cupy runtime compilation failed

Question

I'm new to cupy and try to learn it.

This following code provides an error using cuda11

import numpy
import cupy


def monte_carlo_gpu(n:int, m:int)-> float:
    
    accum = 0
    for i in range(m):
    
        x = cupy.random.random(n, dtype=numpy.float32)
        y = cupy.random.random(n, dtype=numpy.float32)
    
        r = cupy.less(x**2 + y**2, 1.0)
        
        accum += cupy.sum(r)/n

    return 4.0*accum/m

RuntimeError: Runtime compilation failed

How can I resolve this?

I was waiting for pi value (monte carlo GPU)

Answer 1

I do two ways that both works:

1. add J = cp.array(J, dtype=cp.complex128)
2. switch NVIDIA GeForce GTX 1080 Ti to another gpu NVIDIA GeForce RTX 4090

Answer 2

I also get this error on a 1080 Ti running cuda version 12.3 and cupy-cuda12x-13.0.0. By changing versions to cupy-cuda12x-12.3.0, it works like a charm (and other code that didn't error but took forever to compile now works like a charm!).