Dear SYCL Community,
I am currently exploring the offloading of mathematical operations—specifically BLAS and LAPACK routines—using the Intel oneMKL library. However, I’ve come to understand that Intel oneMKL only supports Intel CPUs and GPUs, which limits its portability across heterogeneous hardware platforms.
To align with SYCL’s philosophy of application portability and hardware abstraction, I’m looking for alternative libraries or approaches that support BLAS and LAPACK routines and are backend-agnostic—capable of running on NVIDIA, AMD, and Intel devices.
Could anyone kindly suggest portable solutions or share experiences in achieving this?
Thank you for your time and insights.
Best regards,
Arun