Was trying to run this on Ubuntu. I think I have CUDA all set, path to g++ compiler is /usr/bin so I set the CL_PATH = '/usr/bin' manually. I know this is beyond your purview probably, but any idea why compilation would be failing? Getting these errors. I also added the path to nvcc to my PATH.
pycuda.driver.CompileError: nvcc compilation of /tmp/tmp3pt26bw2/kernel.cu failed
[stderr:
kernel.cu(252): error: identifier "uint8_t" is undefined
typedef uint8_t u8;
^
kernel.cu(253): error: identifier "uint16_t" is undefined
typedef uint16_t u16;
^
kernel.cu(254): error: identifier "uint32_t" is undefined
typedef uint32_t u32;
^
kernel.cu(255): error: identifier "uint64_t" is undefined
typedef uint64_t u64;
^
kernel.cu(3828): error: identifier "UINT64_C" is undefined
state[xi + yi + (yi << 2)] ^= UINT64_C(0x01) << ((blockOff & 7) << 3);
^
5 errors detected in the compilation of "kernel.cu".
]
Was trying to run this on Ubuntu. I think I have CUDA all set, path to g++ compiler is /usr/bin so I set the CL_PATH = '/usr/bin' manually. I know this is beyond your purview probably, but any idea why compilation would be failing? Getting these errors. I also added the path to nvcc to my PATH.