[BLAS, cuBLAS backend, hipSYCL] Add official hipsycl cuBLAS support

[sbalint98]

As discussed under PR #144 hipSYCL rocBLAS support comes with automatic support for cuBLAS as well. This PR adds this information to the README.

Furthermore, this PR differentiates the FindcuBLAS.cmake file between hipSYCL and Intel LLVM. In the case of hipSYCL, there is no need to have the OPENCL_PATH set. Moreover, the CUDA_CUDA_LIBRARY environment variable is also not strictly necessary since finding the libcuda.so is handled by hipSYCL.

onemkl-cuda-hipsycl-test.txt

[mmeterel]

@sbalint98 As you are aware, there were several changes since the time this PR is opened. Could you please rebase and check if the tests are all good? Also, I think, we can enable cuRAND backend with hipSYCL with this PR as well. What do you think?

Could you please let me know when everything is ready?

[sbalint98]

Hi @mmeterel thank you for reminding me. I'll have a look and prepare this PR until Monday. I'll let you know if it is ready for review

[sbalint98]

Sorry for the delay. I have rebased the branch and updated the readme. All tests are passing on my sid
cublas_tests_hipsycl.txt
e

[mmeterel]

Sorry for the delay. I have rebased the branch and updated the readme. All tests are passing on my sid cublas_tests_hipsycl.txt e

@sbalint98 Thanks. Do you plan to add cuRAND support in this PR as well?

[sbalint98]

@nilsfriess and @normallytangent are working on the cuRAND support. I believe for them it might be easier to open a new PR. Would this be acceptable for you?

[mmeterel]

@nilsfriess and @normallytangent are working on the cuRAND support. I believe for them it might be easier to open a new PR. Would this be acceptable for you?

Sure, that works if it will be easier on your side.

[mmeterel]

@sbalint98
Thanks for the PR. I tested it locally (had to rebase hipSYCL and build it with LLVM-13) and it works smoothly. :)

[mmeterel]

@mkrainiuk Could you please take a look at this?

[mkrainiuk]

Thank you for the PR! Changes look good to me. I have one minor comment.

[mkrainiuk]

Is there any reasons we cannot use one set_target_properties? I see the only difference: OPENCL_INCLUDE_DIR that can be added with inline check, something like "$<$BOOL:${USE_ADD_SYCL_TO_TARGET_INTEGRATION}: ${OPENCL_INCLUDE_DIR}>"

[mmeterel]

Not sure...but i also see differences between CUDA_CUDART_LIBRARY vs CUDA_CUDA_LIBRARY