Ivan Pribec

@FedericoPerini @chenna1985 What kind of discretizations are you using?

@chenna1985 No but for a Combustion CFD solver in Fortran…. Ask me anything!

Do any of the genius AI prompt engineers have the prompts for Claude or ChatGPT for writing an FEM solver in C++ for multiphase flows using the phase field approach?

@TeX64AI @lpfeed Yes, in LaTeX these all work as expected; it is in #Typst where I struggled to find the equivalents of a few things.

Both are solvable! For sub-figures, \usepackage{subcaption} is the modern approach — \begin{subfigure}[b]{0.45\textwidth} inside a figure env handles side-by-side panels with separate captions. For math, \usepackage{amsmath} covers most needs: align for multi-line equations with & as the alignment point, align* to suppress numbering. LaTeX's math ecosystem is hard to beat for publications.

The massive response to my #Typst article* suggests that Typst is gaining ground much more rapidly than even I expected: news.ycombinator.com/item?id=453938… #latex #physics #mathematics #typsetting *lwn.net/Articles/10375…

One more week and #Fortran is almost on par with C/C++, at least on my M1 Max.

@FedericoPerini It's also available in flang.

And of course, a LOOP is always faster than an ARRAY🤦‍♂️... we'll at least soon get UNSIGNED, it's experimental on GCC.

@HiroNishikawa You can call C from Fortran too.

This is my first time to write a complete implicit solver in C. The main motivation was parallelization. There is a library that I can use to parallelize a C code relatively easily. I'll have to find time to learn the library and parallelize the C code.

@HiroNishikawa If every researcher published their tests as an abstract class, any third party could override the numerical part and run the same test under the same settings.

@HiroNishikawa But how do you deal fairly with differences between methods? Some methods favor the use of certain norms when calculating the error. Some methods use the grid, others don't.

In fact, there are so many things I’ve always wanted to do but could never find the time for. One is to write a testbed code that implements every algorithm proposed in the literature and compare them for common test cases. Developers can use it to demonstrate the advantages of their new algorithms, and use the results to write a paper. Maybe we can use AI to write such a system.

@lfortranorg Amazing work!

LFortran can now compile the Fortran Package Manager (fpm): lfortran.org/blog/2026/02/l…. Big milestone for us!

As any respectable Fortran compiler must do, LFortran now compiles LAPACK, probably the most famous Fortran library. All tests pass, single/double precision, real/complex, 32-bit/64-bit integers. lfortran.org/blog/2026/02/l…

I thought this was a good read about software: spectrum.ieee.org/it-management-…

@Waffl3x @ChShersh That's a good one! :)

@ChShersh Generics (Fortran has these coming)

Fortran is: 1. High performance 2. Statically and strongly typed 3. Easy to learn and use 4. Versatile 5. Natively parallel Sounds like a perfect language. What am I missing???

@eliasboegel @oandrepoffo @ChShersh @majorm79 I checked a few more variants and all of them use the VPCMPEQD instruction in the inner hot loop: godbolt.org/z/faKKqz513 The C-like version vectorizes with icpx, but not the other variants. So it looks a bit unreliable in general.

@IvanPribec @oandrepoffo @ChShersh @majorm79 Brief test shows that it does on newest GCC and ICX, but not on newest Clang and newest MSVC

C devs will tell you this is peak API and Programming Languages design, And nothing better has been invented since 1972.

@oandrepoffo @ChShersh @majorm79 Does it vectorize?

@ChShersh @majorm79 ultra modern

@ChShersh In Fortran: res = findloc(items(1:size), x)

@fithisux @ChShersh I'm glad you think that way.

@ChShersh BTW the Fortran discourse (like D discord) is an example of politeness.

@RueNahcMohr Nice work! The code is also discussed here: fortran-lang.discourse.group/t/cooley-tukey…

Bored? Want to help recover some ancient code? The code in this PDF needs lots of human care to come back to life. I'll take pull requests and things :] github.com/ruenahcmohr/Fo…

@DoGreatScience I wanted to refer a colleague some of your essays on verification, but the blog is down now. Any plans to publish your writing elsewhere?

wjrider.wordpress.com/2024/06/21/ins… My first blog post in six years! Somehow it all connects. Enjoy!

@johndisandonato @ChShersh Some programmers do it (fortran-lang.discourse.group/t/advent-of-co…). Feel welcome to follow the forum for the 2025 thread.

@ChShersh Fortran is actually fascinating to me. Did dabble into it forever ago. Considering it underpins workhorses such as BLAS and LAPACK it's enough to pique my interest. Maybe I could do AoC 2025 in it

C++ lost this year to C But at least Fortran is no longer in the top 10 most popular languages. I'll take it.