[X86][XOP] Add VPROT instruction opcodes

Added X86ISD opcodes for VPROT vector rotate by variable and by immediate.

llvm-svn: 250620

5 files changed, 32 insertions, 33 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 1794c1097cc..446098a0711 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -19992,6 +19992,8 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::RDSEED:             return "X86ISD::RDSEED";
   case X86ISD::VPMADDUBSW:         return "X86ISD::VPMADDUBSW";
   case X86ISD::VPMADDWD:           return "X86ISD::VPMADDWD";
+  case X86ISD::VPROT:              return "X86ISD::VPROT";
+  case X86ISD::VPROTI:             return "X86ISD::VPROTI";
   case X86ISD::VPSHA:              return "X86ISD::VPSHA";
   case X86ISD::VPSHL:              return "X86ISD::VPSHL";
   case X86ISD::VPCOM:              return "X86ISD::VPCOM";
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 0028eb0c82c..5ca5b5bcf78 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -412,6 +412,8 @@ namespace llvm {
       /// SSE4A Extraction and Insertion.
       EXTRQI, INSERTQI,
 
+      // XOP variable/immediate rotations
+      VPROT, VPROTI,
       // XOP arithmetic/logical shifts
       VPSHA, VPSHL,
       // XOP signed/unsigned integer comparisons
diff --git a/llvm/lib/Target/X86/X86InstrFragmentsSIMD.td b/llvm/lib/Target/X86/X86InstrFragmentsSIMD.td
index f6ef1483c50..ca67a2cff72 100644
--- a/llvm/lib/Target/X86/X86InstrFragmentsSIMD.td
+++ b/llvm/lib/Target/X86/X86InstrFragmentsSIMD.td
@@ -215,6 +215,13 @@ def X86vshli   : SDNode<"X86ISD::VSHLI", SDTIntShiftOp>;
 def X86vsrli   : SDNode<"X86ISD::VSRLI", SDTIntShiftOp>;
 def X86vsrai   : SDNode<"X86ISD::VSRAI", SDTIntShiftOp>;
 
+def X86vprot   : SDNode<"X86ISD::VPROT",
+                        SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisSameAs<0,1>,
+                                      SDTCisVec<2>]>>;
+def X86vproti  : SDNode<"X86ISD::VPROTI",
+                        SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisSameAs<0,1>,
+                                      SDTCisVT<2, i8>]>>;
+
 def X86vpshl   : SDNode<"X86ISD::VPSHL",
                         SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisSameAs<0,1>,
                                       SDTCisVec<2>]>>;
diff --git a/llvm/lib/Target/X86/X86InstrXOP.td b/llvm/lib/Target/X86/X86InstrXOP.td
index 77b07461f65..df9d906f4d9 100644
--- a/llvm/lib/Target/X86/X86InstrXOP.td
+++ b/llvm/lib/Target/X86/X86InstrXOP.td
@@ -108,6 +108,10 @@ multiclass xop3op<bits<8> opc, string OpcodeStr, SDNode OpNode,
 }
 
 let ExeDomain = SSEPackedInt in {
+  defm VPROTB : xop3op<0x90, "vprotb", X86vprot, v16i8>;
+  defm VPROTD : xop3op<0x92, "vprotd", X86vprot, v4i32>;
+  defm VPROTQ : xop3op<0x93, "vprotq", X86vprot, v2i64>;
+  defm VPROTW : xop3op<0x91, "vprotw", X86vprot, v8i16>;
   defm VPSHAB : xop3op<0x98, "vpshab", X86vpsha, v16i8>;
   defm VPSHAD : xop3op<0x9A, "vpshad", X86vpsha, v4i32>;
   defm VPSHAQ : xop3op<0x9B, "vpshaq", X86vpsha, v2i64>;
@@ -118,49 +122,25 @@ let ExeDomain = SSEPackedInt in {
   defm VPSHLW : xop3op<0x95, "vpshlw", X86vpshl, v8i16>;
 }
 
-multiclass xop3op_int<bits<8> opc, string OpcodeStr, Intrinsic Int> {
-  def rr : IXOP<opc, MRMSrcReg, (outs VR128:$dst),
-           (ins VR128:$src1, VR128:$src2),
-           !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-           [(set VR128:$dst, (Int VR128:$src1, VR128:$src2))]>, XOP_4VOp3;
-  def rm : IXOP<opc, MRMSrcMem, (outs VR128:$dst),
-           (ins VR128:$src1, i128mem:$src2),
-           !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-           [(set VR128:$dst,
-              (Int VR128:$src1, (bitconvert (loadv2i64 addr:$src2))))]>,
-           XOP_4V, VEX_W;
-  def mr : IXOP<opc, MRMSrcMem, (outs VR128:$dst),
-           (ins i128mem:$src1, VR128:$src2),
-           !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-           [(set VR128:$dst,
-              (Int (bitconvert (loadv2i64 addr:$src1)), VR128:$src2))]>,
-             XOP_4VOp3;
-}
-
-let ExeDomain = SSEPackedInt in {
-  defm VPROTW : xop3op_int<0x91, "vprotw", int_x86_xop_vprotw>;
-  defm VPROTQ : xop3op_int<0x93, "vprotq", int_x86_xop_vprotq>;
-  defm VPROTD : xop3op_int<0x92, "vprotd", int_x86_xop_vprotd>;
-  defm VPROTB : xop3op_int<0x90, "vprotb", int_x86_xop_vprotb>;
-}
-
-multiclass xop3opimm<bits<8> opc, string OpcodeStr, Intrinsic Int> {
+multiclass xop3opimm<bits<8> opc, string OpcodeStr, SDNode OpNode,
+                     ValueType vt128> {
   def ri : IXOPi8<opc, MRMSrcReg, (outs VR128:$dst),
            (ins VR128:$src1, u8imm:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-           [(set VR128:$dst, (Int VR128:$src1, imm:$src2))]>, XOP;
+           [(set VR128:$dst,
+              (vt128 (OpNode (vt128 VR128:$src1), imm:$src2)))]>, XOP;
   def mi : IXOPi8<opc, MRMSrcMem, (outs VR128:$dst),
            (ins i128mem:$src1, u8imm:$src2),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
            [(set VR128:$dst,
-             (Int (bitconvert (loadv2i64 addr:$src1)), imm:$src2))]>, XOP;
+              (vt128 (OpNode (vt128 (bitconvert (loadv2i64 addr:$src1))), imm:$src2)))]>, XOP;
 }
 
 let ExeDomain = SSEPackedInt in {
-  defm VPROTW : xop3opimm<0xC1, "vprotw", int_x86_xop_vprotwi>;
-  defm VPROTQ : xop3opimm<0xC3, "vprotq", int_x86_xop_vprotqi>;
-  defm VPROTD : xop3opimm<0xC2, "vprotd", int_x86_xop_vprotdi>;
-  defm VPROTB : xop3opimm<0xC0, "vprotb", int_x86_xop_vprotbi>;
+  defm VPROTB : xop3opimm<0xC0, "vprotb", X86vproti, v16i8>;
+  defm VPROTD : xop3opimm<0xC2, "vprotd", X86vproti, v4i32>;
+  defm VPROTQ : xop3opimm<0xC3, "vprotq", X86vproti, v2i64>;
+  defm VPROTW : xop3opimm<0xC1, "vprotw", X86vproti, v8i16>;
 }
 
 // Instruction where second source can be memory, but third must be register
diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
index 01003eb4cc9..74d6e4e6d73 100644
--- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
@@ -1718,6 +1718,14 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(xop_vpcomuq,       INTR_TYPE_3OP, X86ISD::VPCOMU, 0),
   X86_INTRINSIC_DATA(xop_vpcomuw,       INTR_TYPE_3OP, X86ISD::VPCOMU, 0),
   X86_INTRINSIC_DATA(xop_vpcomw,        INTR_TYPE_3OP, X86ISD::VPCOM, 0),
+  X86_INTRINSIC_DATA(xop_vprotb,        INTR_TYPE_2OP, X86ISD::VPROT, 0),
+  X86_INTRINSIC_DATA(xop_vprotbi,       INTR_TYPE_2OP, X86ISD::VPROTI, 0),
+  X86_INTRINSIC_DATA(xop_vprotd,        INTR_TYPE_2OP, X86ISD::VPROT, 0),
+  X86_INTRINSIC_DATA(xop_vprotdi,       INTR_TYPE_2OP, X86ISD::VPROTI, 0),
+  X86_INTRINSIC_DATA(xop_vprotq,        INTR_TYPE_2OP, X86ISD::VPROT, 0),
+  X86_INTRINSIC_DATA(xop_vprotqi,       INTR_TYPE_2OP, X86ISD::VPROTI, 0),
+  X86_INTRINSIC_DATA(xop_vprotw,        INTR_TYPE_2OP, X86ISD::VPROT, 0),
+  X86_INTRINSIC_DATA(xop_vprotwi,       INTR_TYPE_2OP, X86ISD::VPROTI, 0),
   X86_INTRINSIC_DATA(xop_vpshab,        INTR_TYPE_2OP, X86ISD::VPSHA, 0),
   X86_INTRINSIC_DATA(xop_vpshad,        INTR_TYPE_2OP, X86ISD::VPSHA, 0),
   X86_INTRINSIC_DATA(xop_vpshaq,        INTR_TYPE_2OP, X86ISD::VPSHA, 0),