Athas

47 ms·

451.
by Athas 9y ago
Lex and yacc are still frequently used (alongside similar tools like ANTLR). However, most production compilers use hand-written parsers because it permits better error messages and support for partial parses. The latter is useful for e.g
452.
by Athas 9y ago
While there are some good bits in the dragon book, some of its advice on the engineering of a compiler is pretty dated. For example, most of the parsing chapter(s) seem to have the notion that there is a single (global?) symbol table to wh
453.
by Athas 9y ago
It was this: https://web.archive.org/web/20170820092504im_/https://common... Not really offensive unless you identify as Smalltalk (called a "toy"), or C++ (called a "cancerous abstraction
454.
by Athas 9y ago
Blast from the past! I used to work on McCLIM and Climacs as an impressionable youth almost ten years ago. I think the battle for coming up with new ways of defining GUIs is lost (and the web is eating everything anyway), but CLIM was def
455.
by Athas 9y ago
APL (and J by extension) are more tricky to parallelise than you might expect. The frequent reliance on boxing leads to irregular pointer structures, and the absence of compile-time type information makes it hard to generate code at all.
456.
by Athas 9y ago
Being desensitised to humans is not necessarily a problem. The problems occur if humans are seen as threats (which they probably will if they are unused to humans), or more importantly, if humans are seen as a food source. That's whe
457.
VBScript Injection via Gnome Thumbnailer(news.dieweltistgarnichtso.net)
1 points by Athas 9y ago|0 comments
458.
by Athas 9y ago
> If you have shared memory between GPU and CPU the copy latency is reduced by orders of magnitude - you're only waiting on the values to be populated in GPU's caches. Shared memory by itself will not have a significant perform
459.
by Athas 9y ago
This can be vectorised for CPU execution (which is nice!), but it will not solve the problem for GPUs. Moving just that scalar-vector multiplication to a GPU would be detrimental to performance, because the run-time would be dominated by t
460.
by Athas 9y ago
It is really difficult to extract parallelism from sequentially written programs. It is even worse when it comes to restricted fine-grained parallelism, as on a GPU. I'm not aware of any really robust auto-parallelising compiler, and
461.
by Athas 9y ago
Yes, this is what genuinely puzzles me the most. I get that mainframes have their place for highly stable IO-heavy processing, but I've never heard anyone claim that they are particularly effective for compute-bound workloads. Especi
462.
by Athas 9y ago
Depends on how much money you make with it. (No, seriously.)
463.
by Athas 9y ago
Can recommend this link; it's a one-click victory in buzzword bingo: > As businesses adapt to capitalize on digital, trust will be the currency that drives this new economy. > With z14 you can apply machine learning to your most
464.
by Athas 9y ago
I don't know the details of Pakistani law, but it seems like it could be a clever judicial technique for fighting corruption or tax evasion in cases where it cannot be directly proven. If you spend more money than you lawfully earn, t
465.
by Athas 9y ago
Why would anyone ever want a pop-under for any reason? I would be satisfied with a solution that turned any pop-under into a pop-up.
466.
by Athas 9y ago
Yes, old fixed-format languages like FORTRAN and RPG (the older versions). Everything had to be written in specific columns, due to their punched card heritage.
467.
by Athas 9y ago
Intel Labs did do some work on functional language implementation, including a Haskell compiler (backend): https://github.com/IntelLabs/flrc I think the focus was on vectorisation rather than task parallelism, though.
468.
by Athas 9y ago
> The time it takes to allocate a chunk is linear in its size which is quite slow. I am not sure why that is, but maybe it is faster using OpenCL instead of Cuda? What were your timings for allocating and disposing raw memory plotted aga
469.
by Athas 9y ago
Thanks for your response! Do you think some of the pain would be alleviated if Futhark had a simple FFI that would allow calling hand-written primitives when they exist? (With the property that these would be black boxes and not fused wit
470.
by Athas 9y ago
Insightful and reasonable, thanks! In particular, I agree that the error messages could probably do with a few hints of what to do next. I find Elm's work on user-friendly error messages rather inspiring. Editor integration would def
471.
by Athas 9y ago
Yes. Since Futhark does not perform full flattening, replication is limited to how much parallelism is exploited , not how much parallelism is available . (That said, memory explosion is still a common symptom when the compiler misjudges
472.
by Athas 9y ago
What would be a good starting point for adding more friendliness?
473.
Streaming Combinators and Extracting Flat Parallelism(futhark-lang.org)
143 points by Athas 9y ago|18 comments
474.
by Athas 9y ago
Fortran isn't really used much for down-to-the-metal optimisation. It's not at all built for the low-level tricks common in C, but rather expects you to program at a higher level (notably with first-class arrays), after which the
475.
by Athas 9y ago
Does this architecture improve on 64-bit integer performance? Have any of the GPU manufacturers said anything about that? At some point it becomes a necessity for address calculations on large arrays.
476.
by Athas 10y ago
Out of curiosity, what are you doing with Accelerate? I've been looking for examples of people using it in anger.
477.
by Athas 10y ago
I have rarely observed that much difference between CUDA and OpenCL. I have, however, noticed that CUDA (at least by default) is more aggressive about picking faster but less accurate instructions for transcendental functions like inverse
478.
by Athas 10y ago
Not really, it's just another example of uneven performance improvements. Peak CPU performance has increased massively, but achieving it requires the use of vector instructions (and multiple cores). It is true that the speed of per
479.
by Athas 10y ago
> Computers are fast, but not if all that speed is wasted. While computers are fast, the performance improvements have been spread very unevenly. CPUs have improved the most, with memory lagging significantly behind, and persistent st
480.
by Athas 10y ago
Somewhere around OpenCL 1.1 and 1.2. It works fine on NVIDIA GPUs - they are our main test platform.

More ›