From d67705faaa526a31feab831ac1e5e15ee37880a1 Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Mon, 11 Apr 2011 21:09:18 +0000
Subject: [PATCH] Look pass copies when determining whether hoisting would end
 up inserting more copies. rdar://9266679

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129297 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/CodeGen/MachineLICM.cpp                   | 25 +++++++++-----
 test/CodeGen/ARM/2011-04-11-MachineLICMBug.ll | 34 +++++++++++++++++++
 2 files changed, 51 insertions(+), 8 deletions(-)
 create mode 100644 test/CodeGen/ARM/2011-04-11-MachineLICMBug.ll

diff --git a/lib/CodeGen/MachineLICM.cpp b/lib/CodeGen/MachineLICM.cpp
index 1c0f6ade856..b315702eef8 100644
--- a/lib/CodeGen/MachineLICM.cpp
+++ b/lib/CodeGen/MachineLICM.cpp
@@ -39,7 +39,6 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
-
 using namespace llvm;
 
 STATISTIC(NumHoisted,
@@ -169,6 +168,10 @@ namespace {
     /// 
     bool IsLoopInvariantInst(MachineInstr &I);
 
+    /// HasAnyPHIUse - Return true if the specified register is used by any
+    /// phi node.
+    bool HasAnyPHIUse(unsigned Reg) const;
+
     /// HasHighOperandLatency - Compute operand latency between a def of 'Reg'
     /// and an use in the current loop, return true if the target considered
     /// it 'high'.
@@ -758,18 +761,25 @@ bool MachineLICM::IsLoopInvariantInst(MachineInstr &I) {
 }
 
 
-/// HasPHIUses - Return true if the specified register has any PHI use.
-static bool HasPHIUses(unsigned Reg, MachineRegisterInfo *MRI) {
+/// HasAnyPHIUse - Return true if the specified register is used by any
+/// phi node.
+bool MachineLICM::HasAnyPHIUse(unsigned Reg) const {
   for (MachineRegisterInfo::use_iterator UI = MRI->use_begin(Reg),
          UE = MRI->use_end(); UI != UE; ++UI) {
     MachineInstr *UseMI = &*UI;
     if (UseMI->isPHI())
       return true;
+    // Look pass copies as well.
+    if (UseMI->isCopy()) {
+      unsigned Def = UseMI->getOperand(0).getReg();
+      if (TargetRegisterInfo::isVirtualRegister(Def) &&
+          HasAnyPHIUse(Def))
+        return true;
+    }
   }
   return false;
 }
 
-
 /// HasHighOperandLatency - Compute operand latency between a def of 'Reg'
 /// and an use in the current loop, return true if the target considered
 /// it 'high'.
@@ -976,14 +986,13 @@ bool MachineLICM::IsProfitableToHoist(MachineInstr &MI) {
       return false;
   }
 
-  // If result(s) of this instruction is used by PHIs, then don't hoist it.
-  // The presence of joins makes it difficult for current register allocator
-  // implementation to perform remat.
+  // If result(s) of this instruction is used by PHIs outside of the loop, then
+  // don't hoist it if the instruction because it will introduce an extra copy.
   for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = MI.getOperand(i);
     if (!MO.isReg() || !MO.isDef())
       continue;
-    if (HasPHIUses(MO.getReg(), MRI))
+    if (HasAnyPHIUse(MO.getReg()))
       return false;
   }
 
diff --git a/test/CodeGen/ARM/2011-04-11-MachineLICMBug.ll b/test/CodeGen/ARM/2011-04-11-MachineLICMBug.ll
new file mode 100644
index 00000000000..a9dd97182a4
--- /dev/null
+++ b/test/CodeGen/ARM/2011-04-11-MachineLICMBug.ll
@@ -0,0 +1,34 @@
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 | FileCheck %s
+
+; Overly aggressive LICM simply adds copies of constants
+; rdar://9266679
+
+define zeroext i1 @t(i32* nocapture %A, i32 %size, i32 %value) nounwind readonly ssp {
+; CHECK: t:
+entry:
+  br label %for.cond
+
+for.cond:
+  %0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %cmp = icmp ult i32 %0, %size
+  br i1 %cmp, label %for.body, label %return
+
+for.body:
+; CHECK: %for.body
+; CHECK: movs r{{[0-9]+}}, #1
+  %arrayidx = getelementptr i32* %A, i32 %0
+  %tmp4 = load i32* %arrayidx, align 4
+  %cmp6 = icmp eq i32 %tmp4, %value
+  br i1 %cmp6, label %return, label %for.inc
+
+; CHECK: %for.cond
+; CHECK: movs r{{[0-9]+}}, #0
+
+for.inc:
+  %inc = add i32 %0, 1
+  br label %for.cond
+
+return:
+  %retval.0 = phi i1 [ true, %for.body ], [ false, %for.cond ]
+  ret i1 %retval.0
+}
-- 
2.34.1