NASP/cuda-samples
85
0
Fork 0
mirror of https://github.com/NVIDIA/cuda-samples.git synced 2026-10-25 22:28:17 +08:00
Code Issues Packages Projects Releases Wiki Activity

Apply consistent code formatting across the repo. Add clang-format and pre-commit hooks.

Browse Source
This commit is contained in:
Rob Armstrong
2025-03-27 10:30:07 -07:00
parent 2cd58fbc9a
commit ceab6e8bcc
782 changed files with 107230 additions and 106548 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
Samples/1_Utilities/topologyQuery/README.md
View File

@@ -30,4 +30,3 @@ cudaGetDeviceCount, cudaDeviceGetAttribute
Download and install the [CUDA Toolkit 12.5](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
## References (for more details)

68
Samples/1_Utilities/topologyQuery/topologyQuery.cu
View File

@@ -35,48 +35,44 @@
// includes, project
#include <helper_cuda.h>
#include <helper_functions.h> // helper for shared that are common to CUDA Samples
#include <helper_functions.h> // helper for shared that are common to CUDA Samples
int main(int argc, char **argv) {
int deviceCount = 0;
checkCudaErrors(cudaGetDeviceCount(&deviceCount));
int main(int argc, char **argv)
{
int deviceCount = 0;
checkCudaErrors(cudaGetDeviceCount(&deviceCount));
// Enumerates Device <-> Device links
for (int device1 = 0; device1 < deviceCount; device1++) {
for (int device2 = 0; device2 < deviceCount; device2++) {
if (device1 == device2) continue;
// Enumerates Device <-> Device links
for (int device1 = 0; device1 < deviceCount; device1++) {
for (int device2 = 0; device2 < deviceCount; device2++) {
if (device1 == device2)
continue;
int perfRank = 0;
int atomicSupported = 0;
int accessSupported = 0;
int perfRank = 0;
int atomicSupported = 0;
int accessSupported = 0;
checkCudaErrors(cudaDeviceGetP2PAttribute(
&accessSupported, cudaDevP2PAttrAccessSupported, device1, device2));
checkCudaErrors(cudaDeviceGetP2PAttribute(
&perfRank, cudaDevP2PAttrPerformanceRank, device1, device2));
checkCudaErrors(cudaDeviceGetP2PAttribute(
&atomicSupported, cudaDevP2PAttrNativeAtomicSupported, device1,
device2));
checkCudaErrors(
cudaDeviceGetP2PAttribute(&accessSupported, cudaDevP2PAttrAccessSupported, device1, device2));
checkCudaErrors(cudaDeviceGetP2PAttribute(&perfRank, cudaDevP2PAttrPerformanceRank, device1, device2));
checkCudaErrors(
cudaDeviceGetP2PAttribute(&atomicSupported, cudaDevP2PAttrNativeAtomicSupported, device1, device2));
if (accessSupported) {
std::cout << "GPU" << device1 << " <-> GPU" << device2 << ":"
<< std::endl;
std::cout << " * Atomic Supported: "
<< (atomicSupported ? "yes" : "no") << std::endl;
std::cout << " * Perf Rank: " << perfRank << std::endl;
}
if (accessSupported) {
std::cout << "GPU" << device1 << " <-> GPU" << device2 << ":" << std::endl;
std::cout << " * Atomic Supported: " << (atomicSupported ? "yes" : "no") << std::endl;
std::cout << " * Perf Rank: " << perfRank << std::endl;
}
}
}
}
// Enumerates Device <-> Host links
for (int device = 0; device < deviceCount; device++) {
int atomicSupported = 0;
checkCudaErrors(cudaDeviceGetAttribute(
&atomicSupported, cudaDevAttrHostNativeAtomicSupported, device));
std::cout << "GPU" << device << " <-> CPU:" << std::endl;
std::cout << " * Atomic Supported: " << (atomicSupported ? "yes" : "no")
<< std::endl;
}
// Enumerates Device <-> Host links
for (int device = 0; device < deviceCount; device++) {
int atomicSupported = 0;
checkCudaErrors(cudaDeviceGetAttribute(&atomicSupported, cudaDevAttrHostNativeAtomicSupported, device));
std::cout << "GPU" << device << " <-> CPU:" << std::endl;
std::cout << " * Atomic Supported: " << (atomicSupported ? "yes" : "no") << std::endl;
}
return 0;
return 0;
}