[X86] Clear kill flags when rewriting SETCC uses in flag copy lowering.

[craig.topper]

Summary

When we rewrite the setcc we replace set old setcc output register
with the new CondReg. But since CondReg can be shared by other
replacements, we don't know if the kill flags for the old register
are valid for CondReg. So be conservative and remove them.

The test case has a SETCCr and a SETCCm on the same condition so
they end up sharing the same CondReg. The SETCCr had one use with
a kill flag. This kill flag isn't valid after the replacement because
CondReg needs a live range extending to the later SETCCm replacment.

Fixes PR51903.

[RKSimon]

LGTM

[RKSimon]

For reference, this was re-committed at rGb81e26c7f470ee324095f7adfbe0384974db4375

llvm/lib/Target/X86/X86FlagsCopyLowering.cpp

void X86FlagsCopyLoweringPass::rewriteSetCC(MachineBasicBlock &TestMBB,
   if (!CondReg)
     CondReg = promoteCondToReg(TestMBB, TestPos, TestLoc, Cond);
 
   // Rewriting a register def is trivial: we just replace the register and
   // remove the setcc.
   if (!SetCCI.mayStore()) {
     assert(SetCCI.getOperand(0).isReg() &&
            "Cannot have a non-register defined operand to SETcc!");
-    MRI->replaceRegWith(SetCCI.getOperand(0).getReg(), CondReg);
+    Register OldReg = SetCCI.getOperand(0).getReg();
+    // Drop Kill flags on the old register before replacing. CondReg may have
+    // a longer live range.
+    MRI->clearKillFlags(OldReg);
+    MRI->replaceRegWith(OldReg, CondReg);
     SetCCI.eraseFromParent();
     return;
   }
 
   // Otherwise, we need to emit a store.
   auto MIB = BuildMI(*SetCCI.getParent(), SetCCI.getIterator(),
                      SetCCI.getDebugLoc(), TII->get(X86::MOV8mr));
   // Copy the address operands.

llvm/test/CodeGen/X86/flags-copy-lowering.mir

; CHECK-NOT:  COPY{{( killed)?}} $eflags
     %4:gr8 = SETCCr 2, implicit $eflags
     %5:gr8 = SETCCr 4, implicit $eflags
     SETCCm $rsp, 1, $noreg, -16, $noreg, 5, implicit killed $eflags
     MOV8mr $rsp, 1, $noreg, -16, $noreg, killed %3
     MOV8mr $rsp, 1, $noreg, -16, $noreg, killed %4
     MOV8mr $rsp, 1, $noreg, -16, $noreg, killed %5
   ; CHECK-NOT:     $eflags =
   ; CHECK-NOT:             = SET{{.*}}
-  ; CHECK:         MOV8mr {{.*}}, killed %[[A_REG]]
-  ; CHECK-NEXT:    MOV8mr {{.*}}, killed %[[B_REG]]
-  ; CHECK-NEXT:    MOV8mr {{.*}}, killed %[[E_REG]]
-  ; CHECK-NOT:     MOV8mr {{.*}}, killed %[[NE_REG]]
+  ; CHECK:         MOV8mr {{.*}}, %[[A_REG]]
+  ; CHECK-NEXT:    MOV8mr {{.*}}, %[[B_REG]]
+  ; CHECK-NEXT:    MOV8mr {{.*}}, %[[E_REG]]
+  ; CHECK-NOT:     MOV8mr {{.*}}, %[[NE_REG]]
 
     RET 0
 
 ...
 ---
 name:            test_cmov
 # CHECK-LABEL: name: test_cmov
 liveins:         

llvm/test/CodeGen/X86/pr51908.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc %s -o - -mtriple=x86_64-unknown-linux-gnu -run-pass=x86-flags-copy-lowering -verify-machineinstrs | FileCheck %s
+
+--- |
+  ; ModuleID = 'bugpoint-reduced-simplified.ll'
+  source_filename = "test.c"
+  target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+  target triple = "x86_64-unknown-linux-gnu"
+
+  @a = external global i32, align 4
+
+  declare void @e()
+
+  define void @f() {
+  entry:
+    br label %for.body
+
+  for.body:                                         ; preds = %for.body, %entry
+    %0 = load i32, i32* @a, align 4
+    %conv = sext i32 %0 to i64
+    %1 = load i64, i64* undef, align 8
+    %or = or i64 %1, %conv
+    store i64 %or, i64* undef, align 8
+    call void @e()
+    %cmp4 = icmp eq i64 %or, 0
+    %conv5 = zext i1 %cmp4 to i32
+    %conv6 = trunc i32 %conv5 to i8
+    store i8 %conv6, i8* undef, align 1
+    %conv7 = sext i8 %conv6 to i32
+    %bf.cast = trunc i40 undef to i32
+    %xor = xor i32 %conv7, %bf.cast
+    %conv8 = sext i32 %xor to i64
+    store i64 %conv8, i64* undef, align 8
+    br label %for.body
+  }
+
+...
+---
+name:            f
+alignment:       16
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: gr64 }
+  - { id: 1, class: gr64 }
+  - { id: 2, class: gr32 }
+  - { id: 3, class: gr64 }
+  - { id: 4, class: gr32 }
+  - { id: 5, class: gr32 }
+  - { id: 6, class: gr32 }
+  - { id: 7, class: gr64 }
+  - { id: 8, class: gr64 }
+  - { id: 9, class: gr64 }
+  - { id: 10, class: gr64 }
+  - { id: 11, class: gr8 }
+  - { id: 12, class: gr64 }
+frameInfo:
+  maxAlignment:    1
+  hasCalls:        true
+machineFunctionInfo: {}
+body:             |
+  ; CHECK-LABEL: name: f
+  ; CHECK: bb.0.entry:
+  ; CHECK:   successors: %bb.1(0x80000000)
+  ; CHECK:   JMP_1 %bb.1
+  ; CHECK: bb.1.for.body:
+  ; CHECK:   successors: %bb.1(0x80000000)
+  ; CHECK:   [[MOV64rm:%[0-9]+]]:gr64 = MOV64rm $rip, 1, $noreg, target-flags(x86-gotpcrel) @a, $noreg :: (load (s64) from got)
+  ; CHECK:   [[MOVSX64rm32_:%[0-9]+]]:gr64 = MOVSX64rm32 killed [[MOV64rm]], 1, $noreg, 0, $noreg :: (dereferenceable load (s32) from @a)
+  ; CHECK:   [[DEF:%[0-9]+]]:gr64 = IMPLICIT_DEF
+  ; CHECK:   OR64mr [[DEF]], 1, $noreg, 0, $noreg, [[MOVSX64rm32_]], implicit-def $eflags :: (store (s64) into `i64* undef`), (load (s64) from `i64* undef`)
+  ; CHECK:   [[SETCCr:%[0-9]+]]:gr8 = SETCCr 4, implicit $eflags
+  ; CHECK:   ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+  ; CHECK:   CALL64pcrel32 target-flags(x86-plt) @e, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp
+  ; CHECK:   ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+  ; CHECK:   [[MOVZX32rr8_:%[0-9]+]]:gr32 = MOVZX32rr8 [[SETCCr]]
+  ; CHECK:   [[DEF1:%[0-9]+]]:gr64 = IMPLICIT_DEF
+  ; CHECK:   MOV8mr [[DEF1]], 1, $noreg, 0, $noreg, [[SETCCr]] :: (store (s8) into `i8* undef`)
+  ; CHECK:   [[DEF2:%[0-9]+]]:gr32 = IMPLICIT_DEF
+  ; CHECK:   [[XOR32rr:%[0-9]+]]:gr32 = XOR32rr [[MOVZX32rr8_]], [[DEF2]], implicit-def $eflags
+  ; CHECK:   [[MOVSX64rr32_:%[0-9]+]]:gr64 = MOVSX64rr32 [[XOR32rr]]
+  ; CHECK:   [[DEF3:%[0-9]+]]:gr64 = IMPLICIT_DEF
+  ; CHECK:   MOV64mr [[DEF3]], 1, $noreg, 0, $noreg, [[MOVSX64rr32_]] :: (store (s64) into `i64* undef`)
+  ; CHECK:   JMP_1 %bb.1
+  bb.0.entry:
+    JMP_1 %bb.1
+
+  bb.1.for.body:
+    %7:gr64 = MOV64rm $rip, 1, $noreg, target-flags(x86-gotpcrel) @a, $noreg :: (load (s64) from got)
+    %8:gr64 = MOVSX64rm32 killed %7, 1, $noreg, 0, $noreg :: (dereferenceable load (s32) from @a)
+    %9:gr64 = IMPLICIT_DEF
+    OR64mr %9, 1, $noreg, 0, $noreg, %8, implicit-def $eflags :: (store (s64) into `i64* undef`), (load (s64) from `i64* undef`)
+    %10:gr64 = COPY $eflags
+    ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+    CALL64pcrel32 target-flags(x86-plt) @e, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp
+    ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
+    $eflags = COPY %10
+    %11:gr8 = SETCCr 4, implicit $eflags
+    %4:gr32 = MOVZX32rr8 killed %11
+    %12:gr64 = IMPLICIT_DEF
+    SETCCm %12, 1, $noreg, 0, $noreg, 4, implicit $eflags :: (store (s8) into `i8* undef`)
+    %5:gr32 = IMPLICIT_DEF
+    %6:gr32 = XOR32rr %4, %5, implicit-def $eflags
+    %3:gr64 = MOVSX64rr32 %6
+    %0:gr64 = IMPLICIT_DEF
+    MOV64mr %0, 1, $noreg, 0, $noreg, %3 :: (store (s64) into `i64* undef`)
+    JMP_1 %bb.1
+
+...