The CUBLAS and CULA libraries

Day: Monday, October 28
Time: 2:10 PM - 3:00 PM
Place: Snedecor Hall 2113

The CUBLAS library is a CUDA implementation of the Basic Linear Algebra Subroutines (BLAS) library, a standard Fortran/C API for matrix algebra. CULA, the CUDA C implementation of LAPACK, is built on top of CUBLAS. CULA supports more advanced linear algebra calculations than CUBLAS, such as the LU, QR, RQ, QL, SVD, and Cholesky factorizations, solving systems of linear equations for general m by n matrices, ordinary least squares, and eigenvalue problems. In a nutshell, CUBLAS and CULA accelerate common linear algebra routines while taking care of all the GPU parallelism under the hood. This lecture will overview and demonstrate the usage of both CUBLAS and CULA.

Materials

Slides
Video

Example Code

Example2.cpp
ols.cu
mwe.cu
simpleCUBLAS.cu
ae.cu
de.cu
deviceInterface.cu
ll.cu
systemSolve.cu

Makefile
Makefile
Makefile
Makefile
Makefile
Makefile
Makefile
Makefile
Makefile