By: Aaron Spink (aaronspink.delete@this.notearthlink.net), November 21, 2014 3:30 am
Room: Moderated Discussions
Patrick Chase (patrickjchase.delete@this.gmail.com) on November 20, 2014 10:39 pm wrote:
> Responding to your technical point this time: I think that the register stack engine was a fairly straightforward
> outgrowth of Rao's work at Cydrome. Once you have rotating register files for cyclic scheduling (which Cydrome
> did) the idea of extending that to optimize procedure calls isn't much of a stretch.
>
> Itanium was/is very much a VLIW in the "Rao idiom" as opposed to the "Fisher idiom".
>
>
well, personal opinion here is that once you've gone down the register rotation path, you are so far off the reservation that more junk really doesn't matter. I mean, has there ever been a decent CPU architecture with RRF?
> Responding to your technical point this time: I think that the register stack engine was a fairly straightforward
> outgrowth of Rao's work at Cydrome. Once you have rotating register files for cyclic scheduling (which Cydrome
> did) the idea of extending that to optimize procedure calls isn't much of a stretch.
>
> Itanium was/is very much a VLIW in the "Rao idiom" as opposed to the "Fisher idiom".
>
>
well, personal opinion here is that once you've gone down the register rotation path, you are so far off the reservation that more junk really doesn't matter. I mean, has there ever been a decent CPU architecture with RRF?
