CUDA: argument of type "float" is incompatible with parameter of type "void *"

Question

when I defined two float arrays like this:

float* d_A, d_B;
cudaMemcpy(d_B,h_B,size,cudaMemcpyHostToDevice);

I receive an error concerning array d_B

argument of type "float" is incompatible with parameter of type "void *"

if I defined the arrays like this:

float* d_A;
float* d_B;

it works fine and the error disappears. what can be wrong here ? I'm using NSight Visual Studio edition version 3.2

Answer 1

The * specifing the pointer belongs to the variable, not the type. So to initialize it on the same line, use

float *d_A, *d_B;

Putting the * next to the variable with no space makes it much more obvious as well.