High Performance DGEMM on GPUs
Library, ICT,CAS, 2009
Dense matrix operations are important problems in scientific and engineering computing applications. There have been a lot of works on developing high performance libraries for dense matrix operations. Basic Linear Algebra Subprograms (BLAS) is a de facto application programming interface standard for publishing libraries to perform basic linear algebra operations such as vector and matrix multiplication. The first BLAS is released as a building block of LAPACK, which is a performance portable library for implementing dense linear algebra. Hardware vendors (Intel, AMD, IBM, etc.) also provide BLAS librariesy tuned on their own processors, i.e. MKL and ACML. It is well-known that the performance of BLAS depends on the underlying hardware.