stabbles

41 ms·

541.
by stabbles 6y ago
> but again why not use something which already works Sure, but there's always somebody crazy enough to try to implement it :p From the fact that ISPC can generate C++ (with --emit-c++) I would think their compiler is conceptually v
542.
by stabbles 6y ago
A subset of Python for one, and debugging stays awkward with gdb. Makes me think of this little pun: https://github.com/FluxML/Flux.jl/blob/master/src/utils.jl#L...
543.
by stabbles 6y ago
Right, I see what you're getting at. Just taking a slightly less trivial example from Intel's docs, and superficially comparing the assembly they generate to what LoopVectorization.jl can do, I guess there is some hope at least:
544.
by stabbles 6y ago
"SIMD not reliable" is an overstatement itself for a couple reasons: 1. It's very easy to inspect generated code of your kernels where you really need SIMD. For instance: julia> function my_kernel(xs) to
545.
by stabbles 6y ago
I think you're overestimating how many Python developers are willing to write and compile C. Debugging gets more difficult across two languages. Shipping binaries with a Python package correctly can be a hassle too (what about -march=n
546.
by stabbles 6y ago
I would assume scipy uses MKL, and Julia has wrappers for that too: https://github.com/JuliaSparse/MKLSparse.jl . For the GPU CUDA.jl has support for sparse matrices as well.
547.
by stabbles 6y ago
Too bad lichess.org isn't mentioned; unlike chess dot com it's free, open source and blazingly fast.
548.
by stabbles 6y ago
Another aspect is you might get better accuracy with larger input dimensions, but the number of pixels scales quadratically with width/height.
549.
by stabbles 6y ago
> There is evidence (measured using gil_load) that we were throttled by a fundamental Python limitation with multiple threads fighting over the Global Interpreter Lock (GIL). Can anyone comment on how often this is a problem and if this
550.
by stabbles 6y ago
I wouldn't be surprised if CentOS 8 was still around in 2038 running a 4.x linux kernel
551.
by stabbles 6y ago
maybe the author is one of the last people to switch from Windows to Linux as Microsoft embraces Linux
552.
by stabbles 6y ago
Are there any benchmarks that show compilation times using LLVM 11 versus older versions for representative projects?
553.
by stabbles 6y ago
If the big tech firms vanished, and instead there would only be many tech dwarfs, we still would have big data centers that needed the same capacity.
554.
by stabbles 6y ago
I want to add that where Denmark has "hygge" and Norway has "koselig" (both about cosiness, warmness), Finland is the odd one with "sisu", which means perseverance. That has a much more realist take to it than
555.
by stabbles 6y ago
Yes, this is another reason to prefer julia's dispatch over manual dispatch. In this case the cost of the dynamic type being returned is just a _single_ dynamic dispatch. Dynamic dispach means that the correct function foo will be look
556.
by stabbles 6y ago
Manual dispatch can be an antipattern hindering composability though. If package A does: function foo(x) # some lines of code y = x isa Int ? 2.0 : 3x # some other lines that do things with y end Then package
557.
by stabbles 6y ago
Indeed, inspecting typed code, lowered code, llvm ir, and native code from the REPL is great. Combine that with @benchmark for microbenchmarking and it's amazing. Note that the 'manual dispatch' is optimized out in a very ear
558.
by stabbles 6y ago
> Sections that start out collapsed is terrible for searching It doesn't collapse on desktop: https://en.m.wikipedia.org/wiki/Siena_Cathedral
559.
by stabbles 6y ago
Only on mobile, not on desktop.
560.
by stabbles 6y ago
How does this relate to the change in the C++20 standard about two's complement? "signed integers are now defined to be represented using two's complement (signed integer overflow remains undefined behavior)"
561.
by stabbles 6y ago
They could just make the mobile wiki the desktop default too. It is minimal, readable, and clean.
562.
by stabbles 6y ago
Paul Erdős is a well-known name containing an ő.
563.
by stabbles 6y ago
Someone should write a fake event generator that clutters their analytics
564.
by stabbles 6y ago
Now we need a meta static analysis tool that reports from all static analysis tools
565.
by stabbles 6y ago
Today on Hacker News: "Not everyone needs 8 hours of sleep: new research" and "Mindfulness and meditation can worsen depression and anxiety". Or more generally: "Exceptions to the rule exist". What's the v
566.
by stabbles 6y ago
They had a fun talk at JuliaCon 2020: https://www.youtube.com/watch?v=IAF8DjrQSSk :)
567.
by stabbles 6y ago
Is there actually a completely static version of Twitter as well? A 0.0 KB JS version?
568.
by stabbles 6y ago
How to google: make duckduckgo your default search engine, enter "[term] !g site:reddit.com".
569.
by stabbles 6y ago
You can easily do the latter with Gitlab, free for open source projects.
570.
by stabbles 6y ago
Inverting a non-invertible operator impossible, study finds.

More ›