4 ms·

C++ Expression Templates: Eigen, Blaze, Fastor, Armadillo & XTensor (2020)

• gok 4y ago
  It's really unfortunate that C++ keeps adding more functionality that makes one of its critical problems (horrible syntax debugging) worse. All of these libraries generate atrocious compiler errors for even trivial typos.
  • d_tr 4y ago
    At least C++ is now a much nicer language that it was before C++11, isn't it?
    • pjmlp 4y ago
      It is, the main problem is not being able to take cruft away, otherwise we get the Python 3 of C++.
• syockit 4y ago
  Good article. Next time I ever want to revisit using C++ for scientific calculation again I'll keep in mind that Fastor has the fastest compilation time of all. Still, 2 seconds for just a benchmark code is a bit too much. I remember doing this kind of finite difference stencil templating using plain C++. All the performance gains went out of the window as soon as the array size becomes dynamic.
• lostdog 4y ago
  Bonus points for measuring the compile times. Unfortunately matrix libraries are the types of things that really challenge C++, and the compiler devs haven't been able to bring the compile times down to reasonable levels yet.
• Const-me 4y ago
  > We now also have explicit SIMD vectorisation.. So say on an AVX512 enabled architecture you can compute the dot product of two vectors I wonder will it be good? Computing dot product efficiently is harder than it seems, example: https://stackoverflow.com/a/59495197/126995 https://stackoverflow.com/a/59495197/126995 Eigen usually does stuff like that rather well.
• sfpotter 4y ago
  Compile times like these are why I avoid C++ for numerical programming these days.