Integrate tir constant nodes in compilation pipeline

This PR integrates tir.allocate_const to the compilation pipeline to support --link-params.

Change-Id: Ic8d0cb75d596299fcae7078b304598afbf0c5494

Co-authored-by: Giuseppe Rossini <giuseros85@gmail.com>
Change-Id: Id98cc682bbfacfe75c4d8b260fd41658f1f196b2

include/tvm/tir/function.h

@@ -151,42 +151,6 @@ class PrimFunc : public BaseFunc {
   TVM_DEFINE_OBJECT_REF_COW_METHOD(PrimFuncNode);
 };
 
-/*!
- * \brief Describes one parameter that should be linked into the generated module.
- *
- * When parameters are to be linked in with generated code (i.e. on target_host-compatible
- * backends), Relay attaches instances of this object to a global TIR function. Code-generators
- * use the information contained in this node to include the parameter data in the generated
- * module.
- */
-class LinkedParamNode : public Object {
- public:
-  /*! \brief Unique numeric identifier used by runtimes to lookup this parameter. */
-  int64_t id;
-
-  /*! \brief Parameter data which should get linked into the final module. */
-  ::tvm::runtime::NDArray param;
-
-  void VisitAttrs(tvm::AttrVisitor* v) {
-    v->Visit("id", &id);
-    v->Visit("param", &param);
-  }
-
-  static constexpr const char* _type_key = "tir.LinkedParam";
-  TVM_DECLARE_FINAL_OBJECT_INFO(LinkedParamNode, Object);
-};
-
-/*!
- * \brief Managed reference to LinkedParamNode.
- */
-class LinkedParam : public ObjectRef {
- public:
-  TVM_DLL LinkedParam(int64_t id, ::tvm::runtime::NDArray param);
-
-  TVM_DEFINE_OBJECT_REF_METHODS(LinkedParam, ObjectRef, LinkedParamNode);
-  TVM_DEFINE_OBJECT_REF_COW_METHOD(LinkedParamNode);
-};
-
 /*!
  * \brief Specialize parameters of PrimFunc.
  * \param func The PrimFunc to be specialized.

include/tvm/tir/stmt.h

@@ -575,6 +575,42 @@ class Allocate : public Stmt {
   TVM_DEFINE_OBJECT_REF_METHODS(Allocate, Stmt, AllocateNode);
 };
 
+/*!
+ * \brief Describes one parameter that should be linked into the generated module.
+ *
+ * When parameters are to be linked in with generated code (i.e. on target_host-compatible
+ * backends), Relay attaches instances of this object to a global TIR function. Code-generators
+ * use the information contained in this node to include the parameter data in the generated
+ * module.
+ */
+class LinkedParamNode : public Object {
+ public:
+  /*! \brief Unique numeric identifier used by runtimes to lookup this parameter. */
+  int64_t id;
+
+  /*! \brief Parameter data which should get linked into the final module. */
+  ::tvm::runtime::NDArray param;
+
+  void VisitAttrs(tvm::AttrVisitor* v) {
+    v->Visit("id", &id);
+    v->Visit("param", &param);
+  }
+
+  static constexpr const char* _type_key = "tir.LinkedParam";
+  TVM_DECLARE_FINAL_OBJECT_INFO(LinkedParamNode, Object);
+};
+
+/*!
+ * \brief Managed reference to LinkedParamNode.
+ */
+class LinkedParam : public ObjectRef {
+ public:
+  TVM_DLL LinkedParam(int64_t id, ::tvm::runtime::NDArray param);
+
+  TVM_DEFINE_OBJECT_REF_METHODS(LinkedParam, ObjectRef, LinkedParamNode);
+  TVM_DEFINE_OBJECT_REF_COW_METHOD(LinkedParamNode);
+};
+
 /*!
  * \brief Allocate a buffer that can be used in body.
  */

include/tvm/tir/transform.h

@@ -25,10 +25,12 @@
 #define TVM_TIR_TRANSFORM_H_
 
 #include <tvm/ir/transform.h>
+#include <tvm/relay/expr.h>
 #include <tvm/tir/expr.h>
 #include <tvm/tir/function.h>
 
 #include <string>
+#include <vector>
 
 namespace tvm {
 namespace tir {
@@ -431,6 +433,8 @@ TVM_DLL Pass LegalizePackedCalls();
  */
 TVM_DLL Pass FlattenBuffer();
 
+TVM_DLL Pass BindParams(const std::vector<const relay::ConstantNode*>& constants);
+
 }  // namespace transform
 }  // namespace tir
 }  // namespace tvm

python/tvm/te/__init__.py

@@ -31,7 +31,7 @@
 from .tensor import TensorSlice, Tensor
 from .tensor_intrin import decl_tensor_intrin
 from .tag import tag_scope
-from .operation import placeholder, compute, scan, extern, var, size_var
+from .operation import placeholder, compute, scan, extern, var, size_var, const
 from .operation import thread_axis, reduce_axis
 from .operation import create_prim_func
 

python/tvm/te/operation.py

@@ -351,6 +351,28 @@ def var(name="tindex", dtype="int32", span=None):
     return tvm.tir.Var(name, dtype, span)
 
 
+def const(name="tindex", dtype="int32", span=None):
+    """Create a new constant with specified name and dtype
+
+    Parameters
+    ----------
+    name : str
+        The name
+
+    dtype : str
+        The data type
+
+    span : Optional[Span]
+        The location of this variable in the source.
+
+    Returns
+    -------
+    var : Var
+        The result symbolic variable.
+    """
+    return tvm.tir.Const(name, dtype, span)
+
+
 def size_var(name="size", dtype="int32", span=None):
     """Create a new variable represents a tensor shape size, which is non-negative.
 

src/relay/backend/aot_executor_codegen.cc

@@ -540,22 +540,20 @@ class AOTExecutorCodegen : public ExprVisitor {
 
   void VisitExpr_(const ConstantNode* op) override {
     Expr expr = GetRef<Expr>(op);
-    size_t index = params_.size();
-    std::string name = "p" + std::to_string(index);
     StorageInfo& sinfo = storage_device_map_[expr];
-    param_storage_ids_[name] = sinfo->storage_ids[0];
-    params_[name] = op->data;
-    params_by_expr_.Set(expr, name);
+    std::stringstream ss;
+    ss << "constant_" << constant_map_.size();
+
+    tir::Var constant(ss.str(), PointerType(PrimType(DataType(op->data->dtype))));
+    constant_map_[constant.operator->()] = op;
 
     // If the Constant node is an output node we need to copy the content of the parameter to the
     // output A Var node can only produce a single output
     auto output_iter = std::find(return_sid_.begin(), return_sid_.end(), sinfo->storage_ids[0]);
     if (output_iter != return_sid_.end()) {
       int output_index = std::distance(return_sid_.begin(), output_iter);
-      auto param_handle = tvm::tir::Call(DataType::Handle(), tvm::tir::builtin::lookup_param(),
-                                         {tir::StringImm(params_by_expr_[expr])});
-      CopyToOutput(main_signature_[input_vars_.size() + output_index], param_handle, false,
-                   sinfo->storage_sizes_in_bytes[0]);
+      CopyToOutput(main_signature_[input_vars_.size() + output_index], constant,
+                   /* pack_input */ false, sinfo->storage_sizes_in_bytes[0]);
     }
   }
 
@@ -632,6 +630,20 @@ class AOTExecutorCodegen : public ExprVisitor {
       }
     }
 
+    for (auto kv : constant_map_) {
+      auto buffer_var = GetRef<tir::Var>(kv.first);
+      auto dtype = DataType(kv.second->data->dtype);
+
+      int ndim = kv.second->data->ndim;
+      Array<PrimExpr> extents;
+
+      for (int i = 0; i < ndim; i++) {
+        int shape = kv.second->data->shape[i];
+        extents.push_back(tir::make_const(DataType::Int(32), shape));
+      }
+      body = tir::AllocateConst(buffer_var, kv.second->data, dtype, extents, body);
+    }
+
     // Define the attributes
     body = tir::AttrStmt(PrimExpr(), tvm::tir::attr::device_type, 1, body);
     body = tir::AttrStmt(PrimExpr(), tvm::tir::attr::device_id, 0, body);
@@ -680,6 +692,7 @@ class AOTExecutorCodegen : public ExprVisitor {
   Map<Expr, String> params_by_expr_;
   /*! \brief mapping between parameter names ("p0", "p1", etc..) and storage identifiers*/
   std::unordered_map<std::string, int64_t> param_storage_ids_;
+  std::unordered_map<const tir::VarNode*, const ConstantNode*> constant_map_;
 
   /*! \brief plan memory of device result */
   StorageMap storage_device_map_;
@@ -783,6 +796,7 @@ class AOTExecutorCodegen : public ExprVisitor {
     } else {
       ret.lowered_funcs.Set(target_host_str, mod_run);
     }
+
     ret.function_metadata = std::move(function_metadata_);
     ret.metadata = runtime::Metadata(input_vars_.size(), return_sid_.size(),
                                      runtime::kTvmExecutorAot, mod_name);

src/relay/backend/build_module.cc

@@ -392,7 +392,13 @@ class RelayBuildModule : public runtime::ModuleNode {
     }
 
     // Fuse the operations if it is needed.
-    relay_module = transform::FuseOps()(relay_module);
+    if (targets.size() == 1) {
+      const auto& it = targets.begin();
+      With<Target> tctx((*it).second);
+      relay_module = transform::FuseOps()(relay_module);
+    } else {
+      relay_module = transform::FuseOps()(relay_module);
+    }
 
     // Do layout rewrite for auto-scheduler.
     if (backend::IsAutoSchedulerEnabled() && targets.size() == 1) {
@@ -540,28 +546,6 @@ class RelayBuildModule : public runtime::ModuleNode {
 
     auto lowered_funcs = executor_codegen_->GetIRModule();
 
-    // Generate a placeholder function that attaches linked params as its arguments.
-    if (target_host->GetAttr<Bool>("link-params").value_or(Bool(false))) {
-      CHECK(pf != nullptr) << "Unable to link-params with no target_host and no llvm codegen.";
-      auto param_ids = executor_codegen_->GetParamIds();
-      auto link_params = Map<String, tir::LinkedParam>();
-      for (auto param : ret_.params) {
-        link_params.Set(param.first, tir::LinkedParam(param_ids[param.first], param.second));
-      }
-
-      Map<String, ObjectRef> dict;
-      dict.Set(tvm::tir::attr::kLinkedParams, link_params);
-      dict.Set(tvm::attr::kGlobalSymbol, String(::tvm::runtime::symbol::tvm_lookup_linked_param));
-      DictAttrs attrs{dict};
-      auto prim = tir::PrimFunc(Array<tir::Var>(), tir::SeqStmt(Array<tir::Stmt>()), VoidType(),
-                                Map<tir::Var, tir::Buffer>(), attrs);
-      if (lowered_funcs.find(target_host->str()) == lowered_funcs.end()) {
-        lowered_funcs.Set(target_host->str(), IRModule(Map<GlobalVar, BaseFunc>({})));
-      }
-      lowered_funcs[target_host->str()]->Add(
-          GlobalVar(::tvm::runtime::symbol::tvm_lookup_linked_param), prim);
-    }
-
     // When there is no lowered_funcs due to reasons such as optimization.
     if (lowered_funcs.size() == 0) {
       if (target_host.defined() && target_host->kind->name == "llvm") {

src/relay/backend/compile_engine.cc

@@ -35,6 +35,7 @@
 #include <tvm/te/operation.h>
 #include <tvm/te/schedule.h>
 #include <tvm/te/schedule_pass.h>
+#include <tvm/tir/transform.h>
 #include <tvm/topi/tags.h>
 
 #include <functional>
@@ -221,10 +222,17 @@ class CompileEngineImpl : public CompileEngineNode {
     for (te::Tensor arg : cfunc->outputs) {
       all_args.push_back(arg);
     }
+    std::vector<const ConstantNode*> all_consts;
+    for (auto kv : cfunc->constant_tensors) {
+      all_args.push_back(kv.second);
+      all_consts.push_back(kv.first);
+    }
+
     // lower the function
     std::unordered_map<te::Tensor, tir::Buffer> binds;
     auto func_name = cfunc->prim_fn_var->name_hint;
     cfunc->funcs->Update(tvm::LowerSchedule(cfunc->schedule, all_args, func_name, binds));
+    cfunc->funcs->Update(tir::transform::BindParams(all_consts)(cfunc->funcs));
     value->cached_func = cfunc;
 
     return value;

src/relay/backend/te_compiler.cc

@@ -33,6 +33,7 @@
 #include <tvm/te/operation.h>
 #include <tvm/te/schedule.h>
 #include <tvm/te/schedule_pass.h>
+#include <tvm/tir/transform.h>
 #include <tvm/topi/tags.h>
 
 #include <functional>
@@ -222,10 +223,16 @@ class TECompilerImpl : public TECompilerNode {
     for (te::Tensor arg : cfunc->outputs) {
       all_args.push_back(arg);
     }
+    std::vector<const ConstantNode*> all_consts;
+    for (auto kv : cfunc->constant_tensors) {
+      all_args.push_back(kv.second);
+      all_consts.push_back(kv.first);
+    }
 
     std::unordered_map<te::Tensor, tir::Buffer> binds;
     auto func_name = cfunc->prim_fn_var->name_hint;
     cfunc->funcs->Update(tvm::LowerSchedule(cfunc->schedule, all_args, func_name, binds));
+    cfunc->funcs->Update(tir::transform::BindParams(all_consts)(cfunc->funcs));
     value->cached_func = cfunc;
     return value;
   }