Refactor op info parser (#54859)

* add kernel dialect

* change DenseTensorTypeStorage to DenseTensorType

* add test case`

* add first pd_op to kernel dialect

* lower pd op to kernel dialect

* update

* update

* remove useless code

* add attrite print test

* fix bug

* update

* update

* update

* update

* polish code

* fix bug

* polish  code  and add python test

* add test

* fix test error

* add env flag

* fix bug

* revert test env

* change cc_test_old to cc_test

* fix build_static bug

* fix type test error

* udpate cmake

* disable test in windows

* update

* update

* fix bug

* split file

* fix conflict

* polish code and fix conflict

* support place transformer

* finish bug

* add gpu flags

* fix with cuda macro

* add fetch kernel

* support fetch var in new ir

* fix bug

* polish code

* change array equal to np.testing

* support feed in new ir

* update

* fix bug

* try to hack combine op

* add scope guard

* revert atan2 op

* add scope guard

* update

* polish code

* update

* refactor build kernel context

* fix unitest bug

* polish code

* use original order

* remove useless code

* polish code

* fix bug

paddle/fluid/framework/new_executor/interpreter/interpreter_util.cc

@@ -24,6 +24,7 @@
 #include "paddle/fluid/framework/new_executor/interpreter/static_build.h"
 #include "paddle/fluid/ir/dialect/pd_dialect.h"
 #include "paddle/fluid/ir/interface/op_yaml_info.h"
+#include "paddle/fluid/ir/interface/op_yaml_info_parser.h"
 #include "paddle/fluid/ir/phi_kernel_adaptor/phi_kernel_util.h"
 #include "paddle/fluid/memory/stats.h"
 #include "paddle/fluid/operators/controlflow/conditional_block_op_helper.h"
@@ -951,31 +952,27 @@ void BuildOpFuncList(
     auto attr_map = (*it)->attributes();
 
     auto op_name = attr_map.at("op_name").dyn_cast<::ir::StrAttribute>().data();
+    op_func_node.phi_op_name_ = op_name;
 
     if (op_name == "builtin.combine" || op_name == "pd.feed") {
       VLOG(6) << "skip process " << op_name;
       continue;
     }
 
-    op_func_node.phi_op_name_ = op_name;
-
     ::ir::OpInfo op_info = ctx->GetRegisteredOpInfo(op_name);
 
     auto impl =
         op_info.GetInterfaceImpl<paddle::dialect::OpYamlInfoInterface>();
-    auto yaml_info = impl->get_op_info_();
-
-    auto attr_info = std::get<1>(yaml_info);
 
     op_func_node.infer_meta_interface_ =
         op_info.GetInterfaceImpl<paddle::dialect::InferMetaInterface>();
 
     VLOG(6) << "op name" << op_func_node.phi_op_name_;
-
+    dialect::OpYamlInfoParser op_yaml_info_parser(impl->get_op_info_());
     ::ir::BuildInferMetaContext((*it),
                                 value_2_name_map,
                                 scope,
-                                yaml_info,
+                                op_yaml_info_parser,
                                 &(op_func_node.infer_meta_context_));
 
     auto kernel_name =
@@ -996,7 +993,7 @@ void BuildOpFuncList(
     ::ir::BuildPhiKernelContext((*it),
                                 value_2_name_map,
                                 scope,
-                                yaml_info,
+                                op_yaml_info_parser,
                                 &(op_func_node.kernel_context_),
                                 &(op_func_node.input_index),
                                 &(op_func_node.output_index));

paddle/fluid/framework/new_executor/new_ir_interpreter.cc

@@ -116,7 +116,9 @@ void NewIRInterpreter::RunImpl() {
   //   &&
   //       (sync_op_num_ == 0)) {
   VLOG(4) << "Tracing Instruction List";
+
   TraceInstructionList(vec_instruction_);
+
   //   } else {
   //     VLOG(4) << "Non-tracing";
   //     // For the program that only run once, it is no need to
@@ -938,15 +940,6 @@ void NewIRInterpreter::RunOperator(const Instruction& instr_node) {
 }
 
 void NewIRInterpreter::RunInstruction(const Instruction& instr_node) {
-  VLOG(5) << __func__ << " OP id:" << instr_node.Id()
-          << " name:" << instr_node.OpBase()->Type() << " type:"
-          << (instr_node.KernelType() == OpFuncType::kCpuSync
-                  ? "kCpuSync"
-                  : (instr_node.KernelType() == OpFuncType::kGpuSync
-                         ? "kGpuSync"
-                         : "kGpuAsync"))
-          << " runs on " << platform::GetCurrentThreadName();
-
   OperatorBase* op = nullptr;
   if (instr_node.OpBaseValid()) {
     op = instr_node.OpBase();
@@ -1377,8 +1370,9 @@ void NewIRInterpreter::TraceInstructionList(
     }
   }
 
-  for (size_t idx = 0; idx < trace_execute_order_.size(); idx++) {
-    auto instr_id = trace_execute_order_[idx];
+  // TODO(phlrain) use orignal order for now, use better dependecy
+  for (size_t instr_id = 0; instr_id < vec_instruction_.size(); ++instr_id) {
+    /// auto instr_id = trace_execute_order_[idx];
     auto& instr_node = vec_instruction_.at(instr_id);
 
     RunInstruction(instr_node);

paddle/fluid/ir/interface/op_yaml_info_parser.cc

@@ -22,7 +22,7 @@ OpYamlInfoParser::OpYamlInfoParser(const OpInfoTuple& op_info_tuple)
   parse();
 }
 
-bool OpYamlInfoParser::IsTensorArrtibute(size_t index) const {
+bool OpYamlInfoParser::IsTensorAttribute(size_t index) const {
   PADDLE_ENFORCE_LT(
       index,
       InputInfo().size(),
@@ -48,6 +48,21 @@ const std::string& OpYamlInfoParser::AttrTypeName(
   return it->second.type_name;
 }
 
+const std::string& OpYamlInfoParser::TensorAttrTypeName(
+    const std::string& name) const {
+  auto it = map_input_info_.find(name);
+
+  PADDLE_ENFORCE_NE(it,
+                    map_input_info_.end(),
+                    phi::errors::NotFound("Not found [%s] in input map", name));
+
+  PADDLE_ENFORCE_EQ(
+      it->second.is_mutable_attribute,
+      true,
+      phi::errors::PreconditionNotMet("[%s] MUST be a tensor attribute", name));
+  return it->second.type_name;
+}
+
 const std::vector<std::string>& OpYamlInfoParser::InferMetaTensorParams()
     const {
   return vec_infer_meta_tensor_params_;
@@ -62,6 +77,14 @@ const std::vector<std::string>& OpYamlInfoParser::KernelFnAttrParams() const {
   return vec_kernel_fn_attr_params_;
 }
 
+const OpRunTimeInfo& OpYamlInfoParser::OpRuntimeInfo() const {
+  return std::get<3>(op_info_tuple_);
+}
+
+const std::map<std::string, int>& OpYamlInfoParser::Name2Id() const {
+  return map_name2id_;
+}
+
 void OpYamlInfoParser::parse() {
   auto input_info = std::get<0>(op_info_tuple_);
 
@@ -91,7 +114,8 @@ void OpYamlInfoParser::parse() {
   auto runtime_info = std::get<3>(op_info_tuple_);
 
   for (auto& name : runtime_info.infer_meta_param) {
-    if (map_name2id_.count(name)) {
+    if (map_name2id_.count(name) &&
+        !map_input_info_[name].is_mutable_attribute) {
       vec_infer_meta_tensor_params_.push_back(name);
     } else {
       vec_infer_meta_attr_params_.push_back(name);
@@ -99,7 +123,8 @@ void OpYamlInfoParser::parse() {
   }
 
   for (auto& name : runtime_info.kernel_param) {
-    if (map_name2id_.count(name)) {
+    if (map_name2id_.count(name) &&
+        !map_input_info_[name].is_mutable_attribute) {
       vec_kernel_fn_tensor_params_.push_back(name);
     } else {
       vec_kernel_fn_attr_params_.push_back(name);

paddle/fluid/ir/interface/op_yaml_info_parser.h

@@ -25,15 +25,18 @@ class OpYamlInfoParser {
 
   explicit OpYamlInfoParser(const OpInfoTuple& op_info_tuple);
 
-  bool IsTensorArrtibute(size_t index) const;
+  bool IsTensorAttribute(size_t index) const;
   size_t InputTensorNumber() const;
 
   const std::string& AttrTypeName(const std::string& name) const;
+  const std::string& TensorAttrTypeName(const std::string& name) const;
 
   const std::vector<std::string>& InferMetaTensorParams() const;
   const std::vector<std::string>& InferMetaAttrParams() const;
   const std::vector<std::string>& KernelFnTensorParams() const;
   const std::vector<std::string>& KernelFnAttrParams() const;
+  const OpRunTimeInfo& OpRuntimeInfo() const;
+  const std::map<std::string, int>& Name2Id() const;
 
  private:
   void parse();
@@ -41,7 +44,7 @@ class OpYamlInfoParser {
     return std::get<0>(op_info_tuple_);
   }
 
-  const OpInfoTuple& op_info_tuple_;
+  OpInfoTuple op_info_tuple_;
 
   std::map<std::string, int> map_name2id_;
 

paddle/fluid/ir/pass/CMakeLists.txt

@@ -4,4 +4,4 @@ file(GLOB PD_PASS_SRCS "*.cc")
 cc_library(
   pd_op_to_kernel_pass
   SRCS ${PD_PASS_SRCS}
-  DEPS ir phi_utils)
+  DEPS ir phi_utils pd_interface)

paddle/fluid/ir/pass/pd_op_to_kernel_pass.cc

@@ -25,6 +25,7 @@
 #include "paddle/fluid/ir/dialect/pd_dialect.h"
 #include "paddle/fluid/ir/dialect/utils.h"
 #include "paddle/fluid/ir/interface/op_yaml_info.h"
+#include "paddle/fluid/ir/interface/op_yaml_info_parser.h"
 #include "paddle/phi/api/lib/data_transform.h"
 #include "paddle/phi/api/lib/kernel_dispatch.h"
 #include "paddle/phi/common/place.h"
@@ -38,7 +39,8 @@ const int init_on_gpu_threashold = 1000;
 phi::KernelKey GetKernelKey(
     ir::Operation* op,
     const phi::Place& place,
-    const std::unordered_map<ir::Value, ir::OpResult>& map_value_pair) {
+    const std::unordered_map<ir::Value, ir::OpResult>& map_value_pair,
+    const dialect::OpYamlInfoParser* op_info_parser = nullptr) {
   if (op->name() == "pd.feed") {
     // NOTE, for now feed op don't need a kernel, so the data type from Op
     // Result the next op use base program datatype
@@ -51,40 +53,18 @@ phi::KernelKey GetKernelKey(
   phi::DataLayout kernel_layout = phi::DataLayout::UNDEFINED;
   phi::DataType kernel_data_type = phi::DataType::UNDEFINED;
 
-  paddle::dialect::OpYamlInfoInterface op_info_interface =
-      op->dyn_cast<paddle::dialect::OpYamlInfoInterface>();
-  std::vector<paddle::dialect::OpInputInfo> input_info;
-  if (op_info_interface) {
-    auto op_info_res = op_info_interface.GetOpInfo();
-
-    input_info = std::get<0>(op_info_res);
-
+  if (op_info_parser != nullptr) {
     // only suppurt non vector input for now
-    std::map<std::string, int> input_map;
-    int index = 0;
-    int tensor_input_number = 0;
-    for (auto& t : input_info) {
-      // todo filter attribute tensor
-      input_map[t.name] = index++;
-
-      if (!t.is_mutable_attribute) {
-        tensor_input_number += 1;
-      }
-    }
-
-    std::map<std::string, std::string> attr_type_map;
-    auto attr_info = std::get<1>(op_info_res);
-    for (auto& t : attr_info) {
-      VLOG(6) << t.name << "\t" << t.type_name;
-      attr_type_map[t.name] = t.type_name;
-    }
-    auto runtime_info = std::get<3>(op_info_res);
+    int tensor_input_number = op_info_parser->InputTensorNumber();
 
     auto attr_map = op->attributes();
-    auto data_type_info = runtime_info.kernel_key_dtype;
+    auto& data_type_info = op_info_parser->OpRuntimeInfo().kernel_key_dtype;
+
     if (data_type_info.size() > 0 && data_type_info[0] != "") {
       // only support single input and attribute
       auto slot_name = data_type_info[0];
+      auto& input_map = op_info_parser->Name2Id();
+
       if (input_map.count(slot_name)) {
         // parse from input
         int in_index = input_map.at(slot_name);
@@ -95,10 +75,16 @@ phi::KernelKey GetKernelKey(
                 .dyn_cast<paddle::dialect::DenseTensorType>();
         kernel_data_type = TransToPhiDataType(type.dtype());
       } else {
-        PADDLE_ENFORCE_EQ(attr_type_map.count(slot_name),
+        PADDLE_ENFORCE_EQ(attr_map.count(slot_name),
                           true,
                           phi::errors::PreconditionNotMet(
-                              "[%s] MUST in attr map", slot_name));
+                              "[%s] MUST in attribute map", slot_name));
+
+        auto attr_type = op_info_parser->AttrTypeName(slot_name);
+        PADDLE_ENFORCE_EQ(attr_type,
+                          "paddle::dialect::DataTypeAttribute",
+                          phi::errors::PreconditionNotMet(
+                              "Type of [%s] should be DataType", slot_name));
         kernel_data_type = attr_map.at(slot_name)
                                .dyn_cast<paddle::dialect::DataTypeAttribute>()
                                .data();
@@ -140,10 +126,11 @@ phi::KernelKey GetKernelKey(
     paddle::experimental::detail::KernelKeyParser kernel_key_parser;
 
     for (size_t i = 0; i < op->num_operands(); ++i) {
-      // todo filter attribute tensor
-      if ((input_info.size() > i) && input_info[i].is_mutable_attribute) {
+      // NOTE, only op with OpYamlInfo can have TensorArr
+      if (op_info_parser != nullptr && op_info_parser->IsTensorAttribute(i)) {
         continue;
       }
+
       auto input_tmp = op->operand(i);
       auto new_input_tmp = map_value_pair.at(input_tmp);
 
@@ -203,13 +190,20 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
   std::unordered_map<ir::Operation*, ir::Operation*> map_op_pair;
   std::unordered_map<ir::Value, ir::OpResult> map_value_pair;
 
-  std::string op1_name = paddle::dialect::PhiKernelOp::name();
+  std::string op_name = paddle::dialect::PhiKernelOp::name();
 
-  ir::OpInfo op1_info = ctx->GetRegisteredOpInfo(op1_name);
+  ir::OpInfo op_info = ctx->GetRegisteredOpInfo(op_name);
 
   for (auto it = block->begin(); it != block->end(); ++it) {
     VLOG(6) << "op name " << (*it)->name();
-    auto kernel_key = GetKernelKey(*it, cpu_place, map_value_pair);
+    paddle::dialect::OpYamlInfoInterface op_info_interface =
+        (*it)->dyn_cast<paddle::dialect::OpYamlInfoInterface>();
+    OpYamlInfoParser* op_info_parser = nullptr;
+    if (op_info_interface) {
+      op_info_parser = new OpYamlInfoParser(op_info_interface.GetOpInfo());
+    }
+    auto kernel_key =
+        GetKernelKey(*it, cpu_place, map_value_pair, op_info_parser);
     VLOG(6) << "kernel type " << kernel_key;
     // create new Op
 
@@ -256,15 +250,9 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
     // constuct input
     std::vector<ir::OpResult> vec_inputs;
 
-    paddle::dialect::OpYamlInfoInterface op_info_interface =
-        (*it)->dyn_cast<paddle::dialect::OpYamlInfoInterface>();
     std::string kernel_fn_str;
-    std::vector<paddle::dialect::OpInputInfo> input_info;
-    if (op_info_interface) {
-      auto op_info_res = op_info_interface.GetOpInfo();
-      auto runtime_info = std::get<3>(op_info_res);
-      kernel_fn_str = runtime_info.kernel_func[0];
-      input_info = std::get<0>(op_info_res);
+    if (op_info_parser != nullptr) {
+      kernel_fn_str = op_info_parser->OpRuntimeInfo().kernel_func[0];
     }
 
     if ((*it)->num_operands() > 0) {
@@ -284,9 +272,11 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
                 new_in_type.dyn_cast<dialect::AllocatedDenseTensorType>()
                     .place();
 
-            if ((i < input_info.size()) &&
-                (!input_info[i].is_mutable_attribute) &&
-                (place != phi::TransToPhiPlace(kernel_key.backend()))) {
+            bool need_trans =
+                (op_info_parser != nullptr &&
+                 !op_info_parser->IsTensorAttribute(i)) &&
+                (place != phi::TransToPhiPlace(kernel_key.backend()));
+            if (need_trans) {
               if (paddle::experimental::NeedTransformPlace(
                       place, kernel.InputAt(i).backend, {})) {
                 VLOG(6) << "need trans from " << place << " to "
@@ -294,19 +284,19 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
                 // build memcopy op
                 auto copy_kernel_key = kernel_key;
                 copy_kernel_key.set_backend(phi::Backend::GPU);
-                std::unordered_map<std::string, ir::Attribute> op1_attribute{
+                std::unordered_map<std::string, ir::Attribute> op_attribute{
                     {"op_name", ir::StrAttribute::get(ctx, "pd.memcpy_h2d")},
                     {"kernel_name", ir::StrAttribute::get(ctx, "memcpy_h2d")},
                     {"kernel_key",
                      dialect::KernelAttribute::get(ctx, copy_kernel_key)},
                     {"dst_place_type", ir::Int32Attribute::get(ctx, 1)}};
 
-                ir::Operation* op1 = ir::Operation::Create(
-                    {new_in}, op1_attribute, {new_in_type}, op1_info);
+                ir::Operation* op = ir::Operation::Create(
+                    {new_in}, op_attribute, {new_in_type}, op_info);
 
-                program->block()->push_back(op1);
+                program->block()->push_back(op);
 
-                new_in = op1->result(0);
+                new_in = op->result(0);
               }
             }
           } else if (new_in_type.isa<ir::VectorType>()) {
@@ -320,7 +310,7 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
       }
     }
 
-    std::unordered_map<std::string, ir::Attribute> op1_attribute{
+    std::unordered_map<std::string, ir::Attribute> op_attribute{
         {"op_name", ir::StrAttribute::get(ctx, (*it)->name())},
         {"kernel_name", ir::StrAttribute::get(ctx, kernel_fn_str)},
         {"kernel_key", dialect::KernelAttribute::get(ctx, kernel_key)}};
@@ -328,22 +318,22 @@ std::unique_ptr<ir::Program> PdOpLowerToKernelPass(ir::Program* prog) {
     auto op_attr_map = (*it)->attributes();
 
     for (auto it1 = op_attr_map.begin(); it1 != op_attr_map.end(); ++it1) {
-      op1_attribute.emplace(it1->first, it1->second);
+      op_attribute.emplace(it1->first, it1->second);
     }
 
-    ir::Operation* op1 = ir::Operation::Create(
-        vec_inputs, op1_attribute, op_output_types, op1_info);
+    ir::Operation* op = ir::Operation::Create(
+        vec_inputs, op_attribute, op_output_types, op_info);
 
-    map_op_pair[*it] = op1;
+    map_op_pair[*it] = op;
 
     // only deal with single output
     if ((*it)->num_results() > 0) {
       for (size_t i = 0; i < (*it)->num_results(); ++i) {
-        map_value_pair[(*it)->result(i)] = op1->result(i);
+        map_value_pair[(*it)->result(i)] = op->result(i);
       }
     }
 
-    program->block()->push_back(op1);
+    program->block()->push_back(op);
   }
 
   return program;

paddle/fluid/ir/phi_kernel_adaptor/phi_kernel_adaptor.h

@@ -21,6 +21,7 @@
 #include "paddle/fluid/ir/dialect/utils.h"
 #include "paddle/fluid/ir/interface/infermeta.h"
 #include "paddle/fluid/ir/interface/op_yaml_info.h"
+#include "paddle/fluid/ir/interface/op_yaml_info_parser.h"
 #include "paddle/ir/core/builtin_attribute.h"
 #include "paddle/ir/core/builtin_dialect.h"
 #include "paddle/ir/core/builtin_op.h"
@@ -80,7 +81,9 @@ class PhiKernelAdaptor {
 
       phi::InferMetaContext ctx;
 
-      ir::BuildInferMetaContext((*it), name_map, scope_, yaml_info, &ctx);
+      paddle::dialect::OpYamlInfoParser op_yaml_info_parser(yaml_info);
+      ir::BuildInferMetaContext(
+          (*it), name_map, scope_, op_yaml_info_parser, &ctx);
 
       infer_meta_impl->infer_meta_(&ctx);
 
@@ -96,7 +99,7 @@ class PhiKernelAdaptor {
       phi::KernelContext kernel_ctx(dev_ctx);
 
       ir::BuildPhiKernelContext(
-          (*it), name_map, scope_, yaml_info, &kernel_ctx);
+          (*it), name_map, scope_, op_yaml_info_parser, &kernel_ctx);
       kernel_fn(&kernel_ctx);
 
       auto out_value = (*it)->result(0);

paddle/fluid/ir/phi_kernel_adaptor/phi_kernel_util.cc

@@ -36,6 +36,7 @@
 #include "paddle/fluid/ir/dialect/kernel_attribute.h"
 #include "paddle/fluid/ir/dialect/kernel_type.h"
 #include "paddle/fluid/ir/dialect/pd_attribute.h"
+#include "paddle/fluid/ir/interface/op_yaml_info_parser.h"
 #include "paddle/phi/core/enforce.h"
 
 #include "glog/logging.h"
@@ -178,70 +179,29 @@ void BuildInferMetaContext(
     ir::Operation* op,
     const std::unordered_map<ir::Value, std::string>& name_map,
     paddle::framework::Scope* scope,
-    const OpInfoTuple& op_yaml_info,
+    const paddle::dialect::OpYamlInfoParser& op_yaml_info,
     phi::InferMetaContext* ctx) {
   // inputs include input and mutable attributes
-  auto input_info = std::get<0>(op_yaml_info);
-  std::map<std::string, size_t> input_index_map;
-  std::map<std::string, std::string> mutable_attr_type_map;
-  int input_index = 0;
-  for (auto& t : input_info) {
-    VLOG(6) << t.name << "\t" << t.type_name;
-    input_index_map[t.name] = input_index++;
-    if (t.is_mutable_attribute) {
-      mutable_attr_type_map[t.name] = t.type_name;
-    }
-  }
-
-  auto attr_info = std::get<1>(op_yaml_info);
-  std::map<std::string, std::string> attr_type_map;
-  for (auto& t : attr_info) {
-    VLOG(6) << t.name << "\t" << t.type_name;
-    attr_type_map[t.name] = t.type_name;
-  }
-
   auto attr_map = op->attributes();
-  auto runtime_info = std::get<3>(op_yaml_info);
+  auto& vec_infer_meta_tensor_params = op_yaml_info.InferMetaTensorParams();
 
-  // int input_index = 0;
-
-  std::vector<std::string> vec_param_list = runtime_info.infer_meta_param;
-
-  for (size_t input_index = 0; input_index < vec_param_list.size();
-       input_index++) {
-    auto& t = vec_param_list[input_index];
-    if (input_index_map.count(t)) {
-      // get information from input
-      ir::Value ptr = op->operand(input_index_map[t]);
+  auto& name2id = op_yaml_info.Name2Id();
+  for (auto& t : vec_infer_meta_tensor_params) {
+    PADDLE_ENFORCE_EQ(
+        name2id.count(t),
+        true,
+        phi::errors::NotFound("param [%s] MUST in name2id map", t));
+    auto index = op_yaml_info.Name2Id().at(t);
+    ir::Value ptr = op->operand(index);
     auto in_var_name = name_map.at(ptr);
 
-      if (mutable_attr_type_map.count(t)) {
-        VLOG(6) << "ctx->EmplaceBack mutable attr: " << t << "\t"
-                << in_var_name;
-        if (mutable_attr_type_map[t] == "paddle::dialect::IntArrayAttribute") {
-          phi::Attribute r1 = phi::TensorRef(
-              &(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
-          ctx->EmplaceBackAttr(r1);
-        } else if (mutable_attr_type_map[t] ==
-                   "paddle::dialect::ScalarAttribute") {
-          phi::Attribute r1 = phi::TensorRef(
-              &(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
-
-          ctx->EmplaceBackAttr(r1);
-        } else {
-          PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
-                                                  mutable_attr_type_map[t]));
-        }
-
-      } else {
     VLOG(6) << "ctx->EmplaceBackInput: " << t << "\t" << in_var_name;
     auto var = scope->Var(in_var_name);
     if (var->IsType<phi::DenseTensor>()) {
       const phi::TensorBase* tensor_in = &(var->Get<phi::DenseTensor>());
       ctx->EmplaceBackInput(const_cast<phi::TensorBase*>(tensor_in));
     } else if (var->IsType<paddle::framework::TensorRefArray>()) {
-          paddle::small_vector<phi::MetaTensor, phi::kInputSmallVectorSize>
-              inputs;
+      paddle::small_vector<phi::MetaTensor, phi::kInputSmallVectorSize> inputs;
       auto& tensor_array = var->Get<paddle::framework::TensorRefArray>();
       for (size_t i = 0; i < tensor_array.size(); ++i) {
         inputs.emplace_back(std::move(phi::MetaTensor(*tensor_array[i])));
@@ -253,35 +213,60 @@ void BuildInferMetaContext(
                                               var->Type()));
     }
   }
+
+  auto& vec_infer_meta_attr_params = op_yaml_info.InferMetaAttrParams();
+  for (auto& t : vec_infer_meta_attr_params) {
+    if (name2id.count(t)) {
+      // tensor attribute, get information from input
+      ir::Value ptr = op->operand(name2id.at(t));
+      auto in_var_name = name_map.at(ptr);
+
+      auto& tensor_attr_type = op_yaml_info.TensorAttrTypeName(t);
+
+      VLOG(6) << "ctx->EmplaceBack mutable attr: " << t << "\t" << in_var_name;
+      if (tensor_attr_type == "paddle::dialect::IntArrayAttribute") {
+        phi::Attribute r1 =
+            phi::TensorRef(&(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
+        ctx->EmplaceBackAttr(r1);
+      } else if (tensor_attr_type == "paddle::dialect::ScalarAttribute") {
+        phi::Attribute r1 =
+            phi::TensorRef(&(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
+
+        ctx->EmplaceBackAttr(r1);
+      } else {
+        PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
+                                                tensor_attr_type));
+      }
+
+      continue;
     }
 
-    if (attr_type_map.count(t)) {
-      auto type_name = attr_type_map[t];
-      if (type_name == "paddle::dialect::IntArrayAttribute") {
+    auto& attr_type_name = op_yaml_info.AttrTypeName(t);
+
+    if (attr_type_name == "paddle::dialect::IntArrayAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::IntArrayAttribute>().data());
-      } else if (type_name == "paddle::dialect::DataTypeAttribute") {
+    } else if (attr_type_name == "paddle::dialect::DataTypeAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::DataTypeAttribute>().data());
-      } else if (type_name == "ir::Int32Attribute") {
+    } else if (attr_type_name == "ir::Int32Attribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::Int32Attribute>().data());
-      } else if (type_name == "ir::FloatAttribute") {
+    } else if (attr_type_name == "ir::FloatAttribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::FloatAttribute>().data());
-      } else if (type_name == "ir::BoolAttribute") {
+    } else if (attr_type_name == "ir::BoolAttribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::BoolAttribute>().data());
-      } else if (type_name == "paddle::dialect::PlaceAttribute") {
+    } else if (attr_type_name == "paddle::dialect::PlaceAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::PlaceAttribute>().data());
-      } else if (type_name == "paddle::dialect::ScalarAttribute") {
+    } else if (attr_type_name == "paddle::dialect::ScalarAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::ScalarAttribute>().data());
     } else {
       PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
-                                                type_name));
+                                              attr_type_name));
     }
     VLOG(6) << "ctx->EmplaceBackAttr: " << t;
   }
-  }
 
   // TODO(phlrain): use var type instead of op name
   if (op->attributes().count("op_name") &&
@@ -305,77 +290,30 @@ void BuildPhiKernelContext(
     ir::Operation* op,
     const std::unordered_map<ir::Value, std::string>& name_map,
     paddle::framework::Scope* scope,
-    const OpInfoTuple& op_yaml_info,
+    const paddle::dialect::OpYamlInfoParser& op_yaml_info,
     phi::KernelContext* ctx,
     std::map<std::string, std::vector<int>>* input_map,
     std::map<std::string, std::vector<int>>* output_map) {
   // inputs include input and mutable attributes
-  auto input_info = std::get<0>(op_yaml_info);
-  std::map<std::string, size_t> input_index_map;
-  std::map<std::string, std::string> mutable_attr_type_map;
-  int input_index = 0;
-  for (auto& t : input_info) {
-    VLOG(6) << t.name << "\t" << t.type_name;
-    input_index_map[t.name] = input_index++;
-    if (t.is_mutable_attribute) {
-      mutable_attr_type_map[t.name] = t.type_name;
-    }
-  }
-
-  auto attr_info = std::get<1>(op_yaml_info);
-  std::map<std::string, std::string> attr_type_map;
-  for (auto& t : attr_info) {
-    VLOG(6) << t.name << "\t" << t.type_name;
-    attr_type_map[t.name] = t.type_name;
-  }
 
   auto attr_map = op->attributes();
-  auto runtime_info = std::get<3>(op_yaml_info);
-
-  // int input_index = 0;
-  std::vector<std::string> vec_param_list = runtime_info.kernel_param;
-  for (auto& t : vec_param_list) {
-    if (input_index_map.count(t)) {
-      // get information from input
-      ir::Value ptr = op->operand(input_index_map[t]);
-      auto in_var_name = name_map.at(ptr);
-      if (input_map != nullptr) {
-        // only deal with single input for now, [todo] need support multi input
-        // like concat
-        // TODO(phlrain): OpFuncNode need input_index and output_index,
-        // construct input_index and output_here,  should remove input_index and
-        // output_index from OpFuncNode Each in_var_name named "inner_var_" +
-        // index, len("inner_var_") = 10
-
-        size_t tmp_id = std::atol(in_var_name.substr(4, 100).c_str());
-        (*input_map)[std::to_string(input_index_map.at(t))].push_back(tmp_id);
-      }
 
-      if (mutable_attr_type_map.count(t)) {
-        VLOG(6) << "ctx->EmplaceBack mutable attr: " << t << "\t"
-                << in_var_name;
-        if (mutable_attr_type_map[t] == "paddle::dialect::IntArrayAttribute") {
-          phi::Attribute r1 = phi::TensorRef(
-              &(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
-          ctx->EmplaceBackAttr(r1);
-        } else if (mutable_attr_type_map[t] ==
-                   "paddle::dialect::ScalarAttribute") {
-          phi::Attribute r1 = phi::TensorRef(
-              &(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
-
-          ctx->EmplaceBackAttr(r1);
-        } else {
-          PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
-                                                  mutable_attr_type_map[t]));
-        }
+  auto& vec_kernel_fn_tensor_params = op_yaml_info.KernelFnTensorParams();
 
-      } else {
+  auto& name2id = op_yaml_info.Name2Id();
+  for (auto& t : vec_kernel_fn_tensor_params) {
+    PADDLE_ENFORCE_EQ(
+        name2id.count(t),
+        true,
+        phi::errors::NotFound("param [%s] MUST in name2id map", t));
+    auto index = op_yaml_info.Name2Id().at(t);
+    ir::Value ptr = op->operand(index);
+    auto in_var_name = name_map.at(ptr);
     VLOG(6) << "ctx->EmplaceBackInput: " << t << "\t" << in_var_name;
 
-        PADDLE_ENFORCE_NOT_NULL(
-            scope->FindLocalVar(in_var_name),
-            phi::errors::PreconditionNotMet("can not find var[%s] in scope",
-                                            in_var_name));
+    PADDLE_ENFORCE_NOT_NULL(scope->FindLocalVar(in_var_name),
+                            phi::errors::PreconditionNotMet(
+                                "can not find var[%s] in scope", in_var_name));
 
     auto var = scope->Var(in_var_name);
     if (var->IsType<phi::DenseTensor>()) {
@@ -387,7 +325,7 @@ void BuildPhiKernelContext(
       for (size_t i = 0; i < tensor_array.size(); ++i) {
         inputs.emplace_back(tensor_array[i]);
       }
-
+      std::cerr << "is tensor ref " << std::endl;
       ctx->EmplaceBackInputs(std::move(inputs));
     } else if (var->IsType<paddle::framework::FeedList>()) {
       auto feed_list = var->Get<paddle::framework::FeedList>();
@@ -398,35 +336,70 @@ void BuildPhiKernelContext(
                                               var->Type()));
     }
   }
+
+  auto& vec_kernel_fn_attr_params = op_yaml_info.KernelFnAttrParams();
+  for (auto& t : vec_kernel_fn_attr_params) {
+    if (name2id.count(t)) {
+      // tensor attribute, get information from input
+      ir::Value ptr = op->operand(name2id.at(t));
+
+      auto in_var_name = name_map.at(ptr);
+      if (input_map != nullptr) {
+        // only deal with single input for now, [todo] need support multi input
+        // like concat
+        // TODO(phlrain): OpFuncNode need input_index and output_index,
+        // construct input_index and output_here,  should remove input_index and
+        // output_index from OpFuncNode Each in_var_name named "inner_var_" +
+        // index, len("inner_var_") = 10
+
+        size_t tmp_id = std::atol(in_var_name.substr(4, 100).c_str());
+        (*input_map)[std::to_string(name2id.at(t))].push_back(tmp_id);
       }
 
-    if (attr_type_map.count(t)) {
-      auto type_name = attr_type_map[t];
-      if (type_name == "paddle::dialect::IntArrayAttribute") {
+      auto& tensor_attr_type = op_yaml_info.TensorAttrTypeName(t);
+      VLOG(6) << "ctx->EmplaceBack mutable attr: " << t << "\t" << in_var_name;
+      if (tensor_attr_type == "paddle::dialect::IntArrayAttribute") {
+        phi::Attribute r1 =
+            phi::TensorRef(&(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
+        ctx->EmplaceBackAttr(r1);
+      } else if (tensor_attr_type == "paddle::dialect::ScalarAttribute") {
+        phi::Attribute r1 =
+            phi::TensorRef(&(scope->Var(in_var_name)->Get<phi::DenseTensor>()));
+
+        ctx->EmplaceBackAttr(r1);
+      } else {
+        PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
+                                                tensor_attr_type));
+      }
+
+      continue;
+    }
+
+    auto& attr_type_name = op_yaml_info.AttrTypeName(t);
+    if (attr_type_name == "paddle::dialect::IntArrayAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::IntArrayAttribute>().data());
-      } else if (type_name == "paddle::dialect::DataTypeAttribute") {
+    } else if (attr_type_name == "paddle::dialect::DataTypeAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::DataTypeAttribute>().data());
-      } else if (type_name == "ir::Int32Attribute") {
+    } else if (attr_type_name == "ir::Int32Attribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::Int32Attribute>().data());
-      } else if (type_name == "ir::FloatAttribute") {
+    } else if (attr_type_name == "ir::FloatAttribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::FloatAttribute>().data());
-      } else if (type_name == "ir::BoolAttribute") {
+    } else if (attr_type_name == "ir::BoolAttribute") {
       ctx->EmplaceBackAttr(attr_map[t].dyn_cast<ir::BoolAttribute>().data());
-      } else if (type_name == "paddle::dialect::PlaceAttribute") {
+    } else if (attr_type_name == "paddle::dialect::PlaceAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::PlaceAttribute>().data());
-      } else if (type_name == "paddle::dialect::ScalarAttribute") {
+    } else if (attr_type_name == "paddle::dialect::ScalarAttribute") {
       ctx->EmplaceBackAttr(
           attr_map[t].dyn_cast<paddle::dialect::ScalarAttribute>().data());
     } else {
       PADDLE_THROW(phi::errors::Unimplemented("attr type not support [%s] ",
-                                                type_name));
+                                              attr_type_name));
     }
     VLOG(6) << "ctx->EmplaceBackAttr: " << t;
   }
-  }
 
   // TODO(phlrain): use var type instead of op name
   if (op->attributes().count("op_name") &&

paddle/fluid/ir/phi_kernel_adaptor/phi_kernel_util.h

@@ -33,6 +33,7 @@
 
 #include "paddle/fluid/ir/dialect/kernel_attribute.h"
 #include "paddle/fluid/ir/dialect/pd_attribute.h"
+#include "paddle/fluid/ir/interface/op_yaml_info_parser.h"
 #include "paddle/phi/core/infermeta_utils.h"
 
 #include "glog/logging.h"
@@ -47,14 +48,14 @@ void BuildInferMetaContext(
     ir::Operation* op,
     const std::unordered_map<ir::Value, std::string>& name_map,
     paddle::framework::Scope* scope,
-    const OpInfoTuple& op_yaml_info,
+    const paddle::dialect::OpYamlInfoParser& op_yaml_info,
     phi::InferMetaContext* ctx);
 
 void BuildPhiKernelContext(
     ir::Operation* op,
     const std::unordered_map<ir::Value, std::string>& name_map,
     paddle::framework::Scope* scope,
-    const OpInfoTuple& op_yaml_info,
+    const paddle::dialect::OpYamlInfoParser& op_yaml_info,
     phi::KernelContext* ctx,
     std::map<std::string, std::vector<int>>* input_map = nullptr,
     std::map<std::string, std::vector<int>>* output_map = nullptr);

paddle/phi/api/lib/data_transform.h

@@ -118,7 +118,7 @@ void TransDataBackend(const phi::SelectedRows* tensor,
                       Backend target_backend,
                       phi::SelectedRows* out);
 
-inline bool NeedTransformPlace(const phi::Place& input,
+inline bool NeedTransformPlace(const phi::Place& src_place,
                                const Backend& target,
                                const TransformFlag& transform_flag) {
   // NOTE(dev): The default value of TransformFlag is True, if it is set with
@@ -128,9 +128,9 @@ inline bool NeedTransformPlace(const phi::Place& input,
   if (!transform_flag.need_trans_backend()) {
     return false;
   }
-  bool ret = input.GetType() == AllocationType::GPUPINNED ||
+  bool ret = src_place.GetType() == AllocationType::GPUPINNED ||
              (target != Backend::ALL_BACKEND &&
-              phi::TransToPhiBackend(input) !=
+              phi::TransToPhiBackend(src_place) !=
                   (target != Backend::GPUDNN ? target : Backend::GPU));
   return ret;
 }

test/cpp/ir/core/op_yaml_info_parser_test.cc

@@ -60,14 +60,14 @@ TEST(ir_op_info_test, op_op_info_test) {
   auto kernel_fn_tensor_param = op_yaml_info_parser.KernelFnTensorParams();
   auto kernel_fn_attr_param = op_yaml_info_parser.KernelFnAttrParams();
 
-  EXPECT_EQ(infer_meta_tensor_param.size(), 1u);
-  EXPECT_EQ(infer_meta_attr_param.size(), 1u);
-  EXPECT_EQ(kernel_fn_tensor_param.size(), 3u);
-  EXPECT_EQ(kernel_fn_attr_param.size(), 2u);
+  EXPECT_EQ(infer_meta_tensor_param.size(), 0u);
+  EXPECT_EQ(infer_meta_attr_param.size(), 2u);
+  EXPECT_EQ(kernel_fn_tensor_param.size(), 0u);
+  EXPECT_EQ(kernel_fn_attr_param.size(), 5u);
 
   EXPECT_EQ((op_yaml_info_parser.AttrTypeName("seed") == "ir::Int32Attribute"),
             true);
-  EXPECT_EQ(op_yaml_info_parser.IsTensorArrtibute(0), true);
+  EXPECT_EQ(op_yaml_info_parser.IsTensorAttribute(0), true);
 
   EXPECT_EQ(op_yaml_info_parser.InputTensorNumber(), 0u);
 }