gnufx

17 ms·

241.
by gnufx 4y ago
> an integrated Ethernet interface, which allowed it to interface itself with outside worlds with negligible latency Yes, the MPI ping-pong latency I measured between x2200s was the lowest I ever saw for 1GbE. However, the motherboards
242.
by gnufx 4y ago
There's also https://keys.pub/
243.
by gnufx 4y ago
So, the claim is that RHEL has broken packaging which doesn't reflect ABI changes, and somehow that means you have to reverse engineer its build mechanism from srpms to accommodate it being non-deterministic somehow? Care to give an e
244.
by gnufx 4y ago
Sorry, I don't know why I missed off the sentence. The statement was that putting something under the Berkeley licence used by Singularity made it public domain, i.e. they could strip my employer's copyright, specifically. No an
245.
by gnufx 4y ago
Unfortunately the Linux port was never incorporated and is apparently now abandoned: https://github.com/google/capsicum-linux Then, if you're serious about capabilities, as you should be more-or-less, you might wa
246.
by gnufx 4y ago
Supercomputing... Note that Summit changed from RHEL7 to RHEL8 mis-stream. (IBM insisted the system I work on had to do the same on similar hardware because the RHEL 7.9 distribution I could see didn't exist. It allowed more recent
247.
by gnufx 4y ago
I don't know how RHEL packages are built exactly, since I'm not an employee, but mock is what the Fedora infrastructure runs under koji for builds. I don't see how "a CI" can make the process "nearer to reprod
248.
by gnufx 4y ago
It would frighten me having Kurtzer doing legalities, given his bizarre legal theory on copyright specifically (no longer recorded in the issues of a deleted github repo for me to reference).
249.
by gnufx 4y ago
Well, I can do a mock -r almalinux-8-x86_64 bootstrap build on RHEL7. (I can't remember whether there are currently any problems maintaining for RHEL8 on RHEL7, but it has been impossible to maintain packages for RHEL N+1 on RHEL N in
250.
by gnufx 4y ago
I wonder what is the issue if the company "borrowed" its own stuff to use in a different context. Alma always looked like the more serious operation and, if nothing else, the company supplying kernelcare might be considered usefu
251.
by gnufx 4y ago
There was a lot of that talk about the Newport Wafer Fab being sold to China. I remember it produces in-demand automotive stuff, but not the process size(s), if that's even relevant with the different technology. Shame I can't f
252.
by gnufx 4y ago
Probably https://billwadge.com/lucid-language/ is the closest you'll get.
253.
by gnufx 4y ago
I hadn't heard of the example of rms' technical insight there, which is interesting. Not knowing enough about Ada, I wonder how the specification and implementation compare with Fortran 90 modules.
254.
by gnufx 4y ago
Why argue with the people who actually produce the collectors? "GC uncooperative programs using conservative collection" https://www.ravenbrook.com/project/mps/ I think it's a Bartlett-style "m
255.
by gnufx 4y ago
> In no way is the Boehm GC that I'm aware of is generational or incremental Boehm disagrees: https://www.hboehm.info/gc/#details That thread isn't talking about the same thing, which it says no-one volunt
256.
by gnufx 4y ago
Slow in the same way that the conservative stack scanning was bound to leak. I did actually run Emacs with it after experience elsewhere. Edit: I wonder how Boehm is somehow "most conservative", compared with, say, the Memory Po
257.
by gnufx 4y ago
That work is actually still in the repo: https://git.savannah.gnu.org/cgit/emacs.git/tree/?h=other-br... You may not believe the history, but you weren't there; I don't know how much is in mail ar
258.
by gnufx 4y ago
> That is untrue. Indeed. Emacs actually got a preliminary port to the Boehm generational, incremental, mark-sweep collector many years ago, with a complete lack of interest in pursuing it. Also, Emacs conservative stack scanning came f
259.
by gnufx 4y ago
Just doing a fork may be sufficient for something embarrassingly parallel, but interesting things are tightly coupled. Obviously MPI scales down to a single node (a distributed system anyway, these days), typically as real forked processes
260.
by gnufx 4y ago
Is Cloud [sigh] Haskell still alive? For comparison, two old distributed lexical scope systems were Cardelli's Obliq and Kelsey's(?) Kali Scheme. From what I remember, not like remote forking, though.
261.
by gnufx 4y ago
For what it's worth, at least for HPC-ish distributed computing, this sort of thing turns out not to be terribly worthwhile. We have a standard for distribution of computation, shared memory, i/o, and process starting in MPI (and
262.
by gnufx 4y ago
"more accurate"? The rationale for (typically expensive) tracing in HPC circles is that sampling can miss things you want to pick up. I think that's also covered in the vi-hps.org workshop introductions, for instance. (It
263.
by gnufx 4y ago
Nothing is generally better when it comes to serious performance engineering. The standard introduction to the performance engineering workshops under vi-hps.org stress that you need a variety of tools/techniques available.
264.
by gnufx 4y ago
Instrumentation is widely used in HPC performance engineering, but you may still do sampling, rather than tracing, with the result. Often only a specific set of function calls is instrumented (e.g. MPI, i/o), and often using LD_PRELOA
265.
by gnufx 4y ago
I responded to a post about the practicality of MM multiplication methods, though GEMM is quite fundamental.
266.
by gnufx 4y ago
The number of operations per second varies by at least an order of magnitude on the same hardware depending on the GEMM algorithm you use (reference or Goto), and you quote the performance of the implementation in terms of FLOP/s knowi
267.
by gnufx 4y ago
For something Strassen-ish you could look at https://jianyuhuang.com/papers/sc16.pdf and the GPU implementation https://apps.cs.utexas.edu/apps/sites/default/files/tech_rep...
268.
by gnufx 4y ago
Sort of, but it's Eric Raymond's fallacy, which should be given as little weight as everything else he proclaims. It's also generally wrong in science, speaking from experience in experimental physics.
269.
European Commission prefers breaking privacy to protecting kids(lightbluetouchpaper.org)
361 points by gnufx 4y ago|187 comments
270.
by gnufx 4y ago
I haven't gone back and read the original after all these years to check the specification, but it's perhaps worth noting that at least the McIlroy solution is only valid in the C locale. At least these days you can use character

More ›