Hacker News new | past | comments | ask | show | jobs | submit login

This is one I really like

https://p0nce.github.io/d-idioms/#Parallel-foreach

I use D as a complement to R. If I'm doing a simulation, it's really simple to shift to doing them in parallel. (I know other languages also provide such support, but I'm a fan of the overall D language as a replacement for C in the statistical computing realm.)




D is really really nice for writing code then upping the performance as you go.

We do need better scientific programming libraries, however.


> ** is really really nice for writing code then upping the performance as you go.

We do need better scientific programming libraries, however.

This is pretty much a rephrasing to few of the main goals of Julia. You start prototyping using the dynamic python/lispy, and add some typing or some optimization to got C/Fortran performance. It is really finding it's space in scientific computing. Check out Flux.jl, DifferentialEquations.jl, and SciML.jl for a preview of the cool projects Dynamic Dispatch is enabling.

Bonus points for excellent Macros: MLstyle.jl


Put it this way, I am using D quite a lot for microarchitectural benchmarking because the inline asm is so easy to us, I have access to LLVM and GCC intrinsics, I can easily parameterize my code as a template at compile time etc.

Julia has impressive libraries, I don't really rate the language.

D gives more guarantees to the optimizer, makes it easy to give information to said optimizer and has much safer slices which prevent a lot of bugs.

For example of something cool: I have a library which can run a benchmark over a template - that is, imagine a template taking a parameter N or even a type, my library can iterate through these and perform statistics on them seamlessly.


Yep, I often hear people say their favourite language would be great for scientific programming, it only needs scientific libraries and people to use them. Julia already has those things, they're not easy to get.


We already have a library called Mir which has laid down a lot of the hard stuff to begin with. Its very very fast even relative to the small number of people working on it.


I agree. I have everything I need after 7-8 years, but that's not true for everyone.


We're using Go for our discrete sim as well as our optimization platform and I love it.

A lot of these samples of D are really enticing and I think I'm going to have to spend some time really digging in. I hope the toolchain is nice.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: