Closed KCErb closed 9 years ago
Hi KCErb, You hace clMagma that does that. Check out my github account and I have ported clMagma using CMake. Besides that, you also have ViennaCL to help you with LAPACK style routines
Hi skn123,
Great effort porting clMagma with cmake! I am not sure if you know clMagma is open sourced and hosted on bitbucket at https://bitbucket.org/icl/clmagma ? Have you tried building the latest clmagma source with cmake? One thing I know is they are (or have?) getting rid of cblas dependency.
Sigh! But atleast that gives me scope to update clMagma. Whereas Magma Cuda was regularly updated, clMagma had little work. Thanks for pointing out the link. Let me fork that out on Github and build it. If they are getting rid of clBlas, then it would be even better! clBlas can live independently. Moreover, clMagma is primarily focused on Dense matrices with no effort for sparse matrices. ViennaCL would be a better place to hook up clBlas
I meant they are getting rid of cblas dependency (like OpenBlas) but not clblas.
Thanks for the tips! I've looked through both libraries and they both seem fine. Any idea of important differences? My long-term goal is to get a Ruby FFI doing HSA-based Linear Algebra. I think there's a long road ahead (seeing as how I don't really speak C right now) but can you comment on differences between the two?
KCerb, you would need Ruby bindings for these libraries; I am not sure if they have any. The closest you could get would be python bindings
Hi all,
Just a quick question: are there plans for an openCL implementation of LAPACK (or CLAPACK)? Or is that unnecessary? Seems to me that LAPACK is just calling on BLAS, so if you point it at clBLAS it will get all of the HSA goodness you could want.
I'm not an expert on this stuff so I thought I'd ask the experts.
What do you think?