[llvm] r194102 - Slightly change the way stackmap and patchpoint intrinsics are lowered.

[Andrew Trick]

Author: atrick
Date: Tue Nov  5 16:44:04 2013
New Revision: 194102

URL: http://llvm.org/viewvc/llvm-project?rev=194102&view=rev
Log:
Slightly change the way stackmap and patchpoint intrinsics are lowered.

MorphNodeTo is not safe to call during DAG building. It eagerly
deletes dependent DAG nodes which invalidates the NodeMap. We could
expose a safe interface for morphing nodes, but I don't think it's
worth it. Just create a new MachineNode and replaceAllUsesWith.

My understaning of the SD design has been that we want to support
early target opcode selection. That isn't very well supported, but
generally works. It seems reasonable to rely on this feature even if
it isn't widely used.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
    llvm/trunk/test/CodeGen/X86/patchpoint.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=194102&r1=194101&r2=194102&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Tue Nov  5 16:44:04 2013
@@ -6806,11 +6806,19 @@ void SelectionDAGBuilder::visitStackmap(
   if (hasGlue)
     Ops.push_back(*(Call->op_end()-1));
 
-  // Replace the target specific call node with STACKMAP in-place. This way we
-  // don't have to call ReplaceAllUsesWith and STACKMAP will take the call's
-  // place in the chain.
   SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
-  DAG.SelectNodeTo(Call, TargetOpcode::STACKMAP, NodeTys, &Ops[0], Ops.size());
+
+  // Replace the target specific call node with a STACKMAP node.
+  MachineSDNode *MN = DAG.getMachineNode(TargetOpcode::STACKMAP, getCurSDLoc(),
+                                         NodeTys, Ops);
+
+  // StackMap generates no value, so nothing goes in the NodeMap.
+
+  // Fixup the consumers of the intrinsic. The chain and glue may be used in the
+  // call sequence.
+  DAG.ReplaceAllUsesWith(Call, MN);
+
+  DAG.DeleteNode(Call);
 }
 
 /// \brief Lower llvm.experimental.patchpoint directly to its target opcode.
@@ -6898,12 +6906,22 @@ void SelectionDAGBuilder::visitPatchpoin
   if (hasGlue)
     Ops.push_back(*(Call->op_end()-1));
 
-  // Replace the target specific call node with PATCHPOINT in-place. This
-  // way we don't have to call ReplaceAllUsesWith and PATCHPOINT will
-  // take the call's place in the chain.
   SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
-  DAG.SelectNodeTo(Call, TargetOpcode::PATCHPOINT, NodeTys, &Ops[0],
-                   Ops.size());
+
+  // Replace the target specific call node with a STACKMAP node.
+  MachineSDNode *MN = DAG.getMachineNode(TargetOpcode::PATCHPOINT,
+                                         getCurSDLoc(), NodeTys, Ops);
+
+  // PatchPoint generates no value, so nothing goes in the NodeMap.
+  //
+  // FIXME: with anyregcc calling convention it will need to be in the NodeMap
+  // and replace values.
+
+  // Fixup the consumers of the intrinsic. The chain and glue may be used in the
+  // call sequence.
+  DAG.ReplaceAllUsesWith(Call, MN);
+
+  DAG.DeleteNode(Call);
 }
 
 /// TargetLowering::LowerCallTo - This is the default LowerCallTo

Modified: llvm/trunk/test/CodeGen/X86/patchpoint.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/patchpoint.ll?rev=194102&r1=194101&r2=194102&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/patchpoint.ll (original)
+++ llvm/trunk/test/CodeGen/X86/patchpoint.ll Tue Nov  5 16:44:04 2013
@@ -63,6 +63,26 @@ entry:
   ret void
 }
 
+; Test patchpoints reusing the same TargetConstant.
+; <rdar:15390785> Assertion failed: (CI.getNumArgOperands() >= NumArgs + 4)
+; There is no way to verify this, since it depends on memory allocation.
+; But I think it's useful to include as a working example.
+define i64 @testLowerConstant(i64 %arg, i64 %tmp2, i64 %tmp10, i64* %tmp33, i64 %tmp79) {
+entry:
+  %tmp80 = add i64 %tmp79, -16
+  %tmp81 = inttoptr i64 %tmp80 to i64*
+  %tmp82 = load i64* %tmp81, align 8
+  tail call void (i32, i32, ...)* @llvm.experimental.stackmap(i32 14, i32 5, i64 %arg, i64 %tmp2, i64 %tmp10, i64 %tmp82)
+  tail call void (i32, i32, i8*, i32, ...)* @llvm.experimental.patchpoint.void(i32 15, i32 30, i8* null, i32 3, i64 %arg, i64 %tmp10, i64 %tmp82)
+  %tmp83 = load i64* %tmp33, align 8
+  %tmp84 = add i64 %tmp83, -24
+  %tmp85 = inttoptr i64 %tmp84 to i64*
+  %tmp86 = load i64* %tmp85, align 8
+  tail call void (i32, i32, ...)* @llvm.experimental.stackmap(i32 17, i32 5, i64 %arg, i64 %tmp10, i64 %tmp86)
+  tail call void (i32, i32, i8*, i32, ...)* @llvm.experimental.patchpoint.void(i32 18, i32 30, i8* null, i32 3, i64 %arg, i64 %tmp10, i64 %tmp86)
+  ret i64 10
+}
+
 declare void @llvm.experimental.stackmap(i32, i32, ...)
 declare void @llvm.experimental.patchpoint.void(i32, i32, i8*, i32, ...)
 declare i64 @llvm.experimental.patchpoint.i64(i32, i32, i8*, i32, ...)