[RDA][ARM] collectKilledOperands across multiple blocks

sparker-arm · sparker-arm · commit 5618e9be3750 · 2020-03-03T15:23:05.000Z
Use MIOperand in collectLocalKilledOperands to make the search global, as we already have to search for global uses too. This allows us to delete more dead code when tail predicating. Differential Revision: https://reviews.llvm.org/D75167
diff --git a/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h b/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h
@@ -176,7 +176,7 @@ class ReachingDefAnalysis : public MachineFunctionPass {
 
   /// Assuming MI is dead, recursively search the incoming operands which are
   /// killed by MI and collect those that would become dead.
-  void collectLocalKilledOperands(MachineInstr *MI, InstSet &Dead) const;
+  void collectKilledOperands(MachineInstr *MI, InstSet &Dead) const;
 
   /// Return whether removing this instruction will have no effect on the
   /// program, returning the redundant use-def chain.
diff --git a/llvm/lib/CodeGen/ReachingDefAnalysis.cpp b/llvm/lib/CodeGen/ReachingDefAnalysis.cpp
@@ -33,10 +33,6 @@ bool isValidRegUseOf(const MachineOperand &MO, int PhysReg) {
   return isValidRegUse(MO) && MO.getReg() == PhysReg;
 }
 
-bool isKilledRegUse(const MachineOperand &MO) {
-  return isValidRegUse(MO) && MO.isKill();
-}
-
 bool isValidRegDef(const MachineOperand &MO) {
   return isValidReg(MO) && MO.isDef();
 }
@@ -353,7 +349,7 @@ MachineInstr *ReachingDefAnalysis::getUniqueReachingMIDef(MachineInstr *MI,
                                                           int PhysReg) const {
   // If there's a local def before MI, return it.
   MachineInstr *LocalDef = getReachingLocalMIDef(MI, PhysReg);
-  if (InstIds.lookup(LocalDef) < InstIds.lookup(MI))
+  if (LocalDef && InstIds.lookup(LocalDef) < InstIds.lookup(MI))
     return LocalDef;
 
   SmallPtrSet<MachineBasicBlock*, 4> VisitedBBs;
@@ -544,8 +540,8 @@ ReachingDefAnalysis::isSafeToRemove(MachineInstr *MI, InstSet &Visited,
   return true;
 }
 
-void ReachingDefAnalysis::collectLocalKilledOperands(MachineInstr *MI,
-                                                     InstSet &Dead) const {
+void ReachingDefAnalysis::collectKilledOperands(MachineInstr *MI,
+                                                InstSet &Dead) const {
   Dead.insert(MI);
   auto IsDead = [this, &Dead](MachineInstr *Def, int PhysReg) {
     unsigned LiveDefs = 0;
@@ -568,11 +564,11 @@ void ReachingDefAnalysis::collectLocalKilledOperands(MachineInstr *MI,
   };
 
   for (auto &MO : MI->operands()) {
-    if (!isKilledRegUse(MO))
+    if (!isValidRegUse(MO))
       continue;
-    if (MachineInstr *Def = getReachingLocalMIDef(MI, MO.getReg()))
+    if (MachineInstr *Def = getMIOperand(MI, MO))
       if (IsDead(Def, MO.getReg()))
-        collectLocalKilledOperands(Def, Dead);
+        collectKilledOperands(Def, Dead);
   }
 }
 
diff --git a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
@@ -900,40 +900,53 @@ void ARMLowOverheadLoops::IterationCountDCE(LowOverheadLoop &LoLoop) {
   if (!LoLoop.IsTailPredicationLegal())
     return;
 
+  LLVM_DEBUG(dbgs() << "ARM Loops: Trying DCE on loop iteration count.\n");
+
   MachineInstr *Def = RDA->getMIOperand(LoLoop.Start, 0);
-  if (!Def)
+  if (!Def) {
+    LLVM_DEBUG(dbgs() << "ARM Loops: Couldn't find iteration count.\n");
+    return;
+  }
+
+  // Collect and remove the users of iteration count.
+  SmallPtrSet<MachineInstr*, 4> Killed  = { LoLoop.Start, LoLoop.Dec,
+                                            LoLoop.End, LoLoop.InsertPt };
+  SmallPtrSet<MachineInstr*, 2> Remove;
+  if (RDA->isSafeToRemove(Def, Remove, Killed))
+    LoLoop.ToRemove.insert(Remove.begin(), Remove.end());
+  else {
+    LLVM_DEBUG(dbgs() << "ARM Loops: Unsafe to remove loop iteration count.\n");
     return;
+  }
 
-  // Collect IT blocks.
+  // Collect the dead code and the MBBs in which they reside.
+  RDA->collectKilledOperands(Def, Killed);
+  SmallPtrSet<MachineBasicBlock*, 2> BasicBlocks;
+  for (auto *MI : Killed)
+    BasicBlocks.insert(MI->getParent());
+
+  // Collect IT blocks in all affected basic blocks.
   std::map<MachineInstr *, SmallPtrSet<MachineInstr *, 2>> ITBlocks;
-  std::map<MachineInstr *, MachineInstr *> Predicates;
-  MachineInstr *IT = nullptr;
-  for (auto &MI : *Def->getParent()) {
-    if (MI.getOpcode() == ARM::t2IT)
-      IT = &MI;
-    else if (TII->getPredicate(MI) != ARMCC::AL) {
-      ITBlocks[IT].insert(&MI);
-      Predicates[&MI] = IT;
+  for (auto *MBB : BasicBlocks) {
+    for (auto &MI : *MBB) {
+      if (MI.getOpcode() != ARM::t2IT)
+        continue;
+      RDA->getReachingLocalUses(&MI, ARM::ITSTATE, ITBlocks[&MI]);
     }
   }
 
   // If we're removing all of the instructions within an IT block, then
   // also remove the IT instruction.
   SmallPtrSet<MachineInstr*, 2> ModifiedITs;
-  SmallPtrSet<MachineInstr*, 2> DeadITs;
-  SmallPtrSet<MachineInstr*, 4> Killed;
-  RDA->collectLocalKilledOperands(Def, Killed);
   for (auto *MI : Killed) {
-    if (!Predicates.count(MI))
-      continue;
-
-    MachineInstr *IT = Predicates[MI];
-    auto &CurrentBlock = ITBlocks[IT];
-    CurrentBlock.erase(MI);
-    ModifiedITs.insert(IT);
-    if (CurrentBlock.empty()) {
-      DeadITs.insert(IT);
-      ModifiedITs.erase(IT);
+    if (MachineOperand *MO = MI->findRegisterUseOperand(ARM::ITSTATE)) {
+      MachineInstr *IT = RDA->getMIOperand(MI, *MO);
+      auto &CurrentBlock = ITBlocks[IT];
+      CurrentBlock.erase(MI);
+      if (CurrentBlock.empty())
+        ModifiedITs.erase(IT);
+      else
+        ModifiedITs.insert(IT);
     }
   }
 
@@ -945,14 +958,8 @@ void ARMLowOverheadLoops::IterationCountDCE(LowOverheadLoop &LoLoop) {
                for (auto *MI : Killed)
                  dbgs() << " - " << *MI);
     LoLoop.ToRemove.insert(Killed.begin(), Killed.end());
-  }
-
-  // Collect and remove the users of iteration count.
-  SmallPtrSet<MachineInstr*, 4> Ignore = { LoLoop.Start, LoLoop.Dec,
-                                           LoLoop.End, LoLoop.InsertPt };
-  SmallPtrSet<MachineInstr*, 2> Remove;
-  if (RDA->isSafeToRemove(Def, Remove, Ignore))
-    LoLoop.ToRemove.insert(Remove.begin(), Remove.end());
+  } else
+    LLVM_DEBUG(dbgs() << "ARM Loops: Would need to modify IT block(s).\n");
 }
 
 MachineInstr* ARMLowOverheadLoops::ExpandLoopStart(LowOverheadLoop &LoLoop) {
@@ -1076,8 +1083,8 @@ void ARMLowOverheadLoops::Expand(LowOverheadLoop &LoLoop) {
     MIB.add(End->getOperand(0));
     MIB.add(End->getOperand(1));
     LLVM_DEBUG(dbgs() << "ARM Loops: Inserted LE: " << *MIB);
-    LoLoop.Dec->eraseFromParent();
-    End->eraseFromParent();
+    LoLoop.ToRemove.insert(LoLoop.Dec);
+    LoLoop.ToRemove.insert(End);
     return &*MIB;
   };
 
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/it-block-itercount.mir b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/it-block-itercount.mir
@@ -104,8 +104,6 @@ body:             |
   ; CHECK:   frame-setup CFI_INSTRUCTION offset $lr, -4
   ; CHECK:   frame-setup CFI_INSTRUCTION offset $r7, -8
   ; CHECK:   renamable $r3, dead $cpsr = tLSLri killed renamable $r2, 1, 14 /* CC::al */, $noreg
-  ; CHECK:   dead renamable $r12 = t2MOVi 4, 14 /* CC::al */, $noreg, $noreg
-  ; CHECK:   tCMPi8 renamable $r3, 4, 14 /* CC::al */, $noreg, implicit-def dead $cpsr
   ; CHECK:   renamable $r2 = tLEApcrel %const.0, 14 /* CC::al */, $noreg
   ; CHECK:   renamable $q0 = MVE_VLDRWU32 killed renamable $r2, 0, 0, $noreg :: (load 16 from constant-pool)
   ; CHECK:   $lr = MVE_DLSTP_32 killed renamable $r3
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/it-block-mov.mir b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/it-block-mov.mir
@@ -45,12 +45,7 @@ body:             |
   ; CHECK:   frame-setup CFI_INSTRUCTION def_cfa_offset 8
   ; CHECK:   frame-setup CFI_INSTRUCTION offset $lr, -4
   ; CHECK:   frame-setup CFI_INSTRUCTION offset $r4, -8
-  ; CHECK:   tCMPi8 renamable $r1, 4, 14 /* CC::al */, $noreg, implicit-def $cpsr
-  ; CHECK:   renamable $r3 = t2MOVi 4, 14 /* CC::al */, $noreg, $noreg
-  ; CHECK:   t2IT 11, 8, implicit-def $itstate
-  ; CHECK:   dead $r3 = tMOVr renamable $r1, 11 /* CC::lt */, killed $cpsr, implicit killed renamable $r3, implicit killed $itstate
   ; CHECK:   tCMPi8 renamable $r1, 2, 14 /* CC::al */, $noreg, implicit-def $cpsr
-  ; CHECK:   renamable $r12 = t2MOVi 4, 14 /* CC::al */, $noreg, $noreg
   ; CHECK:   tBcc %bb.2, 2 /* CC::hs */, killed $cpsr
   ; CHECK: bb.1:
   ; CHECK:   liveins: $r2
@@ -59,15 +54,8 @@ body:             |
   ; CHECK:   tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $pc
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x80000000)
-  ; CHECK:   liveins: $r0, $r1, $r2, $r12
-  ; CHECK:   renamable $r4, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
-  ; CHECK:   tCMPi8 renamable $r1, 4, 14 /* CC::al */, $noreg, implicit-def $cpsr
-  ; CHECK:   t2IT 11, 8, implicit-def $itstate
-  ; CHECK:   $r12 = tMOVr renamable $r1, 11 /* CC::lt */, killed $cpsr, implicit killed renamable $r12, implicit killed $itstate
-  ; CHECK:   renamable $r3 = t2SUBrr renamable $r1, killed renamable $r12, 14 /* CC::al */, $noreg, $noreg
-  ; CHECK:   renamable $r3, dead $cpsr = tADDi8 killed renamable $r3, 3, 14 /* CC::al */, $noreg
+  ; CHECK:   liveins: $r0, $r1, $r2
   ; CHECK:   $r12 = tMOVr $r1, 14 /* CC::al */, $noreg
-  ; CHECK:   dead renamable $r4 = nuw nsw t2ADDrs killed renamable $r4, killed renamable $r3, 19, 14 /* CC::al */, $noreg, $noreg
   ; CHECK:   renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, undef renamable $q0
   ; CHECK:   $r3 = tMOVr $r0, 14 /* CC::al */, $noreg
   ; CHECK:   $lr = MVE_DLSTP_32 killed renamable $r12
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/multiple-do-loops.mir b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/multiple-do-loops.mir
@@ -563,7 +563,6 @@ body:             |
   ; CHECK:   early-clobber $sp = frame-setup t2STR_PRE killed $r8, $sp, -4, 14 /* CC::al */, $noreg
   ; CHECK:   frame-setup CFI_INSTRUCTION offset $r8, -24
   ; CHECK:   renamable $r6, dead $cpsr = tMOVi8 0, 14 /* CC::al */, $noreg
-  ; CHECK:   dead renamable $r12 = t2MOVi 1, 14 /* CC::al */, $noreg, $noreg
   ; CHECK:   t2CMPrs killed renamable $r6, renamable $r3, 11, 14 /* CC::al */, $noreg, implicit-def $cpsr
   ; CHECK:   tBcc %bb.3, 0 /* CC::eq */, killed $cpsr
   ; CHECK: bb.1.vector.ph:
@@ -801,7 +800,6 @@ body:             |
   ; CHECK:   successors: %bb.6(0x30000000), %bb.4(0x50000000)
   ; CHECK:   liveins: $r0, $r1, $r2, $r3
   ; CHECK:   renamable $r6, dead $cpsr = tMOVi8 0, 14 /* CC::al */, $noreg
-  ; CHECK:   dead renamable $r8 = t2MOVi 1, 14 /* CC::al */, $noreg, $noreg
   ; CHECK:   t2CMPrs killed renamable $r6, renamable $r3, 11, 14 /* CC::al */, $noreg, implicit-def $cpsr
   ; CHECK:   tBcc %bb.6, 0 /* CC::eq */, killed $cpsr
   ; CHECK: bb.4.vector.ph66:
diff --git a/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll b/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll
@@ -576,15 +576,7 @@ define i32 @arm_nn_mat_mul_core_4x_s8(i32 %row_elements, i32 %offset, i8* %row_b
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r10, lr}
 ; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r10, lr}
-; CHECK-NEXT:    add.w r7, r0, #15
 ; CHECK-NEXT:    ldr.w r12, [sp, #32]
-; CHECK-NEXT:    mov.w lr, #1
-; CHECK-NEXT:    asrs r6, r7, #31
-; CHECK-NEXT:    add.w r4, r7, r6, lsr #28
-; CHECK-NEXT:    asrs r5, r4, #4
-; CHECK-NEXT:    cmp r5, #1
-; CHECK-NEXT:    it gt
-; CHECK-NEXT:    asrgt.w lr, r4, #4
 ; CHECK-NEXT:    cmp r0, #1
 ; CHECK-NEXT:    blt .LBB4_3
 ; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
