[PowerPC] Match vec_revb builtins to P9 instructions.

Power9 has instructions that will reverse the bytes within an element for all
sizes (half-word, word, double-word and quad-word). These can be used for the
vec_revb builtins in altivec.h. However, we implement these to match vector
shuffle nodes as that will cover both the builtins and vector shuffles that
occur in the SDAG through other means.

Differential Revision: https://reviews.llvm.org/D33690

llvm-svn: 305214

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

@@ -1128,6 +1128,7 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const {
   case PPCISD::VPERM:           return "PPCISD::VPERM";
   case PPCISD::XXSPLT:          return "PPCISD::XXSPLT";
   case PPCISD::XXINSERT:        return "PPCISD::XXINSERT";
+  case PPCISD::XXREVERSE:       return "PPCISD::XXREVERSE";
   case PPCISD::XXPERMDI:        return "PPCISD::XXPERMDI";
   case PPCISD::VECSHL:          return "PPCISD::VECSHL";
   case PPCISD::CMPB:            return "PPCISD::CMPB";
@@ -1610,22 +1611,34 @@ bool PPC::isSplatShuffleMask(ShuffleVectorSDNode *N, unsigned EltSize) {
   return true;
 }
 
-// Check that the mask is shuffling N byte elements.
-static bool isNByteElemShuffleMask(ShuffleVectorSDNode *N, unsigned Width) {
+/// Check that the mask is shuffling N byte elements. Within each N byte
+/// element of the mask, the indices could be either in increasing or
+/// decreasing order as long as they are consecutive.
+/// \param[in] N: the shuffle vector SD Node to analyze
+/// \param[in] Width: the element width in bytes, could be 2/4/8/16 (HalfWord/
+/// Word/DoubleWord/QuadWord).
+/// \param[in] StepLen: the delta indices number among the N byte element, if
+/// the mask is in increasing/decreasing order then it is 1/-1.
+/// \return true iff the mask is shuffling N byte elements.
+static bool isNByteElemShuffleMask(ShuffleVectorSDNode *N, unsigned Width,
+                                   int StepLen) {
   assert((Width == 2 || Width == 4 || Width == 8 || Width == 16) &&
          "Unexpected element width.");
+  assert((StepLen == 1 || StepLen == -1) && "Unexpected element width.");
 
   unsigned NumOfElem = 16 / Width;
   unsigned MaskVal[16]; //  Width is never greater than 16
   for (unsigned i = 0; i < NumOfElem; ++i) {
     MaskVal[0] = N->getMaskElt(i * Width);
-    if (MaskVal[0] % Width) {
+    if ((StepLen == 1) && (MaskVal[0] % Width)) {
+      return false;
+    } else if ((StepLen == -1) && ((MaskVal[0] + 1) % Width)) {
       return false;
     }
 
     for (unsigned int j = 1; j < Width; ++j) {
       MaskVal[j] = N->getMaskElt(i * Width + j);
-      if (MaskVal[j] != MaskVal[j-1] + 1) {
+      if (MaskVal[j] != MaskVal[j-1] + StepLen) {
         return false;
       }
     }
@@ -1636,7 +1649,7 @@ static bool isNByteElemShuffleMask(ShuffleVectorSDNode *N, unsigned Width) {
 
 bool PPC::isXXINSERTWMask(ShuffleVectorSDNode *N, unsigned &ShiftElts,
                           unsigned &InsertAtByte, bool &Swap, bool IsLE) {
-  if (!isNByteElemShuffleMask(N, 4))
+  if (!isNByteElemShuffleMask(N, 4, 1))
     return false;
 
   // Now we look at mask elements 0,4,8,12
@@ -1713,7 +1726,7 @@ bool PPC::isXXSLDWIShuffleMask(ShuffleVectorSDNode *N, unsigned &ShiftElts,
                                bool &Swap, bool IsLE) {
   assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8");
   // Ensure each byte index of the word is consecutive.
-  if (!isNByteElemShuffleMask(N, 4))
+  if (!isNByteElemShuffleMask(N, 4, 1))
     return false;
 
   // Now we look at mask elements 0,4,8,12, which are the beginning of words.
@@ -1771,6 +1784,35 @@ bool PPC::isXXSLDWIShuffleMask(ShuffleVectorSDNode *N, unsigned &ShiftElts,
   }
 }
 
+bool static isXXBRShuffleMaskHelper(ShuffleVectorSDNode *N, int Width) {
+  assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8");
+
+  if (!isNByteElemShuffleMask(N, Width, -1))
+    return false;
+
+  for (int i = 0; i < 16; i += Width)
+    if (N->getMaskElt(i) != i + Width - 1)
+      return false;
+
+  return true;
+}
+
+bool PPC::isXXBRHShuffleMask(ShuffleVectorSDNode *N) {
+  return isXXBRShuffleMaskHelper(N, 2);
+}
+
+bool PPC::isXXBRWShuffleMask(ShuffleVectorSDNode *N) {
+  return isXXBRShuffleMaskHelper(N, 4);
+}
+
+bool PPC::isXXBRDShuffleMask(ShuffleVectorSDNode *N) {
+  return isXXBRShuffleMaskHelper(N, 8);
+}
+
+bool PPC::isXXBRQShuffleMask(ShuffleVectorSDNode *N) {
+  return isXXBRShuffleMaskHelper(N, 16);
+}
+
 /// Can node \p N be lowered to an XXPERMDI instruction? If so, set \p Swap
 /// if the inputs to the instruction should be swapped and set \p DM to the
 /// value for the immediate.
@@ -1784,7 +1826,7 @@ bool PPC::isXXPERMDIShuffleMask(ShuffleVectorSDNode *N, unsigned &DM,
   assert(N->getValueType(0) == MVT::v16i8 && "Shuffle vector expects v16i8");
 
   // Ensure each byte index of the double word is consecutive.
-  if (!isNByteElemShuffleMask(N, 8))
+  if (!isNByteElemShuffleMask(N, 8, 1))
     return false;
 
   unsigned M0 = N->getMaskElt(0) / 8;
@@ -7859,6 +7901,26 @@ SDValue PPCTargetLowering::LowerVECTOR_SHUFFLE(SDValue Op,
     return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, PermDI);
   }
 
+  if (Subtarget.hasP9Vector()) {
+     if (PPC::isXXBRHShuffleMask(SVOp)) {
+      SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v8i16, V1);
+      SDValue ReveHWord = DAG.getNode(PPCISD::XXREVERSE, dl, MVT::v8i16, Conv);
+      return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveHWord);
+    } else if (PPC::isXXBRWShuffleMask(SVOp)) {
+      SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v4i32, V1);
+      SDValue ReveWord = DAG.getNode(PPCISD::XXREVERSE, dl, MVT::v4i32, Conv);
+      return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveWord);
+    } else if (PPC::isXXBRDShuffleMask(SVOp)) {
+      SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v2i64, V1);
+      SDValue ReveDWord = DAG.getNode(PPCISD::XXREVERSE, dl, MVT::v2i64, Conv);
+      return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveDWord);
+    } else if (PPC::isXXBRQShuffleMask(SVOp)) {
+      SDValue Conv = DAG.getNode(ISD::BITCAST, dl, MVT::v1i128, V1);
+      SDValue ReveQWord = DAG.getNode(PPCISD::XXREVERSE, dl, MVT::v1i128, Conv);
+      return DAG.getNode(ISD::BITCAST, dl, MVT::v16i8, ReveQWord);
+    }
+  }
+
   if (Subtarget.hasVSX()) {
     if (V2.isUndef() && PPC::isSplatShuffleMask(SVOp, 4)) {
       int SplatIdx = PPC::getVSPLTImmediate(SVOp, 4, DAG);

llvm/lib/Target/PowerPC/PPCISelLowering.h

@@ -86,6 +86,10 @@ namespace llvm {
       ///
       XXINSERT,
 
+      /// XXREVERSE - The PPC VSX reverse instruction
+      ///
+      XXREVERSE,
+
       /// VECSHL - The PPC VSX shift left instruction
       ///
       VECSHL,
@@ -458,6 +462,23 @@ namespace llvm {
     /// for a XXSLDWI instruction.
     bool isXXSLDWIShuffleMask(ShuffleVectorSDNode *N, unsigned &ShiftElts,
                               bool &Swap, bool IsLE);
+
+    /// isXXBRHShuffleMask - Return true if this is a shuffle mask suitable
+    /// for a XXBRH instruction.
+    bool isXXBRHShuffleMask(ShuffleVectorSDNode *N);
+
+    /// isXXBRWShuffleMask - Return true if this is a shuffle mask suitable
+    /// for a XXBRW instruction.
+    bool isXXBRWShuffleMask(ShuffleVectorSDNode *N);
+
+    /// isXXBRDShuffleMask - Return true if this is a shuffle mask suitable
+    /// for a XXBRD instruction.
+    bool isXXBRDShuffleMask(ShuffleVectorSDNode *N);
+
+    /// isXXBRQShuffleMask - Return true if this is a shuffle mask suitable
+    /// for a XXBRQ instruction.
+    bool isXXBRQShuffleMask(ShuffleVectorSDNode *N);
+
     /// isXXPERMDIShuffleMask - Return true if this is a shuffle mask suitable
     /// for a XXPERMDI instruction.
     bool isXXPERMDIShuffleMask(ShuffleVectorSDNode *N, unsigned &ShiftElts,

llvm/lib/Target/PowerPC/PPCInstrInfo.td

@@ -53,6 +53,10 @@ def SDT_PPCVecInsert : SDTypeProfile<1, 3, [ SDTCisVec<0>,
   SDTCisVec<1>, SDTCisVec<2>, SDTCisInt<3>
 ]>;
 
+def SDT_PPCVecReverse: SDTypeProfile<1, 1, [ SDTCisVec<0>,
+  SDTCisVec<1>
+]>;
+
 def SDT_PPCxxpermdi: SDTypeProfile<1, 3, [ SDTCisVec<0>,
   SDTCisVec<1>, SDTCisVec<2>, SDTCisInt<3>
 ]>;
@@ -174,6 +178,7 @@ def PPCaddiDtprelL   : SDNode<"PPCISD::ADDI_DTPREL_L", SDTIntBinOp>;
 def PPCvperm     : SDNode<"PPCISD::VPERM", SDT_PPCvperm, []>;
 def PPCxxsplt    : SDNode<"PPCISD::XXSPLT", SDT_PPCVecSplat, []>;
 def PPCxxinsert  : SDNode<"PPCISD::XXINSERT", SDT_PPCVecInsert, []>;
+def PPCxxreverse : SDNode<"PPCISD::XXREVERSE", SDT_PPCVecReverse, []>;
 def PPCxxpermdi  : SDNode<"PPCISD::XXPERMDI", SDT_PPCxxpermdi, []>;
 def PPCvecshl    : SDNode<"PPCISD::VECSHL", SDT_PPCVecShift, []>;
 

llvm/lib/Target/PowerPC/PPCInstrVSX.td

@@ -2340,6 +2340,16 @@ let AddedComplexity = 400, Predicates = [HasP9Vector] in {
   def XXBRD : XX2_XT6_XO5_XB6<60, 23, 475, "xxbrd", vsrc, []>;
   def XXBRQ : XX2_XT6_XO5_XB6<60, 31, 475, "xxbrq", vsrc, []>;
 
+  // Vector Reverse
+  def : Pat<(v8i16 (PPCxxreverse v8i16 :$A)),
+            (v8i16 (COPY_TO_REGCLASS (XXBRH (COPY_TO_REGCLASS $A, VSRC)), VRRC))>;
+  def : Pat<(v4i32 (PPCxxreverse v4i32 :$A)),
+            (v4i32 (XXBRW $A))>;
+  def : Pat<(v2i64 (PPCxxreverse v2i64 :$A)),
+            (v2i64 (XXBRD $A))>;
+  def : Pat<(v1i128 (PPCxxreverse v1i128 :$A)),
+            (v1i128 (COPY_TO_REGCLASS (XXBRQ (COPY_TO_REGCLASS $A, VSRC)), VRRC))>;
+
   // Vector Permute
   def XXPERM  : XX3_XT5_XA5_XB5<60, 26, "xxperm" , vsrc, vsrc, vsrc,
                                 IIC_VecPerm, []>;

llvm/test/CodeGen/PowerPC/vec_revb.ll

@@ -0,0 +1,54 @@
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64-unknown-linux-gnu -mcpu=pwr9 < %s | FileCheck %s
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr9 < %s | FileCheck %s
+
+define <8 x i16> @testXXBRH(<8 x i16> %a) {
+; CHECK-LABEL: testXXBRH:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    xxbrh 34, 34
+; CHECK-NEXT:    blr
+
+entry:
+  %0 = bitcast <8 x i16> %a to <16 x i8>
+  %1 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %2 = bitcast <16 x i8> %1 to <8 x i16>
+  ret <8 x i16> %2
+}
+
+define <4 x i32> @testXXBRW(<4 x i32> %a) {
+; CHECK-LABEL: testXXBRW:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    xxbrw 34, 34
+; CHECK-NEXT:    blr
+
+entry:
+  %0 = bitcast <4 x i32> %a to <16 x i8>
+  %1 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
+  %2 = bitcast <16 x i8> %1 to <4 x i32>
+  ret <4 x i32> %2
+}
+
+define <2 x double> @testXXBRD(<2 x double> %a) {
+; CHECK-LABEL: testXXBRD:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    xxbrd 34, 34
+; CHECK-NEXT:    blr
+
+entry:
+  %0 = bitcast <2 x double> %a to <16 x i8>
+  %1 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
+  %2 = bitcast <16 x i8> %1 to <2 x double>
+  ret <2 x double> %2
+}
+
+define <1 x i128> @testXXBRQ(<1 x i128> %a) {
+; CHECK-LABEL: testXXBRQ:
+; CHECK:       # BB#0: # %entry
+; CHECK-NEXT:    xxbrq 34, 34
+; CHECK-NEXT:    blr
+
+entry:
+  %0 = bitcast <1 x i128> %a to <16 x i8>
+  %1 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %2 = bitcast <16 x i8> %1 to <1 x i128>
+  ret <1 x i128> %2
+}