I'm in a similar position as you were right now: a PhD student in HPC domain who really wants to/ needs to use C++ for everything ( I love Python but has now come to terms with its limitations). Your last sentence is scary for me to hear- can you please elaborate a bit?

Not sure what the parent meant but I always think of https://travisdowns.github.io/blog/2019/08/26/vector-inc.htm...

Of all the commonly used languages, C++ is the most complex, baroque, and filled with subtle and not-so-subtle footguns. Of course with all that downside comes the upside that C and C++ are the lingua franca for low level programming where performance matters, not only HPC (yes, I hope Rust will be there one day, but I digress..). Fortran is pretty much limited to numerical computing, you're not going to find much Fortran libraries outside that domain.

So if you envision doing programming like that outside HPC after your PhD, C++ might be worth the investment.

But of course a lot comes down to what language other people in your group and your particular subfield, and what your supervisors are using. Doing a PhD is hard enough without being the odd guy out who's using a language nobody else in your community is using.

Ever heard of Perl, PHP, JavaScript, sh or Fortran itself for that matter? Plenty of languages are far more "baroque" than C++.

Ignoring all the cliche comments about C++ complexity I'm gonna say something that is a controversial opinion of sorts: if you're a PhD student in CS and you're doing systems research and you can't write/read passable C/C++ then you shouldn't be allowed to graduate.

CS is NOT computational science or HPC (generally).

>you're doing systems research

My god, though, why Python 10 years ago? There were better languages available then, like Ruby or Perl. Was NumPy just too big a thing even then?

The scientific Python stack (numpy, scipy, matplotlib + a plethora of more specialized things) were definitely a very big thing 10 years ago. Deep learning was just on the cusp of taking off in a major way, but in traditional numerical computing python was very big before that.

Ruby and Perl in HPC? I mean this in the nicest way possible if you’re being sincere, but I can’t tell if you are trolling or not

I worked in an HPC lab at a medium-sized university around 10 years ago. Python was definitely the most popular interpreted language among the students. I remember it was years(?) earlier that MIT had switched much of its curriculum to Python.

Some older folks used Perl in places where others would've just written bash; mostly scripting invocations of actual number-crunching programs. Ruby would've raised some eyebrows for sure!

Yeah the switch was like '08 that MIT went to Python.

I think it is far from a foregone conclusion that Perl and Ruby are better than Python, or were ten years ago.

how are ruby and perl better solutions to how appalled you must be given your first sentence? if anything, they're just _different_ solutions that are based on personal preference and nothing more

You're not going to get traction using Ruby or Perl for a Ph.D. in astrophysics :)

The only ones judgy-er about using the "wrong" language than computer scientists are engineers and other scientists.

I reckon that's very true but it's important in one's closed work environment. If one bypasses accepted practice then one must be very good or have very good reason or both.

perl had some history in bioinformatics but ruby and php never made any inroads into science or HPC to the level Python has. Python has almost entirely replaced perl in this context.

I chose Python over perl about 25 years ago and never would have considered ruby or php (both seemed squarely aimed at web developers not scientists). I believe Numeric Python was released around that time, and it was a revelation- especially to matlab users, who recognized the syntax and behavior.

Yeah I did some bioinformatics work years ago and there were still a few Perl things, but not many.

Python was definitely around and common 10 years ago. I did my Master’s 8 years ago and we had a python bootcamp for people not familiar with it and numpy was part of it.

I love perl but it is awful at signal handling. The threading and shared memory interface are not as good as other languages either.

Python has a low barrier for entry and can be useful immediately to newer programmers. It is also fairly well rounded.

That being said, I am not partial to python as OP was with fortran, and steer around it regularly.

>C++ is truly the most wrong language for this space there could be

What? Why???

I switched from Python to C++ because Cython, Numba, etc. just weren't cutting it for my CPU-intensive research needs (program synthesis), and I've never looked back.

It's bloated beyond recognition and full of subtle and not so subtle footguns that you simply don't have to worry about in Fortran.

Gitgud

C++ and Python. Ypu picked the two worst programming languages to learn.

I can't remember the exact quote right now, but what you say echoes that quote that goes something like: If you make a tool that has such a low barrier to entry that it makes it easy enough for unskilled people to use it and inevitably write low-quality code with it, then that is the definition of a good tool.

I've both worked in HPC and written Fortran. C++ is almost always preferable because Fortran is too limited as a language outside of numerical code. Most HPC stopped being limited by the performance of numerical code 20 years ago. Instead, HPC codes were and are commonly limited by memory bandwidth and locality, two things C++ is almost uniquely suited for.

The kind of C++ written for HPC tends to be much simpler than the kind of C++ used to write e.g. database engines. The complexity of C++ is not that onerous in context and Fortran isn't exactly an exemplar of ease of use.