Subject: Re: [linux-audio-dev] MidiShare Linux
From: Paul Barton-Davis (pbd_AT_Op.Net)
Date: Fri Jul 21 2000 - 01:45:33 EEST
>>>So, please lets discuss about these topics in LAD, and lets refer less to
>>>existing implementations and papers.
>>
>>This is completely ridiculous. Those papers and those implementations
>>embody a huge amount of work, and it can't be duplicated by a few
>>emails back and forth on LAD. If you haven't read those papers, you'd
>>better start there (the GRAME paper will have good references), and
>>then come back with specific questions.
>
>You make it sound like they have 100,000 lines of code for this buffer
>stuff alone. Sorry, but you can decsribe their system in 10 lines, I'm
>sure.
Thats not what I meant at all. Lock free data structures are a
large-ish and complex field of parallel programming. There are many
different approaches, many different solutions. If you want to know
about them, I consider it appropriate to go and read the original
papers.
I doubt that you could describe *any* lock free DS in 10 lines.
rather importantly, as work in this field noted right at its
inception, the implementation and approaches have subtle dependencies
on the precise properties of architecture specific instructions. There
are things that could be done on Motorola chips, for example, that
were (are?) not possible on Intel back in 1994.
Given your zeal for researching original papers on hard disk recording
systems, I am suprised at your lack of willingness to go off and read
about this. Whats the difference ?
--p
This archive was generated by hypermail 2b28 : Fri Jul 21 2000 - 02:22:26 EEST