[lanai] Manually match 0/-1 with R0/R1.

[jpienaar]

Summary

Previously 0 and -1 was matched via tablegen rules. But this could cause problems where a physical register was being used where a virtual register was expected (seen in optimizeSelect and TwoAddressInstructionPass). Instead follow AArch64 and match in DAGToDAGISel.

[jpienaar]

Updated to test the type before matching R0 and R1.

[majnemer]

LGTM

lib/Target/Lanai/LanaiISelDAGToDAG.cpp

void LanaiDAGToDAGISel::Select(SDNode *Node) {
   DEBUG(errs() << "Selecting: "; Node->dump(CurDAG); errs() << "\n");
 
   // If we have a custom node, we already have selected!
   if (Node->isMachineOpcode()) {
     DEBUG(errs() << "== "; Node->dump(CurDAG); errs() << "\n");
     return;
   }
 
-  // Instruction Selection not handled by the auto-generated
-  // tablegen selection should be handled here.
+  // Instruction Selection not handled by the auto-generated tablegen selection
+  // should be handled here.
   switch (Opcode) {
+  case ISD::Constant: {
+    ConstantSDNode *ConstNode = cast<ConstantSDNode>(Node);
+    // Materialize zero constants as copies from R0. This allows the coalescer
+    // to propagate these into other instructions.
+    if (ConstNode->isNullValue()) {
+      SDValue New = CurDAG->getCopyFromReg(CurDAG->getEntryNode(), SDLoc(Node),
+                                           Lanai::R0, MVT::i32);
+      return ReplaceNode(Node, New.getNode());
+    }
+    // Materialize all ones constants as copies from R1. This allows the
+    // coalescer to propagate these into other instructions.
+    if (ConstNode->isAllOnesValue()) {
+      SDValue New = CurDAG->getCopyFromReg(CurDAG->getEntryNode(), SDLoc(Node),
+                                           Lanai::R1, MVT::i32);
+      return ReplaceNode(Node, New.getNode());
+    }
+    break;
+  }
   case ISD::FrameIndex:
     selectFrameIndex(Node);
     return;
   default:
     break;
   }
 
   // Select the default instruction

lib/Target/Lanai/LanaiInstrInfo.td

def TRAILZ : InstSpecial<0b011, (outs GPR:$Rd), (ins GPR:$Rs1),
                          "trailz\t$Rs1, $Rd",
                          [(set GPR:$Rd, (cttz GPR:$Rs1))]>;
 }
 
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns
 //===----------------------------------------------------------------------===//
 
-// i32 0 and R0 can be used interchangeably.
-def : Pat<(i32 0), (i32 R0)>;
-// i32 -1 and R1 can be used interchangeably.
-def : Pat<(i32 -1), (i32 R1)>;
-
 // unsigned 16-bit immediate
 def : Pat<(i32 i32lo16z:$imm), (OR_I_LO (i32 R0), imm:$imm)>;
 
 // arbitrary immediate
 def : Pat<(i32 imm:$imm), (OR_I_LO (MOVHI (HI16 imm:$imm)), (LO16 imm:$imm))>;
 
 // Calls
 def : Pat<(Call tglobaladdr:$dst), (CALL tglobaladdr:$dst)>;

test/CodeGen/Lanai/constant_multiply.ll

 ; CHECK: sh %r6, 0x3, %r{{[0-9]+}}
 ; CHECK: sub %r{{[0-9]+}}, %r{{[0-9]+}}, %rv
 define i32 @fm8(i32 inreg %a) #0 {
   %1 = mul nsw i32 %a, -8
   ret i32 %1
 }
 
 ; CHECK-LABEL: fm9:
-; CHECK: sh %r6, 0x3, %r{{[0-9]+}}
-; CHECK: sub %r{{[0-9]+}}, %r6, %r{{[0-9]+}}
-; CHECK: sub %r{{[0-9]+}}, %r{{[0-9]+}}, %rv
+; CHECK: sub  %r0, %r6, %r{{[0-9]+}}
+; CHECK: sh %r6, 0x3, %r9
+; CHECK: sub  %r{{[0-9]+}}, %r9, %rv
 define i32 @fm9(i32 inreg %a) #0 {
   %1 = mul nsw i32 %a, -9
   ret i32 %1
 }
 
 ; CHECK-LABEL: fm10:
-; CHECK: sh %r6, 0x3, %r{{[0-9]+}}
 ; CHECK: sh %r6, 0x1, %r{{[0-9]+}}
 ; CHECK: sub %r{{[0-9]+}}, %r{{[0-9]+}}, %r{{[0-9]+}}
+; CHECK: sh %r6, 0x3, %r{{[0-9]+}}
 ; CHECK: sub %r{{[0-9]+}}, %r{{[0-9]+}}, %rv
 define i32 @fm10(i32 inreg %a) #0 {
   %1 = mul nsw i32 %a, -10
   ret i32 %1
 }
 
 ; CHECK-LABEL: h1:
 ; CHECK: __mulsi3
 define i32 @h1(i32 inreg %a) #0 {
   %1 = mul i32 %a, -1431655765
   ret i32 %1
 }