MAGMA Forum

We'll track this on the bitbucket issue.
-mark

What BLAS & LAPACK library are you using? What size is your problem? Are you timing just the magma_ssyevd_gpu call, or are you timing the runtime of the entire test program? You can run MAGMA's testers in magma/testing to get both MAGMA and LAPACK timings. There are several versions to try (note the...

Sorry, we don't have batched eigenvalue problems yet.
-mark

I'm not sure what you mean. The download for source code is here:
https://icl.utk.edu/magma/software/index.html

-mark

From scanning the code, it looks like depth=1, a single random butterfly is applied.
-mark

Answering your questions in reverse order: 2. For allocating matrices on the GPU using Fortran, in the latest release (MAGMA 2.5 rc 1) there is magmaf_[sdcz]malloc (notice the f for Fortran in magmaf prefix). Otherwise, you can use the (very old) cuBLAS Fortran interface that comes in CUDA/src/fortr...

It's unclear what is occurring in your example. A sample run using one of the MAGMA testers, or at least sample code, so we can attempt to reproduce the issue is needed. What is the problem size? What is your system — OS, CUDA version, what CPU & GPU? Also, it would help to expanding the "[+] Comput...

Done. It's just adding it as a flag in the Makefile. MAGMA should always run on any CUDA arch >= what it was compiled for. So if you compile MAGMA for sm_30, it should also run on sm_35 (Kepler K40), sm_50 (Maxwell), ..., sm_70 (Volta), sm_75 (Turing), etc. Compiling for a specific architecture migh...

We're looking into it. I don't think removing 3.0 capability was the intention.
-mark

Yes, it compiles on Unix. (I'm actually using macOS.) g++ -Wall -fopenmp \ -I ${MAGMADIR}/include \ -I ${MAGMADIR}/sparse/include \ -o example_sparse example_sparse.cpp \ -L ${MAGMADIR}/lib -lmagma -lmagma_sparse \ -L ${CUDADIR} -lcublas -lcudart -lcusparse \ -L /opt/OpenBLAS/lib -lopenblas From the...