allocator.hpp File Reference

#include <cassert>
#include <cstddef>
#include <iostream>
#include <type_traits>
#include <inttypes.h>
#include "utility.hpp"

Go to the source code of this file.

Classes
class	battery::pool_allocator
class	battery::standard_allocator
class	battery::statistics_allocator< Allocator, InternalAllocator >

Namespaces
namespace	battery

Functions
CUDA void *	operator new (size_t bytes, battery::pool_allocator &p)
CUDA void	operator delete (void *ptr, battery::pool_allocator &p)
CUDA void *	operator new (size_t bytes, battery::standard_allocator &p)
CUDA void	operator delete (void *ptr, battery::standard_allocator &p)

Detailed Description

We provide several allocators compatible with the data structures provided by this library. The allocators are aimed to be used to distinguish in which memory (shared, global, managed or the "standard" C++ memory) we should allocate data. This allows us to provide uniform interfaces for both host (C++) and device (CUDA) code.

As a general comment, be careful to always deallocate the memory from the side you allocated it, e.g., do not allocate on the host then try to deallocate it on the device. To avoid these kind of mistakes, you should use battery::shared_ptr when passing data to a CUDA kernel, see the manual for examples.

Function Documentation

operator new() [1/2]

CUDA void * operator new ( size_t bytes, battery::pool_allocator & p )

inline

operator delete() [1/2]

CUDA void operator delete ( void * ptr, battery::pool_allocator & p )

inline

operator new() [2/2]

CUDA void * operator new ( size_t bytes, battery::standard_allocator & p )

inline

operator delete() [2/2]

CUDA void operator delete ( void * ptr, battery::standard_allocator & p )

inline