MAGMA 1.2.1 Released

[Stan Tomov]

MAGMA 1.2.1 was released on June 29, 2012. This release provides the following new functionalities and improvements:

- Fix bug in [zcsd]getrf_gpu.cpp;
- Fix workspace requirement for SVD in [zcsd]gesvd.cpp;
- Fix a bug in freeing pinned memory (in interface_cuda/alloc.cpp);
- Fix a bug in [zcsd]geqrf_mgpu.cpp;
- Fix zdotc to use cblas for portability;
- Fix uppercase entries in blas/lapack headers;
- Use magma_int_t in blas/lapack headers, and fix sources accordingly;
- Fix magma_is_devptr error handling;
- Add magma_malloc_cpu to allocate CPU memory aligned to 32-byte boundary for performance and reproducibility;
- Fix memory leaks in latrd* and zcgeqrsv_gpu
- Remove dependency on CUDA device driver
- Add QR with pivoting in CPU interface (functions [zcsd]geqp3)
- Add hegst/sygst Fortran interface
- Improve performance of gesv CPU interface by 30%
- Improve performance of ungqr/orgqr CPU and GPU interfaces by 30%; more for small matrices.

Further detail can be found in the ReleaseNotes.

See the MAGMA software homepage for a download link.
http://icl.cs.utk.edu/magma/software/index.html

[sameer_asal]

Hello ,

Thank you for your efforts building such libraries.

I am trying to use MAGMA Blas in my codes. I noticed that you are using cublas instead of cublas_v2 ( the file magmablas.h inculdes cublas.h ) , Is there any reason for this , Is there an easy way to convert the code to use v2 ?

Kind Regards,

[mgates3]

Yes, internally we use cublas (v1). This is just historical -- we plan to update the code to cublas v2. This is actually part of why we introduced wrappers, e.g., magma_zgemm around cublasZgemm. In the meantime, if you need to use cublas v2, you *might* be able to:
1) compile MAGMA as is.
2) copy the magma include headers, to include in your sources, and replace #include <cublas.h> with #include <cublas_v2.h>.
I have not tested this, but I think the MAGMA headers need to include cublas only for some data types like cublasStatus_t and cuDoubleComplex.

-mark