triton_cpp v1.2.1
Header-only C++ wrapper for NVIDIA Triton Inference Server clients
Loading...
Searching...
No Matches
cuda_shm.hpp File Reference
#include <cstdint>
#include <stdexcept>
#include <string>
#include <cuda_runtime_api.h>

Go to the source code of this file.

Classes

class  triton_cpp::CudaSharedMemoryRegion
 Own a CUDA device allocation exportable through a CUDA IPC handle. More...
 

Namespaces

namespace  triton_cpp
 

Functions

void triton_cpp::throw_on_cuda_error (cudaError_t status, const char *operation)
 Throw a descriptive exception when a CUDA Runtime API call fails.
 
bool triton_cpp::LocalCudaSharedMemorySupported (std::string *reason=nullptr)
 Check whether this process can allocate CUDA IPC shared memory.