1
0
Fork 0

Finished doing real safe pointer type for built_in types

This commit is contained in:
Pcornat 2025-12-13 12:34:29 +01:00
commit 81d04d6332
Signed by: Pcornat
GPG key ID: E0326CC678A00BDD

View file

@ -6,18 +6,42 @@
#define RAIISAFECUDA_MALLOC_UNMANAGED_HPP
#include <concepts>
#include <cstdint>
#include <memory>
#include <variant>
#include <cuda_runtime_api.h>
namespace safe_cuda::unmanaged {
using deviceMallocDestroyType = decltype(&cudaFree);
using hostMallocDestroyType = decltype(&cudaFreeHost);
template<typename D>
concept CudaArrayDestroyer = std::is_same_v<D, deviceMallocDestroyType> || std::is_same_v<D, hostMallocDestroyType>;
namespace safe_cuda {
enum class allocType : std::uint8_t {
Unmanaged = 0,
Managed = 1,
Host = 2,
};
template<typename T, CudaArrayDestroyer D> requires std::integral<T> || std::floating_point<T>
template<typename T, allocType alloc_type = allocType::Managed>
struct destroyType {
void operator()(T *ptr) const noexcept {
(void) cudaFree(ptr);
}
};
template<typename T>
struct destroyType<T, allocType::Unmanaged> {
void operator()(T *ptr) const noexcept {
(void) cudaFree(ptr);
}
};
template<typename T>
struct destroyType<T, allocType::Host> {
void operator()(T *ptr) const noexcept {
(void) cudaFreeHost(ptr);
}
};
template<typename T, allocType alloc_type, typename D = destroyType<T, alloc_type> >
requires std::integral<T> || std::floating_point<T>
using safePtrType = std::unique_ptr<T, D>;
/**
@ -25,25 +49,26 @@ namespace safe_cuda::unmanaged {
*
* It can allocate unmanaged memory on device and on Host for pinned memory.
* \tparam T bare and built-in type.
* \tparam D destroyer type. It determines if it allocates on device or pinned host.
* \tparam alloc_type Type of allocation: Managed (default), Unmanage, Host.
* \param byteDataSize
* \return
*/
template<typename T, CudaArrayDestroyer D>
std::variant<safePtrType<T, D>, cudaError_t> cuda_malloc(const std::size_t byteDataSize) noexcept {
template<typename T, allocType alloc_type>
std::pair<safePtrType<T, alloc_type>, cudaError_t>
cuda_malloc(const std::size_t byteDataSize) noexcept {
T *ptr_tmp = nullptr;
if constexpr (std::is_same_v<D, deviceMallocDestroyType>) {
const cudaError_t error = cudaMalloc(&ptr_tmp, byteDataSize);
if (error != cudaSuccess) {
return error;
}
} else if constexpr (std::is_same_v<D, hostMallocDestroyType>) {
const cudaError_t error = cudaMallocHost(&ptr_tmp, byteDataSize);
if (error != cudaSuccess) {
return error;
}
cudaError_t error = cudaSuccess;
switch (alloc_type) {
case allocType::Unmanaged:
error = cudaMalloc(reinterpret_cast<void **>(&ptr_tmp), byteDataSize);
return { safePtrType<T, alloc_type>{ ptr_tmp, destroyType<T, alloc_type>{} }, error };
case allocType::Host:
error = cudaMallocHost(reinterpret_cast<void **>(&ptr_tmp), byteDataSize);
return { safePtrType<T, alloc_type>{ ptr_tmp, destroyType<T, alloc_type>{} }, error };
case allocType::Managed:
error = cudaMallocManaged(reinterpret_cast<void **>(&ptr_tmp), byteDataSize);
return { safePtrType<T, alloc_type>{ ptr_tmp, destroyType<T, alloc_type>{} }, error };
}
return safePtrType<T, D>{ ptr_tmp, cudaFree };
}
}