Skip to content

add data type for zero copy #16313

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
NHZlX merged 1 commit into from
Mar 20, 2019
Merged

add data type for zero copy #16313

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
1 change: 1 addition & 0 deletions paddle/fluid/inference/anakin/engine.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,7 @@ void AnakinEngine<TargetT, PrecisionType, RunType>::Execute(
const std::map<std::string, framework::LoDTensor *> &inputs,
const std::map<std::string, framework::LoDTensor *> &outputs,
cudaStream_t stream) {
cudaDeviceSynchronize();
for (const auto &input : inputs) {
auto *tensor = input.second;
auto *data = tensor->data<float>();
Expand Down
14 changes: 14 additions & 0 deletions paddle/fluid/inference/api/details/zero_copy_tensor.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,6 +74,19 @@ T *ZeroCopyTensor::data(PaddlePlace *place, int *size) const {
return res;
}

PaddleDType ZeroCopyTensor::type() {
EAGER_GET_TENSOR;
auto type = tensor->type();
if (type == framework::proto::VarType::FP32) {
return PaddleDType::FLOAT32;
} else if (type == framework::proto::VarType::INT64) {
return PaddleDType::INT64;
Copy link
Contributor

@luotao1 luotao1 Mar 20, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We support INT32 as well.

Copy link
Contributor Author

@NHZlX NHZlX Mar 20, 2019

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah. At present, int32 will not be used of the VIS model and this branch will then merge to develop.

} else {
LOG(ERROR) << "unknown type, only support float32 and int64 now.";
}
return PaddleDType::FLOAT32;
}

template <typename T>
void ZeroCopyTensor::copy_from_cpu(const T *data) {
EAGER_GET_TENSOR;
Expand Down Expand Up @@ -119,6 +132,7 @@ void ZeroCopyTensor::copy_to_cpu(T *data) {
static_cast<const platform::CUDADeviceContext *>(pool.Get(gpu_place));
memory::Copy(platform::CPUPlace(), static_cast<void *>(data), gpu_place,
t_data, ele_num * sizeof(T), dev_ctx->stream());
cudaDeviceSynchronize();
#else
PADDLE_THROW("Not compile with CUDA, should not reach here.");
#endif
Expand Down
3 changes: 3 additions & 0 deletions paddle/fluid/inference/api/paddle_api.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -176,6 +176,8 @@ class ZeroCopyTensor {
device_ = device;
}

PaddleDType type();

protected:
explicit ZeroCopyTensor(void* scope) : scope_{scope} {}
void SetName(const std::string& name) { name_ = name; }
Expand All @@ -190,6 +192,7 @@ class ZeroCopyTensor {
// performance.
mutable void* tensor_{nullptr};
PaddlePlace place_;
PaddleDType dtype_;
int device_;
};

Expand Down