mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 00:21:07 +01:00
03c72976a5
This PR refactors some ref-counting functionality out of `beginAllocateToPool` and `releasePool`. The ref-counting logic is then used in construction and destruction of `torch.cuda.MemPool`. The `use_count` variable in the CUDACachingAllocator is essentially a refcount of how many context managers are using the pool. Since we are now lifting up the MemPool abstraction to the user, the MemPool object itself now needs to hold a an extra reference as well. Part of https://github.com/pytorch/pytorch/issues/124807. Pull Request resolved: https://github.com/pytorch/pytorch/pull/133600 Approved by: https://github.com/eqy, https://github.com/ezyang
30 lines
1.2 KiB
C++
30 lines
1.2 KiB
C++
#include <torch/csrc/python_headers.h>
|
|
|
|
#include <torch/csrc/jit/python/pybind_utils.h>
|
|
#include <torch/csrc/utils/device_lazy_init.h>
|
|
#include <torch/csrc/utils/pybind.h>
|
|
|
|
#include <c10/cuda/CUDACachingAllocator.h>
|
|
|
|
template <typename T>
|
|
using shared_ptr_class_ = py::class_<T, std::shared_ptr<T>>;
|
|
|
|
void THCPMemPool_init(PyObject* module) {
|
|
auto torch_C_m = py::handle(module).cast<py::module>();
|
|
shared_ptr_class_<::c10::cuda::MemPool>(torch_C_m, "_MemPool")
|
|
.def(
|
|
py::init([](c10::cuda::CUDACachingAllocator::CUDAAllocator* allocator,
|
|
bool is_user_created) {
|
|
torch::utils::device_lazy_init(at::kCUDA);
|
|
return std::make_shared<::c10::cuda::MemPool>(
|
|
allocator, is_user_created);
|
|
}))
|
|
.def_property_readonly("id", &::c10::cuda::MemPool::id)
|
|
.def_property_readonly("allocator", &::c10::cuda::MemPool::allocator)
|
|
.def("use_count", &::c10::cuda::MemPool::use_count);
|
|
shared_ptr_class_<::c10::cuda::MemPoolContext>(torch_C_m, "_MemPoolContext")
|
|
.def(py::init<c10::cuda::MemPool*>())
|
|
.def_static(
|
|
"active_pool", &::c10::cuda::MemPoolContext::getActiveMemPool);
|
|
}
|