Scientific computing

There's this one programming language but i forgot what it is...it's like snake or something

>>8418599
>computing clusters
Python? ;)

>>8418599
Cobra? Anaconda? Adder? No idea what you're on about I don't think there's any language with a snake.

>>8418599
>>8418608
So what makes Python the best for scientific computing. I did FORTRAN on recommendation from my professor at the time but he also offered C++ as an alternative.

Why should you learn Python over those?

R for general data manipulation and graphing. MATLAB for ode modeling of the pathways we study.

>>8419180
You shouldn't

>>8417921

CUDA YOU LITTLE SHIT

>>8419180
http://www.stat.washington.edu/~hoytak/blog/whypython.html

>>8419250
>>8417921

CUDA or any other GPU language if you have access to a GPU cluster (pick one of whatever compiler is already installed on the cluster).

If you have don't have access to a GPU cluster or it has Xeon Phi coproccessors, C or Fortran (I guarentee C and FORTRAN compilers will be already installed on your cluster... unless you're using that brand new Chinese one which uses it's own special software). Personally I preferred FORTRAN over C simply because it's simpler and could do everything I needed it to do in much fewer lines than C (FD CFD codes and KMC nucleation). This will depend on what you are doing though (but if it can be done in FORTRAN, it is almost certain it will be easier to program in FORTRAN than C)...

I've used Mathematica on clusters for image processing... making certain plots from large data sets which couldn't be done off my desktop. Just remember, you'll need to start Xvfb if you want to plot using Mathematica via a non-interactive queue script (or if you have service units to burn, X11 forwarding on an interactive queue session).

I know of people who use Matlab instead of Mathematica for the same purposes... I even know someone who uses a FORTRAN library for plotting (but it looks god awful)... Pick whatever you feel most comfortable with... Outside plotting huge data sets, I can't think of many things you should be running Matlab on a cluster for... Number crunching should be done in C, FORTRAN or some GPU language...

Python is basically a free alternative for Matlab or Mathematica... Nobody I know runs any Python code on a cluster... Usually you run Python/Matlab/Mathematica for short calculations where the time to program is a significant amount of time you've invested in the actual calculation... That or plotting...

Other programming languages will be dependent on what you're doing... e.g. R for stats, IDL for astro, etc... There are heaps of field dependent packages which you or your supervisor should be familiar with.

Hi, I work in high performance computing. I don't often write parallel programs, but I have picked up a bit from others and I've read a couple books.

You can honestly get away with anything. Most people use C or FORTRAN, but we support R, MATLAB, Python, Lua, and just about any other language you can think of. There are full-time staff members at my university who will help you at every step.

As far as I know, everything >>8419283 is correct. Computational chemistry is the biggest user of our resources and they have specific packages they prefer.

Our GPU nodes are fairly new and I haven't had much time to experiment, so I can't comment on those.

>>8419314
Thanks. Thanks also to >>8419283 for his insight.

I actually am a computational chemist. I used Firefly (PC Gamess) for my undergrad thesis. I just felt kind of felt frustrated at how the computational package felt like a black box that I needed to learn the language

Actually, it depends on what exactly you're trying to do. I've seen density functional theory codes in both fortran and python. The thing is, the python codes call blas or lapack routines for important stuff, so if you're interested in actually modifying or writing the performance stuff, you'd need to use C or fortran either way.

can someone tell me where functional languages are used or are they just a meme?