How can I get the kernel code to implement a sample by sample processing instead of an array?

[FHuss10]

__kernel void ADD_TO_ARRAY ( __global const short* restrict k_din,

__global const short* restrict k_din2,

__global short * restrict k_out )

{

for(unsigned i=0; i<20; i++)

k_dout[i] = k_din[i] * k_din2[i];

}

0.5 DSP is correctly detected for a multiplier for Arria 10.

All examples I have seen show a for loop for the block of data processing. Is it possible to process sample by sample?

OpenCl does not allow static int function scope.

I would like to process every N samples to produce a result.

Can I please get a sample code?

[GRN]

Hi FHuss10,

Do you mean parallel execution of each loop iteration?

You can use #pargma unroll for the loop.