Subject: Re: [linux-audio-dev] What I want, to stop using Windows
From: John Lazzaro (lazzaro_AT_CS.Berkeley.EDU)
Date: Thu Sep 28 2000 - 01:18:15 EEST
Answering two postings in one here ...
[Ellis Breen writes]
>SFront is cool - does anybody know of a direct-to-machine code SAOL
>compiler?
>Are potential optimisations lost in the translation to C by SFront/SFX?
>Could one spec up a superior intermediate API for SAOL to machine code
>compilation for which a generic engine and platform-specific API libraries
>(and thus optimisations) could be written?
There's some wins possible in that direction (for a micro-optimization
example, many table operations could go better with ASM-optimized rounding
tricks, for a macro-optimization example, SSE and other floating-vector
CPU sidecards could be well-utilized), but there are at least an equal
number of opportunities left in changing high-level optimization and
architecure issues, and so when I return to optimization I'm more likely
to go up than down in abstraction ...
[Paul Barton-Davis writes]
> yikes. please don't use SAOL. saol was an excellent reformulation of
> Csound, but as a general purpose synthesis/fx language, i consider it
> to be weak and to have a number of specific defects.
Just to add some balance here to Paul's view, I think the situation may
be somewhat comparable to the situation in "natural audio" compression.
Seen in light of the state-of-the-art of psychoacoustic compression,
MPEG 2 Audio (i.e. MP3) is not technically the best you can do at the
present time -- and if you're designing a system for, say, compressing
CD's at the head-end of a cable TV system for distributing them down
to a settop box for "CD's on demand", you'd probably go for the most
leading-edge compression system you can, and gladly pay the extra
patent royalties involved :-).
But on the other hand, there are lots of other systems where the
installed base of MP3 as a standard overwhelms efficiency gains
you'd get by moving to a technically superior format.
Let's take Paul's judgement of SAOL vs Quasimodo as a sound language
as fact for a moment -- I have my own opinions, but its irrelevant to
the argument I'm about to make. MPEG 4 as a system is going to be
deployed in a pretty major way in the next 5-10 years, and because
Structured Audio is used in several places in the MPEG Audio Standard
(both as a standalone synth and as a mixdown language for the whole
system), SA implementations _will_ happen, in lots of different contexts.
I think the right reason to implement SAOL in a system today is that
you think you can leverage the MPEG 4 deployment -- if you can't, then
making your decision on technical matters alone probably makes the most
sense. Since I'm into SAOL for the former reason, I really don't have
a technical analysis to present to help you make the "technical merits
only" decision -- all I can do is point out that there's non-technical
reasons why there may be an SA decoder in every PC and PDA and cell-phone
in 5 years, and that might be relevent to your particular project ...
--jl
-------------------------------------------------------------------------
John Lazzaro -- Research Specialist -- CS Division -- EECS -- UC Berkeley
lazzaro [at] cs [dot] berkeley [dot] edu www.cs.berkeley.edu/~lazzaro
-------------------------------------------------------------------------
This archive was generated by hypermail 2b28 : Thu Sep 28 2000 - 01:46:49 EEST