gnufx

65 ms·

811.
by gnufx 6y ago
The most serious scientific computation gets done on distributed memory systems, not shared. That likely just means bindings to MPI and other appropriate libraries. That will often do pretty well on shared memory too.
812.
by gnufx 6y ago
I wasn't implying confidence, as you might have thought the original Aurora had to be delivered. Everything seems to be labelled "oneAPI" now, like the compiler and MPI.
813.
by gnufx 6y ago
Perhaps I should have said "for what it's worth". I'm not clever enough for C++, but it does seem to be what Livermore, for instance, are committed to. The oneapi propaganda does talk about other bindings, which I haven
814.
by gnufx 6y ago
How was KNL too slow for HPC, like Cori? It does need to be the right sort of HPC that can use the SIMD and/or benefit from the memory bandwidth. Just the large core count (for the time and Intel) helps with MPI communication, keepin
815.
by gnufx 6y ago
They're pushing "One API" for programming. They rather have to deliver this time on the Aurora supercomputer.
816.
by gnufx 6y ago
> >"-Ofast" same as "-O3 -ffast-math" That depends on the language and the GCC version. > -ffast-math makes the compiler violate IEEE, it would make a very bad default. Yes, but unfortunately it's the Intel
817.
by gnufx 6y ago
Dynamic linking has considerable advantages in the facilities it provides, which seem to outweigh any efficiency concerns, e.g. in HPC.
818.
by gnufx 6y ago
Turning on LTO in Fedora hasn't been entirely painless despite SUSE's pioneering.
819.
by gnufx 6y ago
ViolaWWW features: stylesheets, scripting, document embedding, at least.
820.
by gnufx 6y ago
Yes, that does plausible, which would be unfortunate.
821.
by gnufx 6y ago
There were a number of fairly sophisticated hypertext systems around then (like the basis of viola). I was an early WWW user when it was mainly a question of using the line-mode browser for the CERN documentation, and it seemed incredibly
822.
by gnufx 6y ago
> hotspot hotjava? I don't know about memory usage, but I was probably running such things on a 48MB Pentium 1 then.
823.
by gnufx 6y ago
Cello didn't have Viola's salient features, did it?
824.
by gnufx 6y ago
If only... There were a number of browsers with "applets" in different languages later in the '90s, at least: Hotjava (however it was CapitalIzed) with Java, Grail with Python, MMM with Caml (light?), and something for Tcl (
825.
by gnufx 6y ago
It's not about the ISA (I assume). The point is that these systems have essentially all free software firmware, as I understand it. You have remote management, but it's something you can presumably fix if you need to. Apart fro
826.
by gnufx 6y ago
But if you want a small GNU/Linux POWER system, you don't talk to IBM, surely. Ours came from OCF in the UK; they might even operate in Denmark. You won't see a price on the web site, but realistically you don't expect
827.
by gnufx 6y ago
Also you can build for ppc64le if you are, or become, a developer for a GNU/Linux distribution. Unfortunately getting ppc64le back online for Fedora copr has no ETA, but you can develop and build for Fedora/EPEL. I can't re
828.
by gnufx 6y ago
> - Floating point rate comparable to GPUs Where do you get the FP performance, exactly, and for what value of "FP"? It's unclear to me in the slides from El Reg, which appear to be about the MMA specifically, and it&#x27
829.
by gnufx 6y ago
I noticed recently that OpenBLAS has gained some code using the POWER 10 matrix-multiplication units: https://github.com/xianyi/OpenBLAS/tree/develop/kernel/power
830.
by gnufx 6y ago
POWER9 has large cache and can take large memory (at least for when it appeared), but was particularly notable for bandwidth/latency. I think they were the first with PCIe 4, for instance, when PCIe 3 was a bottleneck for HPC intercon
831.
by gnufx 6y ago
If you run an HPC cluster, I'd have thought you have to talk to vendors as a matter of course, but I assume there's no point yet. At least one UK vendor will sell you AC922s (POWER9), I guess as one-off, if you wanted.
832.
by gnufx 6y ago
Something like #pragma simd may actually be worse than letting GCC do its thing. It may not use FMA, just AVX, for instance. You may also want (a subset of) -ffast-math for NEON, reductions, etc., if your code is OK with it.
833.
by gnufx 6y ago
Isn't the first thing to do looking at what the compiler can tell you (-fopt-info for GCC)? That can be helpful without deep knowledge of the target, or knowing its assembler. There are also dumps from stages before generating assemb
834.
by gnufx 6y ago
Not only that, but your micro-architecture-specific hand-tuning probably isn't good for tomorrow's micro-architecture, and definitely not for a new architecture. You still need craft kernels in cases like BLAS, but you really wan
835.
by gnufx 6y ago
Allen's original paper includes loop unrolling as a useful optimization even in the 70s, and there can be good reason for inlining.
836.
by gnufx 6y ago
I don't know what to make of results without any idea how to reproduce them. For what it's worth, BLIS' generic C DGEMM code gave around 60% of the hand-coded kernel on Haswell with GCC (8, from memory) on large matrices with
837.
by gnufx 6y ago
How do people who are thrown by juxtaposition as function application and point-free style cope with Unix shell pipelines?
838.
by gnufx 6y ago
I don't know the applications, but it may well not be suitable (providing the command path, for instance). I feel the pain of this from working on an HPC resource manager and more generally wanting to track executions on the system fo
839.
by gnufx 6y ago
I have to fix up the layout essentially manually every time I have to edit C in someone else's idea of correct layout when they haven't included C Mode rules for it. (That's assuming they want a consistent style for changes,
840.
by gnufx 6y ago
For what it's worth, Python doesn't actually use ISWIM's offside rule, as Haskell does; colons are involved.

More ›