home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!ukma!nntp.msstate.edu!nntp.msstate.edu!linder
- From: linder@ERC.MsState.Edu (Dan Linder)
- Subject: Re: asynchronous DSP project at Berkeley
- In-Reply-To: "Carl Bruggeman"'s message of Tue, 27 Oct 1992 11:52:01 -0500
- Message-ID: <LINDER.92Nov5175441@desperado.ERC.MsState.Edu>
- Sender: news@ra.msstate.edu
- Nntp-Posting-Host: desperado.erc.msstate.edu
- Organization: /home3/linder/.organization
- References: <1992Oct27.115210.14352@news.cs.indiana.edu>
- Distribution: na
- Date: Thu, 5 Nov 1992 23:54:40 GMT
- Lines: 38
-
- > In article <1992Oct27.115210.14352@news.cs.indiana.edu> "Carl Bruggeman"
- > <bruggema@gable.cs.indiana.edu> writes:
-
- > I vaguely recall a reference made on this news group several months ago
- > to a group at Berkeley working on a DSP using asynchronous design
- > methods. Does anyone have a reference to the work done by this group?
-
- You might try:
-
- G. M. Jacobs and R. W. Brodersen, "A Fully Asynchronous Digital Signal
- Processor Using Self-Timed Circuits," IEEE Trans. on Comp., vol. 25,
- no. 6, Dec. 1990, pp. 1526-1537.
-
- Martin also has a processor described in:
-
- A. J. Martin, S. M. Burns, T. K. Lee, D. Borkovic, and P. J. Hazewindus,
- "The Design of an Asynchronous Microprocessor," Advanced Research in
- VLSI, Proceedings of the Decennial Caltech Conference on VLSI, March
- 1989, pp. 351-373.
-
- > I am particularly interested in why they chose a DSP rather than a
- > general purpose CPU. I suspect the answer is that the complexity of the
- > global state machine controlling a typical RISC pipeline (and in
- > particular exception handling and interrupts) is overwhelming for any of
- > the current asynchronous design methodologies. Am I wrong?
-
- Brodersen is well known in the DSP area so this probably influenced their
- decision. It is interesting to note that the DSP processor uses
- microprogramming and Martin's processor only has a few instructions. I
- suspect as well that the complexity of a modern RISC control unit is beyond
- current asynchronous techniques. Are there any asynchronous methods that
- can efficiently combine the generation of a multitude of control signals
- while also making the complex logical decisions associated with big state
- machines?
-
- Dan Linder
- linder@erc.msstate.edu
-
-