I've been playing for 2 days with magma 1.0 rc3, linked with Cuda 3.2 and Intel Mkl 11.1.056, i had to modify by hand few sources to got it compiled but his thread helped (viewtopic.php?f=2&t=120
First of all, i wanted to thank you for your work and for keeping it free to use for everyone.
I've tried to solve an eigen problem with magma. Since my original code was in fortran, i had to modifie magma_df77.cpp and magma_zf77.cpp to be able to call magma_dsyevd and magma_zheevd from it.
It failed in trying to use the double real routine. At beginning, i'vent read properly the comment in it so all my arrays were in double. I encounter some memory trouble which seemed to be linked to the fact magam-dsyevd waits for some complex array.
But, when i put my amatrix and work arrays in complex and call magma_dsevd i can't get my results properly.
with the same arrays, i succeeded when i called zheevd...
Is this known issues and will be fixed before release?