From 307473dec0e2fa966037d04725a40b33669dddc8 Mon Sep 17 00:00:00 2001
From: Manman Ren <mren@apple.com>
Date: Fri, 15 Jun 2012 21:32:12 +0000
Subject: [PATCH] ARM: optimization for sub+abs.

This patch will optimize abs(x-y)
FROM
sub, movs, rsbmi
TO
subs, rsbmi

For abs, we will use cmp instead of movs. This is necessary because we already
have an existing peephole pass which optimizes away cmp following sub.

rdar: 11633193


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158551 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/ARM/ARMISelLowering.cpp  | 17 ++++++-----------
 test/CodeGen/ARM/iabs.ll            | 20 +++++++++++++++++++-
 test/CodeGen/ARM/thumb2-it-block.ll |  4 ++--
 3 files changed, 27 insertions(+), 14 deletions(-)

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index c48fa763b40..bf17f15caf4 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -6808,9 +6808,6 @@ ARMTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
     MachineRegisterInfo &MRI = Fn->getRegInfo();
     // In Thumb mode S must not be specified if source register is the SP or
     // PC and if destination register is the SP, so restrict register class
-    unsigned NewMovDstReg = MRI.createVirtualRegister(isThumb2 ?
-      (const TargetRegisterClass*)&ARM::rGPRRegClass :
-      (const TargetRegisterClass*)&ARM::GPRRegClass);
     unsigned NewRsbDstReg = MRI.createVirtualRegister(isThumb2 ?
       (const TargetRegisterClass*)&ARM::rGPRRegClass :
       (const TargetRegisterClass*)&ARM::GPRRegClass);
@@ -6827,12 +6824,10 @@ ARMTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
     // fall through to SinkMBB
     RSBBB->addSuccessor(SinkBB);
 
-    // insert a movs at the end of BB
-    BuildMI(BB, dl, TII->get(isThumb2 ? ARM::t2MOVr : ARM::MOVr),
-      NewMovDstReg)
-      .addReg(ABSSrcReg, RegState::Kill)
-      .addImm((unsigned)ARMCC::AL).addReg(0)
-      .addReg(ARM::CPSR, RegState::Define);
+    // insert a cmp at the end of BB
+    AddDefaultPred(BuildMI(BB, dl, 
+                           TII->get(isThumb2 ? ARM::t2CMPri : ARM::CMPri))
+                   .addReg(ABSSrcReg).addImm(0));
 
     // insert a bcc with opposite CC to ARMCC::MI at the end of BB
     BuildMI(BB, dl,
@@ -6844,7 +6839,7 @@ ARMTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
     // by if-conversion pass
     BuildMI(*RSBBB, RSBBB->begin(), dl,
       TII->get(isThumb2 ? ARM::t2RSBri : ARM::RSBri), NewRsbDstReg)
-      .addReg(NewMovDstReg, RegState::Kill)
+      .addReg(ABSSrcReg, RegState::Kill)
       .addImm(0).addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
 
     // insert PHI in SinkBB,
@@ -6852,7 +6847,7 @@ ARMTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
     BuildMI(*SinkBB, SinkBB->begin(), dl,
       TII->get(ARM::PHI), ABSDstReg)
       .addReg(NewRsbDstReg).addMBB(RSBBB)
-      .addReg(NewMovDstReg).addMBB(BB);
+      .addReg(ABSSrcReg).addMBB(BB);
 
     // remove ABS instruction
     MI->eraseFromParent();
diff --git a/test/CodeGen/ARM/iabs.ll b/test/CodeGen/ARM/iabs.ll
index 89e309d1606..600a8c29ea9 100644
--- a/test/CodeGen/ARM/iabs.ll
+++ b/test/CodeGen/ARM/iabs.ll
@@ -10,7 +10,25 @@ define i32 @test(i32 %a) {
         %b = icmp sgt i32 %a, -1
         %abs = select i1 %b, i32 %a, i32 %tmp1neg
         ret i32 %abs
-; CHECK:  movs r0, r0
+; CHECK:  cmp
 ; CHECK:  rsbmi r0, r0, #0
 ; CHECK:  bx lr
 }
+
+; rdar://11633193
+;; 3 instructions will be generated for abs(a-b):
+;;   subs
+;;   rsbmi
+;;   bx
+define i32 @test2(i32 %a, i32 %b) nounwind readnone ssp {
+entry:
+; CHECK: test2
+; CHECK: subs
+; CHECK-NEXT: rsbmi
+; CHECK-NEXT: bx
+  %sub = sub nsw i32 %a, %b
+  %cmp = icmp sgt i32 %sub, -1
+  %sub1 = sub nsw i32 0, %sub
+  %cond = select i1 %cmp, i32 %sub, i32 %sub1
+  ret i32 %cond
+}
diff --git a/test/CodeGen/ARM/thumb2-it-block.ll b/test/CodeGen/ARM/thumb2-it-block.ll
index 28fd4696535..a25352c0f03 100644
--- a/test/CodeGen/ARM/thumb2-it-block.ll
+++ b/test/CodeGen/ARM/thumb2-it-block.ll
@@ -3,10 +3,10 @@
 
 define i32 @test(i32 %a, i32 %b) {
 entry:
-; CHECK:        movs.w
+; CHECK:        cmp
 ; CHECK-NEXT:   it    mi
 ; CHECK-NEXT:   rsbmi
-; CHECK-NEXT:   movs.w
+; CHECK-NEXT:   cmp
 ; CHECK-NEXT:   it    mi
 ; CHECK-NEXT:   rsbmi
  %cmp1 = icmp slt i32 %a, 0
-- 
2.34.1