ICLDisco / dplasma

DPLASMA is a highly optimized, accelerator-aware, implementation of a dense linear algebra package for distributed heterogeneous systems. It is designed to deliver sustained performance for distributed systems where each node featuring multiple sockets of multicore processors, and if available, accelerators, using the PaRSEC runtime as a backend.
Other
11 stars 9 forks source link

Use only cublas_v2 #101

Open abouteiller opened 1 year ago

abouteiller commented 1 year ago

Description

Mixing usage of cublas_v1 and v2 forces us to do some very ugly things with .h files to resolve conflicting datatypes. cublas_v1 is deprecated for more than 7 years, its time for it to go away.

Describe the solution you'd like

Replace all invocation of zgemm() with zgemm_v2()

Additional context

Converting cublas_v2 to hipblas/rocsolver is a lot easier than from cublas_v1, since the APIs are a lot closer.