[IR] Introduce dead_on_return attribute

Add `dead_on_return` attribute, which is meant to be taken advantage
by the frontend, and states that the memory pointed to by the argument
is dead upon function return. As with `byval`, it is supposed to be
used for passing aggregates by value. The difference lies in the ABI:
`byval` implies that the pointer is explicitly passed as argument to
the callee (during codegen the copy is emitted as per byval contract),
whereas a `dead_on_return`-marked argument implies that the copy
already exists in the IR, is located at a specific stack offset within
the caller, and this memory will not be read further by the caller upon
callee return – or otherwise poison, if read before being written.

RFC: https://discourse.llvm.org/t/rfc-add-dead-on-return-attribute/86871.

Author: antoniofrighetto

llvm/docs/LangRef.rst

@@ -1741,6 +1741,23 @@ Currently, only the following parameter attributes are defined:
 
     This attribute cannot be applied to return values.
 
+``dead_on_return``
+    This attribute indicates that the memory pointed to by the argument is dead
+    upon function return, both upon normal return and if the calls unwinds, meaning
+    that the caller will not depend on its contents. Stores that would be observable
+    either on the return path or on the unwind path may be elided.
+
+    Specifically, the behavior is as-if any memory written through the pointer
+    during the execution of the function is overwritten with a poison value
+    upon function return. The caller may access the memory, but any load
+    not preceded by a store will return poison.
+
+    This attribute does not imply aliasing properties. For pointer arguments that
+    do not alias other memory locations, ``noalias`` attribute may be used in
+    conjunction. Conversely, this attribute always implies ``dead_on_unwind``.
+
+    This attribute cannot be applied to return values.
+
 ``range(<ty> <a>, <b>)``
     This attribute expresses the possible range of the parameter or return value.
     If the value is not in the specified range, it is converted to poison.

llvm/include/llvm/Bitcode/LLVMBitCodes.h

@@ -798,6 +798,7 @@ enum AttributeKindCodes {
   ATTR_KIND_NO_DIVERGENCE_SOURCE = 100,
   ATTR_KIND_SANITIZE_TYPE = 101,
   ATTR_KIND_CAPTURES = 102,
+  ATTR_KIND_DEAD_ON_RETURN = 103,
 };
 
 enum ComdatSelectionKindCodes {

llvm/include/llvm/IR/Argument.h

@@ -78,6 +78,9 @@ class Argument final : public Value {
   /// Return true if this argument has the byval attribute.
   LLVM_ABI bool hasByValAttr() const;
 
+  /// Return true if this argument has the dead_on_return attribute.
+  LLVM_ABI bool hasDeadOnReturnAttr() const;
+
   /// Return true if this argument has the byref attribute.
   LLVM_ABI bool hasByRefAttr() const;
 

llvm/include/llvm/IR/Attributes.td

@@ -198,6 +198,9 @@ def NoFree : EnumAttr<"nofree", IntersectAnd, [FnAttr, ParamAttr]>;
 /// Argument is dead if the call unwinds.
 def DeadOnUnwind : EnumAttr<"dead_on_unwind", IntersectAnd, [ParamAttr]>;
 
+/// Argument is dead upon function return.
+def DeadOnReturn : EnumAttr<"dead_on_return", IntersectAnd, [ParamAttr]>;
+
 /// Disable implicit floating point insts.
 def NoImplicitFloat : EnumAttr<"noimplicitfloat", IntersectPreserve, [FnAttr]>;
 

llvm/lib/Bitcode/Reader/BitcodeReader.cpp

@@ -2244,6 +2244,8 @@ static Attribute::AttrKind getAttrFromCode(uint64_t Code) {
     return Attribute::NoExt;
   case bitc::ATTR_KIND_CAPTURES:
     return Attribute::Captures;
+  case bitc::ATTR_KIND_DEAD_ON_RETURN:
+    return Attribute::DeadOnReturn;
   }
 }
 

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

@@ -938,6 +938,8 @@ static uint64_t getAttrKindEncoding(Attribute::AttrKind Kind) {
     return bitc::ATTR_KIND_NO_EXT;
   case Attribute::Captures:
     return bitc::ATTR_KIND_CAPTURES;
+  case Attribute::DeadOnReturn:
+    return bitc::ATTR_KIND_DEAD_ON_RETURN;
   case Attribute::EndAttrKinds:
     llvm_unreachable("Can not encode end-attribute kinds marker.");
   case Attribute::None:

llvm/lib/IR/Attributes.cpp

@@ -2424,7 +2424,8 @@ AttributeMask AttributeFuncs::typeIncompatible(Type *Ty, AttributeSet AS,
           .addAttribute(Attribute::Writable)
           .addAttribute(Attribute::DeadOnUnwind)
           .addAttribute(Attribute::Initializes)
-          .addAttribute(Attribute::Captures);
+          .addAttribute(Attribute::Captures)
+          .addAttribute(Attribute::DeadOnReturn);
     if (ASK & ASK_UNSAFE_TO_DROP)
       Incompatible.addAttribute(Attribute::Nest)
           .addAttribute(Attribute::SwiftError)

llvm/lib/IR/Function.cpp

@@ -130,6 +130,12 @@ bool Argument::hasByValAttr() const {
   return hasAttribute(Attribute::ByVal);
 }
 
+bool Argument::hasDeadOnReturnAttr() const {
+  if (!getType()->isPointerTy())
+    return false;
+  return hasAttribute(Attribute::DeadOnReturn);
+}
+
 bool Argument::hasByRefAttr() const {
   if (!getType()->isPointerTy())
     return false;

llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp

@@ -1017,10 +1017,10 @@ struct DSEState {
       }
     }
 
-    // Treat byval or inalloca arguments the same as Allocas, stores to them are
-    // dead at the end of the function.
+    // Treat byval, inalloca or dead on return arguments the same as Allocas,
+    // stores to them are dead at the end of the function.
     for (Argument &AI : F.args())
-      if (AI.hasPassPointeeByValueCopyAttr())
+      if (AI.hasPassPointeeByValueCopyAttr() || AI.hasDeadOnReturnAttr())
         InvisibleToCallerAfterRet.insert({&AI, true});
 
     // Collect whether there is any irreducible control flow in the function.

llvm/lib/Transforms/Utils/CodeExtractor.cpp

@@ -1020,6 +1020,7 @@ Function *CodeExtractor::constructFunctionDeclaration(
       case Attribute::EndAttrKinds:
       case Attribute::EmptyKey:
       case Attribute::TombstoneKey:
+      case Attribute::DeadOnReturn:
         llvm_unreachable("Not a function attribute");
       }