Athas

28 ms·

271.
by Athas 7y ago
More or less AMD CUDA: a collection of compilers, tools, and libraries. The main difference is that NVIDIA markets CUDA strongly based on CUDA C++. You know the way you write CUDA is through nvcc and the language it accepts, unless you
272.
by Athas 7y ago
This contains an OpenCL implementation (with support for 2.0 at that!) and the kitchen sink. It's a bit unclear what AMD actually suggests you use for programming their GPUs.
273.
by Athas 7y ago
For anyone else using NixOS, https://github.com/nixos-rocm/nixos-rocm by Anthony Cowley works very well. While ROCm works well, I think it is yet another example of what AMD gets wrong software-wise. I like to think I
274.
by Athas 7y ago
This is a textbook for a university course. The intended audience for this book needs a bare minimum compiler, because that is all they will have time to write. You can't just say "use a parser generator". Who is supposed t
275.
by Athas 7y ago
I don't think it matters that a third of the pages are spent on parsing - that only covers 20% of the chapters. It's just that parsing takes somewhat more space to explain because the figures, tables, and examples take up relativ
276.
by Athas 7y ago
It probably depends a lot on the specifics of the situation. For my PhD, the funding was secure up front. This is not unusual in Europe, but I hear the USA is sometimes different. I was given a lot of freedom, and ended up spending a lot
277.
by Athas 7y ago
Can't speak for everyone, but for my PhD I was trained for doing research. What else would it be? That kind of research is (presumably) also useful in R&D positions in industry, and I certainly see most of the PhD graduates here
278.
by Athas 7y ago
I was taught from, and have since taught with, this book. I like that it's so concise, where I think e.g. the Dragon Book contains way too much irrelevant information (and also dubious and obsolete implementation advice, like global s
279.
by Athas 7y ago
Can you compile a header to an object file so that its definitions don't have to be recompiled every time the including file is recompiled?
280.
by Athas 7y ago
A PhD is not required to do research, nor does research necessarily lead to a PhD. A PhD is ultimately like any other degree: a school programme where, at the end, some instituting is staking part of its reputation on giving the world a pr
281.
by Athas 7y ago
The FUSE maintainer (and this article) implies that kernel driver signing certificates are so difficult to get that forking is not realistic, even without considering the technical issues. I don't know if that is true, but if it is, th
282.
by Athas 7y ago
Actually, my experience is that APL in the wild (as practiced by those who are just trying to do their job) is not very terse at all. It looks like an old-school procedural language. Since APL is an incredibly proprietary community, it ca
283.
by Athas 7y ago
Exactly what are the rules for the "special comparison" in APL? That sounds horrifying to me.
284.
by Athas 7y ago
"Derivation" is just Nix slang for "package". I mean that I contributed some packages to the Nixpkgs collection.
285.
by Athas 7y ago
While Homebrew is perhaps technically crude and somewhat inflexible compared to other and older package managers, I think it deserves real credit for being so easy to add packages to. I contributed Homebrew packages after a few weeks of us
286.
by Athas 7y ago
> Stop this evangelist wars, your language can be great due to some other features (ease of use, knowlidge needed to be proficient in it, forgiveness of mistakes,..), you dont need to compare it to c/c++. It just doesnt make any sen
287.
by Athas 7y ago
The corresponding AMD infrastructure and most related tools are fully open source. The driver is even upstreamed in the Linux kernel!
288.
by Athas 7y ago
Last year, I supervised a student project that involved writing a Vulkan backend for a compiler that already possessed CUDA and OpenCL backends. While I recall that everything ended up working, it didn't run all that fast. The vast m
289.
by Athas 7y ago
It has been for a long time. Only the very earliest versions of CUDA had an nvcc based on Open64.
290.
by Athas 7y ago
OpenCL is decent, and although it's deprecated, it's probably going to continue working for years. It's not as nice as CUDA for direct programming, but don't most people access CUDA through higher-level libraries anyway
291.
by Athas 7y ago
I've never used K, but isn't Dyalog mostly parallel through isolates? Last I checked, basic array operations were not automatically executed in parallel. Did this change?
292.
by Athas 7y ago
APL really is a wonderful core notation, even though the full language is rather crufty. I've long been saddened that its promises of parallelism never seemed to work out.
293.
by Athas 7y ago
>It's not for beginners. I don't consider myself a beginner to Unix or computers (I even have a PhD in the damn things), but I do consider myself a fairly inept and inexperienced systems administrator, with no great desire to s
294.
A Futhark Implementation of Ray Tracing: The Next Week(github.com)
4 points by Athas 7y ago|0 comments
295.
by Athas 7y ago
Lisp machines, definitely. With respect to Forth, I'm not sure. Has there ever been a Forth operating system that provides the facilities we would expect of a modern operating system (and using Forth as the interface)? There are lot
296.
by Athas 7y ago
None, but in C you use the same language mechanism for following an object reference (safe in the absence of manual memory management), indexing an array, and doing crazy address arithmetic. In Oberon, safe and unsafe operations are clearl
297.
by Athas 7y ago
When people ask "how would you write an operating system without C?", Oberon is what I point them to. It's a very different style than Unix, in that the Oberon language is a relatively integrated part of the operating system
298.
by Athas 7y ago
No; the code we would need to generate would be rather different. Would you expect a significant difference? When we did research on nvrtc before implementing this, we couldn't find any concrete information that nvrtc should generate
299.
by Athas 7y ago
CUDA is ultimately an API. AMD even has a converter for transforming CUDA cuda to something more portable[0]. While it would be better in a democratic sense for GPUs to be accessed using a fully free API, having an easily usable proprietar
300.
by Athas 7y ago
Are you certain that the story is as simple as "CUDA performs better"? It's common folklore, but I have seen little evidence. The only situations I know of when CUDA performs better is when CUDA-specific features are used (

More ›