Skip to content

Avoid copying TF string into std:string. #1663

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dlibenzi merged 1 commit into from
Feb 19, 2020
+10 −13
Merged

Avoid copying TF string into std:string. #1663

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 2 additions & 7 deletions third_party/xla_client/record_reader.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,18 +20,13 @@ RecordReader::RecordReader(std::string path, const string& compression,
reader_.reset(new tensorflow::io::RecordReader(file_.get(), options));
}

bool RecordReader::Read(std::string* value) {
// We need to pass a tensorflow::tstring here, which will ultimately result in
// making a copy. Hopefully the tensorflow string story will end with a nice
// outcome.
tensorflow::tstring tvalue;
bool RecordReader::Read(Data* value) {
std::lock_guard<std::mutex> slock(lock_);
xla::Status status = reader_->ReadRecord(&offset_, &tvalue);
xla::Status status = reader_->ReadRecord(&offset_, value);
if (tensorflow::errors::IsOutOfRange(status)) {
return false;
}
XLA_CHECK_OK(status) << path_ << " offset " << offset_;
*value = tvalue;
return true;
}

Expand Down
4 changes: 3 additions & 1 deletion third_party/xla_client/record_reader.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,12 +13,14 @@ namespace util {

class RecordReader {
public:
using Data = tensorflow::tstring;

RecordReader(std::string path, const std::string& compression,
int64 buffer_size);

const std::string& path() const { return path_; }

bool Read(std::string* value);
bool Read(Data* value);

private:
std::string path_;
Expand Down
10 changes: 5 additions & 5 deletions torch_xla/csrc/init_python_bindings.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -275,7 +275,7 @@ std::shared_ptr<xla::util::RecordReader> CreateRecordReader(
}

bool RecordRead(const std::shared_ptr<xla::util::RecordReader>& reader,
std::string* value) {
xla::util::RecordReader::Data* value) {
NoGilSection nogil;
return reader->Read(value);
}
Expand All @@ -286,12 +286,12 @@ py::object RecordReadExample(
return std::vector<int64_t>({size});
};

std::string value;
xla::util::RecordReader::Data value;
if (!RecordRead(reader, &value)) {
return py::none();
}
tensorflow::Example exmsg;
if (!exmsg.ParseFromString(value)) {
if (!exmsg.ParseFromArray(value.data(), value.size())) {
XLA_ERROR() << "Unable to parse TF example from " << reader->path();
}
auto example = py::dict();
Expand Down Expand Up @@ -592,11 +592,11 @@ void InitXlaModuleBindings(py::module m) {
m.def(
"_xla_tfrecord_read",
[](const std::shared_ptr<xla::util::RecordReader>& reader) -> py::object {
std::string record;
xla::util::RecordReader::Data record;
if (!RecordRead(reader, &record)) {
return py::none();
}
return py::bytes(record);
return py::bytes(record.data(), record.size());
});
m.def("_xla_tfexample_read",
[](const std::shared_ptr<xla::util::RecordReader>& reader) {
Expand Down