Than you for the praise!
BLAS is generally just an API standard, so using CuBLAS should be a matter of compiling Matrex with the CuBLAS library linked.
Do I need an NVIDIA accelerator to test CuBLAS or not?
1 Like
Thank you for your reply.
I was impressed.
I will do the CuBlas test.
I don’t want to burden you.
I have a GeForce-equipped computer and an NVIDIA library.
I want to learn technology from your Elixir code.
Thank you very much for your reply.
Kenichi Sasagawa
1 Like
No problem!
Feel free to ask any questions, and thanks for the inspiration!
1 Like
Do you plan to add set_row function?
There is set_column function but there is no way to set a row.
1 Like
Love Matrex. I got it working on my work laptop, but my desktop has been a hiccup! Tinkering around with it in Arch. Running corei3 with nvidia. I can’t seem to get it to working. It keeps complaining about not recognizing the symbol cblas_ssgemm in my deps/matrex/priv/matrix_nifs.so file. I suspect this has something to do with my choice of BLAS, but I’ve tried Atlas, Blas, and OpenBlas to no avail. Any tips?
To respond to my own question and let others know who may come across this. If you’re using a system close to my specs and not having luck with ATLAS, install openblas-lapack in pacman (but follow the ATLAS profiling setup instructions for locking frequency through cpupower.
Awesome program implementation, Versilov! Can’t wait to fold it into my project =)
EDIT:
here were some useful sources for setup:
https://www.astr.tohoku.ac.jp/~akhlaghi/ATLAS_install.html
https://www.archlinux.org/packages/core/x86_64/gcc-fortran/
https://aur.archlinux.org/packages/openblas-lapack/
** and be sure to make sure you have the latest pacman PGP signatures ** (this wasted a ton of my time)
1 Like
