Get SM count with cudaDeviceGetAttribute in KernelHardwareInfo (#927)
This commit is contained in:
parent
df02482f1d
commit
fe2f491dd7
@ -56,15 +56,16 @@ struct KernelHardwareInfo {
|
|||||||
<< cudaGetErrorString(result));
|
<< cudaGetErrorString(result));
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
cudaDeviceProp properties;
|
int multiprocessor_count;
|
||||||
result = cudaGetDeviceProperties(&properties, device_id);
|
result = cudaDeviceGetAttribute(&multiprocessor_count,
|
||||||
|
cudaDevAttrMultiProcessorCount, device_id);
|
||||||
if (result != cudaSuccess) {
|
if (result != cudaSuccess) {
|
||||||
CUTLASS_TRACE_HOST(
|
CUTLASS_TRACE_HOST(
|
||||||
" cudaGetDeviceProperties() returned error "
|
" cudaDeviceGetAttribute() returned error "
|
||||||
<< cudaGetErrorString(result));
|
<< cudaGetErrorString(result));
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
return properties.multiProcessorCount;
|
return multiprocessor_count;
|
||||||
}
|
}
|
||||||
};
|
};
|
||||||
|
|
||||||
|
Loading…
Reference in New Issue
Block a user