[mips] Fix compact branch hazard detection

[sdardis]

Summary

In certain cases it is possible that transient instructions such as
%reg = IMPLICIT_DEF as a single instruction in a basic block to reach
the MipsHazardSchedule pass. This patch teaches MipsHazardSchedule to
properly look through such cases.

[sdardis]

Zoran, could you take a look at this?

Thanks.

[zoran.jovanovic]

LGTM with a nit.

test/CodeGen/Mips/compactbranches/compact-branch-implicit-def.mir

11

Nit: Unnecessary white space.
There are some other unnecessary white spaces in the test case.

[sdardis]

Thanks for the review.

lib/Target/Mips/MipsHazardSchedule.cpp

 char MipsHazardSchedule::ID = 0;
 } // end of anonymous namespace
 
 /// Returns a pass that clears pipeline hazards.
 FunctionPass *llvm::createMipsHazardSchedule() {
   return new MipsHazardSchedule();
 }
 
-// Find the next real instruction from the current position.
-static Iter getNextMachineInstr(Iter Position) {
+// Find the next real instruction from the current position in current basic
+// block.
+static Iter getNextMachineInstrInBB(Iter Position) {
   Iter I = Position, E = Position->getParent()->end();
-  I = std::find_if_not(I, E, [](const Iter &Insn) { return Insn->isTransient(); });
-  assert(I != E);
+  I = std::find_if_not(I, E,
+                       [](const Iter &Insn) { return Insn->isTransient(); });
+
   return I;
 }
 
+// Find the next real instruction from the current position, looking through
+// basic block boundaries.
+static Iter getNextMachineInstr(Iter Position) {
+  if (std::next(Position) == Position->getParent()->end()) {
+    const MachineBasicBlock * MBB = (&*Position)->getParent();
+    for (auto *Succ : MBB->successors()) {
+      if (MBB->isLayoutSuccessor(Succ)) {
+        Iter I = Succ->begin();
+        Iter Next = getNextMachineInstrInBB(I);
+        if (Next == Succ->end()) {
+          return getNextMachineInstr(I);
+        } else {
+          return I;
+        }
+      }
+    }
+    llvm_unreachable("Should have identified the end of the function earlier!");
+  }
+
+  return getNextMachineInstrInBB(Position);
+}
+
 bool MipsHazardSchedule::runOnMachineFunction(MachineFunction &MF) {
 
   const MipsSubtarget *STI =
       &static_cast<const MipsSubtarget &>(MF.getSubtarget());
 
-  // Forbidden slot hazards are only defined for MIPSR6.
+  // Forbidden slot hazards are only defined for MIPSR6 but not microMIPSR6.
   if (!STI->hasMips32r6() || STI->inMicroMipsMode())
     return false;
 
   bool Changed = false;
   const MipsInstrInfo *TII = STI->getInstrInfo();
 
   for (MachineFunction::iterator FI = MF.begin(); FI != MF.end(); ++FI) {
     for (Iter I = FI->begin(); I != FI->end(); ++I) {
 
       // Forbidden slot hazard handling. Use lookahead over state.
       if (!TII->HasForbiddenSlot(*I))
         continue;
 
-      bool InsertNop = false;
-      // Next instruction in the basic block.
-      if (std::next(I) != FI->end() &&
-          !TII->SafeInForbiddenSlot(*getNextMachineInstr(std::next(I)))) {
-        InsertNop = true;
+      Iter Inst;
+      bool LastInstInFunction =
+          std::next(I) == FI->end() && std::next(FI) == MF.end();
+      if (!LastInstInFunction) {
+        if (std::next(I) != FI->end()) {
+          // Start looking from the next instruction in the basic block.
+          Inst = getNextMachineInstr(std::next(I));
         } else {
           // Next instruction in the physical successor basic block.
-        for (auto *Succ : FI->successors()) {
-          if (FI->isLayoutSuccessor(Succ) &&
-              getNextMachineInstr(Succ->begin()) != Succ->end() &&
-              !TII->SafeInForbiddenSlot(*getNextMachineInstr(Succ->begin()))) {
-            InsertNop = true;
-            break;
-          }
+          Inst = getNextMachineInstr(I);
         }
       }
 
-      if (InsertNop) {
+      if (LastInstInFunction || !TII->SafeInForbiddenSlot(*Inst)) {
         Changed = true;
-        MIBundleBuilder(&*I).append(
-            BuildMI(MF, I->getDebugLoc(), TII->get(Mips::NOP)));
+        MIBundleBuilder(&*I)
+            .append(BuildMI(MF, I->getDebugLoc(), TII->get(Mips::NOP)));
         NumInsertedNops++;
       }
     }
   }
   return Changed;
 }

test/CodeGen/Mips/compactbranches/compact-branch-implicit-def.mir

+# RUN: llc -march=mips64 -mcpu=mips64r6 -start-after=block-placement -o - %s | FileCheck %s
+
+# Check that MipsHazardSchedule sees through basic blocks with transient instructions.
+# The mir code in this file isn't representative of the llvm-ir.
+
+--- |
+  ; ModuleID = 'test.ll'
+  source_filename = "test.c"
+  target datalayout = "E-m:e-i8:8:32-i16:16:32-i64:64-n32:64-S128"
+  target triple = "mips64-img-linux-gnu"
+  
+  ; Function Attrs: nounwind
+  define i32 @f(i32 signext %a) { 
+  entry:
+    %retval = alloca i32, align 4
+    %a.addr = alloca i32, align 4
+    store i32 %a, i32* %a.addr, align 4
+    %0 = load i32, i32* %a.addr, align 4
+    %cmp = icmp sgt i32 %0, 5
+    br i1 %cmp, label %if.then, label %if.else
+  
+  if.then:                                          ; preds = %entry
+    %1 = load i32, i32* %a.addr, align 4
+    %2 = load i32, i32* %a.addr, align 4
+    %add = add nsw i32 %1, %2
+    store i32 %add, i32* %retval, align 4
+    br label %return
+  
+  if.else:                                          ; preds = %entry
+    %3 = load i32, i32* %a.addr, align 4
+    %call = call i32 @g(i32 signext %3)
+    store i32 %call, i32* %retval, align 4
+    br label %return
+  
+  return:                                           ; preds = %if.else, %if.then
+    %4 = load i32, i32* %retval, align 4
+    ret i32 %4
+  }
+  
+  declare i32 @g(i32 signext)
+  
+  ; Function Attrs: nounwind
+  declare void @llvm.stackprotector(i8*, i8**)
+  
+  !llvm.ident = !{!0}
+  
+  !0 = !{!"clang version 4.0.0 "}
+
+...
+---
+# CHECK-LABEL: f:
+# CHECK: bgtzc
+# CHECK-NEXT: nop
+# CHECK: bltzc
+# CHECK-NEXT: nop
+# CHECK: blezc
+name:            f
+alignment:       3
+exposesReturnsTwice: false
+legalized:       false
+regBankSelected: false
+selected:        false
+tracksRegLiveness: true
+liveins:         
+  - { reg: '%a0_64' }
+  - { reg: '%t9_64' }
+calleeSavedRegisters: [ '%fp', '%gp', '%ra', '%d12', '%d13', '%d14', '%d15', 
+                        '%f24', '%f25', '%f26', '%f27', '%f28', '%f29', 
+                        '%f30', '%f31', '%fp_64', '%f_hi24', '%f_hi25', 
+                        '%f_hi26', '%f_hi27', '%f_hi28', '%f_hi29', '%f_hi30', 
+                        '%f_hi31', '%gp_64', '%ra_64', '%s0', '%s1', '%s2', 
+                        '%s3', '%s4', '%s5', '%s6', '%s7', '%d24_64', '%d25_64', 
+                        '%d26_64', '%d27_64', '%d28_64', '%d29_64', '%d30_64', 
+                        '%d31_64', '%s0_64', '%s1_64', '%s2_64', '%s3_64', 
+                        '%s4_64', '%s5_64', '%s6_64', '%s7_64' ]
+frameInfo:       
+  isFrameAddressTaken: false
+  isReturnAddressTaken: false
+  hasStackMap:     false
+  hasPatchPoint:   false
+  stackSize:       32
+  offsetAdjustment: 0
+  maxAlignment:    8
+  adjustsStack:    true
+  hasCalls:        true
+  maxCallFrameSize: 0
+  hasOpaqueSPAdjustment: false
+  hasVAStart:      false
+  hasMustTailInVarArgFunc: false
+stack:           
+  - { id: 0, name: retval, offset: -28, size: 4, alignment: 4 }
+  - { id: 1, name: a.addr, offset: -32, size: 4, alignment: 4 }
+  - { id: 2, type: spill-slot, offset: -8, size: 8, alignment: 8, callee-saved-register: '%ra_64' }
+  - { id: 3, type: spill-slot, offset: -16, size: 8, alignment: 8, callee-saved-register: '%fp_64' }
+  - { id: 4, type: spill-slot, offset: -24, size: 8, alignment: 8, callee-saved-register: '%gp_64' }
+body:             |
+  bb.0.entry:
+    successors: %bb.1.if.then(0x40000000), %bb.5.if.else(0x40000000)
+    liveins: %a0_64, %t9_64, %ra_64, %fp_64, %gp_64
+
+    %sp_64 = DADDiu %sp_64, -32
+    CFI_INSTRUCTION def_cfa_offset 32
+    SD killed %ra_64, %sp_64, 24 :: (store 8 into %stack.2)
+    SD killed %fp_64, %sp_64, 16 :: (store 8 into %stack.3)
+    SD killed %gp_64, %sp_64, 8 :: (store 8 into %stack.4)
+    CFI_INSTRUCTION offset %ra_64, -8
+    CFI_INSTRUCTION offset %fp_64, -16
+    CFI_INSTRUCTION offset %gp_64, -24
+    %fp_64 = frame-setup OR64 %sp_64, %zero_64
+    CFI_INSTRUCTION def_cfa_register %fp_64
+    %at_64 = LUi64 @f
+    %v0_64 = DADDu killed %at_64, %t9_64
+    SW %a0, %fp_64, 0 :: (store 4 into %ir.a.addr)
+    BGTZC %a0, %bb.5.if.else, implicit-def %at
+
+  bb.1.if.then:
+    successors: %bb.6.return(0x40000000), %bb.2.if.then(0x40000000)
+    liveins: %a0
+
+    BLTZC %a0, %bb.6.return, implicit-def %at
+
+  bb.2.if.then:
+    successors: %bb.3.if.else(0x80000000)
+    %t8 = IMPLICIT_DEF
+
+  bb.3.if.else:
+    successors: %bb.6.return(0x40000000), %bb.4.if.else(0x40000000)
+    liveins: %t8
+
+    BLEZC %t8, %bb.6.return, implicit-def %at
+
+  bb.4.if.else:
+    successors: %bb.6.return(0x80000000)
+    liveins: %t8
+
+    %at = LW %fp_64, 0 :: (dereferenceable load 4 from %ir.a.addr)
+    %at = ADDu killed %at, %t8
+    SW killed %at, %fp_64, 4 :: (store 4 into %ir.retval)
+    J %bb.6.return, implicit-def dead %at
+  
+  bb.5.if.else:
+    successors: %bb.6.return(0x80000000)
+    liveins: %v0_64
+  
+    %gp_64 = DADDiu killed %v0_64, @f
+    %a0_64 = LW64 %fp_64, 0 :: (dereferenceable load 4 from %ir.a.addr)
+    %t9_64 = LD %gp_64, @g :: (load 8 from call-entry @g)
+    JALR64Pseudo %t9_64, csr_n64, implicit-def dead %ra, implicit %a0_64, implicit %gp_64, implicit-def %sp, implicit-def %v0
+    SW killed %v0, %fp_64, 4 :: (store 4 into %ir.retval)
+  
+  bb.6.return:
+    %v0 = LW %fp_64, 4 :: (dereferenceable load 4 from %ir.retval)
+    %sp_64 = OR64 %fp_64, %zero_64
+    %gp_64 = LD %sp_64, 8 :: (load 8 from %stack.4)
+    %fp_64 = LD %sp_64, 16 :: (load 8 from %stack.3)
+    %ra_64 = LD %sp_64, 24 :: (load 8 from %stack.2)
+    %sp_64 = DADDiu %sp_64, 32
+    PseudoReturn64 %ra_64
+
+...