Article: AMD's Mobile Strategy
By: mpx (mpx.delete@this.nomail.pl), December 16, 2011 11:06 am
Room: Moderated Discussions
Linus Torvalds (torvalds@linux-foundation.org) on 12/15/11 wrote:
---------------------------
>For all we know, it will have the same problems with
>huge I$ footprints etc.
Why is that supposed to be a problem when ARM64 arrives? Clocks move up slowly, process density still goes forward fast, therefore with time for the same latency in cycles you get larger array.
Additionally with ARM you don't need decoded instruction caches, predecode bits etc. It's not only saving silicon but most importantly simplyfing things. Companies have limited budgets, it's better that they can then put their best people to things that matter, rather than "deciphering" strange instruction decodings.
>Sure, it may be easy to decode, but so was alpha.
EV8 would do 8 instructions per clock, thanks to easy decoding.
> Nobody
>cares. That's not where the problems spots tend to be.
Bulldozer has the greatest x86 decoder ever built. So what?
---------------------------
>For all we know, it will have the same problems with
>huge I$ footprints etc.
Why is that supposed to be a problem when ARM64 arrives? Clocks move up slowly, process density still goes forward fast, therefore with time for the same latency in cycles you get larger array.
Additionally with ARM you don't need decoded instruction caches, predecode bits etc. It's not only saving silicon but most importantly simplyfing things. Companies have limited budgets, it's better that they can then put their best people to things that matter, rather than "deciphering" strange instruction decodings.
>Sure, it may be easy to decode, but so was alpha.
EV8 would do 8 instructions per clock, thanks to easy decoding.
> Nobody
>cares. That's not where the problems spots tend to be.
Bulldozer has the greatest x86 decoder ever built. So what?
