I have a question about CLBlast library (OpenCL).
I have installed CLBlast and I have been experimented with the tuning of the routines.
I have run the tunning processes that execute a routine (SGEMM let’s say) on my device for a given problem size and it outputs the best choise of the parameters (GEMMK, KREG, KWG, KWI, MDIMA, MDIMC, MWG, NDIMB, NDIMC, NWG, SA, SB, STRM, STRN, VWM, VWN).
After having found the best parameters, how do I run the SGEMM (call the CLBlastSgemm routine) with the chosen parameters?
Do I have to define those parameters inside the sgemm.opencl kernel and recompile it or there is any easier or automated way?
Thanks in advance.