Computing Engine Research
Adrian Freed
The exploration and development of new computational engines is critical
to CNMAT's mission. We persue three goals simultaneously in this area:
- support of new paradigms,
- greater computational performance, and
- and lower development and ownership cost.
Platform
| SGI | Other UNIX | Power PC | Pentium Pro | DSP
| |
Performance | High Multiply/Add rate | | | | | |
Productivity | Tools (e.g. Matlab) | | | |
| . |
Good HLL Code Generation | | | 96 | 96
| . |
Standard DSP library | | . | . | . | |
Large
(>1Mbyte) Memory | | | | | .
|
Protected Address
Space | | | 96 | 96
| . |
Music and Interactive
Multimedia | Real-time Threads | | | 96
| 96 | .
|
Low Latency I/O (<10mS)
| |
.
| . |
. | |
Audio | Reliable Audio I/O | |
. | . | .
| |
On
board 16-bit Audio | | . | |
96 | |
On
board 4 channel Audio | | .
| . | .
| . |
On board Digital I/O | | .
| . | .
| |
Performance Projections
Year | Clocks
/Multiply/Add | RISC Clock Rate (MHz) | DSP Clock Rate (MHz) |
1995 | 2 | 150 | 80 |
1996 | 1 | 250 | 120 |
Up to the late 1970's, computer music engines were special purpose computers
built from large numbers of moderately integrated components. They were
accordingly expensive to develop, expensive to own, and not very portable.
In the 1980's the development of completely integrated digital signal processing
(DSP) chips, afforded the development of more integrated and easier to program
systems. Our contribution during this period was the Reson8, a machine that
combines 8 DSP56001 chips [Barrière et al. 1989]. Similar systems
are available commercially, e.g. Kyma, and Protools and their popularity
is now peaking .
In the 1990's our efforts moved away from integrating multiple DSP chips
to the adaptation of high performance computer workstations and personal
computers. Although the Reson8 is built from a small number of readily available
elements, the performance of those elements is no longer competitive with
RISC processors. An additional advantage of workstations and personal computers
is that a complete computer music environment can be created by simply adding
software to standard "off-the-shelf" systems manufactured and
supported by large well-funded companies. The result of our efforts in this
area, the HTM software environment [Freed 1992,1994a], forms the standard
computing substrate for our work on new algorithms for musical sound synthesis
and control.
Our latest work explores the reframing of older techniques of the 1960's
and 1970's in modern, low power VLSI technology [Freed 1994b,1995].
References
[Barrière et al. 1989] Barrière, J-B, Baisnee, P-F, Freed,
A., Baudot, M-D, 1989, "A Digital Signal
Multiprocessor and its Musical Application", Proceedings
of the 15th International Computer Music Conference, Ohio State University,
CMA, San Francisco, CA.
[Freed 1992a] Freed, A. "Codevelopment
of User Interface, Control and Digital Signal Processing with the HTM Environment,"
Proceedings of The International Conference on Signal Processing
Applications & Technology, Dallas, Texas, 1994
[Freed 1994a] Freed, A. "New Tools
for Rapid Prototyping of Musical Sound Synthesis Algorithms and Control
Strategies." , Proceedings of The International Conference
on Signal Processing Applications & Technology, Dallas, Texas, 1994.
[Freed 1994b] Freed, A. "The Rebirth
of Computer Music by Analog Signal Processing", ICMC 1994,
[Freed 1995] Freed, A. Custom Hardware for Synthesis of Hundreds of
Sinusoidal Partials, CNMAT internal, 1994, 1995.