apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 09:07:10 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Backend] TRT backend & PP-Infer backend support pinned memory (#403)

Browse Source 
* TRT backend use pinned memory

* refine fd tensor pinned memory logic

* TRT enable pinned memory configurable

* paddle inference support pinned memory

* pinned memory pybindings

Co-authored-by: Jason <jiangjiajun@baidu.com>
This commit is contained in:
Wang Xinyu
2022-10-21 18:51:36 +08:00
committed by GitHub
parent 8dbc1f1d10
commit 43d86114d8
14 changed files with 120 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
fastdeploy/backends/tensorrt/trt_backend.h
View File

@@ -70,6 +70,7 @@ struct TrtBackendOption {
std::map<std::string, std::vector<int32_t>> min_shape;
std::map<std::string, std::vector<int32_t>> opt_shape;
std::string serialize_file = "";
bool enable_pinned_memory = false;
// inside parameter, maybe remove next version
bool remove_multiclass_nms_ = false;
@@ -118,8 +119,8 @@ class TrtBackend : public BaseBackend {
std::vector<void*> bindings_;
std::vector<TrtValueInfo> inputs_desc_;
std::vector<TrtValueInfo> outputs_desc_;
std::map<std::string, FDDeviceBuffer> inputs_buffer_;
std::map<std::string, FDDeviceBuffer> outputs_buffer_;
std::map<std::string, FDDeviceBuffer> inputs_device_buffer_;
std::map<std::string, FDDeviceBuffer> outputs_device_buffer_;
std::string calibration_str_;