[Info-vax] Poulson info from Dave Cantor

[Michael Kraemer]

JF Mezei schrieb:

> 
> They buy chips, but also send a huge wad of money (along with engineers)
> to Intel to help IA64. This was in the billions of dollars disguised as
> some development fund or what not.

Iirc the engineers were transferred once and for all times,
and those billions were investments to develop Itanic servers and
software, distributed over several years.
With BCS' revenues around three or four billions annually only,
there's not much room for extra funding.

> 
> In the end, HP ends up paying nearly 100% of the chip's development
> costs only to get a somewhat palatable architecture that doesn't give HP
> any edge over competitors.
> 
> The fact that Intel develops IA64 at a leasurely pace is an indication
> that it would lose money if it developped it at a normal pace as it does
> its other architectures.

The "other architectures" (i.e. x86) are intel's main business
and are under fierce competition from AMD.
The Itanic is no longer in competition because its sole customers,
VMS and HP-UX, can be blackmailed to swallow everything intel
throws at them (well, HP-UX customers could at least go AIX or Linux),
so intel has no reason to hurry up.

> 
> EPIC may be good to compute pi to infinity since the order of execution
> is pretty much fixed so the compiler can accurately predict what can and
> cannot be made parralel, 

provided the algorithm is parallel. If it's recursive,
Itanics fail here as well.

> but for every day use where logic depends on
> data/input, EPIC is not so great.

I still wonder who was the idiot who talked HP
into the notion that programs are infinitely parallelizable
and it would only need an ultra-smart compiler to do that.
And I wonder even more who was the idiot at HP
who bought that idea. Belluzzo?