Skip to content

GPU: Prevent CUDA 13 from breaking the GPUbenchmark #14706

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

Merged
mconcas merged 3 commits into from
Sep 30, 2025
Merged

GPU: Prevent CUDA 13 from breaking the GPUbenchmark #14706

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4cb6ffe
GPU: Prevent CUDA 13 from breaking the GPUbenchmark
mconcas Sep 30, 2025
8a00fbb
Please consider the following formatting changes (#14707)
alibuild Sep 30, 2025
cf6ce9a
Add cooperativeMultiDevice variable in Kernels.cu
mconcas Sep 30, 2025
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
33 changes: 27 additions & 6 deletions GPU/GPUbenchmark/cuda/Kernels.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -301,6 +301,21 @@ void printDeviceProp(int32_t deviceId)
cudaDeviceProp props;
GPUCHECK(cudaGetDeviceProperties(&props, deviceId));

int32_t clockRateKHz = 0;
int32_t memoryClockRateKHz = 0;
int32_t computeMode = 0;
int32_t cooperativeMultiDevice = 0;

#if (CUDART_VERSION >= 13000)
GPUCHECK(cudaDeviceGetAttribute(&clockRateKHz, cudaDevAttrClockRate, deviceId));
GPUCHECK(cudaDeviceGetAttribute(&memoryClockRateKHz, cudaDevAttrMemoryClockRate, deviceId));
GPUCHECK(cudaDeviceGetAttribute(&computeMode, cudaDevAttrComputeMode, deviceId));
#else
clockRateKHz = props.clockRate;
memoryClockRateKHz = props.memoryClockRate;
computeMode = props.computeMode;
cooperativeMultiDevice = props.cooperativeMultiDeviceLaunch;
#endif
std::cout << std::setw(w1) << "Name: " << props.name << std::endl;
std::cout << std::setw(w1) << "pciBusID: " << props.pciBusID << std::endl;
std::cout << std::setw(w1) << "pciDeviceID: " << props.pciDeviceID << std::endl;
Expand All @@ -309,11 +324,16 @@ void printDeviceProp(int32_t deviceId)
std::cout << std::setw(w1) << "maxThreadsPerMultiProcessor: " << props.maxThreadsPerMultiProcessor
<< std::endl;
std::cout << std::setw(w1) << "isMultiGpuBoard: " << props.isMultiGpuBoard << std::endl;
std::cout << std::setw(w1) << "clockRate: " << (float)props.clockRate / 1000.0 << " Mhz" << std::endl;
std::cout << std::setw(w1) << "memoryClockRate: " << (float)props.memoryClockRate / 1000.0 << " Mhz"

// Use the variables we populated above for the moved properties
std::cout << std::setw(w1) << "clockRate: " << (float)clockRateKHz / 1000.0 << " Mhz" << std::endl;
std::cout << std::setw(w1) << "memoryClockRate: " << (float)memoryClockRateKHz / 1000.0 << " Mhz"
<< std::endl;

std::cout << std::setw(w1) << "memoryBusWidth: " << props.memoryBusWidth << std::endl;
std::cout << std::setw(w1) << "clockInstructionRate: " << (float)props.clockRate / 1000.0

// clockInstructionRate is just another name for clockRate in this context
std::cout << std::setw(w1) << "clockInstructionRate: " << (float)clockRateKHz / 1000.0
<< " Mhz" << std::endl;
std::cout << std::setw(w1) << "totalGlobalMem: " << std::fixed << std::setprecision(2)
<< bytesToGB(props.totalGlobalMem) << " GB" << std::endl;
Expand All @@ -332,7 +352,10 @@ void printDeviceProp(int32_t deviceId)
std::cout << std::setw(w1) << "regsPerBlock: " << props.regsPerBlock << std::endl;
std::cout << std::setw(w1) << "warpSize: " << props.warpSize << std::endl;
std::cout << std::setw(w1) << "l2CacheSize: " << props.l2CacheSize << std::endl;
std::cout << std::setw(w1) << "computeMode: " << props.computeMode << std::endl;

// Use the variable for computeMode
std::cout << std::setw(w1) << "computeMode: " << computeMode << std::endl;

std::cout << std::setw(w1) << "maxThreadsPerBlock: " << props.maxThreadsPerBlock << std::endl;
std::cout << std::setw(w1) << "maxThreadsDim.x: " << props.maxThreadsDim[0] << std::endl;
std::cout << std::setw(w1) << "maxThreadsDim.y: " << props.maxThreadsDim[1] << std::endl;
Expand All @@ -343,8 +366,6 @@ void printDeviceProp(int32_t deviceId)
std::cout << std::setw(w1) << "major: " << props.major << std::endl;
std::cout << std::setw(w1) << "minor: " << props.minor << std::endl;
std::cout << std::setw(w1) << "concurrentKernels: " << props.concurrentKernels << std::endl;
std::cout << std::setw(w1) << "cooperativeLaunch: " << props.cooperativeLaunch << std::endl;
std::cout << std::setw(w1) << "cooperativeMultiDeviceLaunch: " << props.cooperativeMultiDeviceLaunch << std::endl;
#if defined(__HIPCC__)
std::cout << std::setw(w1) << "arch.hasGlobalInt32Atomics: " << props.arch.hasGlobalInt32Atomics << std::endl;
std::cout << std::setw(w1) << "arch.hasGlobalFloatAtomicExch: " << props.arch.hasGlobalFloatAtomicExch
Expand Down