3 ms·

I’m not sure if that holds. Every university I’ve ever done research in has no issues paying for software like MATLAB, ANSYS, Mathematica, Intel MKL, Intel Comp

I’m not sure if that holds. Every university I’ve ever done research in has no issues paying for software like MATLAB, ANSYS, Mathematica, Intel MKL, Intel Compiler Suite, and the like. The people I know that still use Fortran (including myself) are all heavily focused on computational numerics and the fact that ifortran is closed source vs gfortran being open just does not even factor into the equation. If it’s available to me, and I can show that a program compiled by ifortran and MKL is faster than one compiled by gfortran and BLAS then I will use the faster one.

colechristensen 6y ago
I left that world a long time ago, but what held is there was no problem buying site licenses for big software, not as much "I'm a research assistant and I'd like to buy X for $Y,000 for my project" The difference being buying something with broad appeal used by lots of people within the university, and buying something that I'd like to try for my project. Especially when there's already an established method which doesn't involve spending your PI's grant money.
tikej 6y ago
You must have done research at well funded universities :) My experience is quite different. Although after some hassles I was able to get license but it’s online only. Overall I’m trying to move away from commercial software (I don’t teach any, only use foss) but yeah Mathematica, MKL + ifort are unbeatable in some respects (Matlab is slowly going to be replaced except some specialised toolboxes, don’t use ansys) Anyway, proprietary and closed source software is in my opinion harmful for academia. Sometimes a lot, sometimes a little but in general it’s no good, although I understand that it funds some development that wouldn’t be done otherwise.
- beezle 6y ago
  Not sure how I could proprietary/closed software being 'harmful'. If the product is superior and allows researchers/students to do a better or more efficient job then why not (absent severe budget constraints)? It is highly unlikely that those end users would ever modify the code or even need to view the source.
  - tikej 6y ago
    Simply via inability to reproduce or check research code without proprietary software. Science shouldn't be built on something like that.
    - walshemj 6y ago
      back checking / testing is not just run the same "codes"
gnufx 6y ago
The fact that universities splurge money on these proprietary tools rather than spending it on long-term improvements isn't a good thing, even without the pain of actually supporting them. The typical differences between stuff using ifort/MKL and gfortran/OpenBLAS/BLIS are at least similar, and probably smaller, than the sorts of variation you see anyhow on HPC systems, even without the important large-scale optimizations.

colechristensen 6y ago

I left that world a long time ago, but what held is there was no problem buying site licenses for big software, not as much "I'm a research assistant and I'd like to buy X for $Y,000 for my project" The difference being buying something with broad appeal used by lots of people within the university, and buying something that I'd like to try for my project. Especially when there's already an established method which doesn't involve spending your PI's grant money.

tikej 6y ago

You must have done research at well funded universities :) My experience is quite different. Although after some hassles I was able to get license but it’s online only. Overall I’m trying to move away from commercial software (I don’t teach any, only use foss) but yeah Mathematica, MKL + ifort are unbeatable in some respects (Matlab is slowly going to be replaced except some specialised toolboxes, don’t use ansys) Anyway, proprietary and closed source software is in my opinion harmful for academia. Sometimes a lot, sometimes a little but in general it’s no good, although I understand that it funds some development that wouldn’t be done otherwise.

beezle 6y ago
Not sure how I could proprietary/closed software being 'harmful'. If the product is superior and allows researchers/students to do a better or more efficient job then why not (absent severe budget constraints)? It is highly unlikely that those end users would ever modify the code or even need to view the source.
- tikej 6y ago
  Simply via inability to reproduce or check research code without proprietary software. Science shouldn't be built on something like that.
  - walshemj 6y ago
    back checking / testing is not just run the same "codes"

gnufx 6y ago

The fact that universities splurge money on these proprietary tools rather than spending it on long-term improvements isn't a good thing, even without the pain of actually supporting them. The typical differences between stuff using ifort/MKL and gfortran/OpenBLAS/BLIS are at least similar, and probably smaller, than the sorts of variation you see anyhow on HPC systems, even without the important large-scale optimizations.