[M68k] Handle 16 bit MOVs to and from CCR (#114714)

Builds on @TechnoElf 's CCR MOV pr
#107591 and adds some tests.

Fixes #106210.

---------

Co-authored-by: TechnoElf <[email protected]>

Author: knickish

llvm/lib/Target/M68k/M68kInstrData.td

@@ -419,6 +419,8 @@ class MxMoveFromCCR_M<MxOperand MEMOp, MxEncMemOp DST_ENC>
 
 class MxMoveFromCCRPseudo<MxOperand MEMOp>
     : MxPseudo<(outs), (ins MEMOp:$dst, CCRC:$src)>;
+class MxMoveFromCCR_RPseudo<MxOperand MEMOp>
+    : MxPseudo<(outs MEMOp:$dst), (ins CCRC:$src)>;
 } // let Uses = [CCR]
 
 let mayStore = 1 in
@@ -432,7 +434,7 @@ foreach AM = MxMoveSupportedAMs in {
 
 // Only data register is allowed.
 def MOV16dc : MxMoveFromCCR_R;
-def MOV8dc  : MxMoveFromCCRPseudo<MxOp8AddrMode_d.Op>;
+def MOV8dc  : MxMoveFromCCR_RPseudo<MxOp8AddrMode_d.Op>;
 
 //===----------------------------------------------------------------------===//
 // LEA

llvm/lib/Target/M68k/M68kInstrInfo.cpp

@@ -23,7 +23,9 @@
 #include "llvm/CodeGen/LivePhysRegs.h"
 #include "llvm/CodeGen/LiveVariables.h"
 #include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGenTypes/MachineValueType.h"
 #include "llvm/MC/TargetRegistry.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/Regex.h"
@@ -571,6 +573,12 @@ bool M68kInstrInfo::ExpandPUSH_POP(MachineInstrBuilder &MIB,
 }
 
 bool M68kInstrInfo::ExpandCCR(MachineInstrBuilder &MIB, bool IsToCCR) const {
+  if (MIB->getOpcode() == M68k::MOV8cd) {
+    // Promote used register to the next class
+    MachineOperand &Opd = MIB->getOperand(1);
+    Opd.setReg(getRegisterInfo().getMatchingSuperReg(
+        Opd.getReg(), M68k::MxSubRegIndex8Lo, &M68k::DR16RegClass));
+  }
 
   // Replace the pseudo instruction with the real one
   if (IsToCCR)
@@ -579,11 +587,6 @@ bool M68kInstrInfo::ExpandCCR(MachineInstrBuilder &MIB, bool IsToCCR) const {
     // FIXME M68010 or later is required
     MIB->setDesc(get(M68k::MOV16dc));
 
-  // Promote used register to the next class
-  auto &Opd = MIB->getOperand(1);
-  Opd.setReg(getRegisterInfo().getMatchingSuperReg(
-      Opd.getReg(), M68k::MxSubRegIndex8Lo, &M68k::DR16RegClass));
-
   return true;
 }
 
@@ -757,13 +760,27 @@ void M68kInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
   bool ToSR = DstReg == M68k::SR;
 
   if (FromCCR) {
-    assert(M68k::DR8RegClass.contains(DstReg) &&
-           "Need DR8 register to copy CCR");
-    Opc = M68k::MOV8dc;
+    if (M68k::DR8RegClass.contains(DstReg)) {
+      Opc = M68k::MOV8dc;
+    } else if (M68k::DR16RegClass.contains(DstReg)) {
+      Opc = M68k::MOV16dc;
+    } else if (M68k::DR32RegClass.contains(DstReg)) {
+      Opc = M68k::MOV16dc;
+    } else {
+      LLVM_DEBUG(dbgs() << "Cannot copy CCR to " << RI.getName(DstReg) << '\n');
+      llvm_unreachable("Invalid register for MOVE from CCR");
+    }
   } else if (ToCCR) {
-    assert(M68k::DR8RegClass.contains(SrcReg) &&
-           "Need DR8 register to copy CCR");
-    Opc = M68k::MOV8cd;
+    if (M68k::DR8RegClass.contains(SrcReg)) {
+      Opc = M68k::MOV8cd;
+    } else if (M68k::DR16RegClass.contains(SrcReg)) {
+      Opc = M68k::MOV16cd;
+    } else if (M68k::DR32RegClass.contains(SrcReg)) {
+      Opc = M68k::MOV16cd;
+    } else {
+      LLVM_DEBUG(dbgs() << "Cannot copy " << RI.getName(SrcReg) << " to CCR\n");
+      llvm_unreachable("Invalid register for MOVE to CCR");
+    }
   } else if (FromSR || ToSR)
     llvm_unreachable("Cannot emit SR copy instruction");
 

llvm/lib/Target/M68k/M68kRegisterInfo.h

@@ -101,7 +101,7 @@ class M68kRegisterInfo : public M68kGenRegisterInfo {
   const TargetRegisterClass *
   getCrossCopyRegClass(const TargetRegisterClass *RC) const override {
     if (RC == &M68k::CCRCRegClass)
-      return &M68k::DR32RegClass;
+      return &M68k::DR16RegClass;
     return RC;
   }
 

llvm/test/CodeGen/M68k/Control/non-cmov-switch.ll

@@ -0,0 +1,126 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc < %s -mtriple=m68k-linux  -mcpu=M68020 --verify-machineinstrs | FileCheck %s
+
+define internal void @select_i32(i32 %self, ptr nonnull %value) {
+; CHECK-LABEL: select_i32:
+; CHECK:         .cfi_startproc
+; CHECK-NEXT:  ; %bb.0: ; %start
+; CHECK-NEXT:    suba.l #4, %sp
+; CHECK-NEXT:    .cfi_def_cfa_offset -8
+; CHECK-NEXT:    movem.l %d2, (0,%sp) ; 8-byte Folded Spill
+; CHECK-NEXT:    cmpi.l #0, (8,%sp)
+; CHECK-NEXT:    move.w %ccr, %d2
+; CHECK-NEXT:    sne %d1
+; CHECK-NEXT:    move.l (12,%sp), %d0
+; CHECK-NEXT:    move.w %d2, %ccr
+; CHECK-NEXT:    bne .LBB0_2
+; CHECK-NEXT:  ; %bb.1: ; %start
+; CHECK-NEXT:    and.l #255, %d1
+; CHECK-NEXT:    cmpi.l #0, %d1
+; CHECK-NEXT:    bne .LBB0_3
+; CHECK-NEXT:  .LBB0_2: ; %null
+; CHECK-NEXT:    suba.l %a0, %a0
+; CHECK-NEXT:    move.l %d0, (%a0)
+; CHECK-NEXT:  .LBB0_3: ; %exit
+; CHECK-NEXT:    movem.l (0,%sp), %d2 ; 8-byte Folded Reload
+; CHECK-NEXT:    adda.l #4, %sp
+; CHECK-NEXT:    rts
+start:
+  %2 = icmp eq i32 %self, 0
+  %3 = select i1 %2, i32 0, i32 1
+  switch i32 %3, label %exit [
+    i32 0, label %nonnull
+    i32 1, label %null
+  ]
+
+nonnull:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+null:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+exit:                                              ; preds = %nonnull, %null
+  ret void
+}
+
+define internal void @select_i16(i16 %self, ptr nonnull %value) {
+; CHECK-LABEL: select_i16:
+; CHECK:         .cfi_startproc
+; CHECK-NEXT:  ; %bb.0: ; %start
+; CHECK-NEXT:    suba.l #4, %sp
+; CHECK-NEXT:    .cfi_def_cfa_offset -8
+; CHECK-NEXT:    movem.l %d2, (0,%sp) ; 8-byte Folded Spill
+; CHECK-NEXT:    cmpi.w #0, (10,%sp)
+; CHECK-NEXT:    move.w %ccr, %d2
+; CHECK-NEXT:    sne %d1
+; CHECK-NEXT:    move.l (12,%sp), %d0
+; CHECK-NEXT:    move.w %d2, %ccr
+; CHECK-NEXT:    bne .LBB1_2
+; CHECK-NEXT:  ; %bb.1: ; %start
+; CHECK-NEXT:    and.l #255, %d1
+; CHECK-NEXT:    cmpi.w #0, %d1
+; CHECK-NEXT:    bne .LBB1_3
+; CHECK-NEXT:  .LBB1_2: ; %null
+; CHECK-NEXT:    suba.l %a0, %a0
+; CHECK-NEXT:    move.l %d0, (%a0)
+; CHECK-NEXT:  .LBB1_3: ; %exit
+; CHECK-NEXT:    movem.l (0,%sp), %d2 ; 8-byte Folded Reload
+; CHECK-NEXT:    adda.l #4, %sp
+; CHECK-NEXT:    rts
+start:
+  %2 = icmp eq i16 %self, 0
+  %3 = select i1 %2, i16 0, i16 1
+  switch i16 %3, label %exit [
+    i16 0, label %nonnull
+    i16 1, label %null
+  ]
+
+nonnull:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+null:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+exit:                                              ; preds = %nonnull, %null
+  ret void
+}
+
+define internal void @select_i8(i8 %self, ptr nonnull %value) {
+; CHECK-LABEL: select_i8:
+; CHECK:         .cfi_startproc
+; CHECK-NEXT:  ; %bb.0: ; %start
+; CHECK-NEXT:    move.l (8,%sp), %d0
+; CHECK-NEXT:    cmpi.b #0, (7,%sp)
+; CHECK-NEXT:    sne %d1
+; CHECK-NEXT:    bne .LBB2_2
+; CHECK-NEXT:  ; %bb.1: ; %start
+; CHECK-NEXT:    cmpi.b #0, %d1
+; CHECK-NEXT:    bne .LBB2_3
+; CHECK-NEXT:  .LBB2_2: ; %null
+; CHECK-NEXT:    suba.l %a0, %a0
+; CHECK-NEXT:    move.l %d0, (%a0)
+; CHECK-NEXT:  .LBB2_3: ; %exit
+; CHECK-NEXT:    rts
+start:
+  %2 = icmp eq i8 %self, 0
+  %3 = select i1 %2, i8 0, i8 1
+  switch i8 %3, label %exit [
+    i8 0, label %nonnull
+    i8 1, label %null
+  ]
+
+nonnull:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+null:                                              ; preds = %start
+  store ptr %value, ptr null, align 2
+  br label %exit
+
+exit:                                              ; preds = %nonnull, %null
+  ret void
+}