Migration of cublasHandle_t in __global__ or __device__ function is not supported. You may need to redesign the code.

Detailed Help#

The cublasHandle_t is migrated to the cl::sycl::queue. Because SYCL* 2020 does not support device-side enqueue of the kernel, the usage of the cl::sycl::queue is not possible in SYCL kernels.

Suggestions to Fix#

Redesign the code to submit all kernels by the host program instead of device-side enqueue.

Refer to DPCT1020.