allocator.h

//===-- include/flang/Runtime/CUDA/allocator.h ------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
 
#ifndef FORTRAN_RUNTIME_CUDA_ALLOCATOR_H_
#define FORTRAN_RUNTIME_CUDA_ALLOCATOR_H_
 
#include "common.h"
#include "flang/Runtime/descriptor-consts.h"
#include "flang/Runtime/entry-names.h"
 
#include "cuda_runtime.h"
 
namespace Fortran::runtime {
class Lock;
}
 
namespace Fortran::runtime::cuda {
 
extern "C" {
cudaStream_t RTDECL(CUFGetAssociatedStream)(void *);
int RTDECL(CUFSetAssociatedStream)(void *, cudaStream_t);
void RTDECL(CUFRegisterAllocator)();
}
 
void CUFResetStream(cudaStream_t stream);
 
extern Lock asyncDeviceAllocationTableLock;
 
int findAsyncDeviceAllocation(void *ptr);
void insertAsyncDeviceAllocation(
    void *ptr, std::size_t size, cudaStream_t stream);
cudaStream_t getAsyncDeviceAllocationStream(int pos);
void eraseAsyncDeviceAllocation(int pos);
 
void *CUFAllocPinned(std::size_t, std::size_t, std::int64_t *);
void CUFFreePinned(void *);
 
void *CUFAllocDevice(std::size_t, std::size_t, std::int64_t *);
void CUFFreeDevice(void *);
 
void *CUFAllocManaged(std::size_t, std::size_t, std::int64_t *);
void CUFFreeManaged(void *);
 
void *CUFAllocUnified(std::size_t, std::size_t, std::int64_t *);
void CUFFreeUnified(void *);
 
} // namespace Fortran::runtime::cuda
#endif // FORTRAN_RUNTIME_CUDA_ALLOCATOR_H_