Code inspection: A '__shared__' variable cannot have an initializer
This inspection reports an initializer on a __shared__ variable. CUDA does not initialize shared memory when the variable is declared. If you need an initial value, assign it after the declaration.
Example
__global__ void kernel()
{
__shared__ int value = 0;
}
__global__ void kernel()
{
__shared__ int value;
value = 0;
}
27 October 2026