[PowerPC] Eliminate integer compare instructions - vol. 1

This patch is the first in a series of patches to provide code gen for
doing compares in GPRs when the compare result is required in a GPR.

It adds the infrastructure to select GPR sequences for i1->i32 and i1->i64
extensions. This first patch handles equality comparison on i32 operands with
the result sign or zero extended.

Differential Revision: https://reviews.llvm.org/D31847


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@302810 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp

@@ -95,7 +95,8 @@ void PPCInstPrinter::printInst(const MCInst *MI, raw_ostream &O,
     return;
   }
 
-  if (MI->getOpcode() == PPC::RLDICR) {
+  if (MI->getOpcode() == PPC::RLDICR ||
+      MI->getOpcode() == PPC::RLDICR_32) {
     unsigned char SH = MI->getOperand(2).getImm();
     unsigned char ME = MI->getOperand(3).getImm();
     // rldicr RA, RS, SH, 63-SH == sldi RA, RS, SH

lib/Target/PowerPC/PPCFastISel.cpp

@@ -2246,6 +2246,7 @@ bool PPCFastISel::tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
     }
 
     case PPC::EXTSW:
+    case PPC::EXTSW_32:
     case PPC::EXTSW_32_64: {
       if (VT != MVT::i32 && VT != MVT::i16 && VT != MVT::i8)
         return false;

lib/Target/PowerPC/PPCISelDAGToDAG.cpp

@@ -54,6 +54,7 @@
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Target/TargetInstrInfo.h"
 #include "llvm/Target/TargetRegisterInfo.h"
+#include "llvm/ADT/Statistic.h"
 #include <algorithm>
 #include <cassert>
 #include <cstdint>
@@ -68,6 +69,14 @@ using namespace llvm;
 
 #define DEBUG_TYPE "ppc-codegen"
 
+STATISTIC(NumSextSetcc,
+          "Number of (sext(setcc)) nodes expanded into GPR sequence.");
+STATISTIC(NumZextSetcc,
+          "Number of (zext(setcc)) nodes expanded into GPR sequence.");
+STATISTIC(SignExtensionsAdded,
+          "Number of sign extensions for compare inputs added.");
+STATISTIC(ZeroExtensionsAdded,
+          "Number of zero extensions for compare inputs added.");
 // FIXME: Remove this once the bug has been fixed!
 cl::opt<bool> ANDIGlueBug("expose-ppc-andi-glue-bug",
 cl::desc("expose the ANDI glue bug on PPC"), cl::Hidden);
@@ -252,7 +261,28 @@ namespace {
 #include "PPCGenDAGISel.inc"
 
 private:
+    // Conversion type for interpreting results of a 32-bit instruction as
+    // a 64-bit value or vice versa.
+    enum ExtOrTruncConversion { Ext, Trunc };
+
+    // Modifiers to guide how an ISD::SETCC node's result is to be computed
+    // in a GPR.
+    // ZExtOrig - use the original condition code, zero-extend value
+    // ZExtInvert - invert the condition code, zero-extend value
+    // SExtOrig - use the original condition code, sign-extend value
+    // SExtInvert - invert the condition code, sign-extend value
+    enum SetccInGPROpts { ZExtOrig, ZExtInvert, SExtOrig, SExtInvert };
+
     bool trySETCC(SDNode *N);
+    bool tryEXTEND(SDNode *N);
+    SDValue signExtendInputIfNeeded(SDValue Input);
+    SDValue zeroExtendInputIfNeeded(SDValue Input);
+    SDValue addExtOrTrunc(SDValue NatWidthRes, ExtOrTruncConversion Conv);
+    SDValue get32BitZExtCompare(SDValue LHS, SDValue RHS, ISD::CondCode CC,
+                                int64_t RHSValue, SDLoc dl);
+    SDValue get32BitSExtCompare(SDValue LHS, SDValue RHS, ISD::CondCode CC,
+                                int64_t RHSValue, SDLoc dl);
+    SDValue getSETCCInGPR(SDValue Compare, SetccInGPROpts ConvOpts);
 
     void PeepholePPC64();
     void PeepholePPC64ZExt();
@@ -2471,6 +2501,225 @@ bool PPCDAGToDAGISel::trySETCC(SDNode *N) {
   return true;
 }
 
+/// If this node is a sign/zero extension of an integer comparison,
+/// it can usually be computed in GPR's rather than using comparison
+/// instructions and ISEL. We only do this on 64-bit targets for now
+/// as the code is specialized for 64-bit (it uses 64-bit instructions
+/// and assumes 64-bit registers).
+bool PPCDAGToDAGISel::tryEXTEND(SDNode *N) {
+  if (TM.getOptLevel() == CodeGenOpt::None || !TM.isPPC64())
+    return false;
+  assert((N->getOpcode() == ISD::ZERO_EXTEND ||
+          N->getOpcode() == ISD::SIGN_EXTEND) &&
+          "Expecting a zero/sign extend node!");
+
+  if (N->getOperand(0).getOpcode() != ISD::SETCC)
+    return false;
+
+  SDValue WideRes =
+    getSETCCInGPR(N->getOperand(0),
+                  N->getOpcode() == ISD::SIGN_EXTEND ?
+                  SetccInGPROpts::SExtOrig : SetccInGPROpts::ZExtOrig);
+
+  if (!WideRes)
+    return false;
+
+  SDLoc dl(N);
+  bool Inputs32Bit = N->getOperand(0).getOperand(0).getValueType() == MVT::i32;
+  bool Output32Bit = N->getValueType(0) == MVT::i32;
+
+  NumSextSetcc += N->getOpcode() == ISD::SIGN_EXTEND ? 1 : 0;
+  NumZextSetcc += N->getOpcode() == ISD::SIGN_EXTEND ? 0 : 1;
+
+  SDValue ConvOp = WideRes;
+  if (Inputs32Bit != Output32Bit)
+    ConvOp = addExtOrTrunc(WideRes, Inputs32Bit ? ExtOrTruncConversion::Ext :
+                           ExtOrTruncConversion::Trunc);
+  ReplaceNode(N, ConvOp.getNode());
+
+  return true;
+}
+
+/// If the value isn't guaranteed to be sign-extended to 64-bits, extend it.
+/// Useful when emitting comparison code for 32-bit values without using
+/// the compare instruction (which only considers the lower 32-bits).
+SDValue PPCDAGToDAGISel::signExtendInputIfNeeded(SDValue Input) {
+  assert(Input.getValueType() == MVT::i32 &&
+         "Can only sign-extend 32-bit values here.");
+  unsigned Opc = Input.getOpcode();
+
+  // The value was sign extended and then truncated to 32-bits. No need to
+  // sign extend it again.
+  if (Opc == ISD::TRUNCATE &&
+      (Input.getOperand(0).getOpcode() == ISD::AssertSext ||
+       Input.getOperand(0).getOpcode() == ISD::SIGN_EXTEND))
+    return Input;
+
+  LoadSDNode *InputLoad = dyn_cast<LoadSDNode>(Input);
+  // The input is a sign-extending load. No reason to sign-extend.
+  if (InputLoad && InputLoad->getExtensionType() == ISD::SEXTLOAD)
+    return Input;
+
+  ConstantSDNode *InputConst = dyn_cast<ConstantSDNode>(Input);
+  // We don't sign-extend constants and already sign-extended values.
+  if (InputConst || Opc == ISD::AssertSext || Opc == ISD::SIGN_EXTEND_INREG ||
+      Opc == ISD::SIGN_EXTEND)
+    return Input;
+
+  SDLoc dl(Input);
+  SignExtensionsAdded++;
+  return SDValue(CurDAG->getMachineNode(PPC::EXTSW_32, dl, MVT::i32, Input), 0);
+}
+
+/// If the value isn't guaranteed to be zero-extended to 64-bits, extend it.
+/// Useful when emitting comparison code for 32-bit values without using
+/// the compare instruction (which only considers the lower 32-bits).
+SDValue PPCDAGToDAGISel::zeroExtendInputIfNeeded(SDValue Input) {
+  assert(Input.getValueType() == MVT::i32 &&
+         "Can only zero-extend 32-bit values here.");
+  LoadSDNode *InputLoad = dyn_cast<LoadSDNode>(Input);
+  unsigned Opc = Input.getOpcode();
+
+  // No need to zero-extend loaded values (unless they're loaded with
+  // a sign-extending load).
+  if (InputLoad && InputLoad->getExtensionType() != ISD::SEXTLOAD)
+    return Input;
+
+  ConstantSDNode *InputConst = dyn_cast<ConstantSDNode>(Input);
+  bool InputZExtConst = InputConst && InputConst->getSExtValue() >= 0;
+  // An ISD::TRUNCATE will be lowered to an EXTRACT_SUBREG so we have
+  // to conservatively actually clear the high bits. We also don't need to
+  // zero-extend constants or values that are already zero-extended.
+  if (InputZExtConst || Opc == ISD::AssertZext || Opc == ISD::ZERO_EXTEND)
+    return Input;
+
+  SDLoc dl(Input);
+  ZeroExtensionsAdded++;
+  return SDValue(CurDAG->getMachineNode(PPC::RLDICL_32, dl, MVT::i32, Input,
+                                        getI64Imm(0, dl), getI64Imm(32, dl)),
+                 0);
+}
+
+// Handle a 32-bit value in a 64-bit register and vice-versa. These are of
+// course not actual zero/sign extensions that will generate machine code,
+// they're just a way to reinterpret a 32 bit value in a register as a
+// 64 bit value and vice-versa.
+SDValue PPCDAGToDAGISel::addExtOrTrunc(SDValue NatWidthRes,
+                                       ExtOrTruncConversion Conv) {
+  SDLoc dl(NatWidthRes);
+
+  // For reinterpreting 32-bit values as 64 bit values, we generate
+  // INSERT_SUBREG IMPLICIT_DEF:i64, <input>, TargetConstant:i32<1>
+  if (Conv == ExtOrTruncConversion::Ext) {
+    SDValue ImDef(CurDAG->getMachineNode(PPC::IMPLICIT_DEF, dl, MVT::i64), 0);
+    SDValue SubRegIdx =
+      CurDAG->getTargetConstant(PPC::sub_32, dl, MVT::i32);
+    return SDValue(CurDAG->getMachineNode(PPC::INSERT_SUBREG, dl, MVT::i64,
+                                          ImDef, NatWidthRes, SubRegIdx), 0);
+  }
+
+  assert(Conv == ExtOrTruncConversion::Trunc &&
+         "Unknown convertion between 32 and 64 bit values.");
+  // For reinterpreting 64-bit values as 32-bit values, we just need to
+  // EXTRACT_SUBREG (i.e. extract the low word).
+  SDValue SubRegIdx =
+    CurDAG->getTargetConstant(PPC::sub_32, dl, MVT::i32);
+  return SDValue(CurDAG->getMachineNode(PPC::EXTRACT_SUBREG, dl, MVT::i32,
+                                        NatWidthRes, SubRegIdx), 0);
+}
+
+/// Produces a zero-extended result of comparing two 32-bit values according to
+/// the passed condition code.
+SDValue PPCDAGToDAGISel::get32BitZExtCompare(SDValue LHS, SDValue RHS,
+                                             ISD::CondCode CC,
+                                             int64_t RHSValue, SDLoc dl) {
+  bool IsRHSZero = RHSValue == 0;
+  switch (CC) {
+  default: return SDValue();
+  case ISD::SETEQ: {
+    // (zext (setcc %a, %b, seteq)) -> (lshr (cntlzw (xor %a, %b)), 5)
+    // (zext (setcc %a, 0, seteq))  -> (lshr (cntlzw %a), 5)
+    SDValue Xor = IsRHSZero ? LHS :
+      SDValue(CurDAG->getMachineNode(PPC::XOR, dl, MVT::i32, LHS, RHS), 0);
+    SDValue Clz =
+      SDValue(CurDAG->getMachineNode(PPC::CNTLZW, dl, MVT::i32, Xor), 0);
+    SDValue ShiftOps[] = { Clz, getI32Imm(27, dl), getI32Imm(5, dl),
+      getI32Imm(31, dl) };
+    return SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32,
+                                          ShiftOps), 0);
+  }
+  }
+}
+
+/// Produces a sign-extended result of comparing two 32-bit values according to
+/// the passed condition code.
+SDValue PPCDAGToDAGISel::get32BitSExtCompare(SDValue LHS, SDValue RHS,
+                                             ISD::CondCode CC,
+                                             int64_t RHSValue, SDLoc dl) {
+  bool IsRHSZero = RHSValue == 0;
+  switch (CC) {
+  default: return SDValue();
+  case ISD::SETEQ: {
+    // (sext (setcc %a, %b, seteq)) ->
+    //   (ashr (shl (ctlz (xor %a, %b)), 58), 63)
+    // (sext (setcc %a, 0, seteq)) ->
+    //   (ashr (shl (ctlz %a), 58), 63)
+    SDValue CountInput = IsRHSZero ? LHS :
+      SDValue(CurDAG->getMachineNode(PPC::XOR, dl, MVT::i32, LHS, RHS), 0);
+    SDValue Cntlzw =
+      SDValue(CurDAG->getMachineNode(PPC::CNTLZW, dl, MVT::i32, CountInput), 0);
+    SDValue SHLOps[] = { Cntlzw, getI32Imm(58, dl), getI32Imm(0, dl) };
+    SDValue Sldi =
+      SDValue(CurDAG->getMachineNode(PPC::RLDICR_32, dl, MVT::i32, SHLOps), 0);
+    return SDValue(CurDAG->getMachineNode(PPC::SRADI_32, dl, MVT::i32, Sldi,
+                                          getI32Imm(63, dl)), 0);
+  }
+  }
+}
+
+/// Returns an equivalent of a SETCC node but with the result the same width as
+/// the inputs. This can nalso be used for SELECT_CC if either the true or false
+/// values is a power of two while the other is zero.
+SDValue PPCDAGToDAGISel::getSETCCInGPR(SDValue Compare,
+                                       SetccInGPROpts ConvOpts) {
+  assert((Compare.getOpcode() == ISD::SETCC ||
+          Compare.getOpcode() == ISD::SELECT_CC) &&
+         "An ISD::SETCC node required here.");
+
+  SDValue LHS = Compare.getOperand(0);
+  SDValue RHS = Compare.getOperand(1);
+
+  // The condition code is operand 2 for SETCC and operand 4 for SELECT_CC.
+  int CCOpNum = Compare.getOpcode() == ISD::SELECT_CC ? 4 : 2;
+  ISD::CondCode CC =
+    cast<CondCodeSDNode>(Compare.getOperand(CCOpNum))->get();
+  EVT InputVT = LHS.getValueType();
+  if (InputVT != MVT::i32)
+    return SDValue();
+
+  SDLoc dl(Compare);
+  ConstantSDNode *RHSConst = dyn_cast<ConstantSDNode>(RHS);
+  int64_t RHSValue = RHSConst ? RHSConst->getSExtValue() : INT64_MAX;
+
+  if (ConvOpts == SetccInGPROpts::ZExtInvert ||
+      ConvOpts == SetccInGPROpts::SExtInvert)
+    CC = ISD::getSetCCInverse(CC, true);
+
+  if (ISD::isSignedIntSetCC(CC)) {
+    LHS = signExtendInputIfNeeded(LHS);
+    RHS = signExtendInputIfNeeded(RHS);
+  } else if (ISD::isUnsignedIntSetCC(CC)) {
+    LHS = zeroExtendInputIfNeeded(LHS);
+    RHS = zeroExtendInputIfNeeded(RHS);
+  }
+
+  bool IsSext = ConvOpts == SetccInGPROpts::SExtOrig ||
+    ConvOpts == SetccInGPROpts::SExtInvert;
+  if (IsSext)
+    return get32BitSExtCompare(LHS, RHS, CC, RHSValue, dl);
+  return get32BitZExtCompare(LHS, RHS, CC, RHSValue, dl);
+}
+
 void PPCDAGToDAGISel::transferMemOperands(SDNode *N, SDNode *Result) {
   // Transfer memoperands.
   MachineSDNode::mmo_iterator MemOp = MF->allocateMemRefsArray(1);
@@ -2508,6 +2757,12 @@ void PPCDAGToDAGISel::Select(SDNode *N) {
     }
     break;
 
+  case ISD::ZERO_EXTEND:
+  case ISD::SIGN_EXTEND:
+    if (tryEXTEND(N))
+      return;
+    break;
+
   case ISD::SETCC:
     if (trySETCC(N))
       return;

lib/Target/PowerPC/PPCInstr64Bit.td

@@ -634,10 +634,19 @@ let Interpretation64Bit = 1, isCodeGenOnly = 1 in
 defm EXTSW_32_64 : XForm_11r<31, 986, (outs g8rc:$rA), (ins gprc:$rS),
                              "extsw", "$rA, $rS", IIC_IntSimple,
                              [(set i64:$rA, (sext i32:$rS))]>, isPPC64;
+let isCodeGenOnly = 1 in
+def EXTSW_32 : XForm_11<31, 986, (outs gprc:$rA), (ins gprc:$rS),
+                        "extsw $rA, $rS", IIC_IntSimple,
+                        []>, isPPC64;
 
 defm SRADI  : XSForm_1rc<31, 413, (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH),
                          "sradi", "$rA, $rS, $SH", IIC_IntRotateDI,
                          [(set i64:$rA, (sra i64:$rS, (i32 imm:$SH)))]>, isPPC64;
+// For fast-isel:
+let isCodeGenOnly = 1 in
+def SRADI_32  : XSForm_1<31, 413, (outs gprc:$rA), (ins gprc:$rS, u6imm:$SH),
+                         "sradi $rA, $rS, $SH", IIC_IntRotateDI, []>, isPPC64;
+
 defm CNTLZD : XForm_11r<31,  58, (outs g8rc:$rA), (ins g8rc:$rS),
                         "cntlzd", "$rA, $rS", IIC_IntGeneral,
                         [(set i64:$rA, (ctlz i64:$rS))]>;
@@ -721,15 +730,26 @@ defm RLDICL : MDForm_1r<30, 0,
 // For fast-isel:
 let isCodeGenOnly = 1 in
 def RLDICL_32_64 : MDForm_1<30, 0,
-                           (outs g8rc:$rA),
-                           (ins gprc:$rS, u6imm:$SH, u6imm:$MBE),
-                           "rldicl $rA, $rS, $SH, $MBE", IIC_IntRotateDI,
-                           []>, isPPC64;
+                            (outs g8rc:$rA),
+                            (ins gprc:$rS, u6imm:$SH, u6imm:$MBE),
+                            "rldicl $rA, $rS, $SH, $MBE", IIC_IntRotateDI,
+                            []>, isPPC64;
 // End fast-isel.
+let isCodeGenOnly = 1 in
+def RLDICL_32 : MDForm_1<30, 0,
+                         (outs gprc:$rA),
+                         (ins gprc:$rS, u6imm:$SH, u6imm:$MBE),
+                         "rldicl $rA, $rS, $SH, $MBE", IIC_IntRotateDI,
+                         []>, isPPC64;
 defm RLDICR : MDForm_1r<30, 1,
                         (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH, u6imm:$MBE),
                         "rldicr", "$rA, $rS, $SH, $MBE", IIC_IntRotateDI,
                         []>, isPPC64;
+let isCodeGenOnly = 1 in
+def RLDICR_32 : MDForm_1<30, 1,
+                         (outs gprc:$rA), (ins gprc:$rS, u6imm:$SH, u6imm:$MBE),
+                         "rldicr $rA, $rS, $SH, $MBE", IIC_IntRotateDI,
+                         []>, isPPC64;
 defm RLDIC  : MDForm_1r<30, 2,
                         (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH, u6imm:$MBE),
                         "rldic", "$rA, $rS, $SH, $MBE", IIC_IntRotateDI,

lib/Target/PowerPC/PPCInstrInfo.td

@@ -4166,6 +4166,8 @@ def : InstAlias<"rotldi. $rA, $rS, $n", (RLDICLo g8rc:$rA, g8rc:$rS, u6imm:$n, 0
 def : InstAlias<"rotld $rA, $rS, $rB", (RLDCL g8rc:$rA, g8rc:$rS, gprc:$rB, 0)>;
 def : InstAlias<"rotld. $rA, $rS, $rB", (RLDCLo g8rc:$rA, g8rc:$rS, gprc:$rB, 0)>;
 def : InstAlias<"clrldi $rA, $rS, $n", (RLDICL g8rc:$rA, g8rc:$rS, 0, u6imm:$n)>;
+def : InstAlias<"clrldi $rA, $rS, $n",
+                (RLDICL_32 gprc:$rA, gprc:$rS, 0, u6imm:$n)>;
 def : InstAlias<"clrldi. $rA, $rS, $n", (RLDICLo g8rc:$rA, g8rc:$rS, 0, u6imm:$n)>;
 
 def RLWINMbm : PPCAsmPseudo<"rlwinm $rA, $rS, $n, $b",