[llvm] r345163 - [CodeGen] skip lifetime end marker in isInTailCallPosition

[Robert Lougher via llvm-commits]

Author: rlougher
Date: Wed Oct 24 10:03:19 2018
New Revision: 345163

URL: http://llvm.org/viewvc/llvm-project?rev=345163&view=rev
Log:
[CodeGen] skip lifetime end marker in isInTailCallPosition

A lifetime end intrinsic between a tail call and the return should not
prevent the call from being tail call optimized.

Differential Revision: https://reviews.llvm.org/D53519


Added:
    llvm/trunk/test/CodeGen/X86/tailcall-lifetime-end.ll
Modified:
    llvm/trunk/lib/CodeGen/Analysis.cpp

Modified: llvm/trunk/lib/CodeGen/Analysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Analysis.cpp?rev=345163&r1=345162&r2=345163&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/Analysis.cpp (original)
+++ llvm/trunk/lib/CodeGen/Analysis.cpp Wed Oct 24 10:03:19 2018
@@ -496,6 +496,10 @@ bool llvm::isInTailCallPosition(Immutabl
       // Debug info intrinsics do not get in the way of tail call optimization.
       if (isa<DbgInfoIntrinsic>(BBI))
         continue;
+      // A lifetime end intrinsic should not stop tail call optimization.
+      if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(BBI))
+        if (II->getIntrinsicID() == Intrinsic::lifetime_end)
+          continue;
       if (BBI->mayHaveSideEffects() || BBI->mayReadFromMemory() ||
           !isSafeToSpeculativelyExecute(&*BBI))
         return false;

Added: llvm/trunk/test/CodeGen/X86/tailcall-lifetime-end.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcall-lifetime-end.ll?rev=345163&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/tailcall-lifetime-end.ll (added)
+++ llvm/trunk/test/CodeGen/X86/tailcall-lifetime-end.ll Wed Oct 24 10:03:19 2018
@@ -0,0 +1,27 @@
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -o - %s | FileCheck %s
+
+; A lifetime end intrinsic should not prevent a call from being tail call
+; optimized.
+
+define void @foobar() {
+; CHECK-LABEL: foobar
+; CHECK: pushq	%rax
+; CHECK: leaq	4(%rsp), %rdi
+; CHECK: callq	foo
+; CHECK: popq	%rax
+; CHECK: jmp	bar
+entry:
+  %i = alloca i32
+  %0 = bitcast i32* %i to i8*
+  call void @llvm.lifetime.start.p0i8(i64 4, i8* nonnull %0)
+  call void @foo(i32* nonnull %i)
+  tail call void @bar()
+  call void @llvm.lifetime.end.p0i8(i64 4, i8* nonnull %0)
+  ret void
+}
+
+declare void @foo(i32* nocapture %p)
+declare void @bar()
+
+declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture)
+declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture)