[AArch64] Don't form paired loads from epilogue operations on Windows

[efriedma]

Summary

AArch64LoadStoreOptimizer has a bunch of different guards to avoid corrupting Windows SEH prologues/epilogues, but apparently we missed the case of merging two instructions where the first instruction isn't part of the epilogue, but the second instruction is.

Fixes issue discovered at https://reviews.llvm.org/D130049#3704064

[mstorsjo]

LGTM, this does seem to resolve most issues for me.

I did some build testing with D131394, and did find a couple of more issues (less severe ones) that we still have to fix, before we can enable such checking to the same level of strictness as we have on ARM - I'll post patches (or spin off discussion) for those last bits when I have time to brush it up.

llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp

do {
 
     // If we encounter a store aliased with the load, return early.
     if (MI.mayStore() && LoadMI.mayAlias(AA, MI, /*UseTBAA*/ false))
       return false;
   } while (MBBI != B && Count < Limit);
   return false;
 }
 
+static bool needsWinCFI(const MachineFunction *MF) {
+  return MF->getTarget().getMCAsmInfo()->usesWindowsCFI() &&
+         MF->getFunction().needsUnwindTableEntry();
+}
+
 // Returns true if FirstMI and MI are candidates for merging or pairing.
 // Otherwise, returns false.
 static bool areCandidatesToMergeOrPair(MachineInstr &FirstMI, MachineInstr &MI,
                                        LdStPairFlags &Flags,
                                        const AArch64InstrInfo *TII) {
   // If this is volatile or if pairing is suppressed, not a candidate.
   if (MI.hasOrderedMemoryRef() || TII->isLdStPairSuppressed(MI))
     return false;
 
   // We should have already checked FirstMI for pair suppression and volatility.
   assert(!FirstMI.hasOrderedMemoryRef() &&
          !TII->isLdStPairSuppressed(FirstMI) &&
          "FirstMI shouldn't get here if either of these checks are true.");
 
+  if (needsWinCFI(MI.getMF()) && (MI.getFlag(MachineInstr::FrameSetup) ||
+                                  MI.getFlag(MachineInstr::FrameDestroy)))
+    return false;
+
   unsigned OpcA = FirstMI.getOpcode();
   unsigned OpcB = MI.getOpcode();
 
   // Opcodes match: If the opcodes are pre ld/st there is nothing more to check.
   if (OpcA == OpcB)
     return !AArch64InstrInfo::isPreLdSt(FirstMI);
 
   // Try to match a sign-extended load/store with a zero-extended load/store.
case AArch64::ADDXri:
     // we're adding to the register.
     if (!Offset || Offset == UpdateOffset)
       return true;
     break;
   }
   return false;
 }
 
-static bool needsWinCFI(const MachineFunction *MF) {
-  return MF->getTarget().getMCAsmInfo()->usesWindowsCFI() &&
-         MF->getFunction().needsUnwindTableEntry();
-}
-
 MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
     MachineBasicBlock::iterator I, int UnscaledOffset, unsigned Limit) {
   MachineBasicBlock::iterator E = I->getParent()->end();
   MachineInstr &MemMI = *I;
   MachineBasicBlock::iterator MBBI = I;
 
   Register BaseReg = AArch64InstrInfo::getLdStBaseOp(MemMI).getReg();
   int MIUnscaledOffset = AArch64InstrInfo::getLdStOffsetOp(MemMI).getImm() *

llvm/test/CodeGen/AArch64/wineh9.mir

+# RUN: llc -o - %s -mtriple=aarch64-windows -run-pass=aarch64-ldst-opt \
+# RUN:   | FileCheck %s
+
+# CHECK: $x0 = LDRXui $sp, 1
+# CHECK: $x19 = frame-destroy LDRXui $sp, 2
+
+--- |
+  target triple = "aarch64-pc-windows-gnu"
+  
+  define dso_local noundef i64 @_Z1fPFvPxEx(ptr nocapture noundef readonly %g, i64 noundef %x) uwtable {
+  entry:
+    %x.addr = alloca i64, align 8
+    store i64 %x, ptr %x.addr, align 8
+    call void %g(ptr noundef nonnull %x.addr)
+    call void asm sideeffect "", "~{x19}"()
+    %0 = load i64, ptr %x.addr, align 8
+    ret i64 %0
+  }
+...
+---
+name:            _Z1fPFvPxEx
+alignment:       4
+exposesReturnsTwice: false
+legalized:       false
+regBankSelected: false
+selected:        false
+failedISel:      false
+tracksRegLiveness: true
+hasWinCFI:       true
+callsEHReturn:   false
+callsUnwindInit: false
+hasEHCatchret:   false
+hasEHScopes:     false
+hasEHFunclets:   false
+failsVerification: false
+tracksDebugUserValues: true
+registers:       []
+liveins:
+  - { reg: '$x0', virtual-reg: '' }
+  - { reg: '$x1', virtual-reg: '' }
+frameInfo:
+  isFrameAddressTaken: false
+  isReturnAddressTaken: false
+  hasStackMap:     false
+  hasPatchPoint:   false
+  stackSize:       32
+  offsetAdjustment: 0
+  maxAlignment:    8
+  adjustsStack:    true
+  hasCalls:        true
+  stackProtector:  ''
+  functionContext: ''
+  maxCallFrameSize: 0
+  cvBytesOfCalleeSavedRegisters: 0
+  hasOpaqueSPAdjustment: false
+  hasVAStart:      false
+  hasMustTailInVarArgFunc: false
+  hasTailCall:     false
+  localFrameSize:  8
+  savePoint:       ''
+  restorePoint:    ''
+fixedStack:      []
+stack:
+  - { id: 0, name: x.addr, type: default, offset: -24, size: 8, alignment: 8, 
+      stack-id: default, callee-saved-register: '', callee-saved-restored: true, 
+      local-offset: -8, debug-info-variable: '', debug-info-expression: '', 
+      debug-info-location: '' }
+  - { id: 1, name: '', type: spill-slot, offset: -8, size: 8, alignment: 8, 
+      stack-id: default, callee-saved-register: '$lr', callee-saved-restored: true, 
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+  - { id: 2, name: '', type: spill-slot, offset: -16, size: 8, alignment: 8, 
+      stack-id: default, callee-saved-register: '$x19', callee-saved-restored: true, 
+      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
+callSites:       []
+debugValueSubstitutions: []
+constants:       []
+machineFunctionInfo:
+  hasRedZone:      false
+body:             |
+  bb.0.entry:
+    liveins: $x0, $x1, $x19, $lr
+  
+    $sp = frame-setup SUBXri $sp, 32, 0
+    frame-setup SEH_StackAlloc 32
+    frame-setup STRXui killed $x19, $sp, 2 :: (store (s64) into %stack.2)
+    frame-setup SEH_SaveReg 19, 16
+    frame-setup STRXui killed $lr, $sp, 3 :: (store (s64) into %stack.1)
+    frame-setup SEH_SaveReg 30, 24
+    frame-setup SEH_PrologEnd
+    renamable $x8 = COPY $x0
+    STRXui killed renamable $x1, $sp, 1 :: (store (s64) into %ir.x.addr)
+    $x0 = ADDXri $sp, 8, 0
+    BLR killed renamable $x8, csr_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit $x0, implicit-def $sp
+    INLINEASM &"", 1 /* sideeffect attdialect */, 12 /* clobber */, implicit-def dead early-clobber $x19
+    renamable $x0 = LDRXui $sp, 1 :: (dereferenceable load (s64) from %ir.x.addr)
+    frame-destroy SEH_EpilogStart
+    $lr = frame-destroy LDRXui $sp, 3 :: (load (s64) from %stack.1)
+    frame-destroy SEH_SaveReg 30, 24
+    $x19 = frame-destroy LDRXui $sp, 2 :: (load (s64) from %stack.2)
+    frame-destroy SEH_SaveReg 19, 16
+    $sp = frame-destroy ADDXri $sp, 32, 0
+    frame-destroy SEH_StackAlloc 32
+    frame-destroy SEH_EpilogEnd
+    RET_ReallyLR implicit $x0
+
+...