[AArch64] merge index address with large offset into base address

vfdff · vfdff · commit 33190490c667 · 2024-08-15T18:22:52.000+08:00
A case for this transformation, https://gcc.godbolt.org/z/nhYcWq1WE Fold mov w8, llvm#56952 movk w8, rust-lang#15, lsl rust-lang#16 ldrb w0, [x0, x8] into add x0, x0, 1036288 ldrb w0, [x0, 3704] Only LDRBBroX is supported for the first time. Fix llvm#71917 Note: This PR is try relanding the commit 32878c2 with fix crash for PR79756 this crash is exposes when there is MOVKWi instruction in the head of a block, but without MOVZWi
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -4516,6 +4516,16 @@ AArch64InstrInfo::getLdStOffsetOp(const MachineInstr &MI) {
   return MI.getOperand(Idx);
 }
 
+const MachineOperand &
+AArch64InstrInfo::getLdStAmountOp(const MachineInstr &MI) {
+  switch (MI.getOpcode()) {
+  default:
+    llvm_unreachable("Unexpected opcode");
+  case AArch64::LDRBBroX:
+    return MI.getOperand(4);
+  }
+}
+
 static const TargetRegisterClass *getRegClass(const MachineInstr &MI,
                                               Register Reg) {
   if (MI.getParent() == nullptr)
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.h b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
@@ -254,6 +254,9 @@ class AArch64InstrInfo final : public AArch64GenInstrInfo {
   /// Returns whether the physical register is FP or NEON.
   static bool isFpOrNEON(Register Reg);
 
+  /// Returns the shift amount operator of a load/store.
+  static const MachineOperand &getLdStAmountOp(const MachineInstr &MI);
+
   /// Returns whether the instruction is FP or NEON.
   static bool isFpOrNEON(const MachineInstr &MI);
 
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -64,6 +64,8 @@ STATISTIC(NumZeroStoresPromoted, "Number of narrow zero stores promoted");
 STATISTIC(NumLoadsFromStoresPromoted, "Number of loads from stores promoted");
 STATISTIC(NumFailedAlignmentCheck, "Number of load/store pair transformation "
                                    "not passed the alignment check");
+STATISTIC(NumConstOffsetFolded,
+          "Number of const offset of index address folded");
 
 DEBUG_COUNTER(RegRenamingCounter, DEBUG_TYPE "-reg-renaming",
               "Controls which pairs are considered for renaming");
@@ -77,6 +79,11 @@ static cl::opt<unsigned> LdStLimit("aarch64-load-store-scan-limit",
 static cl::opt<unsigned> UpdateLimit("aarch64-update-scan-limit", cl::init(100),
                                      cl::Hidden);
 
+// The LdStConstLimit limits how far we search for const offset instructions
+// when we form index address load/store instructions.
+static cl::opt<unsigned> LdStConstLimit("aarch64-load-store-const-scan-limit",
+                                        cl::init(10), cl::Hidden);
+
 // Enable register renaming to find additional store pairing opportunities.
 static cl::opt<bool> EnableRenaming("aarch64-load-store-renaming",
                                     cl::init(true), cl::Hidden);
@@ -173,6 +180,13 @@ struct AArch64LoadStoreOpt : public MachineFunctionPass {
   findMatchingUpdateInsnForward(MachineBasicBlock::iterator I,
                                 int UnscaledOffset, unsigned Limit);
 
+  // Scan the instruction list to find a register assigned with a const
+  // value that can be combined with the current instruction (a load or store)
+  // using base addressing with writeback. Scan backwards.
+  MachineBasicBlock::iterator
+  findMatchingConstOffsetBackward(MachineBasicBlock::iterator I, unsigned Limit,
+                                  unsigned &Offset);
+
   // Scan the instruction list to find a base register update that can
   // be combined with the current instruction (a load or store) using
   // pre or post indexed addressing with writeback. Scan backwards.
@@ -184,11 +198,19 @@ struct AArch64LoadStoreOpt : public MachineFunctionPass {
   bool isMatchingUpdateInsn(MachineInstr &MemMI, MachineInstr &MI,
                             unsigned BaseReg, int Offset);
 
+  bool isMatchingMovConstInsn(MachineInstr &MemMI, MachineInstr &MI,
+                              unsigned IndexReg, unsigned &Offset);
+
   // Merge a pre- or post-index base register update into a ld/st instruction.
   MachineBasicBlock::iterator
   mergeUpdateInsn(MachineBasicBlock::iterator I,
                   MachineBasicBlock::iterator Update, bool IsPreIdx);
 
+  MachineBasicBlock::iterator
+  mergeConstOffsetInsn(MachineBasicBlock::iterator I,
+                       MachineBasicBlock::iterator Update, unsigned Offset,
+                       int Scale);
+
   // Find and merge zero store instructions.
   bool tryToMergeZeroStInst(MachineBasicBlock::iterator &MBBI);
 
@@ -201,6 +223,9 @@ struct AArch64LoadStoreOpt : public MachineFunctionPass {
   // Find and merge a base register updates before or after a ld/st instruction.
   bool tryToMergeLdStUpdate(MachineBasicBlock::iterator &MBBI);
 
+  // Find and merge an index ldr/st instruction into a base ld/st instruction.
+  bool tryToMergeIndexLdSt(MachineBasicBlock::iterator &MBBI, int Scale);
+
   bool optimizeBlock(MachineBasicBlock &MBB, bool EnableNarrowZeroStOpt);
 
   bool runOnMachineFunction(MachineFunction &Fn) override;
@@ -483,6 +508,16 @@ static unsigned getPreIndexedOpcode(unsigned Opc) {
   }
 }
 
+static unsigned getBaseAddressOpcode(unsigned Opc) {
+  // TODO: Add more index address loads/stores.
+  switch (Opc) {
+  default:
+    llvm_unreachable("Opcode has no base address equivalent!");
+  case AArch64::LDRBBroX:
+    return AArch64::LDRBBui;
+  }
+}
+
 static unsigned getPostIndexedOpcode(unsigned Opc) {
   switch (Opc) {
   default:
@@ -724,6 +759,20 @@ static bool isMergeableLdStUpdate(MachineInstr &MI) {
   }
 }
 
+// Make sure this is a reg+reg Ld/St
+static bool isMergeableIndexLdSt(MachineInstr &MI, int &Scale) {
+  unsigned Opc = MI.getOpcode();
+  switch (Opc) {
+  default:
+    return false;
+  // Scaled instructions.
+  // TODO: Add more index address loads/stores.
+  case AArch64::LDRBBroX:
+    Scale = 1;
+    return true;
+  }
+}
+
 static bool isRewritableImplicitDef(unsigned Opc) {
   switch (Opc) {
   default:
@@ -2053,6 +2102,63 @@ AArch64LoadStoreOpt::mergeUpdateInsn(MachineBasicBlock::iterator I,
   return NextI;
 }
 
+MachineBasicBlock::iterator
+AArch64LoadStoreOpt::mergeConstOffsetInsn(MachineBasicBlock::iterator I,
+                                          MachineBasicBlock::iterator Update,
+                                          unsigned Offset, int Scale) {
+  assert((Update->getOpcode() == AArch64::MOVKWi) &&
+         "Unexpected const mov instruction to merge!");
+  MachineBasicBlock::iterator E = I->getParent()->end();
+  MachineBasicBlock::iterator NextI = next_nodbg(I, E);
+  MachineBasicBlock::iterator PrevI = prev_nodbg(Update, E);
+  MachineInstr &MemMI = *I;
+  unsigned Mask = (1 << 12) * Scale - 1;
+  unsigned Low = Offset & Mask;
+  unsigned High = Offset - Low;
+  Register BaseReg = AArch64InstrInfo::getLdStBaseOp(MemMI).getReg();
+  Register IndexReg = AArch64InstrInfo::getLdStOffsetOp(MemMI).getReg();
+  MachineInstrBuilder AddMIB, MemMIB;
+
+  // Add IndexReg, BaseReg, High (the BaseReg may be SP)
+  AddMIB =
+      BuildMI(*I->getParent(), I, I->getDebugLoc(), TII->get(AArch64::ADDXri))
+          .addDef(IndexReg)
+          .addUse(BaseReg)
+          .addImm(High >> 12) // shifted value
+          .addImm(12);        // shift 12
+  (void)AddMIB;
+  // Ld/St DestReg, IndexReg, Imm12
+  unsigned NewOpc = getBaseAddressOpcode(I->getOpcode());
+  MemMIB = BuildMI(*I->getParent(), I, I->getDebugLoc(), TII->get(NewOpc))
+               .add(getLdStRegOp(MemMI))
+               .add(AArch64InstrInfo::getLdStOffsetOp(MemMI))
+               .addImm(Low / Scale)
+               .setMemRefs(I->memoperands())
+               .setMIFlags(I->mergeFlagsWith(*Update));
+  (void)MemMIB;
+
+  ++NumConstOffsetFolded;
+  LLVM_DEBUG(dbgs() << "Creating base address load/store.\n");
+  LLVM_DEBUG(dbgs() << "    Replacing instructions:\n    ");
+  LLVM_DEBUG(PrevI->print(dbgs()));
+  LLVM_DEBUG(dbgs() << "    ");
+  LLVM_DEBUG(Update->print(dbgs()));
+  LLVM_DEBUG(dbgs() << "    ");
+  LLVM_DEBUG(I->print(dbgs()));
+  LLVM_DEBUG(dbgs() << "  with instruction:\n    ");
+  LLVM_DEBUG(((MachineInstr *)AddMIB)->print(dbgs()));
+  LLVM_DEBUG(dbgs() << "    ");
+  LLVM_DEBUG(((MachineInstr *)MemMIB)->print(dbgs()));
+  LLVM_DEBUG(dbgs() << "\n");
+
+  // Erase the old instructions for the block.
+  I->eraseFromParent();
+  PrevI->eraseFromParent();
+  Update->eraseFromParent();
+
+  return NextI;
+}
+
 bool AArch64LoadStoreOpt::isMatchingUpdateInsn(MachineInstr &MemMI,
                                                MachineInstr &MI,
                                                unsigned BaseReg, int Offset) {
@@ -2100,6 +2206,34 @@ bool AArch64LoadStoreOpt::isMatchingUpdateInsn(MachineInstr &MemMI,
   return false;
 }
 
+bool AArch64LoadStoreOpt::isMatchingMovConstInsn(MachineInstr &MemMI,
+                                                 MachineInstr &MI,
+                                                 unsigned IndexReg,
+                                                 unsigned &Offset) {
+  // The update instruction source and destination register must be the
+  // same as the load/store index register.
+  if (MI.getOpcode() == AArch64::MOVKWi &&
+      TRI->isSuperOrSubRegisterEq(IndexReg, MI.getOperand(1).getReg())) {
+
+    // movz + movk hold a large offset of a Ld/St instruction.
+    MachineBasicBlock::iterator B = MI.getParent()->begin();
+    MachineBasicBlock::iterator MBBI = &MI;
+    // Skip the scene when the MI is the first instruction of a block.
+    if (MBBI == B)
+      return false;
+    MBBI = prev_nodbg(MBBI, B);
+    MachineInstr &MovzMI = *MBBI;
+    if (MovzMI.getOpcode() == AArch64::MOVZWi) {
+      unsigned Low = MovzMI.getOperand(1).getImm();
+      unsigned High = MI.getOperand(2).getImm() << MI.getOperand(3).getImm();
+      Offset = High + Low;
+      // 12-bit optionally shifted immediates are legal for adds.
+      return Offset >> 24 == 0;
+    }
+  }
+  return false;
+}
+
 MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
     MachineBasicBlock::iterator I, int UnscaledOffset, unsigned Limit) {
   MachineBasicBlock::iterator E = I->getParent()->end();
@@ -2255,6 +2389,60 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnBackward(
   return E;
 }
 
+MachineBasicBlock::iterator
+AArch64LoadStoreOpt::findMatchingConstOffsetBackward(
+    MachineBasicBlock::iterator I, unsigned Limit, unsigned &Offset) {
+  MachineBasicBlock::iterator B = I->getParent()->begin();
+  MachineBasicBlock::iterator E = I->getParent()->end();
+  MachineInstr &MemMI = *I;
+  MachineBasicBlock::iterator MBBI = I;
+
+  // If the load is the first instruction in the block, there's obviously
+  // not any matching load or store.
+  if (MBBI == B)
+    return E;
+
+  // Make sure the IndexReg is killed and the shift amount is zero.
+  // TODO: Relex this restriction to extend, simplify processing now.
+  if (!AArch64InstrInfo::getLdStOffsetOp(MemMI).isKill() ||
+      !AArch64InstrInfo::getLdStAmountOp(MemMI).isImm() ||
+      (AArch64InstrInfo::getLdStAmountOp(MemMI).getImm() != 0))
+    return E;
+
+  Register IndexReg = AArch64InstrInfo::getLdStOffsetOp(MemMI).getReg();
+
+  // Track which register units have been modified and used between the first
+  // insn (inclusive) and the second insn.
+  ModifiedRegUnits.clear();
+  UsedRegUnits.clear();
+  unsigned Count = 0;
+  do {
+    MBBI = prev_nodbg(MBBI, B);
+    MachineInstr &MI = *MBBI;
+
+    // Don't count transient instructions towards the search limit since there
+    // may be different numbers of them if e.g. debug information is present.
+    if (!MI.isTransient())
+      ++Count;
+
+    // If we found a match, return it.
+    if (isMatchingMovConstInsn(*I, MI, IndexReg, Offset)) {
+      return MBBI;
+    }
+
+    // Update the status of what the instruction clobbered and used.
+    LiveRegUnits::accumulateUsedDefed(MI, ModifiedRegUnits, UsedRegUnits, TRI);
+
+    // Otherwise, if the index register is used or modified, we have no match,
+    // so return early.
+    if (!ModifiedRegUnits.available(IndexReg) ||
+        !UsedRegUnits.available(IndexReg))
+      return E;
+
+  } while (MBBI != B && Count < Limit);
+  return E;
+}
+
 bool AArch64LoadStoreOpt::tryToPromoteLoadFromStore(
     MachineBasicBlock::iterator &MBBI) {
   MachineInstr &MI = *MBBI;
@@ -2443,6 +2631,34 @@ bool AArch64LoadStoreOpt::tryToMergeLdStUpdate
   return false;
 }
 
+bool AArch64LoadStoreOpt::tryToMergeIndexLdSt(MachineBasicBlock::iterator &MBBI,
+                                              int Scale) {
+  MachineInstr &MI = *MBBI;
+  MachineBasicBlock::iterator E = MI.getParent()->end();
+  MachineBasicBlock::iterator Update;
+
+  // Don't know how to handle unscaled pre/post-index versions below, so bail.
+  if (TII->hasUnscaledLdStOffset(MI.getOpcode()))
+    return false;
+
+  // Look back to try to find a const offset for index LdSt instruction. For
+  // example,
+  // mov x8, #LargeImm   ; = a * (1<<12) + imm12
+  // ldr x1, [x0, x8]
+  // merged into:
+  // add x8, x0, a * (1<<12)
+  // ldr x1, [x8, imm12]
+  unsigned Offset;
+  Update = findMatchingConstOffsetBackward(MBBI, LdStConstLimit, Offset);
+  if (Update != E && (Offset & (Scale - 1)) == 0) {
+    // Merge the imm12 into the ld/st.
+    MBBI = mergeConstOffsetInsn(MBBI, Update, Offset, Scale);
+    return true;
+  }
+
+  return false;
+}
+
 bool AArch64LoadStoreOpt::optimizeBlock(MachineBasicBlock &MBB,
                                         bool EnableNarrowZeroStOpt) {
 
@@ -2521,6 +2737,22 @@ bool AArch64LoadStoreOpt::optimizeBlock(MachineBasicBlock &MBB,
       ++MBBI;
   }
 
+  // 5) Find a register assigned with a const value that can be combined with
+  // into the load or store. e.g.,
+  //        mov x8, #LargeImm   ; = a * (1<<12) + imm12
+  //        ldr x1, [x0, x8]
+  //        ; becomes
+  //        add x8, x0, a * (1<<12)
+  //        ldr x1, [x8, imm12]
+  for (MachineBasicBlock::iterator MBBI = MBB.begin(), E = MBB.end();
+       MBBI != E;) {
+    int Scale;
+    if (isMergeableIndexLdSt(*MBBI, Scale) && tryToMergeIndexLdSt(MBBI, Scale))
+      Modified = true;
+    else
+      ++MBBI;
+  }
+
   return Modified;
 }
 
diff --git a/llvm/test/CodeGen/AArch64/arm64-addrmode.ll b/llvm/test/CodeGen/AArch64/arm64-addrmode.ll
@@ -214,9 +214,8 @@ define void @t17(i64 %a) {
 define i8 @LdOffset_i8(ptr %a)  {
 ; CHECK-LABEL: LdOffset_i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #56952 // =0xde78
-; CHECK-NEXT:    movk w8, #15, lsl #16
-; CHECK-NEXT:    ldrb w0, [x0, x8]
+; CHECK-NEXT:    add x8, x0, #253, lsl #12 // =1036288
+; CHECK-NEXT:    ldrb w0, [x8, #3704]
 ; CHECK-NEXT:    ret
   %arrayidx = getelementptr inbounds i8, ptr %a, i64 1039992
   %val = load i8, ptr %arrayidx, align 1
@@ -227,9 +226,8 @@ define i8 @LdOffset_i8(ptr %a)  {
 define i32 @LdOffset_i8_zext32(ptr %a)  {
 ; CHECK-LABEL: LdOffset_i8_zext32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #56952 // =0xde78
-; CHECK-NEXT:    movk w8, #15, lsl #16
-; CHECK-NEXT:    ldrb w0, [x0, x8]
+; CHECK-NEXT:    add x8, x0, #253, lsl #12 // =1036288
+; CHECK-NEXT:    ldrb w0, [x8, #3704]
 ; CHECK-NEXT:    ret
   %arrayidx = getelementptr inbounds i8, ptr %a, i64 1039992
   %val = load i8, ptr %arrayidx, align 1
@@ -255,9 +253,8 @@ define i32 @LdOffset_i8_sext32(ptr %a)  {
 define i64 @LdOffset_i8_zext64(ptr %a)  {
 ; CHECK-LABEL: LdOffset_i8_zext64:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #56952 // =0xde78
-; CHECK-NEXT:    movk w8, #15, lsl #16
-; CHECK-NEXT:    ldrb w0, [x0, x8]
+; CHECK-NEXT:    add x8, x0, #253, lsl #12 // =1036288
+; CHECK-NEXT:    ldrb w0, [x8, #3704]
 ; CHECK-NEXT:    ret
   %arrayidx = getelementptr inbounds i8, ptr %a, i64 1039992
   %val = load i8, ptr %arrayidx, align 1
diff --git a/llvm/test/CodeGen/AArch64/large-offset-ldr-merge.mir b/llvm/test/CodeGen/AArch64/large-offset-ldr-merge.mir
