Iron-Stark

void invert_device(float *src, float *dst, int n) {
		int batchSize = 1;
		int *P, *INFO;
		cudaMalloc<int>(&P, n * batchSize * sizeof(int));
		cudaMalloc<int>(&INFO, batchSize * sizeof(int));
		int lda = n;
		float *A[] = {src};
		float ** A_d;
		cudaMalloc<float*>(&A_d, sizeof(A));
		cudaMemcpy(A_d, A, sizeof(A), cudaMemcpyHostToDevice);