Off topic - G5 question
hahn at physics.mcmaster.ca
Thu Jun 26 17:39:09 EDT 2003
> unit represents filled. It actually looked like a thoughtful
> architecture at least at the block device/bus level.
it's dandy. the problem is Apple's marketing nonsense.
SIMD is great, too. cache is great. but please don't claim that your
in-cache SIMD FLOPS is a general/true FLOPS rate for the machine,
since it's NOT, for anything other than Mandelbrot surfing.
incidentally, I don't see *any* off-chip cache in the G5's.
> Of course, this makes a LOT of things very difficult to properly
I don't agree at all. I want to see precisely two sets of numbers
when evaluating new hardware performance:
- fairly clean microbenchmarks like Stream, streaming
sequential file IO, zero-byte packet latency, etc.
- best-case performance on real codes - for the compute
cluster world, that means SpecFP and SpecFPrate. other fields
may want tuned-BLAS or 45-filter photoshop.
well, and price, but that's normally boringly predictable
(I say "way too expensive!" a lot ;)
> suck on it. Also one needs to use SSE2 instrumented compilers and
> instructions to get fair comparisons from alternative architectures, for
I find that many of my users spend most of their serious time inside
tuned library functions, so the compiler is less relevant. though Intel's
compilers *do* auto-vectorize onto SSE2. gcc can use SSE2, but doesn't
currently autovectorize. incidentally, gcc 3.3 *does* also support
intrinsics for Altivec vector operations; I'm not sure whether similar
intrinsics are available for SSE2.
> flags on both systems works for me, at least, as this is likely to be
> the extent of the effort I want to make tuning per architecture, ever.
I find this irresponsible (sorry rgb!). if you're going to consume
more than a few CPU hours, and those cycles are coming from a shared
resource, failing to optimize is just plain hostile.
I still see users fling their code at a machine with
g77 prog.f && ./a.out
and wonder why it's so slow. usually, I stifle the obvious answer,
and (hey, this is Canada) apologize for not providing the information
they need to make supra-moronic use of the machine...
regards, mark 'surly' hahn.
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf