thrust::d evice_vector 未正确更新

>我正在尝试将 CUDA 内核方法的结果保存在设备向量中 下面是我的内核代码，当我打印它时，数据将进入 int* out 和 int* outTwo 数组。

template <typename T>
struct KernelArray
{
T*  _array;
int _size;
};
template <typename T>
KernelArray<T> convertToKernel(thrust::device_vector<T>& dVec)
{
KernelArray<T> kArray;
kArray._array = thrust::raw_pointer_cast(&dVec[0]);
kArray._size = (int)dVec.size();
return kArray;
}
__global__ void CompareVector(KernelArray<CollisionSet> inArrayA,    KernelArray<CollisionSet> inArrayB, 
int* out,int* outTwo, unsigned int numColVec)
{
unsigned int index = __umul24(blockIdx.x, blockDim.x) + threadIdx.x;
if (index >= numColVec) return;
for (int i = 0; i < numColVec; i++) {
if (inArrayA._array[index].one == inArrayB._array[i].one)
{
if (inArrayA._array[index].two == inArrayB._array[i].two) {
out[index]= inArrayA._array[i].one;
outTwo[index] = inArrayA._array[i].two; 
printf("out %d, outTwo %d, index %dn", out[1], outTwo[1], index);
}
}
}
}

但是，它不会更新为device_vector

void SweepandPrune() {
device_vector<CollisionSet> ColVectorX, ColVectorY, ColVectorZ;
device_vector<int> one, two;
one.push_back(-1);
two.push_back(-1);
int* d_one = thrust::raw_pointer_cast(&one[0]);
int* d_two = thrust::raw_pointer_cast(&two[0]);
ColVectorX = ProjectVector(0);
ColVectorY = ProjectVector(1);
ColVectorZ = ProjectVector(2);
CompareVector << <1, 256>> > (convertToKernel(ColVectorX),  convertToKernel(ColVectorY), d_one, d_two, ColVectorY.size());
}

当我打印device_vector尺寸时，它仍然是 1(因为我最初放入了 -1(

我已经搜索了这个 https://gist.github.com/docwhite/843f17e33e4c1f2b531a14a4bdfe90ec 我认为raw_pointer_cast应该有效。我错过了什么？

提前谢谢。