@DATABASE mp3enc.guide @WIDTH 80 @AUTHOR{Michael Cheng} @NODE MAIN "mp3enc.guide" mp3enc v0.1 Dodgy docs written by @{" Michael Cheng " link "Author"} @{" Introduction " link "Introduction"} @{" Requirements " link "Requirements"} @{" Usage " link "Usage"} @{" Speed " link "Speed"} @{" Problems " link "Problems"} @{" Source Code " link "Source"} @{" Future " link "Future"} @{" Acknowledgements " link "Acknowledgements"} @{" Author " link "Author"} @{" Appendix " link "Appendix"} @ENDNODE @NODE "Introduction" "Introduction" This is the first amiga release version of ISO MPEG's public C source code for mpeg audio layer 3 encoding. The source if freely available from ISO MPEG, but they do not make any testimony as to its usefulness, completeness or quickness. This amiga release is very alpha-ish. Make sure you read the @{"Problems" link "Problems"} section for info. @ENDNODE @NODE "Requirements" "Requirements" REQUIREMENTS - An Amiga with an FPU. Encoding without an FPU is worse than your worst nightmare. - The fastest CPU you can find :) @ENDNODE @NODE "Usage" "Usage" USAGE The 'encode' program may either be run interactively at the command line, or may be given exact command line arguments to run it without interaction. Interactive mode is quite self explanatory. The following command line arguments are taken from the source code. Use encode -h to view the command options. -l is followed by the layer number (1,2,3) -m is followed by the mode ([s]tereo [j]ointstereo [d]ualchannel [m]ono) -p is followed by the psychoacoustic model number (1,2) -s is followed by the sampling rate -b is followed by the total bitrate, irrespective of the mode -d is followed by the emphasis flag ([n]one [5]0/15microseconds [c]citt) -c is followed by the copyright/no_copyright flag -o is followed by the original/not_original flag -e is followed by the error_protection on/off flag If the input file is in AIFF format, the sampling frequency is read from the AIFF header. (NB: AIFF reading is untested) LIMITATIONS - Only psychoacoustic model 2 is available for mp3 encoding. - Joint Stereo encoding is not available for mp3 encoding. EXAMPLE encode -l 3 -p 2 -s 44.1 -b 128 -m m spot1.raw The source file is "spot1.raw" which is a 44khz 16bit mono pcm file. It is being encoded to layer 3 (-l 3) Using psychoacoustic model 2 (-p 2) Source frequency is 44.1khz (-s 44.1) Total encoded bitrate is 128 (-b 128) mp3 mode is mono (-m m) The output file is spot1.raw.mpg (by default) @ENDNODE @NODE "Problems" "Problems" PROBLEMS - Layer I and Layer II audio encoding work fine, but you'd be much better off getting musicin.lha which is nicely optimized (but doesn't yet do layer III) - There are weird memory problems with encoding stereo samples to mp3. This problem doesn't arise when encoding mono samples to mp3. The problem can cause invalidated partitions and a spectacularly big crash of your amiga. So until you're confident, work on a scratch partition, or encode to ram: - Stephane Tavenard (mpega author) told me that there's a bug in part of the encoder (the bit allocation?) - Doesn't yet encode in the common "joint stereo" mp3 format. - Program seems to only like 44.1khz 16bit PCM samples. Let me know if you have success with other frequencies. @ENDNODE @NODE "Speed" "Speed" Encoding Layer 3 mpeg audio files is very CPU intensive. Just to give you an idea on what sort of times you can expect I've included a "standard" sound file which I have been using to time the encoding. [spot1.raw is 1 second of 44.1khz 16bit mono pcm sound] My 030/881/25Mhz takes 190 seconds to encode this to 128bit/s mp3 An 060/50 takes 20 seconds. Since this was just a mono sound file, approximately double the time for encoding stereo sound. So encoding a 44.1khz stereo song of 3 minutes will take about 2 hours at least on an 060. This is just a guess mind you, as I haven't met anyone masochistic enough to do full scale mp3 encoding on an amiga. @ENDNODE @NODE "Source" "Source" The complete mpeg audio demonstration C source is available from ftp://ftp.tnt.uni-hannover.de/pub/MPEG/audio/mpeg2/software/technical_report/dist10.tar.gz Any changes I have made to this default source are included in the /src directory of this archive. You can simply copy my files over the default distribution. @ENDNODE @NODE "Future" "Future" FUTURE - Figure out that @{"memory corruption bug" link "Problems"} - @{"Optimize" link "Optimize"} code - Add in joint stereo encoding. - Keep up to date with the @{"net effort" link "Net"} @ENDNODE @NODE "Net" "Net" The mailing list for amp has had some discussion recently on improving the ISO MPEG source-code. These guys are pretty good programmers and have lots of ideas for optimization and speedier code. As improvements and optimizations become known to me, I'll release new versions. @ENDNODE @NODE "Optimize" "Optimize" The code is written in ANSI C and is meant to be demonstrative C source into how to implement the mpeg audio standard in software. It was not meant to be fast, tight or optimized for any particular processor. Easy C optimizations don't seem that apparent. Significant speedups will probably come from ASM replacements for the CPU intensive routines. Included below is a profile of the encode program. Lots of time is spent in 3 or 4 main routines. So if you feel like having a go at optimizing in C or ASM, concentrate on those functions ie fft, filter_subband, quantize. NB: I'm not at all experienced in putting ASM code in C code using the gcc compiler. Any pointers would be appreciated. Flat profile: Each sample counts as 0.02 seconds. % cumulative self self total time seconds seconds calls ns/call ns/call name 24.83 51.58 51.58 fft 18.82 90.68 39.10 filter_subband 14.45 120.70 30.02 quantize 8.97 139.34 18.64 mdct 7.17 154.23 14.89 mcount 5.17 164.97 10.74 L3psycho_anal 4.62 174.57 9.60 HuffmanCode 3.97 182.81 8.24 nint 3.71 190.51 7.70 window_subband 1.02 192.63 2.12 calc_noise 0.82 194.33 1.70 count_bit 0.75 195.89 1.56 quantanf_init 0.69 197.33 1.44 mdct_sub 0.64 198.65 1.32 956280 1380.35 1380.35 mcount 0.39 199.47 0.82 calc_scfsi 0.32 200.13 0.66 xr_max 0.31 200.78 0.65 putbits 0.26 201.32 0.54 BF_PartLength 0.25 201.84 0.52 count_bits 0.22 202.30 0.46 BF_addEntry 0.19 202.70 0.40 BF_addElement 0.19 203.10 0.40 get_audio 0.19 203.49 0.39 _mcleanup 0.18 203.87 0.38 iteration_loop 0.17 204.23 0.36 L3para_read 0.17 204.59 0.36 calc_xmin 0.16 204.93 0.34 new_choose_table 0.13 205.19 0.26 II_bits_for_nonoise 0.11 205.41 0.22 I_bits_for_nonoise 0.10 205.61 0.20 BF_resizePartHolder 0.09 205.79 0.18 III_FlushBitstream 0.09 205.97 0.18 huffman_decoder 0.08 206.13 0.16 III_format_bitstream 0.08 206.29 0.16 count1_bitcount 0.07 206.43 0.14 create_ana_filter 0.07 206.57 0.14 ix_max 0.07 206.71 0.14 putMyBits 0.06 206.83 0.12 BF_NewHolderFromBitstreamPart 0.06 206.95 0.12 read_decoder_table 0.06 207.07 0.12 read_huffcodetab 0.05 207.17 0.10 abs_and_sign 0.04 207.25 0.08 bigv_bitcount 0.04 207.33 0.08 huffman_coder 0.03 207.39 0.06 BF_newPartHolder 0.03 207.45 0.06 ConvertToIeeeSingle 0.02 207.49 0.04 I_scale_factor_calc 0.02 207.53 0.04 rewindNbytes 0.01 207.55 0.02 BF_LoadHolderFromBitstreamPart 0.01 207.57 0.02 BF_freePartHolder 0.01 207.59 0.02 BitrateIndex 0.01 207.61 0.02 ConvertFromIeeeSingle 0.01 207.63 0.02 aint 0.01 207.65 0.02 calc_runlen 0.01 207.67 0.02 free 0.01 207.69 0.02 fwrite 0.01 207.71 0.02 read_ana_window 0.01 207.73 0.02 sscanf 0.01 207.75 0.02 subdivide 0.00 207.76 0.01 getbits % the percentage of the total running time of the time program used by this function. cumulative a running sum of the number of seconds accounted seconds for by this function and those listed above it. self the number of seconds accounted for by this seconds function alone. This is the major sort for this listing. calls the number of times this function was invoked, if this function is profiled, else blank. self the average number of milliseconds spent in this ms/call function per call, if this function is profiled, else blank. total the average number of milliseconds spent in this ms/call function and its descendents per call, if this function is profiled, else blank. name the name of the function. This is the minor sort for this listing. The index shows the location of the function in the gprof listing. If the index is in parenthesis it shows where it would appear in the gprof listing if it were to be printed. Call graph (explanation follows) granularity: each sample hit covers 4 byte(s) for 0.01% of 191.16 seconds index % time self children called name [1] 27.0 51.58 0.00 fft [1] 0.00 0.00 2/956280 mcount (57) ----------------------------------------------- [2] 20.5 39.10 0.00 filter_subband [2] 0.00 0.00 2/956280 mcount (57) ----------------------------------------------- [3] 15.7 30.02 0.00 quantize [3] 0.00 0.00 432576/956280 mcount (57) ----------------------------------------------- [4] 9.8 18.64 0.00 mdct [4] ----------------------------------------------- [5] 5.6 10.74 0.00 L3psycho_anal [5] 0.00 0.00 2354/956280 mcount (57) ----------------------------------------------- [6] 5.0 9.60 0.00 HuffmanCode [6] ----------------------------------------------- [7] 4.3 8.24 0.00 nint [7] ----------------------------------------------- [8] 4.0 7.70 0.00 window_subband [8] 0.00 0.00 3/956280 mcount (57) ----------------------------------------------- [9] 1.1 2.12 0.00 calc_noise [9] ----------------------------------------------- [10] 0.9 1.70 0.00 count_bit [10] 0.00 0.00 296344/956280 mcount (57) ----------------------------------------------- [11] 0.8 1.56 0.00 quantanf_init [11] 0.00 0.00 75/956280 mcount (57) ----------------------------------------------- [12] 0.8 1.44 0.00 mdct_sub [12] 0.00 0.00 2496/956280 mcount (57) ----------------------------------------------- [13] 0.4 0.82 0.00 calc_scfsi [13] 0.00 0.00 78/956280 mcount (57) ----------------------------------------------- [14] 0.3 0.66 0.00 xr_max [14] ----------------------------------------------- [15] 0.3 0.65 0.00 putbits [15] 0.00 0.00 12279/956280 mcount (57) ----------------------------------------------- [16] 0.3 0.54 0.00 BF_PartLength [16] 0.00 0.00 22670/956280 mcount (57) ----------------------------------------------- [17] 0.3 0.52 0.00 count_bits [17] 181170887527902135823437523267474075217788929994682683463763719356416.00 0.00 2532/956280 mcount (57) ----------------------------------------------- [18] 0.2 0.46 0.00 BF_addEntry [18] 605818841847439428249236568578918254354336256457716771825440177277036634835629933150429943074526587700645054172271344758075074219307229338132676649943040.00 185012683372606779184939189948247748191489885389347812729919754993664.00 47264/956280 mcount (57) ----------------------------------------------- [19] 0.2 0.40 0.00 BF_addElement [19] 0.00 0.00 12/956280 mcount (57) ----------------------------------------------- [20] 0.2 0.40 0.00 get_audio [20] 4047476907590138774493894118156762147196630027686710400253952.00 186528604498704993296828330197696015318492646116357259408375526981632.00 43/956280 mcount (57) ----------------------------------------------- [21] 0.2 0.38 0.00 iteration_loop [21] 0.00 0.00 777/956280 mcount (57) ----------------------------------------------- [22] 0.2 0.36 0.00 L3para_read [22] 0.00 0.00 1516/956280 mcount (57) ----------------------------------------------- [23] 0.2 0.36 0.00 calc_xmin [23] ----------------------------------------------- [24] 0.2 0.34 0.00 new_choose_table [24] 0.00 0.00 3542/956280 mcount (57) ----------------------------------------------- [25] 0.1 0.26 0.00 II_bits_for_nonoise [25] ----------------------------------------------- [26] 0.1 0.22 0.00 I_bits_for_nonoise [26] ----------------------------------------------- [27] 0.1 0.20 0.00 BF_resizePartHolder [27] ----------------------------------------------- [28] 0.1 0.18 0.00 III_FlushBitstream [28] 0.00 0.00 45225/956280 mcount (57) ----------------------------------------------- [29] 0.1 0.18 0.00 huffman_decoder [29] ----------------------------------------------- [30] 0.1 0.16 0.00 III_format_bitstream [30] 0.00 0.00 179/956280 mcount (57) ----------------------------------------------- [31] 0.1 0.16 0.00 count1_bitcount [31] 0.00 0.00 29640/956280 mcount (57) ----------------------------------------------- [32] 0.1 0.14 0.00 create_ana_filter [32] ----------------------------------------------- [33] 0.1 0.14 0.00 ix_max [33] ----------------------------------------------- [34] 0.1 0.14 0.00 putMyBits [34] 0.00 0.00 23668/956280 mcount (57) ----------------------------------------------- [35] 0.1 0.12 0.00 BF_NewHolderFromBitstreamPart [35] ----------------------------------------------- [36] 0.1 0.12 0.00 read_decoder_table [36] ----------------------------------------------- [37] 0.1 0.12 0.00 read_huffcodetab [37] 72036449917790047203119181568699121193716387992961993939503963630018535499313810933233172723059338501907572632602295074500600661725914190418899158432327893052179950275114164516833180574811284745742205321216.00 0.00 3162/956280 mcount (57) ----------------------------------------------- [38] 0.1 0.10 0.00 abs_and_sign [38] ----------------------------------------------- [39] 0.0 0.08 0.00 bigv_bitcount [39] 0.00 0.00 9494/956280 mcount (57) ----------------------------------------------- [40] 0.0 0.08 0.00 huffman_coder [40] ----------------------------------------------- [41] 0.0 0.06 0.00 BF_newPartHolder [41] 0.00 0.00 93/956280 mcount (57) ----------------------------------------------- [42] 0.0 0.06 0.00 ConvertToIeeeSingle [42] ----------------------------------------------- [43] 0.0 0.04 0.00 I_scale_factor_calc [43] ----------------------------------------------- [44] 0.0 0.04 0.00 rewindNbytes [44] ----------------------------------------------- [45] 0.0 0.02 0.00 BF_LoadHolderFromBitstreamPart [45] 0.00 63831252999598997616031938012764816835702467370965125406544302975248479091727950234001828619861383636281895469195849632052327734654757268413895341528040058105560535902771711594484156597258142353929044623657979067897665039957491712.00 1765/956280 mcount (57) ----------------------------------------------- [46] 0.0 0.02 0.00 BF_freePartHolder [46] 0.00 0.00 39/956280 mcount (57) ----------------------------------------------- [47] 0.0 0.02 0.00 BitrateIndex [47] ----------------------------------------------- [48] 0.0 0.02 0.00 ConvertFromIeeeSingle [48] ----------------------------------------------- [49] 0.0 0.02 0.00 aint [49] ----------------------------------------------- [50] 0.0 0.02 0.00 calc_runlen [50] ----------------------------------------------- [51] 0.0 0.02 0.00 free [51] ----------------------------------------------- [52] 0.0 0.02 0.00 fwrite [52] ----------------------------------------------- [53] 0.0 0.02 0.00 read_ana_window [53] 0.00 0.00 386/956280 mcount (57) ----------------------------------------------- [54] 0.0 0.02 0.00 sscanf [54] ----------------------------------------------- [55] 0.0 0.02 0.00 subdivide [55] ----------------------------------------------- [56] 0.0 0.01 0.00 getbits [56] ----------------------------------------------- This table describes the call tree of the program, and was sorted by the total amount of time spent in each function and its children. Each entry in this table consists of several lines. The line with the index number at the left hand margin lists the current function. The lines above it list the functions that called this function, and the lines below it list the functions this one called. This line lists: index A unique number given to each element of the table. Index numbers are sorted numerically. The index number is printed next to every function name so it is easier to look up where the function in the table. % time This is the percentage of the `total' time that was spent in this function and its children. Note that due to different viewpoints, functions excluded by options, etc, these numbers will NOT add up to 100%. self This is the total amount of time spent in this function. children This is the total amount of time propagated into this function by its children. called This is the number of times the function was called. If the function called itself recursively, the number only includes non-recursive calls, and is followed by a `+' and the number of recursive calls. name The name of the current function. The index number is printed after it. If the function is a member of a cycle, the cycle number is printed between the function's name and the index number. For the function's parents, the fields have the following meanings: self This is the amount of time that was propagated directly from the function into this parent. children This is the amount of time that was propagated from the function's children into this parent. called This is the number of times this parent called the function `/' the total number of times the function was called. Recursive calls to the function are not included in the number after the `/'. name This is the name of the parent. The parent's index number is printed after it. If the parent is a member of a cycle, the cycle number is printed between the name and the index number. If the parents of the function cannot be determined, the word `' is printed in the `name' field, and all the other fields are blank. For the function's children, the fields have the following meanings: self This is the amount of time that was propagated directly from the child into the function. children This is the amount of time that was propagated from the child's children to the function. called This is the number of times the function called this child `/' the total number of times the child was called. Recursive calls by the child are not listed in the number after the `/'. name This is the name of the child. The child's index number is printed after it. If the child is a member of a cycle, the cycle number is printed between the name and the index number. If there are any cycles (circles) in the call graph, there is an entry for the cycle-as-a-whole. This entry shows who called the cycle (as parents) and the members of the cycle (as children.) The `+' recursive calls entry shows the number of function calls that were internal to the cycle, and the calls entry for each member shows, for that member, how many times it was called from other members of the cycle. Index by function name [45] BF_LoadHolderFromBitstreamPart [38] abs_and_sign [33] ix_max [35] BF_NewHolderFromBitstreamPart [49] aint (217) mcount [16] BF_PartLength [39] bigv_bitcount [4] mdct [19] BF_addElement [9] calc_noise [12] mdct_sub [18] BF_addEntry [50] calc_runlen [24] new_choose_table [46] BF_freePartHolder [13] calc_scfsi [7] nint [41] BF_newPartHolder [23] calc_xmin [34] putMyBits [27] BF_resizePartHolder [31] count1_bitcount [15] putbits [47] BitrateIndex [10] count_bit [11] quantanf_init [48] ConvertFromIeeeSingle [17] count_bits [3] quantize [42] ConvertToIeeeSingle [32] create_ana_filter [53] read_ana_window [6] HuffmanCode [1] fft [36] read_decoder_table [28] III_FlushBitstream [2] filter_subband [37] read_huffcodetab [30] III_format_bitstream [51] free [44] rewindNbytes [25] II_bits_for_nonoise [52] fwrite [54] sscanf [26] I_bits_for_nonoise [20] get_audio [55] subdivide [43] I_scale_factor_calc [56] getbits [8] window_subband [22] L3para_read [40] huffman_coder [14] xr_max [5] L3psycho_anal [29] huffman_decoder (57) mcount (154) _mcleanup [21] iteration_loop @ENDNODE @NODE "Acknowledgements" "Acknowledgements" THANKS - Stéphane TAVENARD (tavenard@xiii.univ-angers.fr)for mpega, the best amiga mpeg audio player. - Troels and SimD and Lazaruz and others i've forgotten on #amiga for helping me test. - ISO MPEG Audio Subgroup Software Simulation Group for making the source available. @ENDNODE @NODE "Author" "Author" mike cheng Cstar on #amiga mikecheng@cryogen.com http://www.netforward.com/cryogen/?mikecheng http://mikecheng.dyn.ml.org/ @ENDNODE @NODE "Appendix" "Appendix" @ENDNODE