gpu_vector.hpp File Reference

Simple GPU memory container with RAII management. More...

#include <cstddef>
#include <stdexcept>
#include <vector>

Include dependency graph for gpu_vector.hpp:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
class	pfc::gpu::GPUVector< T >
	Simple GPU memory container (RAII) More...

Macros
#define	PFC_GPU_CUDA_AVAILABLE   0

Detailed Description

Simple GPU memory container with RAII management.

GPUVector provides a simple, RAII-style container for GPU memory that automatically manages CUDA memory allocation and deallocation. It's designed to be similar to std::vector in interface, but memory lives on the GPU.

Key features:

• Automatic memory management (RAII)
• Fixed size (no resizing)
• CPU-GPU transfer methods
• Move semantics supported
• Copy semantics disabled (would require deep copy)

This header must compile on systems without CUDA (e.g., AMD systems). CUDA-specific code is guarded with #ifdef OpenPFC_ENABLE_CUDA.

#ifdef OpenPFC_ENABLE_CUDA
    pfc::gpu::GPUVector<double> vec(100);
    std::vector<double> host_data(100, 1.0);
    vec.copy_from_host(host_data);
    // ... use GPU memory ...
    std::vector<double> result = vec.to_host();
#endif

See also

gpu/kernels_simple.hpp for GPU kernel operations

tungsten-gpu-implementation-plan.md for usage examples

Author

OpenPFC Development Team

Date

2025