[ARMv8] Prevent generation of deprecated IT blocks on ARMv8 in Thumb mode.

IT blocks can only be one instruction lonf, and can only contain a subset of
the 16 instructions.

Patch by Artyom Skrobov!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190309 91177308-0d34-0410-b5e6-96231b3b80d8

15 files changed, 428 insertions, 59 deletions

diff --git a/lib/CodeGen/IfConversion.cpp b/lib/CodeGen/IfConversion.cpp
index 1ae7e3b..418994d 100644
--- a/lib/CodeGen/IfConversion.cpp
+++ b/lib/CodeGen/IfConversion.cpp
@@ -666,32 +666,29 @@ void IfConverter::ScanInstructions(BBInfo &BBI) {
     bool isPredicated = TII->isPredicated(I);
     bool isCondBr = BBI.IsBrAnalyzable && I->isConditionalBranch();
 
-    if (!isCondBr) {
-      if (!isPredicated) {
-        BBI.NonPredSize++;
-        unsigned ExtraPredCost = 0;
-        unsigned NumCycles = TII->getInstrLatency(InstrItins, &*I,
-                                                  &ExtraPredCost);
-        if (NumCycles > 1)
-          BBI.ExtraCost += NumCycles-1;
-        BBI.ExtraCost2 += ExtraPredCost;
-      } else if (!AlreadyPredicated) {
-        // FIXME: This instruction is already predicated before the
-        // if-conversion pass. It's probably something like a conditional move.
-        // Mark this block unpredicable for now.
-        BBI.IsUnpredicable = true;
-        return;
-      }
+    // A conditional branch is not predicable, but it may be eliminated.
+    if (isCondBr)
+      continue;
+
+    if (!isPredicated) {
+      BBI.NonPredSize++;
+      unsigned ExtraPredCost = 0;
+      unsigned NumCycles = TII->getInstrLatency(InstrItins, &*I,
+                                                &ExtraPredCost);
+      if (NumCycles > 1)
+        BBI.ExtraCost += NumCycles-1;
+      BBI.ExtraCost2 += ExtraPredCost;
+    } else if (!AlreadyPredicated) {
+      // FIXME: This instruction is already predicated before the
+      // if-conversion pass. It's probably something like a conditional move.
+      // Mark this block unpredicable for now.
+      BBI.IsUnpredicable = true;
+      return;
     }
 
     if (BBI.ClobbersPred && !isPredicated) {
       // Predicate modification instruction should end the block (except for
       // already predicated instructions and end of block branches).
-      if (isCondBr) {
-        // A conditional branch is not predicable, but it may be eliminated.
-        continue;
-      }
-
       // Predicate may have been modified, the subsequent (currently)
       // unpredicated instructions cannot be correctly predicated.
       BBI.IsUnpredicable = true;
diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 62e8063..696d039 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -513,6 +513,74 @@ bool ARMBaseInstrInfo::DefinesPredicate(MachineInstr *MI,
   return Found;
 }
 
+static bool isV8EligibleForIT(MachineInstr *MI) {
+  switch (MI->getOpcode()) {
+  default:
+    return false;
+  case ARM::tADC:
+  case ARM::tADDi3:
+  case ARM::tADDi8:
+  case ARM::tADDrSPi:
+  case ARM::tADDrr:
+  case ARM::tAND:
+  case ARM::tASRri:
+  case ARM::tASRrr:
+  case ARM::tBIC:
+  case ARM::tCMNz:
+  case ARM::tCMPi8:
+  case ARM::tCMPr:
+  case ARM::tEOR:
+  case ARM::tLDRBi:
+  case ARM::tLDRBr:
+  case ARM::tLDRHi:
+  case ARM::tLDRHr:
+  case ARM::tLDRSB:
+  case ARM::tLDRSH:
+  case ARM::tLDRi:
+  case ARM::tLDRr:
+  case ARM::tLDRspi:
+  case ARM::tLSLri:
+  case ARM::tLSLrr:
+  case ARM::tLSRri:
+  case ARM::tLSRrr:
+  case ARM::tMOVi8:
+  case ARM::tMUL:
+  case ARM::tMVN:
+  case ARM::tORR:
+  case ARM::tROR:
+  case ARM::tRSB:
+  case ARM::tSBC:
+  case ARM::tSTRBi:
+  case ARM::tSTRBr:
+  case ARM::tSTRHi:
+  case ARM::tSTRHr:
+  case ARM::tSTRi:
+  case ARM::tSTRr:
+  case ARM::tSTRspi:
+  case ARM::tSUBi3:
+  case ARM::tSUBi8:
+  case ARM::tSUBrr:
+  case ARM::tTST:
+    return true;
+// there are some "conditionally deprecated" opcodes
+  case ARM::tADDspr:
+    return MI->getOperand(2).getReg() != ARM::PC;
+  case ARM::tADDrSP:
+  case ARM::tBX:
+  case ARM::tBLXr:
+  // ADD PC, SP and BLX PC were always unpredictable,
+  // now on top of it they're deprecated
+    return MI->getOperand(0).getReg() != ARM::PC;
+  case ARM::tADDhirr:
+    return MI->getOperand(0).getReg() != ARM::PC &&
+           MI->getOperand(2).getReg() != ARM::PC;
+  case ARM::tCMPhir:
+  case ARM::tMOVr:
+    return MI->getOperand(0).getReg() != ARM::PC &&
+           MI->getOperand(1).getReg() != ARM::PC;
+  }
+}
+
 /// isPredicable - Return true if the specified instruction can be predicated.
 /// By default, this returns true for every instruction with a
 /// PredicateOperand.
@@ -520,11 +588,17 @@ bool ARMBaseInstrInfo::isPredicable(MachineInstr *MI) const {
   if (!MI->isPredicable())
     return false;
 
-  if ((MI->getDesc().TSFlags & ARMII::DomainMask) == ARMII::DomainNEON) {
-    ARMFunctionInfo *AFI =
-      MI->getParent()->getParent()->getInfo<ARMFunctionInfo>();
-    return AFI->isThumb2Function();
+  ARMFunctionInfo *AFI =
+    MI->getParent()->getParent()->getInfo<ARMFunctionInfo>();
+
+  if (AFI->isThumb2Function()) {
+    if (getSubtarget().hasV8Ops())
+      return isV8EligibleForIT(MI);
+  } else { // non-Thumb
+    if ((MI->getDesc().TSFlags & ARMII::DomainMask) == ARMII::DomainNEON)
+      return false;
   }
+
   return true;
 }
 
diff --git a/lib/Target/ARM/ARMFastISel.cpp b/lib/Target/ARM/ARMFastISel.cpp
index 3cbb9a8..f3a74c7 100644
--- a/lib/Target/ARM/ARMFastISel.cpp
+++ b/lib/Target/ARM/ARMFastISel.cpp
@@ -254,10 +254,10 @@ bool ARMFastISel::DefinesOptionalPredicate(MachineInstr *MI, bool *CPSR) {
 bool ARMFastISel::isARMNEONPred(const MachineInstr *MI) {
   const MCInstrDesc &MCID = MI->getDesc();
 
-  // If we're a thumb2 or not NEON function we were handled via isPredicable.
+  // If we're a thumb2 or not NEON function we'll be handled via isPredicable.
   if ((MCID.TSFlags & ARMII::DomainMask) != ARMII::DomainNEON ||
        AFI->isThumb2Function())
-    return false;
+    return MI->isPredicable();
 
   for (unsigned i = 0, e = MCID.getNumOperands(); i != e; ++i)
     if (MCID.OpInfo[i].isPredicate())
@@ -278,7 +278,7 @@ ARMFastISel::AddOptionalDefs(const MachineInstrBuilder &MIB) {
   // Do we use a predicate? or...
   // Are we NEON in ARM mode and have a predicate operand? If so, I know
   // we're not predicable but add it anyways.
-  if (TII.isPredicable(MI) || isARMNEONPred(MI))
+  if (isARMNEONPred(MI))
     AddDefaultPred(MIB);
 
   // Do we optionally set a predicate?  Preds is size > 0 iff the predicate
diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 1ba78e4..c78db54 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -196,8 +196,13 @@ bool ARMPassConfig::addPreSched2() {
   addPass(createARMExpandPseudoPass());
 
   if (getOptLevel() != CodeGenOpt::None) {
-    if (!getARMSubtarget().isThumb1Only())
+    if (!getARMSubtarget().isThumb1Only()) {
+      // in v8, IfConversion depends on Thumb instruction widths
+      if (getARMSubtarget().hasV8Ops() &&
+          !getARMSubtarget().prefers32BitThumb())
+        addPass(createThumb2SizeReductionPass());
       addPass(&IfConverterID);
+    }
   }
   if (getARMSubtarget().isThumb2())
     addPass(createThumb2ITBlockPass());
diff --git a/lib/Target/ARM/Thumb2ITBlockPass.cpp b/lib/Target/ARM/Thumb2ITBlockPass.cpp
index d8596d7..8f1ae2e 100644
--- a/lib/Target/ARM/Thumb2ITBlockPass.cpp
+++ b/lib/Target/ARM/Thumb2ITBlockPass.cpp
@@ -28,6 +28,7 @@ namespace {
     static char ID;
     Thumb2ITBlockPass() : MachineFunctionPass(ID) {}
 
+    bool hasV8Ops;
     const Thumb2InstrInfo *TII;
     const TargetRegisterInfo *TRI;
     ARMFunctionInfo *AFI;
@@ -192,37 +193,41 @@ bool Thumb2ITBlockPass::InsertITInstructions(MachineBasicBlock &MBB) {
     // Form IT block.
     ARMCC::CondCodes OCC = ARMCC::getOppositeCondition(CC);
     unsigned Mask = 0, Pos = 3;
-    // Branches, including tricky ones like LDM_RET, need to end an IT
-    // block so check the instruction we just put in the block.
-    for (; MBBI != E && Pos &&
-           (!MI->isBranch() && !MI->isReturn()) ; ++MBBI) {
-      if (MBBI->isDebugValue())
-        continue;
-
-      MachineInstr *NMI = &*MBBI;
-      MI = NMI;
-
-      unsigned NPredReg = 0;
-      ARMCC::CondCodes NCC = getITInstrPredicate(NMI, NPredReg);
-      if (NCC == CC || NCC == OCC) {
-        Mask |= (NCC & 1) << Pos;
-        // Add implicit use of ITSTATE.
-        NMI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
-                                               true/*isImp*/, false/*isKill*/));
-        LastITMI = NMI;
-      } else {
-        if (NCC == ARMCC::AL &&
-            MoveCopyOutOfITBlock(NMI, CC, OCC, Defs, Uses)) {
-          --MBBI;
-          MBB.remove(NMI);
-          MBB.insert(InsertPos, NMI);
-          ++NumMovedInsts;
+
+    // v8 IT blocks are limited to one conditional op: skip the loop
+    if (!hasV8Ops) {
+      // Branches, including tricky ones like LDM_RET, need to end an IT
+      // block so check the instruction we just put in the block.
+      for (; MBBI != E && Pos &&
+             (!MI->isBranch() && !MI->isReturn()) ; ++MBBI) {
+        if (MBBI->isDebugValue())
           continue;
+
+        MachineInstr *NMI = &*MBBI;
+        MI = NMI;
+
+        unsigned NPredReg = 0;
+        ARMCC::CondCodes NCC = getITInstrPredicate(NMI, NPredReg);
+        if (NCC == CC || NCC == OCC) {
+          Mask |= (NCC & 1) << Pos;
+          // Add implicit use of ITSTATE.
+          NMI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
+                                                 true/*isImp*/, false/*isKill*/));
+          LastITMI = NMI;
+        } else {
+          if (NCC == ARMCC::AL &&
+              MoveCopyOutOfITBlock(NMI, CC, OCC, Defs, Uses)) {
+            --MBBI;
+            MBB.remove(NMI);
+            MBB.insert(InsertPos, NMI);
+            ++NumMovedInsts;
+            continue;
+          }
+          break;
         }
-        break;
+        TrackDefUses(NMI, Defs, Uses, TRI);
+        --Pos;
       }
-      TrackDefUses(NMI, Defs, Uses, TRI);
-      --Pos;
     }
 
     // Finalize IT mask.
@@ -250,6 +255,7 @@ bool Thumb2ITBlockPass::runOnMachineFunction(MachineFunction &Fn) {
   AFI = Fn.getInfo<ARMFunctionInfo>();
   TII = static_cast<const Thumb2InstrInfo*>(TM.getInstrInfo());
   TRI = TM.getRegisterInfo();
+  hasV8Ops = TM.getSubtarget<ARMSubtarget>().hasV8Ops();
 
   if (!AFI->isThumbFunction())
     return false;
diff --git a/test/CodeGen/ARM/2013-05-05-IfConvertBug.ll b/test/CodeGen/ARM/2013-05-05-IfConvertBug.ll
index 2eeebac..8bc8cb1 100644
--- a/test/CodeGen/ARM/2013-05-05-IfConvertBug.ll
+++ b/test/CodeGen/ARM/2013-05-05-IfConvertBug.ll
@@ -1,4 +1,5 @@
 ; RUN: llc < %s -mtriple=thumbv7-apple-ios -mcpu=cortex-a8 | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck -check-prefix=CHECK-V8 %s
 ; rdar://13782395
 
 define i32 @t1(i32 %a, i32 %b, i8** %retaddr) {
@@ -100,6 +101,27 @@ KBBlockZero.exit:                                 ; preds = %bb2.i
 ; CHECK: [[LABEL]]:
 ; CHECK-NEXT: subs r0, r1, r0
 ; CHECK-NEXT: bx lr
+
+; CHECK-V8-LABEL: wrapDistance:
+; CHECK-V8: cmp r1, #59
+; CHECK-V8-NEXT: bgt
+; CHECK-V8-NEXT: %if.then
+; CHECK-V8-NEXT: subs r0, r2, #1
+; CHECK-V8-NEXT: bx lr
+; CHECK-V8-NEXT: %if.else
+; CHECK-V8-NEXT: subs [[REG:r[0-9]+]], #120
+; CHECK-V8-NEXT: cmp [[REG]], r1
+; CHECK-V8-NEXT: bge
+; CHECK-V8-NEXT: %if.else
+; CHECK-V8-NEXT: cmp r0, #119
+; CHECK-V8-NEXT: bgt
+; CHECK-V8-NEXT: %if.then4
+; CHECK-V8-NEXT: adds r0, r1, #1
+; CHECK-V8-NEXT: bx lr
+; CHECK-V8-NEXT: %if.end5
+; CHECK-V8-NEXT: subs r0, r1, r0
+; CHECK-V8-NEXT: bx lr
+
 define i32 @wrapDistance(i32 %tx, i32 %sx, i32 %w) {
 entry:
   %cmp = icmp slt i32 %sx, 60
diff --git a/test/CodeGen/ARM/arm-and-tst-peephole.ll b/test/CodeGen/ARM/arm-and-tst-peephole.ll
index 0762070..88d797e 100644
--- a/test/CodeGen/ARM/arm-and-tst-peephole.ll
+++ b/test/CodeGen/ARM/arm-and-tst-peephole.ll
@@ -1,6 +1,7 @@
 ; RUN: llc < %s -march=arm | FileCheck -check-prefix=ARM %s
 ; RUN: llc < %s -march=thumb | FileCheck -check-prefix=THUMB %s
 ; RUN: llc < %s -march=thumb -mattr=+thumb2 | FileCheck -check-prefix=T2 %s
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck -check-prefix=V8 %s
 
 ; FIXME: The -march=thumb test doesn't change if -disable-peephole is specified.
 
@@ -39,6 +40,17 @@ tailrecurse:                                      ; preds = %sw.bb, %entry
   br i1 %tst, label %sw.bb, label %tailrecurse.switch
 
 tailrecurse.switch:                               ; preds = %tailrecurse
+; V8-LABEL: %tailrecurse.switch
+; V8: cmp
+; V8-NEXT: beq
+; V8-NEXT: %tailrecurse.switch
+; V8: cmp
+; V8-NEXT: beq
+; V8-NEXT: %tailrecurse.switch
+; V8: cmp
+; V8-NEXT: beq
+; V8-NEXT: b	
+; The trailing space in the last line checks that the branch is unconditional
   switch i32 %and, label %sw.epilog [
     i32 1, label %sw.bb
     i32 3, label %sw.bb6
@@ -73,6 +85,7 @@ sw.epilog:                                        ; preds = %tailrecurse.switch
 ; ARM: bar
 ; THUMB: bar
 ; T2: bar
+; V8-LABEL: bar:
 define internal zeroext i8 @bar(%struct.S* %x, %struct.S* nocapture %y) nounwind readonly {
 entry:
   %0 = getelementptr inbounds %struct.S* %x, i32 0, i32 1, i32 0
@@ -81,22 +94,32 @@ entry:
 ; ARM: ands
 ; THUMB: ands
 ; T2: ands
+; V8: ands
+; V8-NEXT: beq
   %3 = and i32 %2, 112
   %4 = icmp eq i32 %3, 0
   br i1 %4, label %return, label %bb
 
 bb:                                               ; preds = %entry
+; V8-NEXT: %bb
   %5 = getelementptr inbounds %struct.S* %y, i32 0, i32 1, i32 0
   %6 = load i8* %5, align 1
   %7 = zext i8 %6 to i32
 ; ARM: andsne
 ; THUMB: ands
 ; T2: andsne
+; V8: ands
+; V8-NEXT: beq
   %8 = and i32 %7, 112
   %9 = icmp eq i32 %8, 0
   br i1 %9, label %return, label %bb2
 
 bb2:                                              ; preds = %bb
+; V8-NEXT: %bb2
+; V8-NEXT: cmp
+; V8-NEXT: it	ne
+; V8-NEXT: cmpne
+; V8-NEXT: bne
   %10 = icmp eq i32 %3, 16
   %11 = icmp eq i32 %8, 16
   %or.cond = or i1 %10, %11
diff --git a/test/CodeGen/ARM/fast-isel-select.ll b/test/CodeGen/ARM/fast-isel-select.ll
index dbec1c6..40f8807 100644
--- a/test/CodeGen/ARM/fast-isel-select.ll
+++ b/test/CodeGen/ARM/fast-isel-select.ll
@@ -1,6 +1,7 @@
 ; RUN: llc < %s -O0 -verify-machineinstrs -fast-isel-abort -relocation-model=dynamic-no-pic -mtriple=armv7-apple-ios | FileCheck %s --check-prefix=ARM
 ; RUN: llc < %s -O0 -verify-machineinstrs -fast-isel-abort -relocation-model=dynamic-no-pic -mtriple=armv7-linux-gnueabi | FileCheck %s --check-prefix=ARM
 ; RUN: llc < %s -O0 -verify-machineinstrs -fast-isel-abort -relocation-model=dynamic-no-pic -mtriple=thumbv7-apple-ios | FileCheck %s --check-prefix=THUMB
+; RUN: llc < %s -O0 -fast-isel-abort -relocation-model=dynamic-no-pic -mtriple=thumbv8-apple-ios | FileCheck %s --check-prefix=THUMB
 
 define i32 @t1(i1 %c) nounwind readnone {
 entry:
diff --git a/test/CodeGen/ARM/indirectbr.ll b/test/CodeGen/ARM/indirectbr.ll
index 99e84a6..1aeeb91 100644
--- a/test/CodeGen/ARM/indirectbr.ll
+++ b/test/CodeGen/ARM/indirectbr.ll
@@ -1,6 +1,7 @@
 ; RUN: llc < %s -relocation-model=pic -mtriple=armv6-apple-darwin | FileCheck %s -check-prefix=ARM
 ; RUN: llc < %s -relocation-model=pic -mtriple=thumbv6-apple-darwin | FileCheck %s -check-prefix=THUMB
 ; RUN: llc < %s -relocation-model=static -mtriple=thumbv7-apple-darwin | FileCheck %s -check-prefix=THUMB2
+; RUN: llc < %s -relocation-model=static -mtriple=thumbv8-apple-darwin | FileCheck %s -check-prefix=THUMB2
 
 @nextaddr = global i8* null                       ; <i8**> [#uses=2]
 @C.0.2070 = private constant [5 x i8*] [i8* blockaddress(@foo, %L1), i8* blockaddress(@foo, %L2), i8* blockaddress(@foo, %L3), i8* blockaddress(@foo, %L4), i8* blockaddress(@foo, %L5)] ; <[5 x i8*]*> [#uses=1]
@@ -48,14 +49,17 @@ L2:                                               ; preds = %L3, %bb2
 
 L1:                                               ; preds = %L2, %bb2
   %res.3 = phi i32 [ %phitmp, %L2 ], [ 2, %bb2 ]  ; <i32> [#uses=1]
+; ARM-LABEL: %L1
 ; ARM: ldr [[R1:r[0-9]+]], LCPI
 ; ARM: add [[R1b:r[0-9]+]], pc, [[R1]]
 ; ARM: str [[R1b]]
+; THUMB-LABEL: %L1
 ; THUMB: ldr
 ; THUMB: add
 ; THUMB: ldr [[R2:r[0-9]+]], LCPI
 ; THUMB: add [[R2]], pc
 ; THUMB: str [[R2]]
+; THUMB2-LABEL: %L1
 ; THUMB2: ldr [[R2:r[0-9]+]], LCPI
 ; THUMB2-NEXT: str{{(.w)?}} [[R2]]
   store i8* blockaddress(@foo, %L5), i8** @nextaddr, align 4
diff --git a/test/CodeGen/ARM/thumb2-it-block.ll b/test/CodeGen/ARM/thumb2-it-block.ll
index a25352c..47c5dcc 100644
--- a/test/CodeGen/ARM/thumb2-it-block.ll
+++ b/test/CodeGen/ARM/thumb2-it-block.ll
@@ -1,14 +1,15 @@
 ; RUN: llc < %s -march=thumb -mattr=+thumb2 | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck %s
 ; PR11107
 
 define i32 @test(i32 %a, i32 %b) {
 entry:
 ; CHECK:        cmp
 ; CHECK-NEXT:   it    mi
-; CHECK-NEXT:   rsbmi
+; CHECK-NEXT:   rsb{{s?}}mi
 ; CHECK-NEXT:   cmp
 ; CHECK-NEXT:   it    mi
-; CHECK-NEXT:   rsbmi
+; CHECK-NEXT:   rsb{{s?}}mi
  %cmp1 = icmp slt i32 %a, 0
  %sub1 = sub nsw i32 0, %a
  %abs1 = select i1 %cmp1, i32 %sub1, i32 %a
diff --git a/test/CodeGen/Thumb2/v8_IT_1.ll b/test/CodeGen/Thumb2/v8_IT_1.ll
new file mode 100644
index 0000000..e33845d
--- /dev/null
+++ b/test/CodeGen/Thumb2/v8_IT_1.ll
@@ -0,0 +1,19 @@
+; RUN: llc < %s -mtriple=thumbv8 -mattr=+neon | FileCheck %s
+
+;CHECK-LABEL: select_s_v_v:
+;CHECK: beq	.LBB0_2
+;CHECK-NEXT: @ BB#1:
+;CHECK-NEXT: vmov.i32
+;CHECK-NEXT: .LBB0_2:
+;CHECK: bx
+define <16 x i8> @select_s_v_v(i32 %avail, i8* %bar) {
+entry:
+  %vld1 = call <16 x i8> @llvm.arm.neon.vld1.v16i8(i8* %bar, i32 1)
+  %and = and i32 %avail, 1
+  %tobool = icmp eq i32 %and, 0
+  %vld1. = select i1 %tobool, <16 x i8> %vld1, <16 x i8> zeroinitializer
+  ret <16 x i8> %vld1.
+}
+
+declare <16 x i8> @llvm.arm.neon.vld1.v16i8(i8* , i32 )
+
diff --git a/test/CodeGen/Thumb2/v8_IT_2.ll b/test/CodeGen/Thumb2/v8_IT_2.ll
new file mode 100644
index 0000000..fe88316
--- /dev/null
+++ b/test/CodeGen/Thumb2/v8_IT_2.ll
@@ -0,0 +1,37 @@
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck %s
+
+	%struct.quad_struct = type { i32, i32, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct*, %struct.quad_struct* }
+
+define fastcc i32 @CountTree(%struct.quad_struct* %tree) {
+entry:
+; CHECK-LABEL: CountTree:
+; CHECK: bne
+; CHECK: cmp
+; CHECK: it eq
+; CHECK: cmpeq
+; CHECK: bne
+; CHECK: mov
+; CHECK: pop
+	br label %tailrecurse
+
+tailrecurse:		; preds = %bb, %entry
+	%tmp6 = load %struct.quad_struct** null		; <%struct.quad_struct*> [#uses=1]
+	%tmp9 = load %struct.quad_struct** null		; <%struct.quad_struct*> [#uses=2]
+	%tmp12 = load %struct.quad_struct** null		; <%struct.quad_struct*> [#uses=1]
+	%tmp14 = icmp eq %struct.quad_struct* null, null		; <i1> [#uses=1]
+	%tmp17 = icmp eq %struct.quad_struct* %tmp6, null		; <i1> [#uses=1]
+	%tmp23 = icmp eq %struct.quad_struct* %tmp9, null		; <i1> [#uses=1]
+	%tmp29 = icmp eq %struct.quad_struct* %tmp12, null		; <i1> [#uses=1]
+	%bothcond = and i1 %tmp17, %tmp14		; <i1> [#uses=1]
+	%bothcond1 = and i1 %bothcond, %tmp23		; <i1> [#uses=1]
+	%bothcond2 = and i1 %bothcond1, %tmp29		; <i1> [#uses=1]
+	br i1 %bothcond2, label %return, label %bb
+
+bb:		; preds = %tailrecurse
+	%tmp41 = tail call fastcc i32 @CountTree( %struct.quad_struct* %tmp9 )		; <i32> [#uses=0]
+	br label %tailrecurse
+
+return:		; preds = %tailrecurse
+	ret i32 0
+}
+
diff --git a/test/CodeGen/Thumb2/v8_IT_3.ll b/test/CodeGen/Thumb2/v8_IT_3.ll
new file mode 100644
index 0000000..4d90891
--- /dev/null
+++ b/test/CodeGen/Thumb2/v8_IT_3.ll
@@ -0,0 +1,75 @@
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv8 -relocation-model=pic | FileCheck %s --check-prefix=CHECK-PIC
+
+%struct.FF = type { i32 (i32*)*, i32 (i32*, i32*, i32, i32, i32, i32)*, i32 (i32, i32, i8*)*, void ()*, i32 (i32, i8*, i32*)*, i32 ()* }
+%struct.BD = type { %struct.BD*, i32, i32, i32, i32, i64, i32 (%struct.BD*, i8*, i64, i32)*, i32 (%struct.BD*, i8*, i32, i32)*, i32 (%struct.BD*, i8*, i64, i32)*, i32 (%struct.BD*, i8*, i32, i32)*, i32 (%struct.BD*, i64, i32)*, [16 x i8], i64, i64 }
+
+@FuncPtr = external hidden unnamed_addr global %struct.FF*
+@.str1 = external hidden unnamed_addr constant [6 x i8], align 4
+@G = external unnamed_addr global i32
+@.str2 = external hidden unnamed_addr constant [58 x i8], align 4
+@.str3 = external hidden unnamed_addr constant [58 x i8], align 4
+
+define i32 @test() nounwind optsize ssp {
+entry:
+; CHECK-LABEL: test:
+; CHECK: push
+; CHECK-NOT: push
+  %block_size = alloca i32, align 4
+  %block_count = alloca i32, align 4
+  %index_cache = alloca i32, align 4
+  store i32 0, i32* %index_cache, align 4
+  %tmp = load i32* @G, align 4
+  %tmp1 = call i32 @bar(i32 0, i32 0, i32 %tmp) nounwind
+  switch i32 %tmp1, label %bb8 [
+    i32 0, label %bb
+    i32 536870913, label %bb4
+    i32 536870914, label %bb6
+  ]
+
+bb:
+  %tmp2 = load i32* @G, align 4
+  %tmp4 = icmp eq i32 %tmp2, 0
+  br i1 %tmp4, label %bb1, label %bb8
+
+bb1:
+; CHECK: %bb6
+; CHECK: it	eq
+; CHECK-NEXT: ldreq
+; CHECK-NEXT: it	eq
+; CHECK-NEXT: cmpeq
+; CHECK: %bb1
+  %tmp5 = load i32* %block_size, align 4
+  %tmp6 = load i32* %block_count, align 4
+  %tmp7 = call %struct.FF* @Get() nounwind
+  store %struct.FF* %tmp7, %struct.FF** @FuncPtr, align 4
+  %tmp10 = zext i32 %tmp6 to i64
+  %tmp11 = zext i32 %tmp5 to i64
+  %tmp12 = mul nsw i64 %tmp10, %tmp11
+  %tmp13 = call i32 @foo(i8* getelementptr inbounds ([6 x i8]* @.str1, i32 0, i32 0), i64 %tmp12, i32 %tmp5) nounwind
+  br label %bb8
+
+bb4:
+; CHECK-PIC: cmp
+; CHECK-PIC: cmp
+; CHECK-PIC-NEXT: bne
+; CHECK-PIC-NEXT: %bb4
+; CHECK-PIC-NEXT: movs
+; CHECK-PIC-NEXT: add
+; CHECK-PIC-NEXT: pop
+  ret i32 0
+
+bb6:
+  ret i32 1
+
+bb8:
+  ret i32 -1
+}
+
+declare i32 @printf(i8*, ...)
+
+declare %struct.FF* @Get()
+
+declare i32 @foo(i8*, i64, i32)
+
+declare i32 @bar(i32, i32, i32)
diff --git a/test/CodeGen/Thumb2/v8_IT_4.ll b/test/CodeGen/Thumb2/v8_IT_4.ll
new file mode 100644
index 0000000..45c79f4
--- /dev/null
+++ b/test/CodeGen/Thumb2/v8_IT_4.ll
@@ -0,0 +1,43 @@
+; RUN: llc < %s -mtriple=thumbv8-eabi -float-abi=hard | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv8-eabi -float-abi=hard -regalloc=basic | FileCheck %s
+
+%"struct.__gnu_cxx::__normal_iterator<char*,std::basic_string<char, std::char_traits<char>, std::allocator<char> > >" = type { i8* }
+%"struct.__gnu_cxx::new_allocator<char>" = type <{ i8 }>
+%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >" = type { %"struct.__gnu_cxx::__normal_iterator<char*,std::basic_string<char, std::char_traits<char>, std::allocator<char> > >" }
+%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >::_Rep" = type { %"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >::_Rep_base" }
+%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >::_Rep_base" = type { i32, i32, i32 }
+
+
+define weak arm_aapcs_vfpcc i32 @_ZNKSs7compareERKSs(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this, %"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %__str) {
+; CHECK-LABEL: _ZNKSs7compareERKSs:
+; CHECK:      cbnz	r0,
+; CHECK-NEXT: %bb
+; CHECK-NEXT: sub{{(.w)?}} r0, r{{[0-9]+}}, r{{[0-9]+}}
+; CHECK-NEXT: %bb1
+; CHECK-NEXT: pop.w
+entry:
+  %0 = tail call arm_aapcs_vfpcc  i32 @_ZNKSs4sizeEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this) ; <i32> [#uses=3]
+  %1 = tail call arm_aapcs_vfpcc  i32 @_ZNKSs4sizeEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %__str) ; <i32> [#uses=3]
+  %2 = icmp ult i32 %1, %0                        ; <i1> [#uses=1]
+  %3 = select i1 %2, i32 %1, i32 %0               ; <i32> [#uses=1]
+  %4 = tail call arm_aapcs_vfpcc  i8* @_ZNKSs7_M_dataEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this) ; <i8*> [#uses=1]
+  %5 = tail call arm_aapcs_vfpcc  i8* @_ZNKSs4dataEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %__str) ; <i8*> [#uses=1]
+  %6 = tail call arm_aapcs_vfpcc  i32 @memcmp(i8* %4, i8* %5, i32 %3) nounwind readonly ; <i32> [#uses=2]
+  %7 = icmp eq i32 %6, 0                          ; <i1> [#uses=1]
+  br i1 %7, label %bb, label %bb1
+
+bb:                                               ; preds = %entry
+  %8 = sub i32 %0, %1                             ; <i32> [#uses=1]
+  ret i32 %8
+
+bb1:                                              ; preds = %entry
+  ret i32 %6
+}
+
+declare arm_aapcs_vfpcc i32 @memcmp(i8* nocapture, i8* nocapture, i32) nounwind readonly
+
+declare arm_aapcs_vfpcc i32 @_ZNKSs4sizeEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this)
+
+declare arm_aapcs_vfpcc i8* @_ZNKSs7_M_dataEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this)
+
+declare arm_aapcs_vfpcc i8* @_ZNKSs4dataEv(%"struct.std::basic_string<char,std::char_traits<char>,std::allocator<char> >"* %this)
diff --git a/test/CodeGen/Thumb2/v8_IT_5.ll b/test/CodeGen/Thumb2/v8_IT_5.ll
new file mode 100644
index 0000000..3866068
--- /dev/null
+++ b/test/CodeGen/Thumb2/v8_IT_5.ll
@@ -0,0 +1,62 @@
+; RUN: llc < %s -mtriple=thumbv8 | FileCheck %s
+; CHECK: it	ne
+; CHECK-NEXT: cmpne
+; CHECK-NEXT: beq
+; CHECK: cmp
+; CHECK-NEXT: beq
+; CHECK-NEXT: %if.else163
+; CHECK-NEXT: mov.w
+; CHECK-NEXT: b
+; CHECK-NEXT: %if.else145
+; CHECK-NEXT: mov.w
+
+%struct.hc = type { i32, i32, i32, i32 }
+
+define i32 @t(i32 %type) optsize {
+entry:
+  br i1 undef, label %if.then, label %if.else
+
+if.then:
+  unreachable
+
+if.else:
+  br i1 undef, label %if.then15, label %if.else18
+
+if.then15:
+  unreachable
+
+if.else18:
+  switch i32 %type, label %if.else173 [
+    i32 3, label %if.then115
+    i32 1, label %if.then102
+  ]
+
+if.then102:
+  br i1 undef, label %cond.true10.i, label %t.exit
+
+cond.true10.i:
+  br label %t.exit
+
+t.exit:
+  unreachable
+
+if.then115:
+  br i1 undef, label %if.else163, label %if.else145
+
+if.else145:
+  %call150 = call fastcc %struct.hc* @foo(%struct.hc* undef, i32 34865152) optsize
+  br label %while.body172
+
+if.else163:
+  %call168 = call fastcc %struct.hc* @foo(%struct.hc* undef, i32 34078720) optsize
+  br label %while.body172
+
+while.body172:
+  br label %while.body172
+
+if.else173:
+  ret i32 -1
+}
+
+declare hidden fastcc %struct.hc* @foo(%struct.hc* nocapture, i32) nounwind optsize
+