Diff of /branches/dev-api-4/xvidcore/ChangeLog

-revision 1181, Fri Oct 17 15:25:44 2003 UTC
+revision 1242, Fri Dec  5 14:44:35 2003 UTC
 Line 2
  # Ed.Gomez: This ChangeLog is generated from a personal tree maintained
  # under the arch revision control tool. That's why dates may be skewed. I
  # also removed all my email adresses from the output because they are not
- # relevent
+ # relevant.
  #
+-12-05 14:35:22 GMT                                         patch-122
+     Summary:
+       Cap quants correctly (the best we can at least)
+     Revision:
+       xvidcore--devapi4--1.0--patch-122
+     Cap quants correctly (the best we can at least)
+     modified files:
+      src/plugins/plugin_single.c
+-12-05 14:33:48 GMT                                         patch-121
+     Summary:
+       Small glitch
+     Revision:
+       xvidcore--devapi4--1.0--patch-121
+     Small glitch
+     modified files:
+      src/motion/vop_type_decision.c
+-12-05 14:06:19 GMT                                         patch-120
+     Summary:
+       KFthresholding changes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-120
+     As user reports proved, the logic behind the min_key_interval was
+/ misleading because the parameter is kfthreshold indeed and not
+         a minimum keyframe interval
+/ the formula was a bit too aggressive (removing 20% of bitrate
+         per frame until distance to next iframe was 1)
+     I posted a RFC to try to settle a decision on what behavior this
+     setting should have. We have still have no clear answer so i prefer
+     just fixing the misleading name right now and wait for a common
+     position about its behavior later.
+     Libraries are *binary* compatible, but *source code* compatibility
+     is broken (rename rc_2pass2_t->min_key_interval to kfthreshold).
+     This is probably the last API change.
+     NB: fixes a type problem during scaling parameter computing which
+         was causing insane pb_iboost_tax_ratio values.
+     modified files:
+      src/plugins/plugin_2pass2.c src/xvid.h vfw/src/codec.c
+      vfw/src/config.c vfw/src/config.h
+-12-05 00:20:28 GMT                                         patch-119
+     Summary:
+       ivop decision tuning.
+     Revision:
+       xvidcore--devapi4--1.0--patch-119
+     ivop decision tuning from sysKin.
+     modified files:
+      src/motion/vop_type_decision.c
+-12-03 18:55:29 GMT                                         patch-118
+     Summary:
+       VOSH header always written.
+     Revision:
+       xvidcore--devapi4--1.0--patch-118
+     * profile is set to sane default value in BitstreamWriteVolHeaders
+     * VOSH is now always written (note that the ending code is never
+       written)
+     * doubled variable removed from bvop estimation file.
+     bitstream version set to 23
+     Version 22 was used in CVS by a fix from michael for VOSH, it was just
+     lacking the sane default value setting when profile is 0x00 from user.
+x00 is a reserved profile ID in the spec so it wasn't making much sense
+     to write 0x00.
+     modified files:
+      src/bitstream/bitstream.c src/motion/estimation_bvop.c
+      src/xvid.h
+-12-03 15:29:30 GMT                                         patch-117
+     Summary:
+       VFW GUI fixes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-117
+     min key was misleading because it was legacy code from dev-api-3.
+     modified files:
+      vfw/src/codec.c vfw/src/config.c vfw/src/resource.rc
+-12-03 15:22:25 GMT                                         patch-116
+     Summary:
+       ME fixes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-116
+     From syskin:
+     - small typo in chroma sad reset
+     - code tweaking
+       + adv diamond search used intead of mainsearch
+       + and some other stuff
+     - thresholds tuned.
+     modified files:
+      src/motion/estimation.h src/motion/estimation_bvop.c
+      src/motion/estimation_pvop.c
+-12-01 11:17:20 GMT                                         patch-115
+     Summary:
+       Small changes and interlacing bugfix.
+     Revision:
+       xvidcore--devapi4--1.0--patch-115
+     * Interlacing bugfix, code got swapped in a very old patch
+       (back in March)
+     * xvid_encraw forces VOSH writing.
+     * VFW GUI fixes.
+     modified files:
+      examples/Makefile examples/xvid_encraw.c
+      src/utils/mbtransquant.c vfw/src/config.c vfw/src/resource.rc
+-11-30 15:47:41 GMT                                         patch-114
+     Summary:
+       Thread safe MPEG4 quantization functions + xvid_bench update
+     Revision:
+       xvidcore--devapi4--1.0--patch-114
+     * Thread safe MPEG4 quantization functions.
+       Cleaned up version of patch provided by Michael
+        - fixed compiling problems on gcc
+        - added const qualifiers every where it was possible to help C compiler
+          optimization.
+        - added the mpeg_quant_matrices param to all ASM function prototype
+          in comments (even if it's not used, that shows we do it deliberatly)
+        - forces m[intra][0][0] = 8, otherwise XviD could write invalid streams.
+     * Added real CRC computing in xvid_bench.c
+     modified files:
+      TODO examples/xvid_bench.c src/bitstream/bitstream.c
+      src/decoder.c src/decoder.h src/encoder.c src/encoder.h
+      src/motion/estimation.h src/motion/estimation_pvop.c
+      src/motion/estimation_rd_based.c src/quant/quant.h
+      src/quant/quant_h263.c src/quant/quant_matrix.c
+      src/quant/quant_matrix.h src/quant/quant_mpeg.c
+      src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_h263_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/utils/mbtransquant.c src/xvid.c
+-11-29 18:10:25 GMT                                         patch-113
+     Summary:
+       Fixed csp asm rules for real?
+     Revision:
+       xvidcore--devapi4--1.0--patch-113
+     Fixed csp asm rules for real?
+     modified files:
+      build/win32/libxvidcore.dsp
+-11-29 17:58:09 GMT                                         patch-112
+     Summary:
+       TODO/ChangeLog updated
+     Revision:
+       xvidcore--devapi4--1.0--patch-112
+     TODO/ChangeLog updated
+     modified files:
+      ChangeLog TODO
+ #########################################################################
+ # 1.0.0 beta1
+ #########################################################################
+-11-29 17:21:08 GMT                                         patch-111
+     Summary:
+       First beta marking
+     Revision:
+       xvidcore--devapi4--1.0--patch-111
+     First beta marking
+     modified files:
+      build/generic/configure.in src/xvid.c src/xvid.h
+-11-29 16:59:14 GMT                                         patch-110
+     Summary:
+       Catching up with CVS.
+     Revision:
+       xvidcore--devapi4--1.0--patch-110
+     Changes from sysKin:
+     * dquant optimization.
+     * CBR fix
+     modified files:
+      src/encoder.c src/plugins/plugin_single.c
+-11-24 22:05:38 GMT                                         patch-109
+     Summary:
+       Big level handling in trellis.
+     Revision:
+       xvidcore--devapi4--1.0--patch-109
+     Trellis was treating big levels exactly the same way as lower ones.
+     In some cases, trellis was doing wild optimizations favoring a 0
+     because the distortion introduced by that big coeff change was acceptable.
+     But visually this could result in some nasty blocks with wrong chroma
+     information or similar brutal changes in other planes as well.
+     Skal added big levels handling where trellis just tries to minimize
+     the cost varying the run value only. No level modification is done
+     anymore.
+     modified files:
+      TODO src/utils/mbtransquant.c
+-11-23 16:42:55 GMT                                         patch-108
+     Summary:
+       Trellis for MPEG.
+     Revision:
+       xvidcore--devapi4--1.0--patch-108
+     * Added trellis support for MPEg quantization type.
+     * Changed RD fixed point precision, should help avoiding overflow
+       (see the constant TL_SHIFT)
+     NB: we still have some problems when trellis optimizes DC for big DC
+         values.
+     modified files:
+      src/utils/mbtransquant.c
+-11-22 00:53:59 GMT                                         patch-107
+     Summary:
+       Win32 lib project fix (bis)
+     Revision:
+       xvidcore--devapi4--1.0--patch-107
+     * nasm >= 0.98.37 support in project file got reverted in a previous
+       patch, push it back. Nota bene /O3 changed to /O2 for proper
+       compilation with msvc compiler (everyone is not supposed to compile
+       stuff with icc)
+     * TODO update.
+     modified files:
+      TODO build/win32/libxvidcore.dsp
+-11-19 21:26:34 GMT                                         patch-106
+     Summary:
+       updated bench crc
+     Revision:
+       xvidcore--devapi4--1.0--patch-106
+     updated bench crc
+     modified files:
+      examples/xvid_bench.c
+-11-19 16:00:00 GMT                                         patch-105
+     Summary:
+       Lumimasking fixes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-105
+     from sysKin:
+     * New plugin hook entry XVID_PLG_FRAME that happens inside FrameCodeIPB
+       when both type and quant are known. Added hook handling in all plugins.
+     * Fixed lumimasking.
+     from me:
+     * small reverse commit in pvop estimation fixed.
+     modified files:
+      src/bitstream/bitstream.c src/bitstream/bitstream.h
+      src/encoder.c src/motion/estimation_pvop.c
+      src/plugins/plugin_2pass1.c src/plugins/plugin_2pass2.c
+      src/plugins/plugin_dump.c src/plugins/plugin_lumimasking.c
+      src/plugins/plugin_psnr.c src/plugins/plugin_single.c
+      src/xvid.h
+-11-19 15:37:16 GMT                                         patch-104
+     Summary:
+       Removed indirections from SearchData structure.
+     Revision:
+       xvidcore--devapi4--1.0--patch-104
+     Patch from sysKin:
+     * removed indirections in SearchData structure. CheckCandidate functions
+       don't use a const SearchData pointer anymore, but they should be a bit
+       faster because of the less numerous indirections.
+     modified files:
+      src/motion/estimation.h src/motion/estimation_bvop.c
+      src/motion/estimation_common.c src/motion/estimation_gmc.c
+      src/motion/estimation_pvop.c src/motion/estimation_rd_based.c
+      src/motion/vop_type_decision.c
+-11-19 15:33:55 GMT                                         patch-103
+     Summary:
+       Formula error in twopass code.
+     Revision:
+       xvidcore--devapi4--1.0--patch-103
+     The reversing bframe formula in 2pass 2 was not right. This was in fact
+     a test code i used when i was maintaining the code on its own branch.
+     I should not have commited it :\
+     Fixed :-)
+     modified files:
+      src/plugins/plugin_2pass2.c
+-11-18 21:41:08 GMT                                         patch-102
+     Summary:
+       Another problem with mis/unitialized reads.
+     Revision:
+       xvidcore--devapi4--1.0--patch-102
+     Michael introduced a fast subpel refine that uses a (iMinSAD2, currentQMV2)
+     couple of data. The problem is that he plugged this in CheckCandidate16_qpel
+     that is used outside this context, thus a if statement was traversed with
+     garbage data in the standard subpel case. For perfection sake, using a
+     iMinSAD=256*4096 value collects correct data even if it will not be used in
+     the normal subpel case.
+     modified files:
+      src/motion/estimation_pvop.c
+-11-16 17:29:39 GMT                                         patch-101
+     Summary:
+       The MEanalysis patch assumed bvops were always used
+     Revision:
+       xvidcore--devapi4--1.0--patch-101
+     The MEanalysis patch assumed bvops were always used
+     modified files:
+      src/encoder.c
+-11-16 15:12:15 GMT                                         patch-100
+     Summary:
+       MEanalysis using wrong mvs + bframe search using wrong mvs
+     Revision:
+       xvidcore--devapi4--1.0--patch-100
+     Valgrind reported lot of unitialized reads. These unitialized reads
+     helped sysKin finding three bugs:
+      - ZeroMacroblock did not reset the cbp field. So for some skipped
+        blocks, a test was done on the cbp value...
+      - MEanalysis was using wrong mvs from the current bvop (unitialized
+        or just wrong in current context). That's because in devapi3, bframes
+        used to share the same mvs array wheras now, it's one array a bvop.
+      - Collocated skipped MBs for a bvop didn't reset mvs[0] and b_mvs[0].
+     modified files:
+      src/encoder.c src/motion/estimation_bvop.c src/motion/motion.h
+      src/motion/motion_inlines.h src/motion/vop_type_decision.c
+-11-15 15:21:09 GMT                                         patch-99
+     Summary:
+       Small fixes
+     Revision:
+       xvidcore--devapi4--1.0--patch-99
+     Small fixes
+     modified files:
+      src/encoder.c src/motion/vop_type_decision.c
+-11-15 15:02:47 GMT                                         patch-98
+     Summary:
+       DShow update + libxvidcore project file update
+     Revision:
+       xvidcore--devapi4--1.0--patch-98
+     From peter:
+     * DShow now links against libxvidcore.lib
+     * Some updates to the libxvidcore project file
+     modified files:
+      TODO build/win32/libxvidcore.dsp dshow/dshow.dsp
+      dshow/src/CXvidDecoder.cpp dshow/src/CXvidDecoder.h
+-11-15 01:51:28 GMT                                         patch-97
+     Summary:
+       Small updates to doc files
+     Revision:
+       xvidcore--devapi4--1.0--patch-97
+     Small updates to doc files
+     modified files:
+      AUTHORS CodingStyle TODO doc/INSTALL doc/README
+     renamed files:
+      .arch-ids/authors.txt.id
+        ==> .arch-ids/AUTHORS.id
+      .arch-ids/todo.txt.id
+        ==> .arch-ids/TODO.id
+      authors.txt
+        ==> AUTHORS
+      todo.txt
+        ==> TODO
+-11-14 11:23:55 GMT                                         patch-96
+     Summary:
+       Updated ChangeLog
+     Revision:
+       xvidcore--devapi4--1.0--patch-96
+     Updated ChangeLog
+     modified files:
+      ChangeLog
+-11-13 23:09:34 GMT                                         patch-95
+     Summary:
+x8 16bit Block SSE optimization.
+     Revision:
+       xvidcore--devapi4--1.0--patch-95
+     MMXed the calculation of SSE for 8x8 16bit blocks. This helps quite
+     a lot VHQ=4 mode.
+     My tests show with trellis:chroma_me:
+      - ~20% speed improvement for vhq=4.
+      - at least 5% when using vhq=1.
+     Of course this speedup vanishes if more CPU intensive features are used.
+     CruNcher who used gmc/qpel, noticed "only" a ~5% speed improvement.
+     NB: i'm of course talking about overall speed improvement. Such a small
+         patch for such a big improvement :-)
+     modified files:
+      src/motion/estimation_rd_based.c src/motion/sad.c
+      src/motion/sad.h src/motion/x86_asm/sad_mmx.asm src/xvid.c
+-11-13 22:34:33 GMT                                         patch-94
+     Summary:
+       Various small bug fixes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-94
+     * encoder.c: GMC code fix in encoder.c. Now gmcval is initialized correctly
+       when using GME.
+     * xvid_decraw.c: Fix elementary stream output.
+     * plugin_2pass2.c: Small parsing bug in stats reading in 2pass2.
+     * decoder.c: Read resync markers in bframes.
+     modified files:
+      examples/xvid_decraw.c src/decoder.c src/encoder.c
+      src/plugins/plugin_2pass2.c
+-11-11 16:24:05 GMT                                         patch-93
+     Summary:
+       VFW update for overflow control
+     Revision:
+       xvidcore--devapi4--1.0--patch-93
+     From Koepi.
+     * Added widget and code for overflow control strength.
+     * Removed widgets for payback options and kfthresholds.
+     * Activated frame stats in DebugOutputView all the time.
+     From me:
+     * Activated static motion detection in cartoon mode.
+     modified files:
+      vfw/src/codec.c vfw/src/config.c vfw/src/resource.h
+      vfw/src/resource.rc
+-11-09 20:47:47 GMT                                         patch-92
+     Summary:
+       New two pass code.
+     Revision:
+       xvidcore--devapi4--1.0--patch-92
+     New two pass code. I may say it's just a fixed version, though it looks
+     more like a "take all the ideas and write it again" version. It performs
+     better with all natural sequences i have and a bit worse with anime.
+     Including it now, allow me improving the code during the beta releases.
+     modified files:
+      src/encoder.c src/plugins/plugin_2pass1.c
+      src/plugins/plugin_2pass2.c src/xvid.h vfw/src/codec.c
+      vfw/src/config.c vfw/src/config.h
+-11-09 17:07:16 GMT                                         patch-91
+     Summary:
+       Fixes for bframe compensation (used in psnr tests).
+     Revision:
+       xvidcore--devapi4--1.0--patch-91
+     * transfer_8to16_sub2_(c|mmx|xmm|3dne) write back the compensated
+       result to current frame pointer.
+     * transfer_8to16_sub2_mmx uses proper rounding (a+b+1)/2. The +1
+       operation was missing.
+     * Blocks skipped in bframes must be compensated for psnr computing.
+     modified files:
+      src/encoder.c src/motion/estimation_bvop.c
+      src/utils/mem_transfer.c
+      src/utils/x86_asm/mem_transfer_3dne.asm
+      src/utils/x86_asm/mem_transfer_mmx.asm
+-11-05 16:05:44 GMT                                         patch-90
+     Summary:
+       Speed improvement not wasting setedges and interpolate calls.
+     Revision:
+       xvidcore--devapi4--1.0--patch-90
+     Patch from syskin.
+     * This patch avoids calling setedges and interpolate for uneeded cases:
+        - setedges is only called once per frame.
+        - interpolate is called only when the previous rounding
+          was different from the one needed.
+     * Interpolation has been optimized a bit for qpel case, we do the
+       hv pass down to top to use the cache more efficiently (hope so).
+     modified files:
+      src/encoder.c src/encoder.h src/image/image.c
+-11-03 19:51:12 GMT                                         patch-89
+     Summary:
+       SSE2 dev16 fix + xvid_bench DCT block alignments.
+     Revision:
+       xvidcore--devapi4--1.0--patch-89
+     * Small error fixed by Skal in his dev16 code (missing pshufd).
+     * Blocks used by DCT tests are now aligned with DECLARE_ALIGNED_MATRIX
+       this avoids the well know segfaults when using SSE2 instructions that
+       suppose data alignment.
+     modified files:
+      examples/xvid_bench.c src/motion/x86_asm/sad_sse2.asm
+-11-03 15:42:23 GMT                                         patch-88
+     Summary:
+       Align .rodata section for non coff objects
+     Revision:
+       xvidcore--devapi4--1.0--patch-88
+     Align .rodata section for non coff objects
+     modified files:
+      src/bitstream/x86_asm/cbp_mmx.asm
+      src/bitstream/x86_asm/cbp_sse2.asm
+      src/dct/x86_asm/fdct_mmx_ffmpeg.asm
+      src/dct/x86_asm/fdct_mmx_skal.asm
+      src/dct/x86_asm/fdct_sse2_skal.asm
+      src/dct/x86_asm/idct_3dne.asm src/dct/x86_asm/idct_mmx.asm
+      src/dct/x86_asm/idct_sse2_dmitry.asm
+      src/dct/x86_asm/simple_idct_mmx.asm
+      src/image/x86_asm/colorspace_rgb_mmx.asm
+      src/image/x86_asm/colorspace_yuyv_mmx.asm
+      src/image/x86_asm/interpolate8x8_3dn.asm
+      src/image/x86_asm/interpolate8x8_3dne.asm
+      src/image/x86_asm/interpolate8x8_mmx.asm
+      src/image/x86_asm/interpolate8x8_xmm.asm
+      src/image/x86_asm/qpel_mmx.asm
+      src/image/x86_asm/reduced_mmx.asm
+      src/motion/x86_asm/sad_3dn.asm src/motion/x86_asm/sad_3dne.asm
+      src/motion/x86_asm/sad_mmx.asm src/motion/x86_asm/sad_sse2.asm
+      src/motion/x86_asm/sad_xmm.asm
+      src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_h263_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/utils/x86_asm/cpuid.asm
+      src/utils/x86_asm/interlacing_mmx.asm
+      src/utils/x86_asm/mem_transfer_3dne.asm
+-11-02 23:01:43 GMT                                         patch-87
+     Summary:
+       SSE2 update
+     Revision:
+       xvidcore--devapi4--1.0--patch-87
+     * Added Dmitry SSE2 iDCT code back.
+     * Plugged Dmitry iDCT as default for SSE2
+     * Fixed a bug in xvid_bench that was making it would test some CPU
+       instruction set w/o host CPU support. xvidcore init was simply
+       discarding irrelevant cpu flags.
+     new files:
+      src/dct/x86_asm/.arch-ids/idct_sse2_dmitry.asm.id
+      src/dct/x86_asm/idct_sse2_dmitry.asm
+     modified files:
+      build/generic/sources.inc build/win32/libxvidcore.dsp
+      examples/xvid_bench.c src/dct/fdct.h src/dct/idct.h src/xvid.c
+-10-31 14:53:26 GMT                                         patch-86
+     Summary:
+       Better handling of old windres versions + GNU make dependency.
+     Revision:
+       xvidcore--devapi4--1.0--patch-86
+     Old versions of GNU windres (<2.14) don't have the same short options.
+     But long options remain the same so it's better to use long option names
+     to have ful compatibility with older versions.
+     The Makefile appears to be dependent on GNU make because shell expansion
+     for retrieving the path of the Makefile is wrong when using `` even with
+     a single expansion assignment :=. It keeps being expanded when used.
+     modified files:
+      doc/INSTALL vfw/bin/Makefile
+-10-29 11:31:28 GMT                                         patch-85
+     Summary:
+       Added sse2 f/iDCT code from skal
+     Revision:
+       xvidcore--devapi4--1.0--patch-85
+     * Added sse2 f/iDCT code from skal
+     * Added hooking in xvid.c
+     new files:
+      src/dct/x86_asm/.arch-ids/fdct_sse2_skal.asm.id
+      src/dct/x86_asm/fdct_sse2_skal.asm
+     modified files:
+      build/generic/sources.inc build/win32/libxvidcore.dsp
+      src/xvid.c
+-10-29 00:19:10 GMT                                         patch-84
+     Summary:
+       Fix the static motion detection
+     Revision:
+       xvidcore--devapi4--1.0--patch-84
+     Fix the static motion detection
+     modified files:
+      src/motion/estimation_pvop.c
+-10-28 23:39:46 GMT                                         patch-83
+     Summary:
+       Added cartoon option handling.
+     Revision:
+       xvidcore--devapi4--1.0--patch-83
+     Added cartoon widgets + handling code.
+     NB: static motion detection is disabled because of crashes on P4
+         cpus.
+     modified files:
+      vfw/src/codec.c vfw/src/config.c vfw/src/config.h
+      vfw/src/resource.h vfw/src/resource.rc
+-10-28 17:44:09 GMT                                         patch-82
+     Summary:
+       ASM cleanups;
+     Revision:
+       xvidcore--devapi4--1.0--patch-82
+     * Applied same style to all asm files
+     * Replaced current sad sse2 operators with skal's ones
+     * Removed old and unused colorspace asm files
+     removed files:
+      src/image/x86_asm/.arch-ids/rgb_to_yv12_mmx.asm.id
+      src/image/x86_asm/.arch-ids/yuv_to_yv12_mmx.asm.id
+      src/image/x86_asm/.arch-ids/yuyv_to_yv12_mmx.asm.id
+      src/image/x86_asm/.arch-ids/yv12_to_rgb24_mmx.asm.id
+      src/image/x86_asm/.arch-ids/yv12_to_rgb32_mmx.asm.id
+      src/image/x86_asm/.arch-ids/yv12_to_yuyv_mmx.asm.id
+      src/image/x86_asm/rgb_to_yv12_mmx.asm
+      src/image/x86_asm/yuv_to_yv12_mmx.asm
+      src/image/x86_asm/yuyv_to_yv12_mmx.asm
+      src/image/x86_asm/yv12_to_rgb24_mmx.asm
+      src/image/x86_asm/yv12_to_rgb32_mmx.asm
+      src/image/x86_asm/yv12_to_yuyv_mmx.asm
+     modified files:
+      build/generic/sources.inc src/bitstream/x86_asm/cbp_3dne.asm
+      src/bitstream/x86_asm/cbp_mmx.asm
+      src/bitstream/x86_asm/cbp_sse2.asm
+      src/dct/x86_asm/fdct_mmx_ffmpeg.asm
+      src/dct/x86_asm/fdct_mmx_skal.asm
+      src/dct/x86_asm/idct_3dne.asm src/dct/x86_asm/idct_mmx.asm
+      src/dct/x86_asm/simple_idct_mmx.asm
+      src/image/x86_asm/colorspace_mmx.inc
+      src/image/x86_asm/colorspace_rgb_mmx.asm
+      src/image/x86_asm/colorspace_yuv_mmx.asm
+      src/image/x86_asm/colorspace_yuyv_mmx.asm
+      src/image/x86_asm/interpolate8x8_3dn.asm
+      src/image/x86_asm/interpolate8x8_3dne.asm
+      src/image/x86_asm/interpolate8x8_mmx.asm
+      src/image/x86_asm/interpolate8x8_xmm.asm
+      src/image/x86_asm/qpel_mmx.asm
+      src/image/x86_asm/reduced_mmx.asm
+      src/motion/x86_asm/sad_3dn.asm src/motion/x86_asm/sad_3dne.asm
+      src/motion/x86_asm/sad_mmx.asm src/motion/x86_asm/sad_sse2.asm
+      src/motion/x86_asm/sad_xmm.asm
+      src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_h263_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/utils/x86_asm/cpuid.asm
+      src/utils/x86_asm/interlacing_mmx.asm
+      src/utils/x86_asm/mem_transfer_3dne.asm
+      src/utils/x86_asm/mem_transfer_mmx.asm
+-10-27 01:13:47 GMT                                         patch-81
+     Summary:
+       d_mv_bits speedup from sysKin
+     Revision:
+       xvidcore--devapi4--1.0--patch-81
+     d_mv_bits speedup from sysKin
+     modified files:
+      src/motion/motion_inlines.h
+-10-27 00:55:51 GMT                                         patch-80
+     Summary:
+       fDCT changes, new asm CodingStyle applied to dct dir
+     Revision:
+       xvidcore--devapi4--1.0--patch-80
+     * Ported the ffmpeg fDCT functions (mmx and xmm).
+     * Modified the skal's versions a bit to allow rolling loops.
+     * Activated Skal's fDCTs (unrolled versions) for mmx _and_ xmm
+       (old code was ignoring xmm versions)
+     * Removed the SSE2 versions (they'll be back later)
+     * .data -> .rodata
+     * Applied announced asm CodingStyle to the dct dir
+       (I'll have to add a section with the said CodingStyle)
+     modified files:
+      build/generic/sources.inc build/win32/libxvidcore.dsp
+      src/dct/fdct.h src/dct/idct.h
+      src/dct/x86_asm/fdct_mmx_ffmpeg.asm
+      src/dct/x86_asm/fdct_mmx_skal.asm
+      src/dct/x86_asm/idct_3dne.asm src/dct/x86_asm/idct_mmx.asm
+      src/dct/x86_asm/simple_idct_mmx.asm src/xvid.c
+     renamed files:
+      src/dct/x86_asm/.arch-ids/fdct_mmx.asm.id
+        ==> src/dct/x86_asm/.arch-ids/fdct_mmx_ffmpeg.asm.id
+      src/dct/x86_asm/.arch-ids/fdct_xmm.asm.id
+        ==> src/dct/x86_asm/.arch-ids/fdct_mmx_skal.asm.id
+      src/dct/x86_asm/fdct_mmx.asm
+        ==> src/dct/x86_asm/fdct_mmx_ffmpeg.asm
+      src/dct/x86_asm/fdct_xmm.asm
+        ==> src/dct/x86_asm/fdct_mmx_skal.asm
+-10-25 13:48:42 GMT                                         patch-79
+     Summary:
+       BQuant->PQuant fix.
+     Revision:
+       xvidcore--devapi4--1.0--patch-79
+     When using closed_gop, a BFrame before an IFrame is turned into a PFrame.
+     Thus is original quant has to be computed back, a rounding was causing these
+     frames to be orig_quant-1. As a consequence we had very big frames before
+     the IFrame, loosing many bits for nearly no visual benefit.
+     modified files:
+      src/encoder.c
+-10-25 10:26:48 GMT                                         patch-78
+     Summary:
+       Added closed gop option to xvid_encraw
+     Revision:
+       xvidcore--devapi4--1.0--patch-78
+     Added closed gop option to xvid_encraw
+     modified files:
+      examples/xvid_encraw.c
+-10-24 17:39:53 GMT                                         patch-77
+     Summary:
+       RD fixes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-77
+     Inter RD optimization relied on buggy functions to predict bitsize.
+     modified files:
+      src/motion/estimation_pvop.c src/motion/motion_inlines.h
+-10-22 15:48:01 GMT                                         patch-76
+     Summary:
+       Small INSTALL update due to previous build patches.
+     Revision:
+       xvidcore--devapi4--1.0--patch-76
+     Small INSTALL update due to previous build patches.
+     modified files:
+      doc/INSTALL
+-10-21 21:27:46 GMT                                         patch-75
+     Summary:
+       Removed unused next_block vars.
+     Revision:
+       xvidcore--devapi4--1.0--patch-75
+     Removed unused next_block vars.
+     modified files:
+      src/decoder.c
+-10-21 21:24:15 GMT                                         patch-74
+     Summary:
+       VFW build changes.
+     Revision:
+       xvidcore--devapi4--1.0--patch-74
+     The build system has been modified to look like the core lib one
+     minus the configure system.
+     modified files:
+      vfw/bin/Makefile vfw/bin/sources.inc vfw/src/config.c
+      vfw/vfw.dsp
+     renamed files:
+      vfw/bin/.arch-ids/Makefile.cygwin.id
+        ==> vfw/bin/.arch-ids/Makefile.id
+      vfw/bin/.arch-ids/Makefile.inc.id
+        ==> vfw/bin/.arch-ids/sources.inc.id
+      vfw/bin/Makefile.cygwin
+        ==> vfw/bin/Makefile
+      vfw/bin/Makefile.inc
+        ==> vfw/bin/sources.inc
+      vfw/src/.arch-ids/config.rc.id
+        ==> vfw/src/.arch-ids/resource.rc.id
+      vfw/src/config.rc
+        ==> vfw/src/resource.rc
+-10-21 17:00:09 GMT                                         patch-73
+     Summary:
+       Decoder cleanups and speedup
+     Revision:
+       xvidcore--devapi4--1.0--patch-73
+     cleanups, speedups from sysKin
+     modified files:
+      src/decoder.c
+-10-21 16:22:15 GMT                                         patch-72
+     Summary:
+       Build fixes for newer nasm versions.
+     Revision:
+       xvidcore--devapi4--1.0--patch-72
+     nasm does not take care of adding trailing slashes to include paths.
+     A patch to upstream authors has been refused because "the backslash()
+     feature has been abandoned to get back to old nasm behavior"
+     Their choice is kinda stupid as nasm is now open to user mistakes... :\
+     So we fix that on ou side.
+     modified files:
+      build/generic/configure.in build/win32/libxvidcore.dsp
+-10-17 15:13:12 GMT                                         patch-71
+     Summary:
+       Updated docs.
+     Revision:
+       xvidcore--devapi4--1.0--patch-71
+     The doc of devapi4 is mostly out dated, i t is much better not to
+     keep it in the repository at the moment. We'll add new docs later.
+     Added a INSTALL doc that explains the build/install process for
+     supported platforms. It's a first try, things may be added later.
+     new files:
+      doc/.arch-ids/INSTALL.id doc/INSTALL
+     removed files:
+      doc/.arch-ids/API.dox.id doc/.arch-ids/Makefile.id
+      doc/.arch-ids/foot.inc.in.id doc/.arch-ids/header.tex.in.id
+      doc/.arch-ids/xvid-decoding.txt.id
+      doc/.arch-ids/xvid-encoder.txt.id doc/API.dox doc/Makefile
+      doc/foot.inc.in doc/header.tex.in doc/xvid-decoding.txt
+      doc/xvid-encoder.txt
+     modified files:
+      CodingStyle README doc/README
+     renamed files:
+      .arch-ids/README.txt.id
+        ==> .arch-ids/README.id
+      .arch-ids/changelog.txt.id
+        ==> .arch-ids/ChangeLog.id
+      README.txt
+        ==> README
+      changelog.txt
+        ==> ChangeLog
+-10-15 13:53:11 GMT                                         patch-70
+     Summary:
+       Better cross compilation handling.
+     Revision:
+       xvidcore--devapi4--1.0--patch-70
+     With this patch it is now possible to cross compile xvid
+     quite easily for win32 platform on a build linux host.
+     Recipe for debian system:
+      $ apt-get install mingw32
+      (or create your own cross compiler/binutils suite and install mingw32
+       header files -- sorry i don't have a recipe for this, this is let as
+       an exercice for the reader)
+      $ cd ${xvidcore}
+      $ cd build/generic
+      $ ./bootstrap.sh
+      $ ./configure --host=i586-mingw32msvc
+       (all occurences of i586-mingw32msvc may be replaced with the right
+        prefix you've choosen for your cross compiler and cross binutils)
+      $ make
+      $ cd ../../vfw/bin
+      $ make -f Makefile.cygwin \
+       CC=i586-mingw32msvc-gcc WINDRES=i586-mingw32msvc-windres
+     Enjoy your win32 xvid.dll build by free software, on a free OS, for a devil
+     OS target.
+     modified files:
+      build/generic/configure.in vfw/bin/Makefile.cygwin
+-10-14 15:17:28 GMT                                         patch-69
+     Summary:
+       Fixed Qpel+Interpolation decoding. Cleaned up mb->mode usage.
+     Revision:
+       xvidcore--devapi4--1.0--patch-69
+     * Fixed interpolate mode + qpel decoding.
+     * MB->mb_type completely replaced by MB->mode
+     modified files:
+      src/decoder.c
+-10-12 21:57:24 GMT                                         patch-68
+     Summary:
+       ac/dc prediction for intra RD search.
+     Revision:
+       xvidcore--devapi4--1.0--patch-68
+     From syskin, added real ac/dc prediction for INTRA's bitcount.
+     modified files:
+      src/motion/estimation_rd_based.c
+-10-09 18:15:50 GMT                                         patch-67
+     Summary:
+       Pigrated asm code to new quant API.
+     Revision:
+       xvidcore--devapi4--1.0--patch-67
+     Many changes that are mostly cosmetic in the asm files.
+      * indent
+      * added xor eax, eax in quant_(h263|mpeg)_intra_.* functions
+        (just to make sure the returned value isn't random)
+      * added xor eax, eax in dequant_(h263|mpeg)_.* functions
+        (just to make sure the returned value isn't random)
+      * synced cpuid.asm XVID_CPU_feature constants with the one
+        defined in the C code (xvid.h)
+      * enabled all cpu tests in xvid_bench.c
+     modified files:
+      examples/xvid_bench.c src/quant/quant_h263.c
+      src/quant/quant_mpeg.c
+      src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_h263_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/utils/x86_asm/cpuid.asm src/xvid.h
+-10-08 21:05:47 GMT                                         patch-66
+     Summary:
+       Updated xvid_bench for quant API changes
+     Revision:
+       xvidcore--devapi4--1.0--patch-66
+     Updated xvid_bench for quant API changes
+     modified files:
+      examples/xvid_bench.c
+-10-07 13:03:51 GMT                                         patch-65
+     Summary:
+       Quant functions API changes (first step)
+     Revision:
+       xvidcore--devapi4--1.0--patch-65
+     In the road to instance safe mpeg quantization, a small cleanup
+     to the quant API was needed. It consists in changing the way we
+     name the functions quant_{mpeg|h263}_{inter|intra}_{arch} and in
+     a move to a more unified API (even intra functions return the sum
+     of coefficients, it can be used as a complexity measure at a later
+     time).
+     This patch touch lot of files, but all changes are trivial.
+     NB: we should check the IA64 asm validity, i changed things but
+         i can't test them.
+     new files:
+      src/quant/.arch-ids/quant.h.id src/quant/quant.h
+     removed files:
+      src/quant/.arch-ids/quant_h263.h.id
+      src/quant/.arch-ids/quant_mpeg4.h.id src/quant/quant_h263.h
+      src/quant/quant_mpeg4.h
+     modified files:
+      build/generic/sources.inc build/win32/libxvidcore.dsp
+      src/decoder.c src/encoder.c src/image/qpel.c
+      src/motion/estimation_rd_based.c
+      src/quant/ia64_asm/quant_h263_ia64.s src/quant/quant_h263.c
+      src/quant/quant_matrix.c src/quant/quant_matrix.h
+      src/quant/quant_mpeg.c
+      src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_h263_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/utils/mbtransquant.c src/xvid.c src/xvid.h
+     renamed files:
+      src/quant/.arch-ids/quant_mpeg4.c.id
+        ==> src/quant/.arch-ids/quant_mpeg.c.id
+      src/quant/quant_mpeg4.c
+        ==> src/quant/quant_mpeg.c
+      src/quant/x86_asm/.arch-ids/quantize4_mmx.asm.id
+        ==> src/quant/x86_asm/.arch-ids/quantize_mpeg_mmx.asm.id
+      src/quant/x86_asm/.arch-ids/quantize4_xmm.asm.id
+        ==> src/quant/x86_asm/.arch-ids/quantize_mpeg_xmm.asm.id
+      src/quant/x86_asm/.arch-ids/quantize_3dne.asm.id
+        ==> src/quant/x86_asm/.arch-ids/quantize_h263_3dne.asm.id
+      src/quant/x86_asm/.arch-ids/quantize_mmx.asm.id
+        ==> src/quant/x86_asm/.arch-ids/quantize_h263_mmx.asm.id
+      src/quant/x86_asm/quantize4_mmx.asm
+        ==> src/quant/x86_asm/quantize_mpeg_mmx.asm
+      src/quant/x86_asm/quantize4_xmm.asm
+        ==> src/quant/x86_asm/quantize_mpeg_xmm.asm
+      src/quant/x86_asm/quantize_3dne.asm
+        ==> src/quant/x86_asm/quantize_h263_3dne.asm
+      src/quant/x86_asm/quantize_mmx.asm
+        ==> src/quant/x86_asm/quantize_h263_mmx.asm
+-10-05 00:15:15 GMT                                         patch-64
+     Summary:
+       Updated ChangeLog
+     Revision:
+       xvidcore--devapi4--1.0--patch-64
+     Updated ChangeLog
+     modified files:
+      changelog.txt
 -10-04 16:04:30 GMT                                         patch-63
      Summary:

 Legend:



Removed from v.1181
 


changed lines


 
Added in v.1242
 Legend:



Removed from v.1181
 


changed lines


 
Added in v.1242
-Removed from v.1181
+Added in v.1242

No admin address has been configured	ViewVC Help
Powered by ViewVC 1.0.4