement.

commit 1b558de42dc08a303c2faf79fc9999b48a876370 r2155
Author: Anton Mitrofanov <>
Date:   Sun Jan 22 22:13:52 2012 +0400

    Use TV range algorithm for bit-depth conversions
    Such sources are more common, so better to be correct for the common case.
    This also produces less error for the case of full range than the previous algorithm produced for the case of TV range.

commit 83c371deba853a4ebb28739e868df86b3153fb3e r2154
Author: Hii <>
Date:   Wed Jan 25 16:29:22 2012 +0800

    Bump dates to 2012

commit a2925c5a707e833c34fa0a64d497c02e6dcfe6e6 r2153
Author: Henrik Gramner <>
Date:   Sat Jan 28 21:38:27 2012 +0100

    Add Windows resource file
    Displays version info in Windows Explorer.

commit 98ade832d053f6bfca4d0dd2ab0cd1c88531721d r2152
Author: Sergey Radionov <>
Date:   Mon Jan 16 13:22:44 2012 -0800

    Fix win32 pthread_cond_signal
    Isn't used by x264 currently, so didn't cause a problem.
    Fix backported from libav.

commit a3f44077dc238dea92c0894d352b5a8723b9201b r2151
Author: Mans Rullgard <>
Date:   Wed Feb 1 15:55:25 2012 -0800

    ARM: align asm functions to 4 bytes.
    Some linkers apparently fail to correctly align ARM functions when mixing with Thumb code.

commit d3a39c92f5c130cad6d45e9daffa5a2beb145ebb r2150
Author: Anton Mitrofanov <>
Date:   Sun Jan 22 13:00:23 2012 +0400

    Fix normalization of colorspace when input is packed YUV 4:2:2

These builds will NOT work with Avidemux 2.5.6 or older. Please update to Avidemux 2.6 r8054 or later now!
You asked me once, what was in Room 101. I told you that you knew the answer already.
Everyone knows it. The thing that is in Room 101 is the worst thing in the world.

MuldeR's OpenSource stuff: