C++ Function Overloading starting...
Device Count: 1
GPU Device 0: "Hopper" with compute capability 9.0

Shared Size:   1024
Constant Size: 0
Local Size:    0
Max Threads Per Block: 1024
Number of Registers: 12
PTX Version: 90
Binary Version: 90
simple_kernel(const int *pIn, int *pOut, int a) PASSED

Shared Size:   2048
Constant Size: 0
Local Size:    0
Max Threads Per Block: 1024
Number of Registers: 14
PTX Version: 90
Binary Version: 90
simple_kernel(const int2 *pIn, int *pOut, int a) PASSED

Shared Size:   2048
Constant Size: 0
Local Size:    0
Max Threads Per Block: 1024
Number of Registers: 14
PTX Version: 90
Binary Version: 90
simple_kernel(const int *pIn1, const int *pIn2, int *pOut, int a) PASSED