Add attrProto.release_s interface (#22977)

### Description Add AttributeProto.release_s interface, which is used to obtain the string in the attribute using move semantics instead of copying it ### Motivation and Context The ep_context node stores a lot of information in attributes, which may cause the memory usage to increase. Use this interface to avoid memory waste --------- Co-authored-by: GenMing Zhong <[email protected]> Co-authored-by: genmingz <[email protected]>
microsoft · Dec 13, 2024 · 62e7e24 · 62e7e24
1 parent 2a36fd4
commit 62e7e24
Show file tree

Hide file tree

Showing 7 changed files with 17 additions and 1 deletion.
diff --git a/onnxruntime/core/providers/shared_library/provider_interfaces.h b/onnxruntime/core/providers/shared_library/provider_interfaces.h
@@ -390,6 +390,7 @@ struct ProviderHost {
   virtual void AttributeProto__set_name(ONNX_NAMESPACE::AttributeProto* p, const ::std::string& value) = 0;
   virtual void AttributeProto__set_type(ONNX_NAMESPACE::AttributeProto* p, ONNX_NAMESPACE::AttributeProto_AttributeType value) = 0;
   virtual ONNX_NAMESPACE::TensorProto* AttributeProto__add_tensors(ONNX_NAMESPACE::AttributeProto* p) = 0;
+  virtual std::string* AttributeProto__release_s(ONNX_NAMESPACE::AttributeProto* p) = 0;
 
   // GraphProto
   virtual std::unique_ptr<ONNX_NAMESPACE::GraphProto> GraphProto__construct() = 0;

diff --git a/onnxruntime/core/providers/shared_library/provider_wrappedtypes.h b/onnxruntime/core/providers/shared_library/provider_wrappedtypes.h
@@ -122,6 +122,7 @@ struct AttributeProto final {
   void set_name(const ::std::string& value) { return g_host->AttributeProto__set_name(this, value); }
   void set_type(AttributeProto_AttributeType value) { return g_host->AttributeProto__set_type(this, value); }
   TensorProto* add_tensors() { return g_host->AttributeProto__add_tensors(this); }
+  std::string* release_s() { return g_host->AttributeProto__release_s(this); }
 
   typedef AttributeProto_AttributeType AttributeType;
   static constexpr AttributeType UNDEFINED = AttributeProto_AttributeType_UNDEFINED;

diff --git a/onnxruntime/core/providers/vitisai/imp/attr_proto.cc b/onnxruntime/core/providers/vitisai/imp/attr_proto.cc
@@ -104,4 +104,8 @@ std::vector<std::string> attr_proto_get_strings(const ONNX_NAMESPACE::AttributeP
   }
   return ret;
 }
+std::string* attr_proto_release_string(ONNX_NAMESPACE::AttributeProto* attr) {
+  vai_assert(attr->type() == ONNX_NAMESPACE::AttributeProto_AttributeType_STRING, attr->name());
+  return attr->release_s();
+}
 }  // namespace vaip
diff --git a/onnxruntime/core/providers/vitisai/imp/attr_proto.h b/onnxruntime/core/providers/vitisai/imp/attr_proto.h
@@ -23,5 +23,6 @@ const ONNX_NAMESPACE::TensorProto& attr_proto_get_tensor(const ONNX_NAMESPACE::A
 gsl::span<const int64_t> attr_proto_get_ints(const ONNX_NAMESPACE::AttributeProto& attr);
 gsl::span<const float> attr_proto_get_floats(const ONNX_NAMESPACE::AttributeProto& attr);
 std::vector<std::string> attr_proto_get_strings(const ONNX_NAMESPACE::AttributeProto& attr);
+std::string* attr_proto_release_string(ONNX_NAMESPACE::AttributeProto* attr);
 
 }  // namespace vaip
diff --git a/onnxruntime/core/providers/vitisai/imp/global_api.cc b/onnxruntime/core/providers/vitisai/imp/global_api.cc
@@ -449,6 +449,13 @@ vaip_core::OrtApiForVaip* create_org_api_hook() {
     return vaip_core::DllSafe(model_proto.SerializeAsString());
   };
   the_global_api.model_proto_delete = [](ONNX_NAMESPACE::ModelProto* p) { delete p; };
+  the_global_api.attr_proto_release_string = [](ONNX_NAMESPACE::AttributeProto* attr) -> vaip_core::DllSafe<std::string> {
+    auto pstr = vaip::attr_proto_release_string(attr);
+    std::string local_str = std::move(*pstr);
+    pstr = nullptr;
+    return vaip_core::DllSafe<std::string>(std::move(local_str));
+  };
+
   if (!s_library_vitisaiep.vaip_get_version) {
     return reinterpret_cast<vaip_core::OrtApiForVaip*>(&(the_global_api.host_));
   } else {

diff --git a/onnxruntime/core/providers/vitisai/include/vaip/vaip_ort_api.h b/onnxruntime/core/providers/vitisai/include/vaip/vaip_ort_api.h
@@ -13,7 +13,7 @@ struct OrtApi;
 
 namespace vaip_core {
 
-#define VAIP_ORT_API_MAJOR (11u)
+#define VAIP_ORT_API_MAJOR (12u)
 #define VAIP_ORT_API_MINOR (0u)
 #define VAIP_ORT_API_PATCH (0u)
 struct OrtApiForVaip {
@@ -234,6 +234,7 @@ struct OrtApiForVaip {
   ModelProto* (*model_to_proto)(Model& model);                                                                                                        // [95]
   DllSafe<std::string> (*model_proto_serialize_as_string)(ModelProto& model_proto);                                                                   // [96]
   void (*model_proto_delete)(ModelProto* p);                                                                                                          // [97]
+  DllSafe<std::string> (*attr_proto_release_string)(AttributeProto* attr);                                                                            // [98]
 };
 
 #ifndef USE_VITISAI

diff --git a/onnxruntime/core/session/provider_bridge_ort.cc b/onnxruntime/core/session/provider_bridge_ort.cc
@@ -497,6 +497,7 @@ struct ProviderHostImpl : ProviderHost {
   void AttributeProto__set_name(ONNX_NAMESPACE::AttributeProto* p, const ::std::string& value) override { return p->set_name(value); }
   void AttributeProto__set_type(ONNX_NAMESPACE::AttributeProto* p, ONNX_NAMESPACE::AttributeProto_AttributeType value) override { return p->set_type(value); }
   ONNX_NAMESPACE::TensorProto* AttributeProto__add_tensors(ONNX_NAMESPACE::AttributeProto* p) override { return p->add_tensors(); }
+  std::string* AttributeProto__release_s(ONNX_NAMESPACE::AttributeProto* p) override { return p->release_s(); }
 
   // GraphProto (wrapped)
   std::unique_ptr<ONNX_NAMESPACE::GraphProto> GraphProto__construct() override { return std::make_unique<ONNX_NAMESPACE::GraphProto>(); }