With Intel’s oneAPI, libmkl_ccgdll has been extended to offload some cluster operations to GPUs via SYCL. The DLL now detects available devices and, for large matrix operations, copies data to GPUs while maintaining MPI consistency across nodes. This hybrid CPU-GPU cluster work is orchestrated through the same interface – the user code does not change, only the underlying MKL runtime.
Check your build logs for exact error. Use dumpbin /dependents myprogram.exe to see which DLLs it expects. libmklccgdll work