pase@ogcvax.UUCP (Douglas M. Pase) (03/09/88)
I'm looking for a good description of tagging in tagged-token dataflow. Specifically I need to understand how the tags are manipulated in loops and recursion to keep the values separate during execution. I have read the treatment in Hwang and Briggs (Computer Architecture and Parallel Processing) and the January 1985 CACM article (The Manchester Prototype Dataflow Computer), but neither coverage is adequate. Both mention that there are three fields <name, iteration, index> and generally what their uses are, but they don't get much farther than "it's possible to do this". This is not intended as criticism; the intent of both writings was to describe dataflow machines in general terms, not discuss implementation details of the token tagging techniques. My question: do you know of a reference which discribes tagging in a lot of detail? If you do would you be willing to send me the title, source, and author? I have a few more articles to try, but it appears they were also intended for general audiences (rated G?). Any assistance would be very much appreciated. -- Doug Pase -- ...ucbvax!tektronix!ogcvax!pase or pase@cse.ogc.edu (CSNet)
segall@steppenwolf.rutgers.edu (Ed Segall) (03/16/88)
pase@ogcvax.UUCP (Douglas M. Pase) writes: >I'm looking for a good description of tagging in tagged-token dataflow. There is a discussion of the U-Interpreter in chapter 2 of MIT/LCS/TR-332 "Resource Management for the Tagged Token Dataflow Architecture" by David E. Culler. This is an MS Thesis from 1985. The coverage is somewhat more extensive than in Hwang and Briggs, and answered a few of my questions about dataflow. In particular, it shows what sort of performance problems the implementors ran into. --Ed Segall -- uucp: ...{harvard, ut-sally, sri-iu, ihnp4!packard}!topaz!caip!segall arpa: segall@CAIP.RUTGERS.EDU