OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

Remove remaining THC code (#69039)

Browse Source 
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/69039

Test Plan: Imported from OSS

Reviewed By: anjali411

Differential Revision: D32872476

Pulled By: ngimel

fbshipit-source-id: 7972aacc24aef9450fb59b707ed6396c501bcb31
This commit is contained in:
Peter Bell 2021-12-08 12:14:23 -08:00 committed by Facebook GitHub Bot
parent 7407e3d6fd
commit e279963eef
35 changed files with 20 additions and 289 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
BUILD.bazel
View File

@ -389,27 +389,6 @@ filegroup(
],
)
filegroup(
name = "thc_srcs_cu",
srcs = [
"aten/src/THC/THCReduceApplyUtils.cu.cc",
"aten/src/THC/THCSortUtils.cu.cc",
"aten/src/THC/THCTensor.cu.cc",
"aten/src/THC/THCTensorCopy.cu.cc",
"aten/src/THC/THCTensorMathScan.cu.cc",
"aten/src/THC/THCTensorScatterGather.cu.cc",
"aten/src/THC/THCTensorSort.cu.cc",
"aten/src/THC/generated/THCTensorSortByte.cu.cc",
"aten/src/THC/generated/THCTensorSortChar.cu.cc",
"aten/src/THC/generated/THCTensorSortDouble.cu.cc",
"aten/src/THC/generated/THCTensorSortFloat.cu.cc",
"aten/src/THC/generated/THCTensorSortHalf.cu.cc",
"aten/src/THC/generated/THCTensorSortInt.cu.cc",
"aten/src/THC/generated/THCTensorSortLong.cu.cc",
"aten/src/THC/generated/THCTensorSortShort.cu.cc",
],
)
filegroup(
name = "aten_srcs_cu",
srcs = [
@ -550,9 +529,7 @@ cc_library(
"aten/src/**/*.h",
"aten/src/**/*.hpp",
"aten/src/TH/**/*.cpp",
"aten/src/THC/**/*.cpp",
"aten/src/THC/*.cuh",
"aten/src/THC/generic/*.cu.cc",
],
exclude = [
"aten/src/ATen/Config.h",
@ -692,7 +669,6 @@ cu_library(
name = "aten_cuda",
srcs = [
":aten_srcs_cu",
":thc_srcs_cu",
],
copts = ATEN_COPTS + torch_cuda_half_options,
visibility = ["//visibility:public"],

3
CONTRIBUTING.md
View File

@ -243,9 +243,6 @@ into the repo directory.
* [aten](aten) - C++ tensor library for PyTorch (no autograd support)
* [src](aten/src) - [README](aten/src/README.md)
* [TH](aten/src/TH)
[THC](aten/src/THC) - Legacy library code from the original
Torch. Try not to add things here; we're slowly porting these to
[native](aten/src/ATen/native).
* generic - Contains actual implementations of operators,
parametrized over `scalar_t`. Files here get compiled N times
per supported scalar type in PyTorch.

4
aten/src/ATen/Context.cpp
View File

@ -20,9 +20,7 @@
namespace at {
Context::Context()
: thc_state(nullptr, [](THCState* p) { /* no-op */ }),
thh_state(nullptr, [](THHState* p) { /* no-op */ }) {}
Context::Context() = default;
// TODO: This could be bad juju if someone calls globalContext() in the
// destructor of an object with static lifetime.

19
aten/src/ATen/Context.h
View File

@ -91,28 +91,19 @@ class TORCH_API Context {
}
// defined in header so that getNonVariableType has ability to inline
// call_once check. getNonVariableType is called fairly frequently
THCState* lazyInitCUDA() {
void lazyInitCUDA() {
std::call_once(thc_init,[&] {
thc_state = detail::getCUDAHooks().initCUDA();
detail::getCUDAHooks().initCUDA();
});
return thc_state.get();
}
THHState* lazyInitHIP() {
void lazyInitHIP() {
std::call_once(thh_init,[&] {
thh_state = detail::getHIPHooks().initHIP();
detail::getHIPHooks().initHIP();
});
return thh_state.get();
}
static const at::cuda::NVRTC& getNVRTC() {
return detail::getCUDAHooks().nvrtc();
}
THCState* getTHCState() {
// AT_ASSERT(thc_state);
return thc_state.get();
}
THHState* getTHHState() {
return thh_state.get();
}
static bool setFlushDenormal(bool on);
@ -261,8 +252,6 @@ class TORCH_API Context {
#endif
bool display_vmap_fallback_warnings_ = false;
c10::optional<at::QEngine> quantized_engine = c10::nullopt;
std::unique_ptr<THCState, void(*)(THCState*)> thc_state;
std::unique_ptr<THHState, void(*)(THHState*)> thh_state;
Allocator* prev_allocator_ptr_{nullptr};
};

14
aten/src/ATen/cuda/detail/CUDAHooks.cpp
View File

@ -16,8 +16,6 @@
#include <c10/util/Exception.h>
#include <c10/cuda/CUDACachingAllocator.h>
#include <THC/THCGeneral.h> // For THCState
#if AT_CUDNN_ENABLED()
#include <ATen/cudnn/cudnn-wrapper.h>
#endif
@ -57,16 +55,8 @@ void set_magma_init_fn(void (*fn)()) {
// NB: deleter is dynamic, because we need it to live in a separate
// compilation unit (alt is to have another method in hooks, but
// let's not if we don't need to!)
std::unique_ptr<THCState, void (*)(THCState*)> CUDAHooks::initCUDA() const {
void CUDAHooks::initCUDA() const {
C10_LOG_API_USAGE_ONCE("aten.init.cuda");
// NOTE: THCState is now an empty struct but this pointer is passed
// to every THC function. So we can't remove it before the rest of THC.
auto thc_state = std::unique_ptr<THCState, void (*)(THCState*)>(
new THCState(),
[](THCState* p) {
delete p;
});
// Force the update to enable unit testing. This code get executed before unit tests
// have a chance to enable vitals.
at::vitals::VitalsAPI.setVital("CUDA", "used", "true", /* force = */ true);
@ -79,8 +69,6 @@ std::unique_ptr<THCState, void (*)(THCState*)> CUDAHooks::initCUDA() const {
TORCH_INTERNAL_ASSERT(magma_init_fn != nullptr, "Cannot initilaize magma, init routine not set");
magma_init_fn();
#endif
return thc_state;
}
const Generator& CUDAHooks::getDefaultCUDAGenerator(DeviceIndex device_index) const {

2
aten/src/ATen/cuda/detail/CUDAHooks.h
View File

@ -19,7 +19,7 @@ TORCH_CUDA_CPP_API c10::optional<int64_t> getDeviceIndexWithPrimaryContext();
// The real implementation of CUDAHooksInterface
struct CUDAHooks : public at::CUDAHooksInterface {
CUDAHooks(at::CUDAHooksArgs) {}
std::unique_ptr<THCState, void(*)(THCState*)> initCUDA() const override;
void initCUDA() const override;
Device getDeviceFromPtr(void* data) const override;
bool isPinnedPtr(void* data) const override;
const Generator& getDefaultCUDAGenerator(DeviceIndex device_index = -1) const override;

5
aten/src/ATen/detail/CUDAHooksInterface.h
View File

@ -10,9 +10,6 @@
#include <functional>
#include <memory>
// Forward-declares THCState
struct THCState;
// Forward-declares at::cuda::NVRTC
namespace at { namespace cuda {
struct NVRTC;
@ -73,7 +70,7 @@ struct TORCH_API CUDAHooksInterface {
virtual ~CUDAHooksInterface() {}
// Initialize THCState and, transitively, the CUDA state
virtual std::unique_ptr<THCState, void (*)(THCState*)> initCUDA() const {
virtual void initCUDA() const {
TORCH_CHECK(false, "Cannot initialize CUDA without ATen_cuda library. ", CUDA_HELP);
}

7
aten/src/ATen/detail/HIPHooksInterface.h
View File

@ -10,9 +10,6 @@
#include <functional>
#include <memory>
// Forward-declares THHState
struct THHState;
namespace at {
class Context;
}
@ -29,8 +26,8 @@ struct TORCH_API HIPHooksInterface {
// squelch -Werror=non-virtual-dtor
virtual ~HIPHooksInterface() {}
// Initialize THHState and, transitively, the HIP state
virtual std::unique_ptr<THHState, void (*)(THHState*)> initHIP() const {
// Initialize the HIP library state
virtual void initHIP() const {
AT_ERROR("Cannot initialize HIP without ATen_hip library.");
}

2
aten/src/ATen/miopen/Exceptions.h
View File

@ -5,8 +5,6 @@
#include <stdexcept>
#include <sstream>
struct THCState;
namespace at { namespace native {
class miopen_exception : public std::runtime_error {

1
aten/src/ATen/miopen/Utils.h
View File

@ -1,7 +1,6 @@
#pragma once
#include <ATen/ATen.h>
#include <aten/src/THH/THH.h>
#include <ATen/miopen/miopen-wrapper.h>
#include <ATen/miopen/Handle.h>

2
aten/src/ATen/native/miopen/Conv_miopen.cpp
View File

@ -114,7 +114,7 @@ std::tuple<at::Tensor,at::Tensor,at::Tensor> miopen_depthwise_convolution_backwa
#include <ATen/native/ConvUtils.h>
#include <c10/util/irange.h>
#include <c10/cuda/CUDACachingAllocator.h>
#include <c10/hip/HIPCachingAllocator.h>
#include <functional>
#include <iterator>

2
aten/src/ATen/native/miopen/RNN_miopen.cpp
View File

@ -37,8 +37,6 @@ namespace at { namespace native {
#else // AT_ROCM_ENABLED()
#include <aten/src/THH/THH.h>
#include <ATen/miopen/miopen-wrapper.h>
#include <ATen/miopen/Descriptors.h>
#include <ATen/miopen/Types.h>

16
aten/src/THC/CMakeLists.txt
View File

@ -2,22 +2,8 @@ set(ATen_CUDA_INCLUDE ${ATen_CUDA_INCLUDE}
"${CMAKE_CURRENT_SOURCE_DIR}"
PARENT_SCOPE)
set(ATen_CUDA_SRCS ${ATen_CUDA_SRCS}
${CMAKE_CURRENT_SOURCE_DIR}/THCTensor.cpp
PARENT_SCOPE)
install(FILES
THC.h
THCGeneral.h
THCGeneral.hpp
THCTensor.h
THCAtomics.cuh
THCDeviceUtils.cuh
THCGenerateByteType.h
# See Note [TH abstraction violation]
THCTensor.hpp
DESTINATION "${ATEN_INSTALL_INCLUDE_SUBDIR}/THC")
install(FILES
generic/THCTensor.cpp
generic/THCTensor.h
DESTINATION "${ATEN_INSTALL_INCLUDE_SUBDIR}/THC/generic")

8
aten/src/THC/THC.h
View File

@ -1,8 +0,0 @@
#ifndef THC_INC
#define THC_INC
#include <THC/THCGeneral.h>
#include <c10/cuda/CUDACachingAllocator.h>
#include <THC/THCTensor.h>
#endif

18
aten/src/THC/THCGeneral.h
View File

@ -1,18 +0,0 @@
#ifndef THC_GENERAL_INC
#define THC_GENERAL_INC
#include <TH/THGeneral.h>
#include <c10/core/Allocator.h>
#include <c10/cuda/CUDAStream.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <cublas_v2.h>
#include <cusparse.h>
/* Global state of THC. */
struct THCState {};
#endif

3
aten/src/THC/THCGeneral.hpp
View File

@ -1,3 +0,0 @@
#pragma once
#include <THC/THCGeneral.h>

14
aten/src/THC/THCTensor.cpp
View File

@ -1,14 +0,0 @@
#include <THC/THCGeneral.h>
#include <THC/THCTensor.hpp>
#include <new>
#include <THC/generic/THCTensor.cpp>
#include <THC/THCGenerateByteType.h>
void THCTensor_setStorage(THCState *state, THCTensor *self, c10::StorageImpl *storage_, ptrdiff_t storageOffset_, at::IntArrayRef size_, at::IntArrayRef stride_)
{
c10::raw::intrusive_ptr::incref(storage_);
THTensor_wrap(self).set_(at::Storage(c10::intrusive_ptr<at::StorageImpl>::reclaim(storage_)),
storageOffset_, size_, stride_);
}

18
aten/src/THC/THCTensor.h
View File

@ -1,18 +0,0 @@
#ifndef THC_TENSOR_INC
#define THC_TENSOR_INC
#include <TH/THTensor.h>
#include <THC/THCGeneral.h>
#define THCTensor_(NAME) TH_CONCAT_4(TH,CReal,Tensor_,NAME)
#define THC_DESC_BUFF_LEN 64
typedef struct TORCH_CUDA_CU_API THCDescBuff {
char str[THC_DESC_BUFF_LEN];
} THCDescBuff;
#include <THC/generic/THCTensor.h>
#include <THC/THCGenerateByteType.h>
#endif

18
aten/src/THC/THCTensor.hpp
View File

@ -1,18 +0,0 @@
#pragma once
// STOP!!! Thinking of including this header directly? Please
// read Note [TH abstraction violation]
#include <THC/THCTensor.h>
#include <TH/THTensor.hpp>
#include <THC/THCGeneral.hpp>
#include <ATen/ATen.h>
TORCH_CUDA_CU_API void THCTensor_setStorage(
THCState* state,
THCTensor* self,
c10::StorageImpl* storage_,
ptrdiff_t storageOffset_,
at::IntArrayRef size_,
at::IntArrayRef stride_);

24
aten/src/THC/generic/THCTensor.cpp
View File

@ -1,24 +0,0 @@
#ifndef THC_GENERIC_FILE
#define THC_GENERIC_FILE "THC/generic/THCTensor.cpp"
#else
#include <ATen/InferSize.h>
#include <ATen/NativeFunctions.h>
/**** creation methods ****/
THCTensor *THCTensor_(newWithStorage1d)(THCState *state, c10::StorageImpl *storage, ptrdiff_t storageOffset,
int64_t size0, int64_t stride0)
{
c10::raw::intrusive_ptr::incref(storage);
THTensor* self = c10::make_intrusive<at::TensorImpl, at::UndefinedTensorImpl>(
c10::intrusive_ptr<at::StorageImpl>::reclaim(storage),
at::DispatchKey::CUDA,
caffe2::TypeMeta::Make<scalar_t>())
.release();
THCTensor_setStorage(state, self, storage, storageOffset, {size0}, {stride0});
return self;
}
#endif

33
aten/src/THC/generic/THCTensor.h
View File

@ -1,33 +0,0 @@
#ifndef THC_GENERIC_FILE
#define THC_GENERIC_FILE "THC/generic/THCTensor.h"
#else
#define THCTensor THTensor
// These used to be distinct types; for some measure of backwards compatibility and documentation
// alias these to the single THCTensor type.
#define THCudaTensor THCTensor
#define THCudaDoubleTensor THCTensor
#define THCudaHalfTensor THCTensor
#define THCudaByteTensor THCTensor
#define THCudaCharTensor THCTensor
#define THCudaShortTensor THCTensor
#define THCudaIntTensor THCTensor
#define THCudaLongTensor THCTensor
#define THCudaBoolTensor THCTensor
#define THCudaBFloat16Tensor THCTensor
#define THCudaComplexFloatTensor THCTensor
#define THCudaComplexDoubleTensor THCTensor
/**** access methods ****/
TORCH_CUDA_CU_API c10::StorageImpl* THCTensor_(
storage)(THCState* state, const THCTensor* self);
/**** creation methods ****/
TORCH_CUDA_CU_API THCTensor* THCTensor_(newWithStorage1d)(
THCState* state,
c10::StorageImpl* storage_,
ptrdiff_t storageOffset_,
int64_t size0_,
int64_t stride0_);
#endif

37
docs/source/notes/windows.rst
View File

@ -63,11 +63,9 @@ Extension
CFFI Extension
^^^^^^^^^^^^^^
The support for CFFI Extension is very experimental. There're
generally two steps to enable it under Windows.
First, specify additional ``libraries`` in ``Extension``
object to make it build on Windows.
The support for CFFI Extension is very experimental. You must specify
additional ``libraries`` in ``Extension`` object to make it build on
Windows.
.. code-block:: python
@ -82,35 +80,6 @@ object to make it build on Windows.
libraries=['ATen', '_C'] # Append cuda libraries when necessary, like cudart
)
Second, here is a workground for "unresolved external symbol
state caused by ``extern THCState *state;``"
Change the source code from C to C++. An example is listed below.
.. code-block:: cpp
#include <THC/THC.h>
#include <ATen/ATen.h>
THCState *state = at::globalContext().thc_state;
extern "C" int my_lib_add_forward_cuda(THCudaTensor *input1, THCudaTensor *input2,
THCudaTensor *output)
{
if (!THCudaTensor_isSameSizeAs(state, input1, input2))
return 0;
THCudaTensor_resizeAs(state, output, input1);
THCudaTensor_cadd(state, output, input1, 1.0, input2);
return 1;
}
extern "C" int my_lib_add_backward_cuda(THCudaTensor *grad_output, THCudaTensor *grad_input)
{
THCudaTensor_resizeAs(state, grad_input, grad_output);
THCudaTensor_fill(state, grad_input, 1);
return 1;
}
Cpp Extension
^^^^^^^^^^^^^

5
test/cpp/c10d/ProcessGroupGlooAsyncTest.cpp
View File

@ -68,9 +68,9 @@ class AsyncInputIsOutputTest : public AsyncTest {
AsyncInputIsOutputTest(const std::string& path, int numTensors)
: AsyncTest(path),
numTensors_(numTensors),
numDevices_(cudaNumDevices()),
state_(::at::globalContext().lazyInitCUDA()) {
numDevices_(cudaNumDevices()) {
// Allocate inputs on available devices in a round robin fashion.
::at::globalContext().lazyInitCUDA();
inputs_.resize(numTensors_);
for (const auto i : c10::irange(numTensors_)) {
inputs_[i] = at::empty(
@ -121,7 +121,6 @@ class AsyncInputIsOutputTest : public AsyncTest {
protected:
const int numTensors_;
const int numDevices_;
THCState* state_;
std::vector<at::Tensor> inputs_;
std::vector<CUDAStream> streams_;
};

3
test/cpp/c10d/ProcessGroupNCCLTest.cpp
View File

@ -57,9 +57,9 @@ class NCCLTest : public NCCLTestBase {
NCCLTest(const std::string& path, int worldSize, std::chrono::milliseconds pgTimeout = kProcessGroupDefaultTimeout)
: NCCLTestBase(path, pgTimeout),
numDevices_(cudaNumDevices()),
state_(::at::globalContext().lazyInitCUDA()),
worldSize_(worldSize) {
// Each device has a single tensor to perf the NCCL op
::at::globalContext().lazyInitCUDA();
tensors_.resize(numDevices_);
inputs_.resize(numDevices_);
outputs_.resize(numDevices_);
@ -163,7 +163,6 @@ class NCCLTest : public NCCLTestBase {
}
const int numDevices_;
THCState* state_;
int worldSize_;
std::vector<at::Tensor> tensors_;
std::vector<std::vector<at::Tensor>> inputs_;

2
test/test_cpp_extensions_jit.py
View File

@ -422,8 +422,6 @@ class TestCppExtensionJIT(common.TestCase):
for the corresponding issue.
"""
cuda_source = """
#include <THC/THCGeneral.h>
template<typename T, typename U>
__global__ void half_test_kernel(const T* input, U* output) {
if (input[0] < input[1] || input[0] >= input[1]) {

4
torch/csrc/DynamicTypes.cpp
View File

@ -20,10 +20,6 @@
#include <unordered_map>
#include <vector>
#ifdef USE_CUDA
#include <THC/THC.h>
#endif
namespace torch {
namespace {
std::unordered_map<at::DeprecatedTypeProperties*, PyTypeObject*> attype_to_py_storage_type;

8
torch/csrc/cuda/Module.cpp
View File

@ -34,7 +34,6 @@
using namespace torch;
THCState *state = nullptr;
static bool in_bad_fork = false; // True for children forked after cuda init
#ifndef WIN32
@ -42,7 +41,6 @@ static bool in_bad_fork = false; // True for children forked after cuda init
static void forked_child() {
in_bad_fork = true;
torch::utils::set_run_yet_variable_to_false();
state = nullptr;
}
#endif
@ -522,7 +520,7 @@ static PyObject * THCPModule_initExtension(PyObject *self, PyObject *noargs)
HANDLE_TH_ERRORS
TORCH_INTERNAL_ASSERT(!in_bad_fork); // Handled at python level
poison_fork();
state = at::globalContext().lazyInitCUDA();
at::globalContext().lazyInitCUDA();
auto m = THPObjectPtr(PyImport_ImportModule("torch.cuda"));
if (!m) throw python_error();
@ -542,10 +540,6 @@ static PyObject * THCPModule_initExtension(PyObject *self, PyObject *noargs)
set_module_attr("has_magma", at::hasMAGMA() ? Py_True : Py_False);
set_module_attr("has_half", has_half ? Py_True : Py_False);
auto _state_cdata = THPObjectPtr(PyLong_FromVoidPtr(state));
if (!_state_cdata) throw python_error();
set_module_attr("_state_cdata", _state_cdata.get());
auto num_gpus = c10::cuda::device_count();
auto default_cuda_generators = PyTuple_New(static_cast<Py_ssize_t>(num_gpus));
for(const auto i : c10::irange(num_gpus)) {

4
torch/csrc/cuda/Module.h
View File

@ -1,10 +1,6 @@
#ifndef THCP_CUDA_MODULE_INC
#define THCP_CUDA_MODULE_INC
#include <ATen/Context.h>
extern THCState *state;
void THCPModule_setDevice(int idx);
PyObject * THCPModule_getDevice_wrap(PyObject *self);
PyObject * THCPModule_setDevice_wrap(PyObject *self, PyObject *arg);

1
torch/csrc/cuda/Storage.cpp
View File

@ -7,7 +7,6 @@
// See Note [TH abstraction violation]
// - Used to get at allocator from storage
#include <TH/THTensor.hpp>
#include <THC/THCTensor.hpp>
#include <torch/csrc/cuda/THCP.h>
#include <torch/csrc/cuda/override_macros.h>

1
torch/csrc/cuda/THCP.h
View File

@ -2,7 +2,6 @@
#define THCP_H
#include <torch/csrc/python_headers.h>
#include <torch/csrc/THP.h>
#include <torch/csrc/cuda/serialization.h>
#include <torch/csrc/cuda/Module.h>

2
torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp
View File

@ -9,8 +9,6 @@
#include <tuple>
#include <unordered_set>
#include <THC/THC.h>
#include <ATen/cuda/CUDAContext.h>
#include <c10/core/DeviceType.h>
#include <c10/cuda/CUDAGraphsC10Utils.h>

1
torch/csrc/generic/utils.h
View File

@ -11,7 +11,6 @@
struct THPStorage;
struct THSPTensor;
typedef class THPPointer<THWTensor> THWTensorPtr;
typedef class THPPointer<THPStorage> THPStoragePtr;
#if (!defined(THC_GENERIC_FILE)) && \

1
torch/csrc/jit/codegen/fuser/cuda/fused_kernel.cpp
View File

@ -7,7 +7,6 @@
#include <ATen/CUDAGeneratorImpl.h>
#include <ATen/cuda/CUDAContext.h>
#include <ATen/cuda/nvrtc_stub/ATenNVRTC.h>
#include <THC/THC.h>
#include <c10/cuda/CUDAGuard.h>
#include <torch/csrc/jit/resource_guard.h>

2
torch/csrc/utils.cpp
View File

@ -179,7 +179,7 @@ bool maybeThrowBackCompatKeepdimWarn(char *func) {
template<>
void THPPointer<THTensor>::free() {
if (ptr) {
THTensor_free(LIBRARY_STATE ptr);
c10::raw::intrusive_ptr::decref(ptr);
}
}

1
torch/csrc/utils.h
View File

@ -10,7 +10,6 @@
#include <torch/csrc/utils/python_compat.h>
#ifdef USE_CUDA
#include <THC/THC.h>
#include <c10/cuda/CUDAStream.h>
#endif