There are two issues:
1. APIs of blas functions in APPML are different from standard BLAS. Please take a loot at clAmdBlas.h
for more details.
2. Arguments passed to BLAS functions in APPML resident on GPU's global memory. Data in HPL are
located on CPU's memory. You need to transfer data to GPU's memory explicitly before calling blas functions
You may write wrapper functions on top of blas functions in APPMNL, whose API are exactly the same as
the standard BLAS's. You need to transfer data between CPU and GPU in these wrapper functions.
Hope it is helpful.