在pyCuda内核中生成单随机数

import pycuda.driver as cuda
import pycuda.autoinit
from pycuda.compiler import SourceModule

mod = SourceModule("""
#include <curand_kernel.h>

extern "C" {
    __global__ void generate_random(float* result) {
        int tid = blockIdx.x * blockDim.x + threadIdx.x;
        curandState state;
        curand_init(0, tid, 0, &state);  // 初始化随机数生成器
        result[tid] = curand_uniform(&state);  // 生成随机数
    }
}
""")

generate_random = mod.get_function("generate_random")

# 定义生成随机数的数组大小和线程块大小
array_size = 1
block_size = 1

# 创建结果数组
result = cuda.mem_alloc(array_size * 4)  # 4字节浮点数

# 调用内核函数生成随机数
generate_random(result, block=(block_size, 1, 1), grid=(array_size, 1))

# 将结果从GPU内存复制到主机内存
output = numpy.empty(array_size, dtype=numpy.float32)
cuda.memcpy_dtoh(output, result)

print("随机数:", output[0])