mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
4cb534f92e
Summary:
This is an automatic change generated by the following script:
```
#!/usr/bin/env python3
from subprocess import check_output, check_call
import os
def get_compiled_files_list():
import json
with open("build/compile_commands.json") as f:
data = json.load(f)
files = [os.path.relpath(node['file']) for node in data]
for idx, fname in enumerate(files):
if fname.startswith('build/') and fname.endswith('.DEFAULT.cpp'):
files[idx] = fname[len('build/'):-len('.DEFAULT.cpp')]
return files
def run_clang_tidy(fname):
check_call(["python3", "tools/clang_tidy.py", "-c", "build", "-x", fname,"-s"])
changes = check_output(["git", "ls-files", "-m"])
if len(changes) == 0:
return
check_call(["git", "commit","--all", "-m", f"NOLINT stubs for {fname}"])
def main():
git_files = check_output(["git", "ls-files"]).decode("ascii").split("\n")
compiled_files = get_compiled_files_list()
for idx, fname in enumerate(git_files):
if fname not in compiled_files:
continue
if fname.startswith("caffe2/contrib/aten/"):
continue
print(f"[{idx}/{len(git_files)}] Processing {fname}")
run_clang_tidy(fname)
if __name__ == "__main__":
main()
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/56892
Reviewed By: H-Huang
Differential Revision: D27991944
Pulled By: malfet
fbshipit-source-id: 5415e1eb2c1b34319a4f03024bfaa087007d7179
54 lines
1.8 KiB
C++
54 lines
1.8 KiB
C++
#include "caffe2/operators/async_net_barrier_op.h"
|
|
|
|
namespace caffe2 {
|
|
|
|
namespace {
|
|
std::pair<std::vector<DeviceOption>, std::vector<DeviceOption>>
|
|
asyncBarrierOpDevInfer(const OperatorDef& def) {
|
|
auto op_device =
|
|
def.has_device_option() ? def.device_option() : DeviceOption();
|
|
ArgumentHelper helper(def);
|
|
auto cross_device = helper.GetSingleArgument<int>("cross_device", 0);
|
|
std::vector<DeviceOption> opt;
|
|
for (int i = 0; i < def.input().size(); ++i) {
|
|
if (cross_device == 1) {
|
|
DeviceOption dev;
|
|
dev.set_device_type(op_device.device_type());
|
|
dev.set_device_id(i);
|
|
opt.push_back(dev);
|
|
} else {
|
|
opt.push_back(op_device);
|
|
}
|
|
}
|
|
return std::make_pair(opt, opt);
|
|
}
|
|
}
|
|
|
|
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
|
|
OPERATOR_SCHEMA(AsyncNetBarrier)
|
|
.NumInputs(1, INT_MAX)
|
|
.NumOutputs(1, INT_MAX)
|
|
.IdenticalTypeAndShape()
|
|
.InputsCanCrossDevices()
|
|
.AllowOneToOneInplace()
|
|
.DeviceInferenceFunction(asyncBarrierOpDevInfer)
|
|
.SetDoc(R"DOC(
|
|
This is a pretty much no-op operator, since it's only purposes is make sure that
|
|
async_scheduling will schedule certian operations earlier than others.
|
|
|
|
Exaple where this operator can work well - mixture of data-parallel and model-
|
|
parallel training, where one wants to force that all copies are started before
|
|
data-parallel part starts.
|
|
)DOC")
|
|
.Arg(
|
|
"cross_device",
|
|
"Specifies either inputs should be across different devices in dev inference options");
|
|
|
|
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
|
|
SHOULD_NOT_DO_GRADIENT(AsyncNetBarrier);
|
|
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
|
|
REGISTER_CPU_OPERATOR(AsyncNetBarrier, AsyncNetBarrierOp<CPUContext>);
|
|
|
|
|
|
} // namespace caffe2
|