I am learning cuda programming. I encounter a performance problem which I couldn’t explain. It is related to streaming.
my PC:
Intel(R) Core™ i7-7700K CPU @ 4.20GHz + GeForce GTX 1060 6GB (4095 MB)
Code 1(without streaming, synchronous memory copy) takes about 70ms
Code 2（with one streaming， asynchronous memory copy） takes about 23ms
What puzzle me is that Code 2 has only one stream, but the performance improves greatly compared to Code 1,
In my opinion, stream contribute to performace by pipeline, but there should be at least two streams.But why
one stream works in my case?
/Code 1 begin****************************************/
#include <stdio.h>
#include “cuda_runtime.h”
#include “device_launch_parameters.h”
#define HANDLE_ERROR( err ) (err)
#define N (10241024)
#define FULL_DATA_SIZE (N20)
global void kernel( int *a, int *b, int *c ) {
int idx = threadIdx.x + blockIdx.x * blockDim.x;
if (idx < N) {
int idx1 = (idx + 1) % 256;
int idx2 = (idx + 2) % 256;
float as = (a[idx] + a[idx1] + a[idx2]) / 3.0f;
float bs = (b[idx] + b[idx1] + b[idx2]) / 3.0f;
c[idx] = (as + bs) / 2;
}
}
int main( void ) {
cudaDeviceProp prop;
int whichDevice;
HANDLE_ERROR( cudaGetDevice( &whichDevice ) );
HANDLE_ERROR( cudaGetDeviceProperties( &prop, whichDevice ) );
if (!prop.deviceOverlap) {
printf( “Device will not handle overlaps, so no speed up from streams\n” );
return 0;
}
cudaEvent_t start, stop;
float elapsedTime;
cudaStream_t stream;
int *host_a, *host_b, *host_c;
int *dev_a, *dev_b, *dev_c;
// start the timers
HANDLE_ERROR( cudaEventCreate( &start ) );
HANDLE_ERROR( cudaEventCreate( &stop ) );
// initialize the stream
// HANDLE_ERROR( cudaStreamCreate( &stream ) );
// allocate the memory on the GPU
HANDLE_ERROR( cudaMalloc( (void**)&dev_a,
N * sizeof(int) ) );
HANDLE_ERROR( cudaMalloc( (void**)&dev_b,
N * sizeof(int) ) );
HANDLE_ERROR( cudaMalloc( (void**)&dev_c,
N * sizeof(int) ) );
// allocate host locked memory, used to stream
host_a= (int*)malloc( FULL_DATA_SIZE * sizeof(int) );
host_b = (int*)malloc(FULL_DATA_SIZE * sizeof(int) );
host_c = (int*)malloc(FULL_DATA_SIZE * sizeof(int) );
for (int i=0; i<FULL_DATA_SIZE; i++) {
host_a[i] = rand();
host_b[i] = rand();
}
HANDLE_ERROR( cudaEventRecord( start, 0 ) );
// now loop over full data, in bite-sized chunks
for (int i=0; i<FULL_DATA_SIZE; i+= N) {
// copy the locked memory to the device, async
HANDLE_ERROR( cudaMemcpy( dev_a, host_a+i,
N * sizeof(int),
cudaMemcpyHostToDevice) );
HANDLE_ERROR( cudaMemcpy( dev_b, host_b+i,
N * sizeof(int),
cudaMemcpyHostToDevice) );
kernel<<<N/256,256>>>( dev_a, dev_b, dev_c );
// copy the data from device to locked memory
HANDLE_ERROR( cudaMemcpy( host_c+i, dev_c,
N * sizeof(int),
cudaMemcpyDeviceToHost) );
}
// copy result chunk from locked to full buffer
//HANDLE_ERROR( cudaStreamSynchronize( stream ) );
HANDLE_ERROR( cudaEventRecord( stop, 0 ) );
HANDLE_ERROR( cudaEventSynchronize( stop ) );
HANDLE_ERROR( cudaEventElapsedTime( &elapsedTime,
start, stop ) );
printf( "Time taken: %3.1f ms\n", elapsedTime );
// cleanup the streams and memory
//HANDLE_ERROR( cudaFreeHost( host_a ) );
//HANDLE_ERROR( cudaFreeHost( host_b ) );
//HANDLE_ERROR( cudaFreeHost( host_c ) );
HANDLE_ERROR( cudaFree( dev_a ) );
HANDLE_ERROR( cudaFree( dev_b ) );
HANDLE_ERROR( cudaFree( dev_c ) );
//HANDLE_ERROR( cudaStreamDestroy( stream ) );
system("pause");
return 0;
}
/Code 1 end****************************************/
/Code 2 begin****************************************/
#include <stdio.h>
#include “cuda_runtime.h”
#include “device_launch_parameters.h”
#define HANDLE_ERROR( err ) (err)
#define N (10241024)
#define FULL_DATA_SIZE (N20)
global void kernel( int *a, int *b, int *c ) {
int idx = threadIdx.x + blockIdx.x * blockDim.x;
if (idx < N) {
int idx1 = (idx + 1) % 256;
int idx2 = (idx + 2) % 256;
float as = (a[idx] + a[idx1] + a[idx2]) / 3.0f;
float bs = (b[idx] + b[idx1] + b[idx2]) / 3.0f;
c[idx] = (as + bs) / 2;
}
}
int main( void ) {
cudaDeviceProp prop;
int whichDevice;
HANDLE_ERROR( cudaGetDevice( &whichDevice ) );
HANDLE_ERROR( cudaGetDeviceProperties( &prop, whichDevice ) );
if (!prop.deviceOverlap) {
printf( “Device will not handle overlaps, so no speed up from streams\n” );
return 0;
}
cudaEvent_t start, stop;
float elapsedTime;
cudaStream_t stream;
int *host_a, *host_b, *host_c;
int *dev_a, *dev_b, *dev_c;
// start the timers
HANDLE_ERROR( cudaEventCreate( &start ) );
HANDLE_ERROR( cudaEventCreate( &stop ) );
// initialize the stream
HANDLE_ERROR( cudaStreamCreate( &stream ) );
// allocate the memory on the GPU
HANDLE_ERROR( cudaMalloc( (void**)&dev_a,
N * sizeof(int) ) );
HANDLE_ERROR( cudaMalloc( (void**)&dev_b,
N * sizeof(int) ) );
HANDLE_ERROR( cudaMalloc( (void**)&dev_c,
N * sizeof(int) ) );
// allocate host locked memory, used to stream
HANDLE_ERROR( cudaHostAlloc( (void**)&host_a,
FULL_DATA_SIZE * sizeof(int),
cudaHostAllocDefault ) );
HANDLE_ERROR( cudaHostAlloc( (void**)&host_b,
FULL_DATA_SIZE * sizeof(int),
cudaHostAllocDefault ) );
HANDLE_ERROR( cudaHostAlloc( (void**)&host_c,
FULL_DATA_SIZE * sizeof(int),
cudaHostAllocDefault ) );
for (int i=0; i<FULL_DATA_SIZE; i++) {
host_a[i] = rand();
host_b[i] = rand();
}
HANDLE_ERROR( cudaEventRecord( start, 0 ) );
// now loop over full data, in bite-sized chunks
for (int i=0; i<FULL_DATA_SIZE; i+= N) {
// copy the locked memory to the device, async
HANDLE_ERROR( cudaMemcpyAsync( dev_a, host_a+i,
N * sizeof(int),
cudaMemcpyHostToDevice,
stream ) );
HANDLE_ERROR( cudaMemcpyAsync( dev_b, host_b+i,
N * sizeof(int),
cudaMemcpyHostToDevice,
stream ) );
kernel<<<N/256,256,0,stream>>>( dev_a, dev_b, dev_c );
// copy the data from device to locked memory
HANDLE_ERROR( cudaMemcpyAsync( host_c+i, dev_c,
N * sizeof(int),
cudaMemcpyDeviceToHost,
stream ) );
}
// copy result chunk from locked to full buffer
HANDLE_ERROR( cudaStreamSynchronize( stream ) );
HANDLE_ERROR( cudaEventRecord( stop, 0 ) );
HANDLE_ERROR( cudaEventSynchronize( stop ) );
HANDLE_ERROR( cudaEventElapsedTime( &elapsedTime,
start, stop ) );
printf( "Time taken: %3.1f ms\n", elapsedTime );
// cleanup the streams and memory
HANDLE_ERROR( cudaFreeHost( host_a ) );
HANDLE_ERROR( cudaFreeHost( host_b ) );
HANDLE_ERROR( cudaFreeHost( host_c ) );
HANDLE_ERROR( cudaFree( dev_a ) );
HANDLE_ERROR( cudaFree( dev_b ) );
HANDLE_ERROR( cudaFree( dev_c ) );
HANDLE_ERROR( cudaStreamDestroy( stream ) );
system("pause");
return 0;
}
/Code 2 end****************************************/