How to call class function inside a parallel_for having queue as a parameter

[Anitha]

Hi,

I tried this

q.parallel_for(nd_range<1>((16), (16)), [=](nd_item<1> item)

{    auto index = item.get_global_id(0);

      classA.func(a[index], q);

});

void classA::func(float *a, queue & my_queue)

{

}

But got this error

Kernel parameter has non_trivially copy constructible class/struct type 'sycl::queue'

[VarshaS_Intel]

Hi,

Thanks for posting in Intel Communities.

Could you please provide us with the OS details and the version of dpcpp you are using?

Could you please provide us with a complete sample reproducer code to investigate more at our end?

Thanks & Regards,

Varsha

[Anitha]

Hi,

OS details:

OS - Redhat enterprise Linux 8.0

Processor - Intel Xeon(R) 16 - COMPUTE UNITS

OS type - 64-bit

ToolKit versions:

Basekit version - 2021.2.0-2883

Hpckit version - 2021.2.0-2997

Dpcpp compiler version - 2021.2.0-610

Code has been attached.

Thanks & Regards,

Anitha

[VarshaS_Intel]

Hi,

We are working on your issue. We will get back to you soon.

Thanks & Regards,

Varsha

[VarshaS_Intel]

Hi,

Could you please let us know the use case of passing the queue as a parameter and calling it inside a kernel?

In your code(fft.cpp) you are submitting a queue inside the function definition which is being called inside a kernel. So, indirectly it became a recursive call. Recursion is not supported by the SYCL.

And also, if you want to use FFT with SYCL kernels please find the path where you can find the samples for it in C:\Program Files (x86)\Intel\oneAPI\mkl\latest\examples\examples_dpcpp.zip\dpcpp\dft\source

Thanks & Regards,

Varsha

[Anitha]

Hi,

Here we are trying to do fft 16 times for different set of input's using for loop.

To get better execution time we thought of making it parallel we used parallel_for and passed queue as a parameter which is required by a fft.

Thanks and Regards,

Anitha

[VarshaS_Intel]

Hi,

>>passed queue as a parameter which is required by a fft.

Could you please try the samples if you want to use FFT with SYCL kernels please find the path where you can find the samples for it in C:\Program Files (x86)\Intel\oneAPI\mkl\latest\examples\examples_dpcpp.zip\dpcpp\dft\source

Could you please let us know what commands are you using at the time of compilation of your code? And also, could you please provide the complete error log?

Thanks & Regards,

Varsha

[VarshaS_Intel]

Hi,

Could you please provide us with the steps/command you have followed ?

Thanks & Regards,

Varsha

[VarshaS_Intel]

Hi,

We have not heard back from you. This thread will no longer be monitored by Intel. If you need further assistance, please post a new question.

Thanks & Regards,

Varsha

[Anitha]

Hi,

We are using eclipse to build the program.

Command : dpcpp

All options : -g -Wall -O0 -l /parallelFor

And using terminal to run.

./parallelFor

But while building itself we are getting error.

Thanks & Regards,

Anitha