mad-user

mad-user@lists.mars.org

303 discussions

S: faster decoder than libmad
by jens-lippmann＠t-online.de 02 Jan '05

02 Jan '05

Hello all! I'm looking for a MP3 decoding lib faster than libmad, probably solely written in Intel assembler. I currently use libmad 0.15.1b to stream MP3 files into WAV sound buffers and I need to improve the performance. The mere task of streaming a single audio file saps about 11 milliseconds (@1.8GHz). libmad uses inline and asm here and there, but the costliest parts (frame decoding and synthing) are written in C, optimized for speed but still C. So, my question is if there's anything freely available that decodes faster than libmad? kind regards Jens Lippmann

3 2

S: faster decoder than libmad (2)
by jens-lippmann＠t-online.de 30 Dec '04

30 Dec '04

> The mere task of streaming a single audio file > saps about 11 milliseconds (@1.8GHz). Sorry I just realized I need to correct this figure. I mean the task to feed a 176 KB/s wave stream in real-time affords frame decoding and synthing of about 44 milliseconds (per second) on a P4 1800. This means 4-5% of your application is needed to keep a single wave stream going, so I'm looking to improve that performance. kind regards Jens Lippmann

1 0

[help] support multiple instance of libmad
by Sangwoon Chung 08 Dec '04

08 Dec '04

Hello there, I'm trying to decoding and playing multiple mp3 files in a time using libmad. But it does not seem to support multiple instance of libmad. I really want to know the libmad support multiple instance ? If not, please let me know how and what should I customize libmad source to support multiple instance. I'll appreciate any Help, P.S Sorry for posting this articel for mad-dev list... Thanks a lot, Sangwoon swchung(a)rtsolutions.co.kr

2 1

Bad sync frame detection
by Hugues Lismonde 11 Nov '04

11 Nov '04

Hi, I've got a little problem with libmad. On some mp3 I have, libmad find a first frame somewhere it probably shouldn't. The problem appeared with mythtv where some files were played in mono instead of stereo and at half the speed. I made some tests with madlld and I get some strange output like this: [...] madlld: recoverable frame level error (forbidden bitrate value) madlld: recoverable frame level error (forbidden bitrate value) madlld: recoverable frame level error (forbidden bit allocation value) madlld: recoverable frame level error (forbidden bit allocation value) madlld: recoverable frame level error (forbidden bit allocation value) madlld: 25000 kb/s audio MPEG layer I stream without CRC, single channel with no emphasis at 44100 Hz sample rate madlld: recoverable frame level error (lost synchronization) madlld: recoverable frame level error (reserved header layer value) [...] I understand that "lost sync" can occur (and it does on good files too so I guess it's ok) but I can't find 25000 kb/s in bitrate_table (frame.c) so I think the "forbidden bit allocation value" errors offsets something by less than a byte. I wrote a small test program which search for sync word and try to decode the frame header without using libmad. Using it I get the following output: Frame at 504 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 508 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 512 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 516 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 520 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 532 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 536 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 540 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 544 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 548 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 560 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 564 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 568 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 572 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 576 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 588 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 592 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 596 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 600 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 604 - MPEG v1 Layer I No CRC Bad frame! (bitrate) Frame at 48330 - MPEG v1 Layer III No CRC Bitrate 192 Samplerate 44 Padding 1 Stereo No emphasis [...] It was written fast so it might be wrong, but it seems correct (the last line is consistent with iTunes or mp3blaster). I haven't found how to get the current location in the buffer from libmad so I can't tell where it get this strange frame. I'll be glad to give any precision or do any tests one might want but I'm a bit lost right now. From what I can see, libmad resync on the next frame, but for mythtv it's too late since it never rebind the channels or sample rate for the current playing song. If it's not the right list, please let me know so I can ask somewhere else. -- Hugues Lismonde < hugues.lismonde(a)hli.be >

1 0

madlld extension for proper 16bit pcm output
by jens-lippmann＠t-online.de 07 Nov '04

07 Nov '04

Hi all! Those acquainted with madlld (libMAD low-level demo) may have noticed the rather poor PCM output which comprises lots of noticeable sound dropouts. This is related to a quick'n'dirty 16bit truncation done by MadFixedToUshort. I spent several hours into gathering information about this problem and analysing the madplay package which provides better algorithms for this task. Finally I wrote a header to be included in madlld.cpp. I just want to keep you up to date with my implementation, so I've attached the header. Please drop me a note if you find it useful. Note that the copyright of the header still is retained by the madplay authors. Good bye and have fun! Jens Lippmann --- How to incorporate the header: 1. Add these lines below the include section in madlld.cpp: #include "madplay.h" struct audio_stats Audio_stats; struct audio_dither Audio_dither; //#define MadFixedToUshort MadFixedToUshortOrig #define MadFixedToUshort(a) (unsigned short)audio_linear_round(16,a,&Audio_stats) //#define MadFixedToUshort(a) (unsigned short)audio_linear_dither(16,a,&Audio_dither,&Audio_stats) 2. Rename the function MadFixedToUshort into MadFixedToUshortOrig. 3. Clear Audio_stats and Audio_dither before you call MpegAudioDecoder (the decoding loop): ZeroMemory(&Audio_stats,sizeof(Audio_stats)); ZeroMemory(&Audio_dither,sizeof(Audio_dither)); 4. Convert your MP3 stream to PCM and listen with awe. ;-> --- The header follows below (I do not want to add it as an enclosure because it might not show up in the list archive then). 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< 8< /* * madplay - MPEG audio decoder and player * Copyright (C) 2000-2004 Robert Leslie * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ struct audio_stats { unsigned long clipped_samples; mad_fixed_t peak_clipping; mad_fixed_t peak_sample; }; struct audio_dither { mad_fixed_t error[3]; mad_fixed_t random; }; /* * NAME: clip() * DESCRIPTION: gather signal statistics while clipping */ static inline void clip(mad_fixed_t *sample, struct audio_stats *stats) { enum { MIN = -MAD_F_ONE, MAX = MAD_F_ONE - 1 }; if (*sample >= stats->peak_sample) { if (*sample > MAX) { ++stats->clipped_samples; if (*sample - MAX > stats->peak_clipping) stats->peak_clipping = *sample - MAX; *sample = MAX; } stats->peak_sample = *sample; } else if (*sample < -stats->peak_sample) { if (*sample < MIN) { ++stats->clipped_samples; if (MIN - *sample > stats->peak_clipping) stats->peak_clipping = MIN - *sample; *sample = MIN; } stats->peak_sample = -*sample; } } /* * NAME: audio_linear_round() * DESCRIPTION: generic linear sample quantize routine */ # if defined(_MSC_VER) extern /* needed to satisfy bizarre MSVC++ interaction with inline */ # endif inline signed long audio_linear_round(unsigned int bits, mad_fixed_t sample, struct audio_stats *stats) { /* round */ sample += (1L << (MAD_F_FRACBITS - bits)); /* clip */ clip(&sample, stats); /* quantize and scale */ return sample >> (MAD_F_FRACBITS + 1 - bits); } /* * NAME: prng() * DESCRIPTION: 32-bit pseudo-random number generator */ static inline unsigned long prng(unsigned long state) { return (state * 0x0019660dL + 0x3c6ef35fL) & 0xffffffffL; } /* * NAME: audio_linear_dither() * DESCRIPTION: generic linear sample quantize and dither routine */ # if defined(_MSC_VER) extern /* needed to satisfy bizarre MSVC++ interaction with inline */ # endif inline signed long audio_linear_dither(unsigned int bits, mad_fixed_t sample, struct audio_dither *dither, struct audio_stats *stats) { unsigned int scalebits; mad_fixed_t output, mask, random; enum { MIN = -MAD_F_ONE, MAX = MAD_F_ONE - 1 }; /* noise shape */ sample += dither->error[0] - dither->error[1] + dither->error[2]; dither->error[2] = dither->error[1]; dither->error[1] = dither->error[0] / 2; /* bias */ output = sample + (1L << (MAD_F_FRACBITS + 1 - bits - 1)); scalebits = MAD_F_FRACBITS + 1 - bits; mask = (1L << scalebits) - 1; /* dither */ random = prng(dither->random); output += (random & mask) - (dither->random & mask); dither->random = random; /* clip */ if (output >= stats->peak_sample) { if (output > MAX) { ++stats->clipped_samples; if (output - MAX > stats->peak_clipping) stats->peak_clipping = output - MAX; output = MAX; if (sample > MAX) sample = MAX; } stats->peak_sample = output; } else if (output < -stats->peak_sample) { if (output < MIN) { ++stats->clipped_samples; if (MIN - output > stats->peak_clipping) stats->peak_clipping = MIN - output; output = MIN; if (sample < MIN) sample = MIN; } stats->peak_sample = -output; } /* quantize */ output &= ~mask; /* error feedback */ dither->error[0] = sample - output; /* scale */ return output >> scalebits; }

1 0

Libmad and libid3tag
by David Barron 26 Oct '04

26 Oct '04

Is it possible to get libmad and libid3tag as windows .dll's? Thanks

1 0

MAD and Jack.....
by andrew＠fnarg.demon.co.uk 08 Sep '04

08 Sep '04

Hi there, I'm writing a playout system which utilises Jack and MAD to play MP3 files. I've written a simple app which under libao (I was using this prior to Jack) will happily play any MP3 file I throw at it. However, upon adding the relevant code to make it a Jack client I've run across a problem. MAD seems to output audio at either 24bit fixed point or 16bit signed whereas Jack expects audio in 32bit floating point. Is there an easy way to make this work, am I being thick? Please help! Cheers, Andy M

2 1

How to manipulate both V1 and V2 tags in the same file
by Cedric Tefft 07 Sep '04

07 Sep '04

I've been playing around with libid3tag 0.15.1b for a couple days now, and I'm at a loss when it comes to manipulating multiple tags in the same file. When reading a file with only one tag (either V1 or V2), reading/parsing the frames and fields works more or less as I'd expect given the API functions in id3tag.h. Where I run into trouble is when I read an MP3 file with both V1 and V2 tags. The first thing I do after opening the file is call id3_file_tag() to get a pointer to a tag. What I find confusing is that even though the tag's ID3_TAG_OPTION_ID3V1 option is set, walking the field list only gives me the frames from the V2 tag: for(i=0;i<tag->nframes;i++) { frame=tag->frames[i]; /* etc. */ } Poking around a bit in file.c it appears the id3_file structure maintains references to all of the individual tags in the file, but looking through the API functions available in id3tag.h, I don't see how to access any tag other than the primary (via the id3_file_tag() function). So my question is, how do I access the fields of a V1 tag when a file has both V1 and V2 tags? Are the V1 fields stored in the primary tag structure along with the V2 fields? If so, why can't I access them when I walk the list of the other (V2) frames? If they're stored in a tag structure other than the primary, how do I access it? Thanks, - Cedric

2 2

Re: [mad-user] MAD and Jack.....
by jens-lippmann＠t-online.de 07 Sep '04

07 Sep '04

Hi! I'm new to the list and I'm not familiar with libmad, but have a look at this function from the madlld sample. You can easily change it to produce 32-bit floating point: float f; if (Fixed & (1<<31)) { // negative Fixed = Fixed & ((1<<31)-1); // strip sign f = -(float)Fixed / (1<<MAD_F_FRACBITS); } else { f = (float)Fixed / (1<<MAD_F_FRACBITS); } return f; hope this helps Jens /**************************************************************************** * Converts a sample from mad's fixed point number format to an unsigned * * short (16 bits). * ****************************************************************************/ static unsigned short MadFixedToUshort(mad_fixed_t Fixed) { /* A fixed point number is formed of the following bit pattern: * * SWWWFFFFFFFFFFFFFFFFFFFFFFFFFFFF * MSB LSB * S ==> Sign (0 is positive, 1 is negative) * W ==> Whole part bits * F ==> Fractional part bits * * This pattern contains MAD_F_FRACBITS fractional bits, one * should alway use this macro when working on the bits of a fixed * point number. It is not guaranteed to be constant over the * different platforms supported by libmad. * * The unsigned short value is formed by the least significant * whole part bit, followed by the 15 most significant fractional * part bits. Warning: this is a quick and dirty way to compute * the 16-bit number, madplay includes much better algorithms. */ Fixed=Fixed>>(MAD_F_FRACBITS-15); return((unsigned short)Fixed); }

1 0

Why too fast song tact by LibMad on RH9 Linux?
by Maxipoint Rep Office 17 Aug '04

17 Aug '04

Do you have a soundcard that will only play at 48KHz? RE: no, because same mp3s on windows play OK Or perhaps it's a MP3 recorded at 22KHz and you've opened the sound device at 44KHz. RE: no you can see in code that I am setRate (44100); all songs are recorded on the 192kbps 44.1kHz. Maybe I must define kbps too? -GP #include <sys/soundcard.h> #include <sys/types.h> #include <sys/stat.h> #include <sys/ioctl.h> #include <fcntl.h> #include <errno.h> #include <unistd.h> #include "audio.h" #include <qstring.h> CAudioSink::CAudioSink () { // initialize device if ((sound_dev = open ("/dev/sound/dsp", O_WRONLY)) == -1) sound_dev = open ("/dev/dsp", O_WRONLY); if (sound_dev == -1) throw tr ("Failed to initialize OSS device"); if (ioctl (sound_dev, SNDCTL_DSP_SYNC, 0) == -1) throw tr ("Failed to sync with DSP"); // TODO: this must be moved to another method to handle mono data channels = 2; ioctl (sound_dev, SNDCTL_DSP_CHANNELS, &channels); ioctl (sound_dev, SNDCTL_DSP_SYNC, 0); int fmt = AFMT_S16_LE; // set 16-bit data in little-endian format ioctl (sound_dev, SNDCTL_DSP_SAMPLESIZE, &fmt); ioctl (sound_dev, SNDCTL_DSP_SYNC, 0); // set default playback rate setRate (44100); } CAudioSink::~CAudioSink () { close (sound_dev); } // set samplerate of data void CAudioSink::setRate (int _rate) { rate = _rate; ioctl (sound_dev, SNDCTL_DSP_SPEED, &rate); } // reset audio device void CAudioSink::flush () { ioctl (sound_dev, SNDCTL_DSP_RESET, 0); } // play data buffer void CAudioSink::play (unsigned char* ptr, int len) { while (len) { int wrote; wrote = write (sound_dev, ptr, len); // this check handles that fact that write syscall can be // interrupted at any time if (wrote == -1) { if (errno == EINTR) continue; else break; } ptr += wrote; len -= wrote; } }

1 0

← Newer
1
...
4
5
6
7
8
9
10
...
31
Older →

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

mad-user