From: Gabor Greif <ggreif@gmail.com>
Date: Wed, 29 Sep 2010 10:12:08 +0000 (+0000)
Subject: improve heuristics to find the 'and' corresponding to 'tst' to also catch opportuniti... 
X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=05642a3eba3f35aa8fdf6aa16d87561560e60af3;p=oota-llvm.git

improve heuristics to find the 'and' corresponding to 'tst' to also catch opportunities on thumb2
added some doxygen on the way

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@115033 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index c38b89534bf..26558d900d8 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -1433,16 +1433,29 @@ AnalyzeCompare(const MachineInstr *MI, unsigned &SrcReg, int &CmpMask,
   return false;
 }
 
-static bool isSuitableForMask(const MachineInstr &MI, unsigned SrcReg,
+/// isSuitableForMask - Identify a suitable 'and' instruction that
+/// operates on the given source register and applies the same mask
+/// as a 'tst' instruction. Provide a limited look-through for copies.
+/// When successful, MI will hold the found instruction.
+static bool isSuitableForMask(MachineInstr *&MI, unsigned SrcReg,
                               int CmpMask, bool CommonUse) {
-  switch (MI.getOpcode()) {
+  switch (MI->getOpcode()) {
     case ARM::ANDri:
     case ARM::t2ANDri:
-      if (CmpMask != MI.getOperand(2).getImm())
+      if (CmpMask != MI->getOperand(2).getImm())
         return false;
-      if (SrcReg == MI.getOperand(CommonUse ? 1 : 0).getReg())
+      if (SrcReg == MI->getOperand(CommonUse ? 1 : 0).getReg())
         return true;
       break;
+    case ARM::COPY: {
+      // Walk down one instruction which is potentially an 'and'.
+      const MachineInstr &Copy = *MI;
+      MachineBasicBlock::iterator AND(next(MachineBasicBlock::iterator(MI)));
+      if (AND == MI->getParent()->end()) return false;
+      MI = AND;
+      return isSuitableForMask(MI, Copy.getOperand(0).getReg(),
+                               CmpMask, true);
+    }
   }
 
   return false;
@@ -1467,16 +1480,15 @@ OptimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg, int CmpMask,
 
   // Masked compares sometimes use the same register as the corresponding 'and'.
   if (CmpMask != ~0) {
-      if (!isSuitableForMask(*MI, SrcReg, CmpMask, false)) {
+    if (!isSuitableForMask(MI, SrcReg, CmpMask, false)) {
       MI = 0;
       for (MachineRegisterInfo::use_iterator UI = MRI.use_begin(SrcReg),
            UE = MRI.use_end(); UI != UE; ++UI) {
         if (UI->getParent() != CmpInstr->getParent()) continue;
-        MachineInstr &PotentialAND = *UI;
+        MachineInstr *PotentialAND = &*UI;
         if (!isSuitableForMask(PotentialAND, SrcReg, CmpMask, true))
           continue;
-        SrcReg = PotentialAND.getOperand(0).getReg();
-        MI = &PotentialAND;
+        MI = PotentialAND;
         break;
       }
       if (!MI) return false;
diff --git a/test/CodeGen/ARM/arm-and-tst-peephole.ll b/test/CodeGen/ARM/arm-and-tst-peephole.ll
index 9085b6804c0..8d3c0e013b9 100644
--- a/test/CodeGen/ARM/arm-and-tst-peephole.ll
+++ b/test/CodeGen/ARM/arm-and-tst-peephole.ll
@@ -1,5 +1,6 @@
-; RUN: llc < %s -mtriple=arm-apple-darwin | FileCheck %s
-; RUN: llc < %s -mtriple=thumbv6-apple-darwin | FileCheck -check-prefix=T2 %s
+; RUN: llc < %s -march=arm | FileCheck %s
+; RUN: llc < %s -march=thumb | FileCheck -check-prefix=THUMB %s
+; RUN: llc < %s -march=thumb -mattr=+thumb2 | FileCheck -check-prefix=T2 %s
 
 %struct.Foo = type { i8* }
 
@@ -18,13 +19,16 @@ tailrecurse:                                      ; preds = %sw.bb, %entry
   %0 = ptrtoint i8* %tmp2 to i32
 
 ; CHECK:      ands r12, r12, #3
-; CHECK-NEXT: beq LBB0_2
+; CHECK-NEXT: beq .LBB0_2
 
-; T2:      movs r5, #3
-; T2-NEXT: mov r6, r4
-; T2-NEXT: ands r6, r5
-; T2-NEXT: tst r4, r5
-; T2-NEXT: beq LBB0_3
+; THUMB:      movs r5, #3
+; THUMB-NEXT: mov r6, r4
+; THUMB-NEXT: ands r6, r5
+; THUMB-NEXT: tst r4, r5
+; THUMB-NEXT: beq .LBB0_3
+
+; T2:      ands r12, r12, #3
+; T2-NEXT: beq .LBB0_3
 
   %and = and i32 %0, 3
   %tst = icmp eq i32 %and, 0