[Beowulf] [rjdyer at vcu.edu: Re: Intel?]

[Mark Hahn]

> Being one of those 'apparent idiots' that followed Apples advice and 
> spent literally hundreds of hours in altivec, I'm a bit peeved at the 
> whole thing.  I must essentially redo _everything_.  And if so, what is 

I'm not replying to this on scitech because it's not 
really helpful for their discussion.

what I'm wondering is: has anyone ever honestly compared 
altivec with mmx/3dnow/sse*?  for instance, PPC and modern x86
both have similar 128b SIMD.  altivec is probably cleaner 
and iirc has more visible registers.  but are the differences 
critical?

for instance, altivec has 32x128 and x86-64 has "only" 16x128.
but how relevant is this?  studies of conventional reg-set usage
show diminishing returns after about 16 regs.

GCC now seems to have reasonably nice support for vector operations,
on top of the intrinsics, and even some auto-vectorization.  any 
preliminary reports of experience with gcc 4?

thanks, mark hahn.