cuda::atomic::fetch_max
Defined in header <cuda/atomic>:
template <typename T, cuda::thread_scope Scope>
__host__ __device__
T cuda::atomic<T, Scope>::fetch_max(T const& val,
cuda::std::memory_order order
= cuda::std::memory_order_seq_cst);
Atomically find the maximum of the value stored in the cuda::atomic
and val. The maximum is found using
cuda::std::max.
Example
#include <cuda/atomic>
__global__ void example_kernel() {
cuda::atomic<int> a(0);
auto x = a.fetch_max(1);
auto y = a.load();
assert(x == 0 && y == 1);
}