changelog

音频编码

	regen by hand.

2005-05-04 17:00  takehiro

	* configure.in (takehiro-2002_05_07-experimental):

	- do not use "-march" option for the case of cross build.
	- added comments about SPARC + IEEE754 hack

2005-05-04 16:58  takehiro

	* libmp3lame/: quantize.c, i386/quantSSE.nas
	  (takehiro-2002_05_07-experimental):

	update x^(3/4) hack.
	- use IEEE754 HACK version of fabs()
	- better "digital silent" detection. not use sum, but use maximum.

2005-05-04 16:57  takehiro

	* libmp3lame/psymodel.c (takehiro-2002_05_07-experimental):

	common expression reduction.

2005-04-26 16:08  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix selection of the band to increase the scalefactor, when all the band have no
	audible noise.

2005-04-25 16:35  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix handling of SCALEFAC_ANYTHING_GOES in calc_noise().
	This sometimes makes assertion failure when with --substep option.

2005-04-24 10:28  takehiro

	* frontend/main.c (takehiro-2002_05_07-experimental):

	fix it skips clipping when with --decode option.

2005-04-23 18:25  takehiro

	* libmp3lame/mpglib_interface.c (takehiro-2002_05_07-experimental):

	clean up.

2005-04-23 18:20  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	Handmade aprox. code to calculate x^(3/4) for IEEE754 machines.
	On MacOSX + G4, the speed gain is arround 14% !

2005-04-23 15:31  takehiro

	* libmp3lame/gain_analysis.c (takehiro-2002_05_07-experimental):

	supress warnings

2005-04-23 15:11  takehiro

	* testcase.mp3, frontend/get_audio.c, frontend/get_audio.h,
	  frontend/gtkanal.c, frontend/main.c, include/lame.h,
	  libmp3lame/lame.c, libmp3lame/mpglib_interface.c
	  (takehiro-2002_05_07-experimental):

	use floating values for the data between frontend and encoding core.

2005-04-23 10:10  takehiro

	* libmp3lame/mpglib_interface.c (takehiro-2002_05_07-experimental):

	prevent to re-initialization.

2005-04-23 10:08  takehiro

	* frontend/get_audio.c (takehiro-2002_05_07-experimental):

	add error check for lame_decoe_init()
	fix comment.

2005-04-23 09:57  takehiro

	* libmp3lame/mpglib_interface.c (takehiro-2002_05_07-experimental):

	clean up

2005-04-23 09:56  takehiro

	* frontend/get_audio.c (takehiro-2002_05_07-experimental):

	clean up endianness.
	fix mp3 decoding.

2005-04-23 08:42  takehiro

	* libmp3lame/tables.c (takehiro-2002_05_07-experimental):

	tuning masking parameter balance for long/short blocks.

2005-04-23 08:11  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix memory copy size in CBR_2nd_iteration().
	This bug causes assertion failure when CBR with -h option.

2005-04-23 08:07  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	reduce memory copy size.

2005-04-23 07:57  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	little bit better initial value for quantization factor search.

2005-04-23 07:55  takehiro

	* libmp3lame/: quantize.c, quantize_pvt.h
	  (takehiro-2002_05_07-experimental):

	reduce memory copy size.

2005-04-22 17:39  takehiro

	* mpglib/interface.c (takehiro-2002_05_07-experimental):

	supress warnings with GCC4.0

2005-04-21 16:36  takehiro

	* libmp3lame/bitstream.c (takehiro-2002_05_07-experimental):

	oops, stupid bug in loop counter... sorry.

2005-04-21 15:42  takehiro

	* libmp3lame/tables.c (takehiro-2002_05_07-experimental):

	improve the spreading function for the lower frequency,
	where the resolution of frequency is not sufficient for phsycho-analysis.

2005-04-21 15:36  takehiro

	* libmp3lame/tables.c (takehiro-2002_05_07-experimental):

	clean up.

2005-04-21 15:03  takehiro

	* libmp3lame/lame.c (takehiro-2002_05_07-experimental):

	verbose the configuration message.

2005-04-17 17:28  takehiro

	* libmp3lame/lame.c (takehiro-2002_05_07-experimental):

	clean up arround lame_print_internals()

2005-04-17 16:16  takehiro

	* frontend/: get_audio.c, get_audio.h, main.c, main.h, parse.c
	  (takehiro-2002_05_07-experimental):

	clean up endianness detection and configuration.

2005-04-17 15:49  takehiro

	* frontend/: get_audio.c, get_audio.h
	  (takehiro-2002_05_07-experimental):

	clean up MAX_U_32_NUM definition

2005-04-17 15:42  takehiro

	* frontend/parse.c (takehiro-2002_05_07-experimental):

	general clean up.

2005-04-17 15:35  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	fix assertion failure when CBR encoding.

2005-04-17 12:49  takehiro

	* frontend/parse.c (takehiro-2002_05_07-experimental):

	clean up options.
	- make some options as experimental.
	- fix floating point/integer argument.

2005-04-17 11:01  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	optimized best_huffman_divide(), by removing needless call of choose_table().

2005-04-16 15:58  takehiro

	* libmp3lame/bitstream.c (takehiro-2002_05_07-experimental):

	general clean up.
	- supress magic numbers
	- use putbits16() instead of putbits24(), where it can.
	- trivial loop counter optimization

2005-04-16 13:55  takehiro

	* testcase.mp3, libmp3lame/psymodel.c
	  (takehiro-2002_05_07-experimental):

	ATH value adjustment for M/S channels.

2005-04-16 13:53  takehiro

	* libmp3lame/bitstream.c, frontend/brhist.c, frontend/main.c
	  (takehiro-2002_05_07-experimental):

	supress warnings

2005-04-16 10:42  takehiro

	* libmp3lame/psymodel.c (takehiro-2002_05_07-experimental):

	reduce conditional jumps

2005-04-16 10:42  takehiro

	* libmp3lame/machine.h (takehiro-2002_05_07-experimental):

	added (FLOAT) cast in fast_log() macros.

2005-04-13 14:12  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	oops, I disabled the new code...

2005-04-13 14:11  takehiro

	* frontend/gtkanal.c (takehiro-2002_05_07-experimental):

	one more fix for the delay.

2005-04-13 13:50  takehiro

	* frontend/gtkanal.c (takehiro-2002_05_07-experimental):

	fix frame delay

2005-04-13 13:41  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	optimize lossless coding efficiency and speed (mainly for low bitrate).

2005-04-09 17:50  takehiro

	* libmp3lame/bitstream.c (takehiro-2002_05_07-experimental):

	optimized bitstream copy code.

2005-04-09 17:29  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	one more reduction of needless divide point search in best_huffman_divide()

2005-04-09 09:58  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	prevent needless search of best_huffman_divide().

2005-04-04 01:20  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	optimize best_scalefac_store()
	- only call noquant_count_bits() when it is really needed.
	- skip the band with the SCALEFAC_ANYTHING_GOES.

2005-04-03 18:28  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	little bit better lossless coding.

2005-04-03 18:27  takehiro

	* libmp3lame/bitstream.c (takehiro-2002_05_07-experimental):

	make assertion condition strictly.

2005-04-03 17:42  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	one more off by one... oops.

2005-04-03 17:24  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	fix off-by-one bug.

2005-04-03 12:25  takehiro

	* libmp3lame/: quantize.c, quantize_pvt.h
	  (takehiro-2002_05_07-experimental):

	use the theoretical minimum value of xr34.

2005-04-03 11:57  takehiro

	* frontend/: gtkanal.c, mp3x.c (takehiro-2002_05_07-experimental):

	fix (part of) mp3x display delay problem

2005-04-03 11:32  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	fix table selection

2005-04-03 05:50  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	reduce calling ixmax() and simplify count_bits()

2005-04-02 18:32  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	potential fix for floating point exception when analysing MPEG2 Layer3 encoding by mp3x.

2005-04-02 17:06  takehiro

	* frontend/timestatus.c (takehiro-2002_05_07-experimental):

	reduce calling GetRealTime()

2005-04-02 16:45  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	Try to find the better quantization factor in the higher frequncy band,
	even when it fails to encode without the noise in the lower frequency band.

2005-04-02 16:21  takehiro

	* libmp3lame/i386/choose_table.nas
	  (takehiro-2002_05_07-experimental):

	simplify the loop condition.

2005-04-02 14:54  takehiro

	* libmp3lame/bitstream.c (takehiro-2002_05_07-experimental):

	do not use putbits24(), where the length to write is lesser than 18.

2005-04-02 14:02  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	fix off-by-one bug in best_huffman_divide()

2005-04-02 12:21  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	fix comments and reduce the table size.

2005-04-02 12:13  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	simplify cont_bit_noESC_from4() and add comments.

2005-04-02 11:56  takehiro

	* libmp3lame/takehiro.c (takehiro-2002_05_07-experimental):

	remove needless memory clear.

2005-04-02 07:28  takehiro

	* libmp3lame/version.h (takehiro-2002_05_07-experimental):

	Now alpha 14.
	- Dramatical VBR speed improvement.
	  10-20% on typical archtecture (i86, Alpha, Sparc), and 95% on G4.
	  CBR also gains but it is little (2% on the typical arch, 30% on G4).
	- New and improved ATH level autoadjustment.
	- Use subblock gain more (CBR and VBR).
	- Fix quantization bug in count1 region, which causes annoying noise in high-freq.

2005-04-02 07:16  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	supress magic numbers.

2005-03-29 15:07  takehiro

	* libmp3lame/: machine.h, quantize.c, tables.c, takehiro.c
	  (takehiro-2002_05_07-experimental):

	supress magic numbers.

2005-03-28 17:26  takehiro

	* libmp3lame/machine.h (takehiro-2002_05_07-experimental):

	it seems thmath.h decreases spped very much on Alpha system.

2005-03-28 16:58  takehiro

	* libmp3lame/lame.c (takehiro-2002_05_07-experimental):

	fix debug/release build is reverted and warnings

2005-03-28 16:57  takehiro

	* libmp3lame/encoder.h (takehiro-2002_05_07-experimental):

	fix built on 64bit machines (I checked it on Alpha and AMD64)

2005-03-28 09:44  takehiro

	* frontend/parse.c (takehiro-2002_05_07-experimental):

	adjust message width

2005-03-28 03:51  takehiro

	* libmp3lame/: quantize.c, takehiro.c
	  (takehiro-2002_05_07-experimental):

	fix SCALEFAC_ANYTHING_GOES bug finally! now it gains speed improvement arround 15%.

	and fix count1 region quantization.

2005-03-28 03:50  takehiro

	* libmp3lame/lame.c (takehiro-2002_05_07-experimental):

	added alignment check when debug build.

2005-03-27 17:48  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix treatment of scale factor band with "anything goes".
	But it seems there're still bugs arround it.

2005-03-27 16:45  takehiro

	* libmp3lame/: machine.h, quantize.c, quantize_pvt.h, tables.c,
	  tables.h, takehiro.c (takehiro-2002_05_07-experimental):

	fix it may use preflag on short blocks.
	use integer value of quantization adjustments.

2005-03-27 15:22  takehiro

	* libmp3lame/i386/quantSSE.nas (takehiro-2002_05_07-experimental):

	fix comments one more.

2005-03-27 11:57  takehiro

	* libmp3lame/i386/quantSSE.nas (takehiro-2002_05_07-experimental):

	fixed comments

2005-03-27 09:32  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	code clean up.

2005-03-26 19:58  takehiro

	* libmp3lame/i386/choose_table.nas
	  (takehiro-2002_05_07-experimental):

	fix comment

2005-03-26 19:54  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	do not call scale_bitcounter() to check the scalefac values is valid or not.
	(spped up arround 1.5%)

2005-03-26 19:37  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix it recognize (scalefac=SCALEFAC_ANYTHING_GOES) + (subblock gain = 2) as
	not an amplified band.

2005-03-26 10:39  takehiro

	* libmp3lame/: quantize.c, quantize_pvt.h
	  (takehiro-2002_05_07-experimental):

	suppress magic numbers

2005-03-26 09:48  takehiro

	* frontend/: main.c, timestatus.c
	  (takehiro-2002_05_07-experimental):

	reduce calling lame_get_frameNum()

2005-03-23 03:08  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	use subblock_gain more when VBR.

2005-03-23 03:06  takehiro

	* libmp3lame/: machine.h, tables.c
	  (takehiro-2002_05_07-experimental):

	cleanup

2005-03-22 16:58  takehiro

	* libmp3lame/: tables.c, takehiro.c, tables.h
	  (takehiro-2002_05_07-experimental):

	make scfsi_band[] as static variable.

2005-03-22 13:42  takehiro

	* libmp3lame/tables.c (takehiro-2002_05_07-experimental):

	remove gabages.
	oops...

2005-03-22 13:40  takehiro

	* libmp3lame/: quantize_pvt.h, tables.c
	  (takehiro-2002_05_07-experimental):

	more precise ROUNDFAC definition.

2005-03-22 07:13  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	fix built problem on old (not C99) compilers.

2005-03-21 16:24  takehiro

	* libmp3lame/: encoder.h, quantize.c, quantize_pvt.h, takehiro.c
	  (takehiro-2002_05_07-experimental):

	reduce memory copy size.

2005-03-21 14:38  takehiro

	* libmp3lame/quantize.c (takehiro-2002_05_07-experimental):

	Fix loop variable initialization.
	This will fix decreasing bitrate too much on VBR.