[GitHub] [incubator-tvm] tqchen commented on a change in pull request #4616: [REFACTOR] Initialize Unified IR Type Data Structures

[GitBox <gi...@apache.org>]

tqchen commented on a change in pull request #4616: [REFACTOR] Initialize Unified IR Type Data Structures
URL: https://github.com/apache/incubator-tvm/pull/4616#discussion_r363012363
 
 

 ##########
 File path: include/tvm/ir/type.h
 ##########
 @@ -0,0 +1,246 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * \file tvm/ir/type.h
+ * \brief IR/AST nodes for the unified type system in TVM.
+ *
+ * We use Relay's type system as the unified type system
+ * throughout the stack.
+ *
+ * This file contains types that are common across IR variants.
+ *
+ * ## Relation between Type and runtime::DataType
+ *
+ * Besides Type, we also store a dtype field in some of the low-level IR's Expr.
+ * runtime::DataType(dtype) provides coarse grained type information
+ * during compile time and runtime. It is eagerly built in
+ * low-level expression construction and can be used for
+ * quick type checking in the low-level IR.
+ * For example, when an Expr's dtype is int32,
+ * we know for sure that its type is also int32.
+ *
+ * On the other hand, Type provides more fine grained information.
+ * For example, a low level expression can have DataType::Handle() as
+ * its dtype and MemRef[float32] as its type.
+ * Types are usually lazily constructed via type checking,
+ * so they may not readily be available during IR construction.
+ *
+ * The unified Type serves as a common bridge across IR dialects.
+ * For example, we require all the functions to have a type signature,
+ * which allow us to build cross dialect function calls.
+ */
+#ifndef TVM_IR_TYPE_H_
+#define TVM_IR_TYPE_H_
+
+#include <tvm/runtime/object.h>
+#include <tvm/node/node.h>
+#include <tvm/node/container.h>
+#include <tvm/ir/span.h>
+#include <string>
+
+namespace tvm {
+
+/*! \brief Base type of all the types. */
+class TypeNode : public Object {
+ public:
+  /*!
+   * \brief Span that points to the original source code.
+   *        Reserved debug information.
+   */
+  mutable Span span;
+
+  static constexpr const char* _type_key = "relay.Type";
+  TVM_DECLARE_BASE_OBJECT_INFO(TypeNode, Object);
+};
+
+/*!
+ * \brief Type is the base type of all types.
+ *
+ * Relay's type system contains following two key concepts:
+ *
+ * - PrimitiveType: type of primitive type values used in the low-level IR.
+ * - TensorType: type of certain Tensor values in the expression.
+ * - FunctionType: the type of the function.
+ *
+ * There are also advanced types to support generic(polymorphic types),
+ * which can be ignored when first reading the code base.
+ */
+class Type : public ObjectRef {
+ public:
+  TVM_DEFINE_OBJECT_REF_METHODS(Type, ObjectRef, TypeNode);
+};
+
+/*! \brief Possible kinds of TypeVars. */
+enum TypeKind : int {
+  kType = 0,
+  /*! \brief Template variable in shape expression. */
+  kShapeVar = 1,
 
 Review comment:
   I don't know as it was migrated from the original file, we can discuss how to improve in subsequent PRs

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services