gnufx

22 ms·

271.
by gnufx 4y ago
I took the liberty of adding an explanation to the title. The significance is: > Mes [...] can be bootstrapped [...] starting from the 357-byte hex0 > binary of the bootstrap-seeds so: > we now have a > Reduced Binary Seed boot
272.
GNU Mes 0.24 released [major bootstrapping achievement](lists.gnu.org)
9 points by gnufx 4y ago|3 comments
273.
by gnufx 4y ago
Algol 68 is interesting if only for some of the things written in it, particularly in the UK that I know of, e.g. https://en.wikipedia.org/wiki/ALGOL_68#Examples_of_use
274.
by gnufx 4y ago
Perhaps related: I understood that the Liverpudlian Fortran 90 compiler done by NA Software was based on an Algol 68 compiler (I don't know which one). I thought that was odd at the time, but it probably wasn't, and presumably s
275.
GCC 12.1 Released(gcc.gnu.org)
14 points by gnufx 4y ago|0 comments
276.
by gnufx 4y ago
Fedora, a volunteer project, currently has packages for four different current and future releases.
277.
by gnufx 4y ago
You can potentially get sponsorship (but the chips4makers web site seems to be down) https://nlnet.nl/project/Chips4Makers/ Long, long ago, UK academics could get fabrication done at the Rutherford lab, as I saw o
278.
by gnufx 4y ago
It was touted (or at least referred to?) as a "supercomputing" device, which it definitely wasn't, though I don't remember the numbers.
279.
by gnufx 4y ago
I'm not sure where it's documented, but I remember Libre-SOC have a scheme to keep everything they do free by having an interface to someone else who deals with the proprietary dirty work. Obviously the problem is still there at
280.
by gnufx 4y ago
I mentioned Libre-SOC under another item; their tools are listed at https://libre-soc.org/HDL_workflow/
281.
by gnufx 4y ago
Perhaps also see the (OpenPOWER-based) Libre-SOC effort https://libre-soc.org/
282.
by gnufx 5y ago
I don't know what's the basis for denying it, but the parent is correct about the intention when it was developed. I don't understand why it should be thought to imply Lisp knowledge. It is useful to lay out the customizati
283.
by gnufx 5y ago
I was responding to a blanket statement alluding to AVX512 (which is rather a can of worms). I don't understand what this is on about, especially with reference to the article. Of course you account for matrix dimensions at run time,
284.
by gnufx 5y ago
What does that have to do with what I wrote? At least compare maths function apples with apples. Most of the point of optimized L3 BLAS is optimizing for memory so it's compute-bound asymptotically but, for instance, header-only libxs
285.
by gnufx 5y ago
JIT-compiled code is way faster than linear algebra kernels which are completely vectorized by GCC et al? How does that work?
286.
by gnufx 5y ago
Yes, as a long-term Lisper in research computing who knows a bit about linear algebra implementation and compilation. I don't see the comparisons I think are relevant to claims of how much more performant it's supposed to be. I
287.
by gnufx 5y ago
The article asks "Which Micro-optimizations matter for BLAS3?", implying small dimensions, but doesn't actually tell me. The problem is well-studied, depending on what you consider "small". The most important thin
288.
by gnufx 5y ago
Yes, but what you'd want to do is look for coincidences between a detector for a cosmic ray shower around (above?) the electronics you're monitoring with whatever it is these days that instruments ECC events. The time resolution
289.
by gnufx 5y ago
> On the other hand, Eigen didn't claim to be faster than any BLAS during my interaction with informational side of their website. It claims to be faster than any free BLAS in the FAQ (but talks about ATLAS and GotoBLAS there). Giv
290.
by gnufx 5y ago
For anyone still reading this: I'm now confused by Eigen reportedly being basically on a par with optimized L3 BLAS, which I assume doesn't just mean punting to optimized BLAS, which it seems it can do. I can't see any indi
291.
by gnufx 5y ago
[0] still looks bad for complex (CGEMM, I assume). It doesn't have comparisons there, but Eigen's claim to be faster than any free BLAS is surely wrong. (I wouldn't doubt its similar if it has the right loop structure and p
292.
by gnufx 5y ago
Correction: That date is when the Netlib email service for the existing collection started.
293.
by gnufx 5y ago
https://github.com/flame/blis/blob/master/docs/Performance.m... but ignore the SKX results with the old OpenBLAS there.
294.
by gnufx 5y ago
It may have improved recently -- I haven't measured -- but serial Eigen seems mostly a little less performant at plateau than optimized BLAS GEMM for reals, and about half as good for complex in results I've seen for v3.3. For mu
295.
by gnufx 5y ago
> Finally, regarding alternatives like GSL, ccmath, or whatever is available on netlib, those projects post-date NR. No, Netlib started in April 1985[1], though as a physicist or engineer you likely had access to NAG or something similar
296.
by gnufx 5y ago
To support the sibling: If we're talking history, the assumption in the early 80s was that software was shared, at least in the technical computing circles I lived in and to the extent it was portable enough. We were suspicious of pe
297.
by gnufx 5y ago
> steelpillow: "We should all trust [vulture hacks on 1 April] why, exactly?" El Reg tends to be a bit slow on the ball, but I wonder if it was cunningly held for a day or two (and if they had to spend $1500). https:/&#x2
298.
by gnufx 5y ago
OK, hands up all those who were about to post something similar.
299.
by gnufx 5y ago
I'd guess that previous machine-specific BLAS used assembly. The fundamental feature of the GotoBLAS approach is the blocking structure; that can provide >80% of the performance of hand-tuned assembly/intrinsics-based kernels
300.
by gnufx 5y ago
You probably only know LINPACK from benchmarks. These days you use LAPACK.

More ›