What is cuda::access_property::shared?

cuda::associate_access_property(&shmem, cuda::access_property::shared{});

I see this here: cuda::access_property | libcu++

Any meaning?

It means this pointer is a pointer to shared memory.

Yeah… I see… But why we write it? What is the usage?

It can be used to tell the compiler to assume that a pointer is global memory or shared memory.

1 Like