OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 12:21:27 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

[2/N] Enable clang-tidy checks in torch/csrc/profiler (#113439)

Browse Source 
Fixes #ISSUE_NUMBER

Pull Request resolved: https://github.com/pytorch/pytorch/pull/113439
Approved by: https://github.com/Skylion007
This commit is contained in:
cyy 2023-11-14 00:39:54 +00:00 committed by PyTorch MergeBot
parent a43c757275
commit ff82dcd8fa
16 changed files with 30 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.lintrunner.toml
View File

@ -266,7 +266,6 @@ exclude_patterns = [
'torch/csrc/jit/serialization/export.cpp',
'torch/csrc/lazy/**/*',
'torch/csrc/onnx/init.cpp',
'torch/csrc/profiler/**/*',
'torch/csrc/quantized/**/*',
'torch/csrc/mps/**/*',
]

14
torch/csrc/profiler/collection.h
View File

@ -47,9 +47,9 @@ struct TORCH_API RawTensorMetadataBase {
explicit RawTensorMetadataBase(const at::Tensor& t);
StorageImplData data_;
c10::ScalarType dtype_;
c10::Layout layout_;
uint32_t dim_;
c10::ScalarType dtype_{c10::ScalarType::Undefined};
c10::Layout layout_{c10::Layout::Strided};
uint32_t dim_{0};
};
// Collected during profiling.
@ -64,8 +64,8 @@ struct TORCH_API RawTensorMetadata : RawTensorMetadataBase {
// Wrap `weak_self_` in `c10::optional` and split device into components to
// keep struct default constructable. (which the std::array initializer needs)
c10::optional<WeakTensor> weak_self_;
c10::DeviceType device_type_;
c10::DeviceIndex device_index_;
c10::DeviceType device_type_{c10::DeviceType::CPU};
c10::DeviceIndex device_index_{-1};
};
// Used during post processing.
@ -367,8 +367,8 @@ struct TORCH_API Result : public std::enable_shared_from_this<Result> {
template <typename T, typename Fn>
void visit_if_base(Fn&& fn) const {
visit([&](const auto& extra_fields) {
using extra_fields_t = typename std::remove_cv<
typename std::remove_reference<decltype(extra_fields)>::type>::type;
using extra_fields_t = typename std::remove_cv_t<
typename std::remove_reference_t<decltype(extra_fields)>>;
if constexpr (std::is_base_of_v<T, extra_fields_t>) {
fn(extra_fields);

3
torch/csrc/profiler/combined_traceback.cpp
View File

@ -90,6 +90,7 @@ SymbolizedTracebacks symbolize(
for (const auto& e : to_symbolize) {
if (e->python_) {
if (cur_python != e->python_ && !cur_py_frames.empty()) {
// NOLINTNEXTLINE(clang-analyzer-core.CallAndMessage)
cur_python->appendSymbolized(cur_py_frames, r);
cur_py_frames.clear();
}
@ -103,11 +104,13 @@ SymbolizedTracebacks symbolize(
}
}
if (!cur_py_frames.empty()) {
// NOLINTNEXTLINE(clang-analyzer-core.CallAndMessage)
cur_python->appendSymbolized(cur_py_frames, r);
cur_py_frames.clear();
}
std::vector<std::vector<uint64_t>> python_frame_fragments =
std::move(r.tracebacks);
r.tracebacks = {};
for (const auto& sc : to_symbolize) {
r.tracebacks.emplace_back();

2
torch/csrc/profiler/combined_traceback.h
View File

@ -29,7 +29,7 @@ struct TORCH_API CapturedTraceback : public c10::GatheredContext {
CapturedTraceback() = default;
CapturedTraceback(const CapturedTraceback&) = delete;
CapturedTraceback& operator=(const CapturedTraceback&) = delete;
~CapturedTraceback();
~CapturedTraceback() override;
using visitproc = int (*)(void* self, void* arg);

3
torch/csrc/profiler/containers.h
View File

@ -7,7 +7,6 @@
#include <forward_list>
#include <new>
#include <utility>
#include <vector>
#include <c10/macros/Macros.h>
#include <c10/util/ArrayRef.h>
@ -49,7 +48,7 @@ class AppendOnlyList {
public:
using array_t = block_t<T, ChunkSize>;
static_assert(
std::is_base_of<std::array<T, ChunkSize>, array_t>::value,
std::is_base_of_v<std::array<T, ChunkSize>, array_t>,
"AppendOnlyList expects raw low level pointer storage.");
static_assert(ChunkSize > 0, "Block cannot be empty.");

1
torch/csrc/profiler/data_flow.cpp
View File

@ -10,6 +10,7 @@ namespace impl {
namespace {
static constexpr TensorImplAddress NoTensorImpl{nullptr};
// NOLINTNEXTLINE(cppcoreguidelines-pro-type-member-init)
struct RawTensorInfo {
TensorImplAddress impl_;
StorageImplData storage_;

1
torch/csrc/profiler/events.h
View File

@ -1,6 +1,7 @@
#pragma once
#include <array>
#include <cstdint>
#include <cstring>
#include <vector>

1
torch/csrc/profiler/kineto_client_interface.cpp
View File

@ -1,7 +1,6 @@
#ifdef USE_KINETO
#include <libkineto.h>
#include <torch/csrc/autograd/profiler_kineto.h>
#include <cstdlib>
// Ondemand tracing is not supported on Apple or edge platform
#if defined(__APPLE__) || defined(EDGE_PROFILER_USE_KINETO)

2
torch/csrc/profiler/kineto_shim.cpp
View File

@ -1,8 +1,6 @@
#include <torch/csrc/profiler/collection.h>
#include <torch/csrc/profiler/kineto_shim.h>
#include <type_traits>
#ifdef USE_KINETO
#include <libkineto.h>
#endif

4
torch/csrc/profiler/orchestration/observer.cpp
View File

@ -94,8 +94,8 @@ ProfilerConfig ProfilerConfig::fromIValue(
// ----------------------------------------------------------------------------
// -- Profiler base class -----------------------------------------------------
// ----------------------------------------------------------------------------
/*explicit*/ ProfilerStateBase::ProfilerStateBase(const ProfilerConfig& config)
: c10::MemoryReportingInfoBase(), config_(config) {}
/*explicit*/ ProfilerStateBase::ProfilerStateBase(ProfilerConfig config)
: c10::MemoryReportingInfoBase(), config_(std::move(config)) {}
ProfilerStateBase::~ProfilerStateBase() {
if (handle_) {

2
torch/csrc/profiler/orchestration/observer.h
View File

@ -108,7 +108,7 @@ struct TORCH_API ProfilerConfig {
// -- Profiler base class -----------------------------------------------------
// ----------------------------------------------------------------------------
struct TORCH_API ProfilerStateBase : public c10::MemoryReportingInfoBase {
explicit ProfilerStateBase(const ProfilerConfig& config);
explicit ProfilerStateBase(ProfilerConfig config);
~ProfilerStateBase() override;
static ProfilerStateBase* get(bool global);

2
torch/csrc/profiler/python/combined_traceback.cpp
View File

@ -2,8 +2,6 @@
#include <torch/csrc/python_headers.h>
#include <torch/csrc/utils/pybind.h>
#include <torch/csrc/utils/pythoncapi_compat.h>
#include <iostream>
namespace py = pybind11;
namespace torch {

1
torch/csrc/profiler/python/init.cpp
View File

@ -56,6 +56,7 @@ PyTypeObject THPCapturedTracebackType = {
nullptr, /* tp_getattro */
nullptr, /* tp_setattro */
nullptr, /* tp_as_buffer */
// NOLINTNEXTLINE(misc-redundant-expression)
Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC, /* tp_flags */
nullptr, /* tp_doc */
(traverseproc)THPCapturedTraceback_traverse, /* tp_traverse */

2
torch/csrc/profiler/standalone/execution_trace_observer.cpp
View File

@ -10,7 +10,6 @@
#endif // _WIN32
#include <fmt/format.h>
#include <algorithm>
#include <chrono>
#include <cmath>
#include <fstream>
@ -19,7 +18,6 @@
#include <mutex>
#include <sstream>
#include <stack>
#include <stdexcept>
#include <vector>
#include <ATen/core/TensorBody.h>

7
torch/csrc/profiler/unwind/unwind.cpp
View File

@ -59,6 +59,7 @@ struct UpgradeExclusive {
}
private:
// NOLINTNEXTLINE(cppcoreguidelines-avoid-const-or-ref-data-members)
std::shared_lock<std::shared_timed_mutex>& rdlock_;
};
@ -97,6 +98,7 @@ struct LibraryInfo {
};
static const char* process_name() {
// NOLINTNEXTLINE(*-c-arrays*)
static char name[PATH_MAX + 1] = "";
if (*name == '\0') {
ssize_t len = readlink("/proc/self/exe", name, PATH_MAX);
@ -145,6 +147,7 @@ struct UnwindCache {
library_name = process_name();
}
auto eh_frame_hdr =
// NOLINTNEXTLINE(performance-no-int-to-ptr)
(void*)(segments[i].p_vaddr + info->dlpi_addr);
self->all_libraries_.emplace_back(
std::move(library_name),
@ -272,6 +275,7 @@ extern "C" void unwind_c(std::vector<void*>* result, int64_t rsp, int64_t rbp);
extern "C" void unwind_c(std::vector<void*>* result, int64_t rsp, int64_t rbp) {
std::shared_lock lock(cache_mutex_);
UnwindState state{};
// NOLINTNEXTLINE(performance-no-int-to-ptr)
state.rip = *(int64_t*)(rsp);
// +8 because we saved rsp after the return address was already pushed
// to the stack
@ -279,6 +283,7 @@ extern "C" void unwind_c(std::vector<void*>* result, int64_t rsp, int64_t rbp) {
state.rbp = rbp;
unwind_cache.checkRefresh(lock);
while (true) { // unwind for _start sets rip as being undefined
// NOLINTNEXTLINE(performance-no-int-to-ptr)
result->push_back((void*)state.rip);
const Unwinder& uw = unwind_cache.unwinderFor(state.rip, lock);
if (uw.terminator()) {
@ -355,6 +360,7 @@ struct Symbolizer {
auto& entry = getOrCreate(maybe_library->first);
entry.queried.push_back(addr);
auto libaddress = maybe_library->second - 1;
// NOLINTNEXTLINE(performance-no-int-to-ptr)
entry.comm->out() << (void*)libaddress << "\n";
// we need to make sure we don't write more than 64k bytes to
// a pipe before reading the results. Otherwise the buffer may
@ -393,6 +399,7 @@ struct Symbolizer {
Entry& getOrCreate(const std::string& name) {
auto it = entries_.find(name);
if (it == entries_.end()) {
// NOLINTNEXTLINE(*-c-arrays*)
const char* args[] = {
"addr2line", "-C", "-f", "-e", name.c_str(), nullptr};
it = entries_

7
torch/csrc/profiler/unwind/unwinder.h
View File

@ -1,7 +1,7 @@
#pragma once
#include <stdint.h>
#include <torch/csrc/profiler/unwind/action.h>
#include <torch/csrc/profiler/unwind/unwind_error.h>
#include <cstdint>
#include <limits>
struct UnwindState {
@ -54,10 +54,13 @@ struct Unwinder {
r.rsp = (reg_ == D_RSP ? cur.rsp : cur.rbp) + off_;
r.rbp = rbp_off_ == std::numeric_limits<int64_t>::max()
? cur.rbp
// NOLINTNEXTLINE(performance-no-int-to-ptr)
: *(int64_t*)(r.rsp + rbp_off_);
if (deref_) {
// NOLINTNEXTLINE(performance-no-int-to-ptr)
r.rsp = *(int64_t*)r.rsp;
}
// NOLINTNEXTLINE(performance-no-int-to-ptr)
r.rip = *(int64_t*)(r.rsp + rip_off_);
return r;
@ -70,5 +73,5 @@ struct Unwinder {
int64_t off_;
int64_t rip_off_;
int64_t rbp_off_;
bool deref_;
bool deref_{false};
};