Reputation: 25
Currently I am working on a kernel, that can be optimized using float16 types. However, I did not find any documentation about converting float16 into float*, because my output variable is a float*. Here is the sample code
_kernel void IncrementMatrix( __global float* Source, __global float* Target, __global float* out )
{
const int globalID = get_global_id(0);
float16 S = vload16( globalID , Source );
float16 T = vload16( globalID , Target );
S = S + T;
out[globalID*16] = (float*)S; // this is not working
}
I already tried this out = (float*)S;, but it is giving invalid type conversion error.
Upvotes: 1
Views: 1256