Thanks to visit codestin.com
Credit goes to github.com

Skip to content

some optimization for the main.cc code #3

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

some optimization for the main.cc code #3

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a924e74
Update main.cu
asimay Aug 11, 2023
9fc2507
Update main.cu
asimay Aug 11, 2023
05628c4
Update main.cu
asimay Aug 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions 06_MemoryBasics/src/main.cu
View file
Open in desktop
Copy link
Author

@asimay asimay Aug 14, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

sorry, maybe the line 76: cudaMemcpyToSymbol(&cFoo, &bar, sizeof(int)); modification is wrong.
please ignore this line.

Original file line number Diff line number Diff line change
Expand Up @@ -23,14 +23,14 @@ __global__ void WriteGlobalMemory(int* __restrict dOutPtr)
*dOutPtr = dFoo * dFoo;
}

__device__ void WriteAndPrintSharedMemory(int* sFoo)
__device__ void WriteAndPrintSharedMemory(int* __restrict sFoo)
{
// Write a computed result to shared memory for other threads to see
sFoo[threadIdx.x] = 42 * (threadIdx.x + 1);
// We make sure that no thread prints while the other still writes (parallelism!)
__syncwarp();
// Print own computed result and result by neighbor
printf("ThreadID: %d, sFoo[0]: %d, sFoo[1]: %d\n", threadIdx.x, sFoo[0], sFoo[1]);
printf("ThreadID: %d, sFoo[%d]: %d \n", threadIdx.x, threadIdx.x, sFoo[threadIdx.x]);
}

__global__ void WriteAndPrintSharedMemoryFixed()
Expand Down Expand Up @@ -73,7 +73,7 @@ int main()
GPU memory. Can be updated with cudaMemcpyToSymbol.
This syntax is unusual, but this is how it should be
*/
cudaMemcpyToSymbol(cFoo, &bar, sizeof(int));
cudaMemcpyToSymbol(&cFoo, &bar, sizeof(int));
ReadConstantMemory<<<1, 1>>>();
cudaDeviceSynchronize();

Expand Down Expand Up @@ -140,4 +140,4 @@ a syncwarp, so that other threads may fail to see it. You might need a block
size larger than 32 threads for this to happen and you may have to let the writing
thread do some "fake" work to delay its write to shared memory. Or it may work
immediately :) A solution should be provided by the following code sample.
*/
*/
6 changes: 3 additions & 3 deletions 08_Reductions/src/main.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -201,8 +201,8 @@ __global__ void reduceFinal(const float* __restrict input, int N)

__shared__ float data[BLOCK_SIZE];
// Already combine two values upon load from global memory.
data[threadIdx.x] = id < N / 2 ? input[id] : 0;
data[threadIdx.x] += id + N/2 < N ? input[id + N / 2] : 0;
data[threadIdx.x] = id < N ? input[id] : 0;
data[threadIdx.x] += (id + N < 2*N) ? input[id + N] : 0;

for (int s = blockDim.x / 2; s > 16; s /= 2)
{
Expand Down Expand Up @@ -312,4 +312,4 @@ Can you observe any difference in terms of speed / computed results?
2) Do you have any other ideas how the reduction could be improved?
Making it even faster should be quite challenging, but if you have
some suggestions, try them out and see how they affect performance!
*/
*/