Expand MULHS for all types

Once MULHS was expanded, this exposed an issue where the condition
register was thought to be 16-bit. This caused an attempt to copy a
16-bit register to an 8-bit register.

Author: shepmaster

lib/Target/AVR/AVRISelLowering.cpp

@@ -137,8 +137,11 @@ AVRTargetLowering::AVRTargetLowering(AVRTargetMachine &tm)
   // Expand 16 bit multiplications.
   setOperationAction(ISD::SMUL_LOHI, MVT::i16, Expand);
   setOperationAction(ISD::UMUL_LOHI, MVT::i16, Expand);
-  setOperationAction(ISD::MULHS, MVT::i16, Expand);
-  setOperationAction(ISD::MULHU, MVT::i16, Expand);
+
+  for (MVT VT : MVT::integer_valuetypes()) {
+    setOperationAction(ISD::MULHS, VT, Expand);
+    setOperationAction(ISD::MULHU, VT, Expand);
+  }
 
   //  Runtime library functions
   {
@@ -232,6 +235,12 @@ const char *AVRTargetLowering::getTargetNodeName(unsigned Opcode) const {
   }
 }
 
+EVT AVRTargetLowering::getSetCCResultType(const DataLayout &DL, LLVMContext &,
+                                          EVT VT) const {
+  assert(!VT.isVector() && "No AVR SetCC type for vectors!");
+  return MVT::i8;
+}
+
 SDValue AVRTargetLowering::LowerShifts(SDValue Op, SelectionDAG &DAG) const {
   //:TODO: this function has to be completely rewritten to produce optimal
   // code, for now it's producing very long but correct code.

lib/Target/AVR/AVRISelLowering.h

@@ -94,6 +94,9 @@ class AVRTargetLowering : public TargetLowering {
 
   bool isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const override;
 
+  EVT getSetCCResultType(const DataLayout &DL, LLVMContext &Context,
+                         EVT VT) const override;
+
   MachineBasicBlock *
   EmitInstrWithCustomInserter(MachineInstr *MI,
                               MachineBasicBlock *MBB) const override;

test/CodeGen/AVR/issue-cannot-select-mulhs.ll

@@ -0,0 +1,31 @@
+; RUN: llc < %s -march=avr | FileCheck %s
+
+define i1 @signed_multiplication_did_overflow(i8, i8) unnamed_addr {
+; CHECK-LABEL: signed_multiplication_did_overflow:
+entry-block:
+  %2 = tail call { i8, i1 } @llvm.smul.with.overflow.i8(i8 %0, i8 %1)
+  %3 = extractvalue { i8, i1 } %2, 1
+  ret i1 %3
+
+; Multiply, fill the low byte with the sign of the low byte via
+; arithmetic shifting, compare it to the high byte.
+;
+; CHECK: muls   r24, r22
+; CHECK: mov    [[HIGH:r[0-9]+]], r1
+; CHECK: mov    [[LOW:r[0-9]+]], r0
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: asr    {{.*}}[[LOW]]
+; CHECK: ldi    [[RET:r[0-9]+]], 1
+; CHECK: cp     {{.*}}[[HIGH]], {{.*}}[[LOW]]
+; CHECK: brne   [[LABEL:LBB[_0-9]+]]
+; CHECK: ldi    {{.*}}[[RET]], 0
+; CHECK: {{.*}}[[LABEL]]
+; CHECK: ret
+}
+
+declare { i8, i1 } @llvm.smul.with.overflow.i8(i8, i8)

test/CodeGen/AVR/smul-with-overflow.ll

@@ -0,0 +1,22 @@
+; RUN: llc < %s -march=avr | FileCheck %s
+
+define i1 @unsigned_multiplication_did_overflow(i8, i8) unnamed_addr {
+; CHECK-LABEL: unsigned_multiplication_did_overflow:
+entry-block:
+  %2 = tail call { i8, i1 } @llvm.umul.with.overflow.i8(i8 %0, i8 %1)
+  %3 = extractvalue { i8, i1 } %2, 1
+  ret i1 %3
+
+; Multiply, return if the high byte is zero
+;
+; CHECK: mul    r{{[0-9]+}}, r{{[0-9]+}}
+; CHECK: mov    [[HIGH:r[0-9]+]], r1
+; CHECK: ldi    [[RET:r[0-9]+]], 1
+; CHECK: cpi    {{.*}}[[HIGH]], 0
+; CHECK: brne   [[LABEL:LBB[_0-9]+]]
+; CHECK: ldi    {{.*}}[[RET]], 0
+; CHECK: {{.*}}[[LABEL]]
+; CHECK: ret
+}
+
+declare { i8, i1 } @llvm.umul.with.overflow.i8(i8, i8)