How to use CAO mode (Compiler Assisted Offload) of MKL

[yu__frank]

Hello everyone

I want use the CAO mode of MKL.

I writed a sample code ,

#pragma offload target(mic) \

    in(n, alpha, incx, incy) \
    in(x:length(n)) \
    in(y:length(n)) \
    out(y:length(n))
{
    cblas_caxpy(n, &alpha, x, incx, y, incy);
}

Compiler ICC printf some error message:

mkl_test.cpp(63): error: function "cblas_caxpy" called in offload region must have been declared with compatible "target" attribute
        cblas_caxpy(n, &alpha, x, incx, y, incy);

So , Is the MKL(2017) support 's CAO mode.

[Loc_N_Intel · Accepted solution]

Yes, CAO is supported in MKL (see https://software.intel.com/en-us/node/528601)

A good example of how to use CAO can be downloaded here: https://software.intel.com/en-us/articles/intel-math-kernel-library-intel-mkl-compiler-assisted-offload-and-automatic-offload-example

Thanks

[yu__frank]

Nguyen .

Thanks for response.

I run the example successful.

There is another important thing to note here. The MKL library of MIC is not installed by default, you need to manually install it.