d65y@vax5.cit.cornell.edu (01/19/91)
It seems to me that the advanced superscalar and pipelined architectures, are beginning to look a lot like dataflow machines internally. Am I completely off-base or has this been noted before? If someone spent as much effort building a 1 million transister dataflow chip, would they get as good or better performance? Sam Paik d65y@vax5.cit.cornell.edu "I cannot tell you anything. I am from the Ministry of Information." - response of a journalist's Iraqi escort
prener@arnor.uucp (01/22/91)
In our early thinking about what ultimately turned into the RISC System/6000, we spent some time discussing the superscalar approach with dataflow enthusiasts. We all agreed that superscalar delivered more performance than dataflow, for a given amount of hardware. Dan Prener (prener @ ibm.com)