[Info-vax] RealWorldTech on Poulson

[Johnny Billquist]

On 2011-07-01 23.11, Neil Rieck wrote:
> http://www.realworldtech.com/page.cfm?ArticleID=RWT051811113343
>
> Quote: General Purpose VLIW is Dead
> Quote: "Poulson is a radical departure from the initial Itanium
> philosophy, and takes into account years of experience, and technology
> and market changes. Poulson abandons the idea of simple hardware
> controlled by the compiler and is the first dynamically scheduled
> Itanium design, with modest out-of-order execution."

Yeah. Funny how that "years of experience, and technology and market 
changes" essentially is "what other companies did 20 years ago". So 
Intel really took a long time to become enlightened.

What I wonder is how they maintain backwards compatibility if they skip 
the VLIW and dynamically schedule instructions out of order. But I must 
admit that it's so long since I looked at this now, that I don't 
remember all the details, and maybe it works fine to just suck in large 
words and reschedule as the CPU sees fit, even though the compiler have 
already scheduled the instructions when creating the VLIW.

The Alpha was explicitly designed to allow out of order execution. I, 
for some reason, seem to remember that Itanium wasn't designed for that.

> Translation: Itanium will become more Alpha-like
>
> Quote: "Itanium's performance lead over SPARC based rivals from Oracle
> and Fujitsu will only grow, as Intel leapfrogs to 32nm."
>
> Comment: that is probably why Oracle said they would no longer develop
> for new Itaniums.

I've yet to see that the width of the traces is a true measurement of 
performance.

	Johnny