Re: IA64

[John Leuner <jewel@pixie.co.za>]

> > As a non-Alpha user (but an aspirant one), I'm curious to know how the
> > 'community' sees IA64? Is it perceived as a threat? Would alpha users
> > be persuaded to switch over easily?
> 
> I, for one, am excited about having a 64-bit Intel chip out there for a
> lot of reasons.  For one, it'll make porting much easier since most
> software authors will know the issues involved (rather than assume that
> the whole world is 32-bit, like they do now) and write things
> properly.  Also, I think the competition may be good for Alpha since it'll
> bring more attention to the 64-bit market.  Right now, clock speed seems
> to be what people look at rather than performance and I really hope that
> changes (in my experience, a similarly-clocked Alpha can easily outperform
> Intel's current offerings in almost every area).

As you say it's good to have more 64-bit awareness, but I'm wondering how
long it will take the IA64 to reach "the masses", ie how long will it
remain a high-end architecture. It will probably remain expensive for at
least a year or two (but we'll see what AMD does too), and since it
emulates IA32 it won't be competitive with existing IA32 solutions.

Personally I'ld like to get hold of a 64-bit machine for operating systems
research, but it's really a significant expense (for a poor student) and
I'ld probably have to first earn the money, making the decision even more
painful. (You can buy a lot of other stuff, eg lots of IA32 boxes, or an
overseas holiday for the price of a 21264!). 
 
> On the downside, however, Intel's clout and manufacturing volume could
> spell trouble for Alpha.  Since we don't really know what the pricepoint
> on the Intels will be, it's hard to tell what effect their introduction
> will have on the Alpha market, but one thing's for sure: Intel's bound to
> try to undercut the Alpha prices.  It's going to be interesting.  I'm also
> curious to see how well FP performance is compared to Alphas (haven't seen
> a side-by-side comparison yet).  Alpha still may reign in the scientific
> market in the long run if the FP performance is still higher than Intel's.

Who fabs the Alpha processors?

> > How do the free software compiler-writer people feel about the
> > architecture?
> 
> I agree with the other statements made about gcc's support for IA64 (at
> least for now).  It's certainly going to be a bit before gcc optimises
> well for IA64, especially since it's just recently gotten better (albeit
> still not great) for 64-bit processors in general.  The unique pipelining
> of the IA64 architecture makes this even uglier, though, so I'm very glad
> to not have gotten involved in that port yet :-P

>From what I've read I understand that there is a lot of work to be done by
the compiler to benefit from IA64's features. It also seems very different
from what we're used to.
 
> SGI's compilers are much better, as previously stated, mostly because of
> SGI's experience with 64-bit issues (WRT to MIPS) and they've actually got
> some really good people in their compiler area.  I just doubt that their
> compiler will be more prominent (or take over) gcc's slot as the premiere
> Linux compiler any time soon.  I'm hoping they donate more to gcc to make
> it better since SGI doesn't turn much business from selling/supporting
> compilers, even on their own machines.
>
> I could go on and on about this, but I won't in the interest of brevity
> :-)

:-). Interesting how the GNU project was centered around GCC, and of
course GCC is very portable, but yet it's very difficult when you start
using radically different architectures. 
 
> At any rate, IA64 is coming and, yes, I wouldn't mind having one, but what
> it does to Alpha remains to be seen, I guess.

Well Alpha has had a decent head start, we'll see what happens.


John Leuner