The GPU is put into hibernation state after GPU processing is repeated at a predetermined interval.
After restarting the process, the processing speed of the GPU significantly slows down.
This phenomenon did not occur in previous CUDA versions.
I would like to know how to make the processing always fast even with newer CUDA versions.
The phenomenon can be generated by repeating the following simple process.
- Sleep for 5 milliseconds every cycle.
- Perform appropriate GPU processing.
- Perform a 5 second sleep on the 1000th loop.
*Repeat the above process.
“CUDA 7.0” and “CUDA 11.8” processing times were compared.
The graph is shown below.
#include "cuda_runtime.h"
#include "device_launch_parameters.h"
#include <stdio.h>
#include <Windows.h>
#include <thread>
cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size, double &processTime);
__global__ void addKernel(int *c, const int *a, const int *b)
{
int i = threadIdx.x;
for (int i = 0; i < 10000; i++) {
c[i] = a[i] + b[i];
}
}
int main()
{
cudaError_t cudaStatus;
{
// warm up
const int arraySize = 5;
const int a[arraySize] = { 1, 2, 3, 4, 5 };
const int b[arraySize] = { 10, 20, 30, 40, 50 };
int c[arraySize] = { 0 };
double processTime;
cudaError_t cudaStatus = addWithCuda(c, a, b, arraySize, processTime);
}
double processTime = 0;
int index = 0;
while (true) {
timeBeginPeriod(1);
Sleep(5);
const int arraySize = 5;
const int a[arraySize] = { 1, 2, 3, 4, 5 };
const int b[arraySize] = { 10, 20, 30, 40, 50 };
int c[arraySize] = { 0 };
cudaError_t cudaStatus = addWithCuda(c, a, b, arraySize, processTime);
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "addWithCuda failed!");
return 1;
}
printf("loop count : %d / processTime : %lf \n", index, processTime);
if (index == 999){
index = 0;
Sleep(5000);
}
else {
index++;
}
}
// cudaDeviceReset must be called before exiting in order for profiling and
// tracing tools such as Nsight and Visual Profiler to show complete traces.
cudaStatus = cudaDeviceReset();
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaDeviceReset failed!");
return 1;
}
return 0;
}
// Helper function for using CUDA to add vectors in parallel.
cudaError_t addWithCuda(int *c, const int *a, const int *b, unsigned int size, double &processTime)
{
LARGE_INTEGER m_liFreq;
LARGE_INTEGER m_start;
QueryPerformanceFrequency(&m_liFreq);
QueryPerformanceCounter(&m_start);
int *dev_a = 0;
int *dev_b = 0;
int *dev_c = 0;
cudaError_t cudaStatus;
// Choose which GPU to run on, change this on a multi-GPU system.
cudaStatus = cudaSetDevice(0);
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaSetDevice failed! Do you have a CUDA-capable GPU installed?");
goto Error;
}
// Allocate GPU buffers for three vectors (two input, one output) .
cudaStatus = cudaMalloc((void**)&dev_c, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMalloc failed!");
goto Error;
}
cudaStatus = cudaMalloc((void**)&dev_a, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMalloc failed!");
goto Error;
}
cudaStatus = cudaMalloc((void**)&dev_b, size * sizeof(int));
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMalloc failed!");
goto Error;
}
// Copy input vectors from host memory to GPU buffers.
cudaStatus = cudaMemcpy(dev_a, a, size * sizeof(int), cudaMemcpyHostToDevice);
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMemcpy failed!");
goto Error;
}
cudaStatus = cudaMemcpy(dev_b, b, size * sizeof(int), cudaMemcpyHostToDevice);
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMemcpy failed!");
goto Error;
}
// Launch a kernel on the GPU with one thread for each element.
addKernel<<<1, size>>>(dev_c, dev_a, dev_b);
// Check for any errors launching the kernel
cudaStatus = cudaGetLastError();
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus));
goto Error;
}
// cudaDeviceSynchronize waits for the kernel to finish, and returns
// any errors encountered during the launch.
cudaStatus = cudaDeviceSynchronize();
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaDeviceSynchronize returned error code %d after launching addKernel!\n", cudaStatus);
goto Error;
}
// Copy output vector from GPU buffer to host memory.
cudaStatus = cudaMemcpy(c, dev_c, size * sizeof(int), cudaMemcpyDeviceToHost);
if (cudaStatus != cudaSuccess) {
fprintf(stderr, "cudaMemcpy failed!");
goto Error;
}
Error:
cudaFree(dev_c);
cudaFree(dev_a);
cudaFree(dev_b);
LARGE_INTEGER m_end;
QueryPerformanceCounter(&m_end);
processTime = 1000.0 * (double)(m_end.QuadPart - m_start.QuadPart) / m_liFreq.QuadPart;
return cudaStatus;
}