[X86] Turn X86ISD::AND nodes that have no flag users back into ISD::AND just before isel to enable test instruction matching

[craig.topper]

Summary

EmitTest sometimes creates X86ISD::AND specifically to hide the AND from DAG combine. But this prevents isel patterns that look for (cmp (and X, Y), 0) from being able to see it. So we end up with an AND and a TEST. The TEST gets removed by compare instruction optimization during the peephole pass.

This patch attempts to fix this by converting X86ISD::AND with no flag users back into ISD::AND during the DAG preprocessing just before isel.

In order to do this correctly I had to make the X86ISD::AND node created by EmitTest in this case really have a flag output. Which arguably it should have had anyway so that the number of operands would be consistent for the opcode in all cases. Then I had to modify the ReplaceAllUsesWith to understand that we might be looking at an instruction with 2 outputs. Though in this case there are no uses to replace since we just created the node, but that's what the code did before so I just made it keep working.

[RKSimon]

LGTM

llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp

void X86DAGToDAGISel::PreprocessISelDAG() {
   OptForSize = MF->getFunction().optForSize();
   OptForMinSize = MF->getFunction().optForMinSize();
   assert((!OptForMinSize || OptForSize) && "OptForMinSize implies OptForSize");
 
   for (SelectionDAG::allnodes_iterator I = CurDAG->allnodes_begin(),
        E = CurDAG->allnodes_end(); I != E; ) {
     SDNode *N = &*I++; // Preincrement iterator to avoid invalidation issues.
 
+    // If this is a target specific AND node with no flag usages, turn it back
+    // into ISD::AND to enable test instruction matching.
+    if (N->getOpcode() == X86ISD::AND && !N->hasAnyUseOfValue(1)) {
+      SDValue Res = CurDAG->getNode(ISD::AND, SDLoc(N), N->getValueType(0),
+                                    N->getOperand(0), N->getOperand(1));
+      --I;
+      CurDAG->ReplaceAllUsesOfValueWith(SDValue(N, 0), Res);
+      ++I;
+      CurDAG->DeleteNode(N);
+    }
+
     if (OptLevel != CodeGenOpt::None &&
         // Only do this when the target can fold the load into the call or
         // jmp.
         !Subtarget->useRetpoline() &&
         ((N->getOpcode() == X86ISD::CALL && !Subtarget->slowTwoMemOps()) ||
          (N->getOpcode() == X86ISD::TC_RETURN &&
           (Subtarget->is64Bit() ||
            !getTargetMachine().isPositionIndependent())))) {

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

switch (WideVal.getOpcode()) {
       case ISD::XOR: ConvertedOp = X86ISD::XOR; break;
     }
 
     if (ConvertedOp) {
       const TargetLowering &TLI = DAG.getTargetLoweringInfo();
       if (TLI.isOperationLegal(WideVal.getOpcode(), WideVT)) {
         SDValue V0 = DAG.getNode(ISD::TRUNCATE, dl, VT, WideVal.getOperand(0));
         SDValue V1 = DAG.getNode(ISD::TRUNCATE, dl, VT, WideVal.getOperand(1));
-        Op = DAG.getNode(ConvertedOp, dl, VT, V0, V1);
+        SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
+        Op = DAG.getNode(ConvertedOp, dl, VTs, V0, V1);
       }
     }
   }
 
   if (Opcode == 0) {
     // Emit KTEST for bit vectors
     if (auto Node = EmitKTEST(Op, DAG, Subtarget))
       return Node;
 
     // Emit a CMP with 0, which is the TEST pattern.
     return DAG.getNode(X86ISD::CMP, dl, MVT::i32, Op,
                        DAG.getConstant(0, dl, Op.getValueType()));
   }
   SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
   SmallVector<SDValue, 4> Ops(Op->op_begin(), Op->op_begin() + NumOperands);
 
   SDValue New = DAG.getNode(Opcode, dl, VTs, Ops);
-  DAG.ReplaceAllUsesWith(Op, New);
+  DAG.ReplaceAllUsesOfValueWith(SDValue(Op.getNode(), 0), New);
   return SDValue(New.getNode(), 1);
 }
 
 /// Emit nodes that will be selected as "cmp Op0,Op1", or something
 /// equivalent.
 SDValue X86TargetLowering::EmitCmp(SDValue Op0, SDValue Op1, unsigned X86CC,
                                    const SDLoc &dl, SelectionDAG &DAG) const {
   if (isNullConstant(Op1))

llvm/trunk/test/CodeGen/X86/2012-08-16-setcc.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
 
 ; rdar://12081007
 
 define i32 @and_1(i8 zeroext %a, i8 zeroext %b, i32 %x) {
 ; CHECK-LABEL: and_1:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    xorl %eax, %eax
-; CHECK-NEXT:    andb %dil, %sil
+; CHECK-NEXT:    testb %dil, %sil
 ; CHECK-NEXT:    cmovnel %edx, %eax
 ; CHECK-NEXT:    retq
   %1 = and i8 %b, %a
   %2 = icmp ne i8 %1, 0
   %3 = select i1 %2, i32 %x, i32 0
   ret i32 %3
 }
 
 define zeroext i1 @and_2(i8 zeroext %a, i8 zeroext %b) {
 ; CHECK-LABEL: and_2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    andb %dil, %sil
+; CHECK-NEXT:    testb %dil, %sil
 ; CHECK-NEXT:    setne %al
 ; CHECK-NEXT:    retq
   %1 = and i8 %b, %a
   %2 = icmp ne i8 %1, 0
   ret i1 %2
 }
 
 define i32 @xor_1(i8 zeroext %a, i8 zeroext %b, i32 %x) {

llvm/trunk/test/CodeGen/X86/jump_sign.ll

 @a = common global i32 0, align 4
 define i32 @func_test1(i32 %p1) nounwind uwtable {
 ; CHECK-LABEL: func_test1:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    movl b, %eax
 ; CHECK-NEXT:    cmpl {{[0-9]+}}(%esp), %eax
 ; CHECK-NEXT:    setb %cl
 ; CHECK-NEXT:    movl a, %eax
-; CHECK-NEXT:    movl %eax, %edx
-; CHECK-NEXT:    andb %cl, %dl
+; CHECK-NEXT:    testb %al, %cl
 ; CHECK-NEXT:    je .LBB18_2
 ; CHECK-NEXT:  # %bb.1: # %if.then
 ; CHECK-NEXT:    decl %eax
 ; CHECK-NEXT:    movl %eax, a
 ; CHECK-NEXT:  .LBB18_2: # %if.end
 ; CHECK-NEXT:    retl
 entry:
   %t0 = load i32, i32* @b, align 4

llvm/trunk/test/CodeGen/X86/test-shrink-bug.ll

 ; CHECK-X86-NEXT:  LBB1_2: ## %yes
 ; CHECK-X86-NEXT:    addl $12, %esp
 ; CHECK-X86-NEXT:    retl
 ;
 ; CHECK-X64-LABEL: fail:
 ; CHECK-X64:       # %bb.0:
 ; CHECK-X64-NEXT:    pushq %rax
 ; CHECK-X64-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-X64-NEXT:    andw $263, %di # imm = 0x107
+; CHECK-X64-NEXT:    testw $263, %di # imm = 0x107
 ; CHECK-X64-NEXT:    je .LBB1_2
 ; CHECK-X64-NEXT:  # %bb.1:
 ; CHECK-X64-NEXT:    pand {{.*}}(%rip), %xmm0
 ; CHECK-X64-NEXT:    pcmpeqd {{.*}}(%rip), %xmm0
 ; CHECK-X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
 ; CHECK-X64-NEXT:    pand %xmm0, %xmm1
 ; CHECK-X64-NEXT:    pextrw $4, %xmm1, %eax
 ; CHECK-X64-NEXT:    testb $1, %al