如果您对性能感兴趣(通常是人们使用 GPU 执行计算任务的原因),则不应使用推力,也不应调用或编写自己的 CUDA 内核。您应该使用 CUBLAS 库。学习练习,如果想学习自己的CUDA内核,可以参考a first-level-optimized CUDA version in the CUDA programming guide in the shared memory section。如果你真的想在一次推力调用中使用推力,这是可能的。
基本思想是使用元素方面的操作,如 here 所述的推力::变换。每个输出数组元素的点积是使用由循环组成的函子计算的。
这是一个考虑 3 种方法的工作示例。您原来的双嵌套循环方法(相对较慢)、单推力调用方法(更快)和 cublas 方法(最快,当然对于较大的矩阵大小)。下面的代码只对 200 或更小的矩阵边尺寸运行方法 1,因为它太慢了。以下是 Tesla P100 的示例:
$ cat t463.cu
#include <thrust/device_vector.h>
#include <thrust/transform.h>
#include <thrust/inner_product.h>
#include <thrust/execution_policy.h>
#include <thrust/equal.h>
#include <thrust/iterator/constant_iterator.h>
#include <cublas_v2.h>
#include <iostream>
#include <time.h>
#include <sys/time.h>
#include <cstdlib>
#define USECPSEC 1000000ULL
long long dtime_usec(unsigned long long start){
timeval tv;
gettimeofday(&tv, 0);
return ((tv.tv_sec*USECPSEC)+tv.tv_usec)-start;
}
struct dp
{
float *A, *B;
int m,n,r;
dp(float *_A, float *_B, int _m, int _n, int _r): A(_A), B(_B), m(_m), n(_n), r(_r) {};
__host__ __device__
float operator()(size_t idx){
float sum = 0.0f;
int row = idx/r;
int col = idx - (row*r); // cheaper modulo
for (int i = 0; i < m; i++)
sum += A[col + row*i] * B[col + row*i];
return sum;}
};
const int dsd = 200;
int main(int argc, char *argv[]){
int ds = dsd;
if (argc > 1) ds = atoi(argv[1]);
const int n = ds;
const int m = ds;
const int r = ds;
// data setup
thrust::device_vector<float> data(n*m,1);
thrust::device_vector<float> other(m*r,1);
thrust::device_vector<float> result(n*r,0);
// method 1
//let's pretend that other is (already) transposed for efficient memory access by thrust
// therefore each dot-product is formed using a row of data and a row of other
long long dt = dtime_usec(0);
if (ds < 201){
for (int i = 0; i < n; ++i)
{
for (int j = 0; j < r;++j)
{
result[i*r+ j] = thrust::inner_product(data.begin()+(i*m), data.begin()+((i+1)*m),other.begin()+(j*m), 0.0f);
}
}
cudaDeviceSynchronize();
dt = dtime_usec(dt);
if (thrust::equal(result.begin(), result.end(), thrust::constant_iterator<float>(m)))
std::cout << "method 1 time: " << dt/(float)USECPSEC << "s" << std::endl;
else
std::cout << "method 1 failure" << std::endl;
}
thrust::fill(result.begin(), result.end(), 0);
cudaDeviceSynchronize();
// method 2
//let's pretend that data is (already) transposed for efficient memory access by thrust
// therefore each dot-product is formed using a column of data and a column of other
dt = dtime_usec(0);
thrust::transform(thrust::counting_iterator<int>(0), thrust::counting_iterator<int>(n*r), result.begin(), dp(thrust::raw_pointer_cast(data.data()), thrust::raw_pointer_cast(other.data()), m, n, r));
cudaDeviceSynchronize();
dt = dtime_usec(dt);
if (thrust::equal(result.begin(), result.end(), thrust::constant_iterator<float>(m)))
std::cout << "method 2 time: " << dt/(float)USECPSEC << "s" << std::endl;
else
std::cout << "method 2 failure" << std::endl;
// method 3
// once again, let's pretend the data is ready to go for CUBLAS
cublasHandle_t h;
cublasCreate(&h);
thrust::fill(result.begin(), result.end(), 0);
float alpha = 1.0f;
float beta = 0.0f;
cudaDeviceSynchronize();
dt = dtime_usec(0);
cublasSgemm(h, CUBLAS_OP_T, CUBLAS_OP_T, n, r, m, &alpha, thrust::raw_pointer_cast(data.data()), n, thrust::raw_pointer_cast(other.data()), m, &beta, thrust::raw_pointer_cast(result.data()), n);
cudaDeviceSynchronize();
dt = dtime_usec(dt);
if (thrust::equal(result.begin(), result.end(), thrust::constant_iterator<float>(m)))
std::cout << "method 3 time: " << dt/(float)USECPSEC << "s" << std::endl;
else
std::cout << "method 3 failure" << std::endl;
}
$ nvcc -o t463 t463.cu -lcublas
$ ./t463
method 1 time: 20.1648s
method 2 time: 6.3e-05s
method 3 time: 5.7e-05s
$ ./t463 1024
method 2 time: 0.008063s
method 3 time: 0.000458s
$
对于默认维度 200 的情况,单推力调用和 cublas 方法相当接近,但比循环方法快得多。对于 1024 的边尺寸,cublas 方法比单推力调用方法快近 20 倍。
请注意,我为所有 3 种方法选择了“最佳”转置配置。对于方法 1,最佳情况下的时间是 inner_product 使用每个输入矩阵中的“行”(实际上是第二个输入矩阵的转置)。对于方法 2,最佳情况下的时间是当函子从每个输入矩阵(实际上是第一个输入矩阵的转置)中遍历一个“列”时。对于方法 3,为两个输入矩阵选择 CUBLAS_OP_T 似乎是最快的。实际上,只有 cublas 方法具有灵活性,可用于各种输入情况并具有良好的性能。