Article: Parallelism at HotPar 2010
By: Rohit (.delete@this..), August 6, 2010 7:24 am
Room: Moderated Discussions
hobold (hobold@vectorizer.org) on 8/6/10 wrote:
---------------------------
>Semantically and concerning their design intent (this is a guess, I cannot read
>minds :-), LRB and Cray's machines are close relatives. But the microarchitectures
>are vastly different. Cray used "deep pipelining" to quickly process vector elements
>one by one. The straight forward parallelism of the vector registers allowed him
>to pipeline deeper and clock higher than what was possible with scalar pipelines.
>LRB and other modern SIMD machines have arrays of parallel ALUs and do computation
>truly in parallel (and they are pipelined on top of that, of course).
You seem to be suggesting that the physical width of ALU's was less than the logical width exposed to programmer on Cray's.
Is that correct?
---------------------------
>Semantically and concerning their design intent (this is a guess, I cannot read
>minds :-), LRB and Cray's machines are close relatives. But the microarchitectures
>are vastly different. Cray used "deep pipelining" to quickly process vector elements
>one by one. The straight forward parallelism of the vector registers allowed him
>to pipeline deeper and clock higher than what was possible with scalar pipelines.
>LRB and other modern SIMD machines have arrays of parallel ALUs and do computation
>truly in parallel (and they are pipelined on top of that, of course).
You seem to be suggesting that the physical width of ALU's was less than the logical width exposed to programmer on Cray's.
Is that correct?