mad-dev November 2003

mad-dev@lists.mars.org

8 participants
8 discussions

random seeking with MAD
by Jedediah Smith 01 Dec '03

01 Dec '03

How would I implement fast, sample-resolution random seeking with MAD? In other words, I need to be able to reposition the stream to a specific sample and get resynced in constant time and without too much overhead. The MP3 also needs to be streamed from disk rather than loaded entirely in RAM. I imagine the main problem would be reconstructing the bit pool at the seek point. From my understanding of the MP3 format, the furthest back bit pool data can be stored is 9 frames. So here is my idea: 1. On initial load, create an index of all the frame offsets in the mp3 and cache it in RAM 2. When seeking to sample N, calculate the containing frame number F as F=N/1152 3. Use the cached frame index to find the file offset of frame F-9 4. Seek to that offset in the file and begin decoding 5. Decode 9 frames silently, ignoring any MAD_ERROR_BADDATAPTR 6. Begin playing at sample N which is within the current frame Will this work? Is there a faster way that avoids decoding the audio data for the extra 9 frames?

2 5

How to unroll the imdct??
by forks.tw 30 Nov '03

30 Nov '03

The IMDCT function in MAD seems not easy to understand. Rob Leslie says that it is custom unrolled and factorized implementation. But I still don't know how ... Would you please tell me about the IMDCT algorithm in MAD or any reference documents? Thank you very much!! -- forks forks.tw(a)yahoo.com.tw --------------------------------- Do you Yahoo!? Free Pop-Up Blocker - Get it now

3 2

error recovery?
by Gerd Knorr 28 Nov '03

28 Nov '03

Hi, Any hints how to recover from decoding errors, i.e. how to continue after mad_frame_decode() returned MAD_ERROR_LOSTSYNC? Just call mad_stream_buffer() with the next data block and then mad_frame_decode() again seems not to work. In the mad.h header file is a function called "mad_stream_sync", should I call that one? Before calling mad_stream_buffer() or thereafter? Gerd -- You have a new virus in /var/mail/kraxel

2 1

RE: [mad-dev] A question for mmap()
by hank.jan＠billionton.com.tw 27 Nov '03

27 Nov '03

Hi, The explanation is very clear. Thanks a lot. Regards, Hank -----Original Message----- From: Jesse McGrew [mailto:jmcgrew@hansprestige.com] Sent: Thursday, November 27, 2003 3:12 PM To: mad-dev(a)lists.mars.org Subject: Re: [mad-dev] A question for mmap() On Thu, 27 Nov 2003 14:57:11 +0800, <hank.jan(a)billionton.com.tw> wrote: > I have a question about mmap(), hope someone can clarify for me. > As I know, mmap() maps the whole file into memory, so OS can access the > source file directly from memory. > However there should be a restriction to memory size on the embedded > system. > > In my test, I compile madplay with mmap supported, and then play a mp3 > file > of 43MB, and it plays well. > But my question is I only get 32MB SDRAM on my development board. > It seems that the memory size is not large enough to store all the data. > How does mmap() work? mmap() maps pages of memory to correspond to the file, but it doesn't actually load the file into memory. The file is loaded on demand. When you access one of the mapped pages, the OS loads the correct part of the file into memory. The effect is similar to virtual memory, where pages are swapped to and from the disk when they're needed. Thus, if you're using a 32 bit CPU, you may be able to map up to 4 GB with mmap(), even if you only have 32 MB of RAM. Adding more RAM will allow you to keep more of the file in memory at once, so random access will be faster, but it doesn't affect the size of the file you can map.

1 0

A question for mmap()
by hank.jan＠billionton.com.tw 27 Nov '03

27 Nov '03

Hi, I have a question about mmap(), hope someone can clarify for me. As I know, mmap() maps the whole file into memory, so OS can access the source file directly from memory. However there should be a restriction to memory size on the embedded system. In my test, I compile madplay with mmap supported, and then play a mp3 file of 43MB, and it plays well. But my question is I only get 32MB SDRAM on my development board. It seems that the memory size is not large enough to store all the data. How does mmap() work? Besides, I want to implement the fast forward/rewind function. But it is a little slow to rewind, because it needs to search the location from the first frame, if mmap() isn't enabled. (Maybe there are other better methods, but I don't know...) If mmap() works, then it should be faster than without mmap(), shouldn't it? Sorry for any unclear expression. Thanks for help. Regards, Hank

2 1

Re: [mad-user] Build problem with cygwin/mingw
by Jedediah Smith 17 Nov '03

17 Nov '03

This doesn't work.. error looks like this: /usr/bin/install -c .libs/libmad.lai c:\source\vc\mixer\libmad/lib/libmad.la /usr/bin/install: cannot create regular file `c:sourcevcmixerlibmad/lib/libmad.la': No such file or directory Those windows paths just aren't making it through everything without getting mangled. How exactly would I install manually? Just copy libmad.a? On Fri, 14 Nov 2003 14:22:02 -0800, Rob Leslie <rob(a)mars.org> wrote: > [Note Reply-To.] > > On Nov 14, 2003, at 12:52 PM, Jedediah Smith wrote: >> Following steps were followed: >> >> ./configure --host=mingw32 --prefix==/cygdrive/c/source/vc/mixer/libmad >> >> Configure step goes fine. The prefix is the dir I want to install to >> e.g. c:\source\vc\mixer\libmad which already exists. > [...] >> ranlib /cygdrive/c/source/vc/mixer/libmad/lib/libmad.a >> c:\mingw\bin\ranlib.exe: >> /cygdrive/c/source/vc/mixer/libmad/lib/libmad.a: No such file or >> directory > [...] >> It looks like ranlib is being passed the cygwin version of the path but >> it doesn't recognize it as it's part of mingw and only allows normal >> Windows paths. What shall I do about this? > > I think both Cygwin and Mingw understand Windows paths, so try: > > ./configure --host=mingw32 --prefix='C:\\source\\vc\\mixer\\libmad' > > If not you can just install the library by hand; libmad doesn't hardcode > any paths from the configured prefix, so it's safe to omit. >

2 1

how must I align buffers passed to mad_stream_buffer?
by Jedediah Smith 17 Nov '03

17 Nov '03

My decoder uses an arbitrary buffer length and passes data blocks of this size, read from the MP3 file, to mad_stream_buffer. However, MAD throws a "lost synchronization" error immediately after returning from the input callback which feeds the data to mad_stream_buffer. The resulting decoded audio has artifacts at the equivelant locations. Clearly mad expects the MP3 data to be aligned, perhaps on frame boundaries, but I can't find any mention of this in the source comments.

2 1

III_decode function question
by Grigory Abuladze 07 Nov '03

07 Nov '03

Hi! I am working with mad-0.15.0 Could someone explain to me this? why in III_decode() - layer3.c the IMDCT processing is split for subBand 0-1 & rest - up to 31? It is look like subBand 0 & 1 is not tested to "nonzero" Is it reason for this split? Thank for Your answers. -- Best regards, Grigory mailto:Ryhor@tut.by

1 0

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-dev November 2003