lib/Target/ARM/ARMScheduleA8.td

   1 //=- ARMScheduleA8.td - ARM Cortex-A8 Scheduling Definitions -*- tablegen -*-=//
   2 //
   3 //                     The LLVM Compiler Infrastructure
   4 //
   5 // This file is distributed under the University of Illinois Open Source
   6 // License. See LICENSE.TXT for details.
   7 //
   8 //===----------------------------------------------------------------------===//
   9 //
  10 // This file defines the itinerary class data for the ARM Cortex A8 processors.
  11 //
  12 //===----------------------------------------------------------------------===//
  13
  14 //
  15 // Scheduling information derived from "Cortex-A8 Technical Reference Manual".
  16 // Functional Units.
  17 def A8_Pipe0   : FuncUnit; // pipeline 0
  18 def A8_Pipe1   : FuncUnit; // pipeline 1
  19 def A8_LSPipe  : FuncUnit; // Load / store pipeline
  20 def A8_NPipe   : FuncUnit; // NEON ALU/MUL pipe
  21 def A8_NLSPipe : FuncUnit; // NEON LS pipe
  22 //
  23 // Dual issue pipeline represented by A8_Pipe0 | A8_Pipe1
  24 //
  25 def CortexA8Itineraries : ProcessorItineraries<
  26   [A8_Pipe0, A8_Pipe1, A8_LSPipe, A8_NPipe, A8_NLSPipe],
  27   [], [
  28   // Two fully-pipelined integer ALU pipelines
  29   //
  30   // No operand cycles
  31   InstrItinData<IIC_iALUx    , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>]>,
  32   //
  33   // Binary Instructions that produce a result
  34   InstrItinData<IIC_iALUi ,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
  35   InstrItinData<IIC_iALUr ,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 2]>,
  36   InstrItinData<IIC_iALUsi,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 1]>,
  37   InstrItinData<IIC_iALUsir,[InstrStage<1,[A8_Pipe0, A8_Pipe1]>], [2, 1, 2]>,
  38   InstrItinData<IIC_iALUsr,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 1, 1]>,
  39   //
  40   // Bitwise Instructions that produce a result
  41   InstrItinData<IIC_iBITi ,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
  42   InstrItinData<IIC_iBITr ,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 2]>,
  43   InstrItinData<IIC_iBITsi,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 1]>,
  44   InstrItinData<IIC_iBITsr,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 1, 1]>,
  45   //
  46   // Unary Instructions that produce a result
  47   InstrItinData<IIC_iUNAr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
  48   InstrItinData<IIC_iUNAsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1]>,
  49   //
  50   // Zero and sign extension instructions
  51   InstrItinData<IIC_iEXTr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1]>,
  52   InstrItinData<IIC_iEXTAr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2, 1]>,
  53   InstrItinData<IIC_iEXTAsr,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>],[2, 2, 1, 1]>,
  54   //
  55   // Compare instructions
  56   InstrItinData<IIC_iCMPi , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2]>,
  57   InstrItinData<IIC_iCMPr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
  58   InstrItinData<IIC_iCMPsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1]>,
  59   InstrItinData<IIC_iCMPsr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1, 1]>,
  60   //
  61   // Test instructions
  62   InstrItinData<IIC_iTSTi , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2]>,
  63   InstrItinData<IIC_iTSTr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
  64   InstrItinData<IIC_iTSTsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1]>,
  65   InstrItinData<IIC_iTSTsr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1, 1]>,
  66   //
  67   // Move instructions, unconditional
  68   InstrItinData<IIC_iMOVi , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1]>,
  69   InstrItinData<IIC_iMOVr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1]>,
  70   InstrItinData<IIC_iMOVsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1]>,
  71   InstrItinData<IIC_iMOVsr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1, 1]>,
  72   InstrItinData<IIC_iMOVix2,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  73                              InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2]>,
  74   InstrItinData<IIC_iMOVix2addpc,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  75                                   InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  76                                   InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [3]>,
  77   InstrItinData<IIC_iMOVix2ld,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  78                                InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  79                                InstrStage<1, [A8_LSPipe]>], [5]>,
  80   //
  81   // Move instructions, conditional
  82   InstrItinData<IIC_iCMOVi , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2]>,
  83   InstrItinData<IIC_iCMOVr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1]>,
  84   InstrItinData<IIC_iCMOVsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1]>,
  85   InstrItinData<IIC_iCMOVsr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 1, 1]>,
  86   InstrItinData<IIC_iCMOVix2,[InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
  87                               InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [3, 1]>,
  88   //
  89   // MVN instructions
  90   InstrItinData<IIC_iMVNi , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1]>,
  91   InstrItinData<IIC_iMVNr , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1]>,
  92   InstrItinData<IIC_iMVNsi, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1]>,
  93   InstrItinData<IIC_iMVNsr, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [1, 1, 1]>,
  94
  95   // Integer multiply pipeline
  96   // Result written in E5, but that is relative to the last cycle of multicycle,
  97   // so we use 6 for those cases
  98   //
  99   InstrItinData<IIC_iMUL16   , [InstrStage<1, [A8_Pipe0]>], [5, 1, 1]>,
 100   InstrItinData<IIC_iMAC16   , [InstrStage<2, [A8_Pipe0]>], [6, 1, 1, 4]>,
 101   InstrItinData<IIC_iMUL32   , [InstrStage<2, [A8_Pipe0]>], [6, 1, 1]>,
 102   InstrItinData<IIC_iMAC32   , [InstrStage<2, [A8_Pipe0]>], [6, 1, 1, 4]>,
 103   InstrItinData<IIC_iMUL64   , [InstrStage<3, [A8_Pipe0]>], [6, 6, 1, 1]>,
 104   InstrItinData<IIC_iMAC64   , [InstrStage<3, [A8_Pipe0]>], [6, 6, 1, 1]>,
 105
 106   // Integer load pipeline
 107   //
 108   // Immediate offset
 109   InstrItinData<IIC_iLoad_i   , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 110                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 111   InstrItinData<IIC_iLoad_bh_i, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 112                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 113   InstrItinData<IIC_iLoad_d_i,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 114                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 115   //
 116   // Register offset
 117   InstrItinData<IIC_iLoad_r   , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 118                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 119   InstrItinData<IIC_iLoad_bh_r, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 120                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 121   InstrItinData<IIC_iLoad_d_r , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 122                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 123   //
 124   // Scaled register offset, issues over 2 cycles
 125   // FIXME: lsl by 2 takes 1 cycle.
 126   InstrItinData<IIC_iLoad_si  , [InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 127                                  InstrStage<1, [A8_LSPipe]>], [4, 1, 1]>,
 128   InstrItinData<IIC_iLoad_bh_si,[InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 129                                  InstrStage<1, [A8_LSPipe]>], [4, 1, 1]>,
 130   //
 131   // Immediate offset with update
 132   InstrItinData<IIC_iLoad_iu  , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 133                                  InstrStage<1, [A8_LSPipe]>], [3, 2, 1]>,
 134   InstrItinData<IIC_iLoad_bh_iu,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 135                                  InstrStage<1, [A8_LSPipe]>], [3, 2, 1]>,
 136   //
 137   // Register offset with update
 138   InstrItinData<IIC_iLoad_ru  , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 139                                  InstrStage<1, [A8_LSPipe]>], [3, 2, 1, 1]>,
 140   InstrItinData<IIC_iLoad_bh_ru,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 141                                  InstrStage<1, [A8_LSPipe]>], [3, 2, 1, 1]>,
 142   InstrItinData<IIC_iLoad_d_ru, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 143                                  InstrStage<1, [A8_LSPipe]>], [3, 2, 1, 1]>,
 144   //
 145   // Scaled register offset with update, issues over 2 cycles
 146   InstrItinData<IIC_iLoad_siu , [InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 147                                  InstrStage<2, [A8_LSPipe]>], [4, 3, 1, 1]>,
 148   InstrItinData<IIC_iLoad_bh_siu,[InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 149                                   InstrStage<2, [A8_LSPipe]>], [4, 3, 1, 1]>,
 150   //
 151   // Load multiple, def is the 5th operand. Pipeline 0 only.
 152   // FIXME: A8_LSPipe cycle time is dynamic, this assumes 3 to 4 registers.
 153   InstrItinData<IIC_iLoad_m  , [InstrStage<2, [A8_Pipe0], 0>,
 154                                 InstrStage<2, [A8_LSPipe]>], [1, 1, 1, 1, 3]>,
 155   //
 156   // Load multiple + update, defs are the 1st and 5th operands.
 157   InstrItinData<IIC_iLoad_mu , [InstrStage<3, [A8_Pipe0], 0>,
 158                                 InstrStage<3, [A8_LSPipe]>], [2, 1, 1, 1, 3]>,
 159   //
 160   // Load multiple plus branch
 161   InstrItinData<IIC_iLoad_mBr, [InstrStage<3, [A8_Pipe0], 0>,
 162                                 InstrStage<3, [A8_LSPipe]>,
 163                                 InstrStage<1, [A8_Pipe0, A8_Pipe1]>],
 164                                [1, 2, 1, 1, 3]>,
 165   //
 166   // Pop, def is the 3rd operand.
 167   InstrItinData<IIC_iPop  ,    [InstrStage<3, [A8_Pipe0], 0>,
 168                                 InstrStage<3, [A8_LSPipe]>], [1, 1, 3]>,
 169   //
 170   // Push, def is the 3th operand.
 171   InstrItinData<IIC_iPop_Br,   [InstrStage<3, [A8_Pipe0], 0>,
 172                                 InstrStage<3, [A8_LSPipe]>,
 173                                 InstrStage<1, [A8_Pipe0, A8_Pipe1]>],
 174                                [1, 1, 3]>,
 175
 176   //
 177   // iLoadi + iALUr for t2LDRpci_pic.
 178   InstrItinData<IIC_iLoadiALU, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 179                                 InstrStage<1, [A8_LSPipe]>,
 180                                 InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [4, 1]>,
 181
 182
 183   // Integer store pipeline
 184   //
 185   // Immediate offset
 186   InstrItinData<IIC_iStore_i  , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 187                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 188   InstrItinData<IIC_iStore_bh_i,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 189                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 190   InstrItinData<IIC_iStore_d_i, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 191                                  InstrStage<1, [A8_LSPipe]>], [3, 1]>,
 192   //
 193   // Register offset
 194   InstrItinData<IIC_iStore_r  , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 195                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 196   InstrItinData<IIC_iStore_bh_r,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 197                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 198   InstrItinData<IIC_iStore_d_r, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 199                                  InstrStage<1, [A8_LSPipe]>], [3, 1, 1]>,
 200   //
 201   // Scaled register offset, issues over 2 cycles
 202   InstrItinData<IIC_iStore_si , [InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 203                                  InstrStage<2, [A8_LSPipe]>], [3, 1, 1]>,
 204   InstrItinData<IIC_iStore_bh_si,[InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 205                                   InstrStage<2, [A8_LSPipe]>], [3, 1, 1]>,
 206   //
 207   // Immediate offset with update
 208   InstrItinData<IIC_iStore_iu , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 209                                  InstrStage<1, [A8_LSPipe]>], [2, 3, 1]>,
 210   InstrItinData<IIC_iStore_bh_iu,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 211                                  InstrStage<1, [A8_LSPipe]>], [2, 3, 1]>,
 212   //
 213   // Register offset with update
 214   InstrItinData<IIC_iStore_ru  , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 215                                   InstrStage<1, [A8_LSPipe]>], [2, 3, 1, 1]>,
 216   InstrItinData<IIC_iStore_bh_ru,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 217                                   InstrStage<1, [A8_LSPipe]>], [2, 3, 1, 1]>,
 218   InstrItinData<IIC_iStore_d_ru, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 219                                   InstrStage<1, [A8_LSPipe]>], [2, 3, 1, 1]>,
 220   //
 221   // Scaled register offset with update, issues over 2 cycles
 222   InstrItinData<IIC_iStore_siu, [InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 223                                  InstrStage<2, [A8_LSPipe]>], [3, 3, 1, 1]>,
 224   InstrItinData<IIC_iStore_bh_siu,[InstrStage<2, [A8_Pipe0, A8_Pipe1], 0>,
 225                                    InstrStage<2, [A8_LSPipe]>], [3, 3, 1, 1]>,
 226   //
 227   // Store multiple. Pipeline 0 only.
 228   // FIXME: A8_LSPipe cycle time is dynamic, this assumes 3 to 4 registers.
 229   InstrItinData<IIC_iStore_m , [InstrStage<2, [A8_Pipe0], 0>,
 230                                 InstrStage<2, [A8_LSPipe]>]>,
 231   //
 232   // Store multiple + update
 233   InstrItinData<IIC_iStore_mu, [InstrStage<2, [A8_Pipe0], 0>,
 234                                 InstrStage<2, [A8_LSPipe]>], [2]>,
 235
 236   //
 237   // Preload
 238   InstrItinData<IIC_Preload, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>], [2, 2]>,
 239
 240   // Branch
 241   //
 242   // no delay slots, so the latency of a branch is unimportant
 243   InstrItinData<IIC_Br      , [InstrStage<1, [A8_Pipe0, A8_Pipe1]>]>,
 244
 245   // VFP
 246   // Issue through integer pipeline, and execute in NEON unit. We assume
 247   // RunFast mode so that NFP pipeline is used for single-precision when
 248   // possible.
 249   //
 250   // FP Special Register to Integer Register File Move
 251   InstrItinData<IIC_fpSTAT , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 252                               InstrStage<1, [A8_NLSPipe]>], [20]>,
 253   //
 254   // Single-precision FP Unary
 255   InstrItinData<IIC_fpUNA32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 256                                InstrStage<1, [A8_NPipe]>], [7, 1]>,
 257   //
 258   // Double-precision FP Unary
 259   InstrItinData<IIC_fpUNA64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 260                                InstrStage<4, [A8_NPipe], 0>,
 261                                InstrStage<4, [A8_NLSPipe]>], [4, 1]>,
 262   //
 263   // Single-precision FP Compare
 264   InstrItinData<IIC_fpCMP32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 265                                InstrStage<1, [A8_NPipe]>], [1, 1]>,
 266   //
 267   // Double-precision FP Compare
 268   InstrItinData<IIC_fpCMP64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 269                                InstrStage<4, [A8_NPipe], 0>,
 270                                InstrStage<4, [A8_NLSPipe]>], [4, 1]>,
 271   //
 272   // Single to Double FP Convert
 273   InstrItinData<IIC_fpCVTSD , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 274                                InstrStage<7, [A8_NPipe], 0>,
 275                                InstrStage<7, [A8_NLSPipe]>], [7, 1]>,
 276   //
 277   // Double to Single FP Convert
 278   InstrItinData<IIC_fpCVTDS , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 279                                InstrStage<5, [A8_NPipe], 0>,
 280                                InstrStage<5, [A8_NLSPipe]>], [5, 1]>,
 281   //
 282   // Single-Precision FP to Integer Convert
 283   InstrItinData<IIC_fpCVTSI , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 284                                InstrStage<1, [A8_NPipe]>], [7, 1]>,
 285   //
 286   // Double-Precision FP to Integer Convert
 287   InstrItinData<IIC_fpCVTDI , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 288                                InstrStage<8, [A8_NPipe], 0>,
 289                                InstrStage<8, [A8_NLSPipe]>], [8, 1]>,
 290   //
 291   // Integer to Single-Precision FP Convert
 292   InstrItinData<IIC_fpCVTIS , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 293                                InstrStage<1, [A8_NPipe]>], [7, 1]>,
 294   //
 295   // Integer to Double-Precision FP Convert
 296   InstrItinData<IIC_fpCVTID , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 297                                InstrStage<8, [A8_NPipe], 0>,
 298                                InstrStage<8, [A8_NLSPipe]>], [8, 1]>,
 299   //
 300   // Single-precision FP ALU
 301   InstrItinData<IIC_fpALU32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 302                                InstrStage<1, [A8_NPipe]>], [7, 1, 1]>,
 303   //
 304   // Double-precision FP ALU
 305   InstrItinData<IIC_fpALU64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 306                                InstrStage<9, [A8_NPipe], 0>,
 307                                InstrStage<9, [A8_NLSPipe]>], [9, 1, 1]>,
 308   //
 309   // Single-precision FP Multiply
 310   InstrItinData<IIC_fpMUL32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 311                                InstrStage<1, [A8_NPipe]>], [7, 1, 1]>,
 312   //
 313   // Double-precision FP Multiply
 314   InstrItinData<IIC_fpMUL64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 315                                InstrStage<11, [A8_NPipe], 0>,
 316                                InstrStage<11, [A8_NLSPipe]>], [11, 1, 1]>,
 317   //
 318   // Single-precision FP MAC
 319   InstrItinData<IIC_fpMAC32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 320                                InstrStage<1, [A8_NPipe]>], [7, 2, 1, 1]>,
 321   //
 322   // Double-precision FP MAC
 323   InstrItinData<IIC_fpMAC64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 324                                InstrStage<19, [A8_NPipe], 0>,
 325                                InstrStage<19, [A8_NLSPipe]>], [19, 2, 1, 1]>,
 326   //
 327   // Single-precision FP DIV
 328   InstrItinData<IIC_fpDIV32 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 329                                InstrStage<20, [A8_NPipe], 0>,
 330                                InstrStage<20, [A8_NLSPipe]>], [20, 1, 1]>,
 331   //
 332   // Double-precision FP DIV
 333   InstrItinData<IIC_fpDIV64 , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 334                                InstrStage<29, [A8_NPipe], 0>,
 335                                InstrStage<29, [A8_NLSPipe]>], [29, 1, 1]>,
 336   //
 337   // Single-precision FP SQRT
 338   InstrItinData<IIC_fpSQRT32, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 339                                InstrStage<19, [A8_NPipe], 0>,
 340                                InstrStage<19, [A8_NLSPipe]>], [19, 1]>,
 341   //
 342   // Double-precision FP SQRT
 343   InstrItinData<IIC_fpSQRT64, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 344                                InstrStage<29, [A8_NPipe], 0>,
 345                                InstrStage<29, [A8_NLSPipe]>], [29, 1]>,
 346
 347   //
 348   // Integer to Single-precision Move
 349   InstrItinData<IIC_fpMOVIS,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 350                                InstrStage<1, [A8_NPipe]>],
 351                               [2, 1]>,
 352   //
 353   // Integer to Double-precision Move
 354   InstrItinData<IIC_fpMOVID,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 355                                InstrStage<1, [A8_NPipe]>],
 356                               [2, 1, 1]>,
 357   //
 358   // Single-precision to Integer Move
 359   InstrItinData<IIC_fpMOVSI,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 360                                InstrStage<1, [A8_NPipe]>],
 361                               [20, 1]>,
 362   //
 363   // Double-precision to Integer Move
 364   InstrItinData<IIC_fpMOVDI,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 365                                InstrStage<1, [A8_NPipe]>],
 366                               [20, 20, 1]>,
 367
 368   //
 369   // Single-precision FP Load
 370   InstrItinData<IIC_fpLoad32, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 371                                InstrStage<1, [A8_NLSPipe], 0>,
 372                                InstrStage<1, [A8_LSPipe]>],
 373                               [2, 1]>,
 374   //
 375   // Double-precision FP Load
 376   InstrItinData<IIC_fpLoad64, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 377                                InstrStage<1, [A8_NLSPipe], 0>,
 378                                InstrStage<1, [A8_LSPipe]>],
 379                               [2, 1]>,
 380   //
 381   // FP Load Multiple
 382   // FIXME: A8_LSPipe cycle time is dynamic, this assumes 3 to 4 registers.
 383   InstrItinData<IIC_fpLoad_m, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 384                                InstrStage<1, [A8_NLSPipe], 0>,
 385                                InstrStage<1, [A8_LSPipe]>,
 386                                InstrStage<1, [A8_NLSPipe], 0>,
 387                                InstrStage<1, [A8_LSPipe]>], [1, 1, 1, 2]>,
 388   //
 389   // FP Load Multiple + update
 390   InstrItinData<IIC_fpLoad_mu,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 391                                InstrStage<1, [A8_NLSPipe], 0>,
 392                                InstrStage<1, [A8_LSPipe]>,
 393                                InstrStage<1, [A8_NLSPipe], 0>,
 394                                InstrStage<1, [A8_LSPipe]>], [2, 1, 1, 1, 2]>,
 395   //
 396   // Single-precision FP Store
 397   InstrItinData<IIC_fpStore32,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 398                                InstrStage<1, [A8_NLSPipe], 0>,
 399                                InstrStage<1, [A8_LSPipe]>],
 400                               [1, 1]>,
 401   //
 402   // Double-precision FP Store
 403   InstrItinData<IIC_fpStore64,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 404                                InstrStage<1, [A8_NLSPipe], 0>,
 405                                InstrStage<1, [A8_LSPipe]>],
 406                               [1, 1]>,
 407   //
 408   // FP Store Multiple
 409   InstrItinData<IIC_fpStore_m,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 410                                InstrStage<1, [A8_NLSPipe], 0>,
 411                                InstrStage<1, [A8_LSPipe]>,
 412                                InstrStage<1, [A8_NLSPipe], 0>,
 413                                InstrStage<1, [A8_LSPipe]>], [1, 1, 1, 1]>,
 414   //
 415   // FP Store Multiple + update
 416   InstrItinData<IIC_fpStore_mu,[InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 417                                 InstrStage<1, [A8_NLSPipe], 0>,
 418                                 InstrStage<1, [A8_LSPipe]>,
 419                                 InstrStage<1, [A8_NLSPipe], 0>,
 420                                 InstrStage<1, [A8_LSPipe]>], [2, 1, 1, 1, 1]>,
 421
 422   // NEON
 423   // Issue through integer pipeline, and execute in NEON unit.
 424   //
 425   // VLD1
 426   InstrItinData<IIC_VLD1,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 427                                InstrStage<2, [A8_NLSPipe], 0>,
 428                                InstrStage<2, [A8_LSPipe]>],
 429                               [2, 1]>,
 430   // VLD1x2
 431   InstrItinData<IIC_VLD1x2,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 432                                InstrStage<2, [A8_NLSPipe], 0>,
 433                                InstrStage<2, [A8_LSPipe]>],
 434                               [2, 2, 1]>,
 435   //
 436   // VLD1x3
 437   InstrItinData<IIC_VLD1x3,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 438                                InstrStage<3, [A8_NLSPipe], 0>,
 439                                InstrStage<3, [A8_LSPipe]>],
 440                               [2, 2, 3, 1]>,
 441   //
 442   // VLD1x4
 443   InstrItinData<IIC_VLD1x4,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 444                                InstrStage<3, [A8_NLSPipe], 0>,
 445                                InstrStage<3, [A8_LSPipe]>],
 446                               [2, 2, 3, 3, 1]>,
 447   //
 448   // VLD1u
 449   InstrItinData<IIC_VLD1u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 450                                InstrStage<2, [A8_NLSPipe], 0>,
 451                                InstrStage<2, [A8_LSPipe]>],
 452                               [2, 2, 1]>,
 453   //
 454   // VLD1x2u
 455   InstrItinData<IIC_VLD1x2u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 456                                InstrStage<2, [A8_NLSPipe], 0>,
 457                                InstrStage<2, [A8_LSPipe]>],
 458                               [2, 2, 2, 1]>,
 459   //
 460   // VLD1x3u
 461   InstrItinData<IIC_VLD1x3u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 462                                InstrStage<3, [A8_NLSPipe], 0>,
 463                                InstrStage<3, [A8_LSPipe]>],
 464                               [2, 2, 3, 2, 1]>,
 465   //
 466   // VLD1x4u
 467   InstrItinData<IIC_VLD1x4u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 468                                InstrStage<3, [A8_NLSPipe], 0>,
 469                                InstrStage<3, [A8_LSPipe]>],
 470                               [2, 2, 3, 3, 2, 1]>,
 471   //
 472   // VLD1ln
 473   InstrItinData<IIC_VLD1ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 474                                InstrStage<3, [A8_NLSPipe], 0>,
 475                                InstrStage<3, [A8_LSPipe]>],
 476                               [3, 1, 1, 1]>,
 477   //
 478   // VLD1lnu
 479   InstrItinData<IIC_VLD1lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 480                                InstrStage<3, [A8_NLSPipe], 0>,
 481                                InstrStage<3, [A8_LSPipe]>],
 482                               [3, 2, 1, 1, 1, 1]>,
 483   //
 484   // VLD1dup
 485   InstrItinData<IIC_VLD1dup,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 486                                InstrStage<2, [A8_NLSPipe], 0>,
 487                                InstrStage<2, [A8_LSPipe]>],
 488                               [2, 1]>,
 489   //
 490   // VLD1dupu
 491   InstrItinData<IIC_VLD1dupu, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 492                                InstrStage<2, [A8_NLSPipe], 0>,
 493                                InstrStage<2, [A8_LSPipe]>],
 494                               [2, 2, 1, 1]>,
 495   //
 496   // VLD2
 497   InstrItinData<IIC_VLD2,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 498                                InstrStage<2, [A8_NLSPipe], 0>,
 499                                InstrStage<2, [A8_LSPipe]>],
 500                               [2, 2, 1]>,
 501   //
 502   // VLD2x2
 503   InstrItinData<IIC_VLD2x2,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 504                                InstrStage<3, [A8_NLSPipe], 0>,
 505                                InstrStage<3, [A8_LSPipe]>],
 506                               [2, 2, 3, 3, 1]>,
 507   //
 508   // VLD2ln
 509   InstrItinData<IIC_VLD2ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 510                                InstrStage<3, [A8_NLSPipe], 0>,
 511                                InstrStage<3, [A8_LSPipe]>],
 512                               [3, 3, 1, 1, 1, 1]>,
 513   //
 514   // VLD2u
 515   InstrItinData<IIC_VLD2u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 516                                InstrStage<2, [A8_NLSPipe], 0>,
 517                                InstrStage<2, [A8_LSPipe]>],
 518                               [2, 2, 2, 1, 1, 1]>,
 519   //
 520   // VLD2x2u
 521   InstrItinData<IIC_VLD2x2u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 522                                InstrStage<3, [A8_NLSPipe], 0>,
 523                                InstrStage<3, [A8_LSPipe]>],
 524                               [2, 2, 3, 3, 2, 1]>,
 525   //
 526   // VLD2lnu
 527   InstrItinData<IIC_VLD2lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 528                                InstrStage<3, [A8_NLSPipe], 0>,
 529                                InstrStage<3, [A8_LSPipe]>],
 530                               [3, 3, 2, 1, 1, 1, 1, 1]>,
 531   //
 532   // VLD2dup
 533   InstrItinData<IIC_VLD2dup,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 534                                InstrStage<2, [A8_NLSPipe], 0>,
 535                                InstrStage<2, [A8_LSPipe]>],
 536                               [2, 2, 1]>,
 537   //
 538   // VLD2dupu
 539   InstrItinData<IIC_VLD2dupu, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 540                                InstrStage<2, [A8_NLSPipe], 0>,
 541                                InstrStage<2, [A8_LSPipe]>],
 542                               [2, 2, 2, 1, 1]>,
 543   //
 544   // VLD3
 545   InstrItinData<IIC_VLD3,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 546                                InstrStage<4, [A8_NLSPipe], 0>,
 547                                InstrStage<4, [A8_LSPipe]>],
 548                               [3, 3, 4, 1]>,
 549   //
 550   // VLD3ln
 551   InstrItinData<IIC_VLD3ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 552                                InstrStage<5, [A8_NLSPipe], 0>,
 553                                InstrStage<5, [A8_LSPipe]>],
 554                               [4, 4, 5, 1, 1, 1, 1, 2]>,
 555   //
 556   // VLD3u
 557   InstrItinData<IIC_VLD3u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 558                                InstrStage<4, [A8_NLSPipe], 0>,
 559                                InstrStage<4, [A8_LSPipe]>],
 560                               [3, 3, 4, 2, 1]>,
 561   //
 562   // VLD3lnu
 563   InstrItinData<IIC_VLD3lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 564                                InstrStage<5, [A8_NLSPipe], 0>,
 565                                InstrStage<5, [A8_LSPipe]>],
 566                               [4, 4, 5, 2, 1, 1, 1, 1, 1, 2]>,
 567   //
 568   // VLD3dup
 569   InstrItinData<IIC_VLD3dup,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 570                                InstrStage<3, [A8_NLSPipe], 0>,
 571                                InstrStage<3, [A8_LSPipe]>],
 572                               [2, 2, 3, 1]>,
 573   //
 574   // VLD3dupu
 575   InstrItinData<IIC_VLD3dupu, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 576                                InstrStage<3, [A8_NLSPipe], 0>,
 577                                InstrStage<3, [A8_LSPipe]>],
 578                               [2, 2, 3, 2, 1, 1]>,
 579   //
 580   // VLD4
 581   InstrItinData<IIC_VLD4,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 582                                InstrStage<4, [A8_NLSPipe], 0>,
 583                                InstrStage<4, [A8_LSPipe]>],
 584                               [3, 3, 4, 4, 1]>,
 585   //
 586   // VLD4ln
 587   InstrItinData<IIC_VLD4ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 588                                InstrStage<5, [A8_NLSPipe], 0>,
 589                                InstrStage<5, [A8_LSPipe]>],
 590                               [4, 4, 5, 5, 1, 1, 1, 1, 2, 2]>,
 591   //
 592   // VLD4u
 593   InstrItinData<IIC_VLD4u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 594                                InstrStage<4, [A8_NLSPipe], 0>,
 595                                InstrStage<4, [A8_LSPipe]>],
 596                               [3, 3, 4, 4, 2, 1]>,
 597   //
 598   // VLD4lnu
 599   InstrItinData<IIC_VLD4lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 600                                InstrStage<5, [A8_NLSPipe], 0>,
 601                                InstrStage<5, [A8_LSPipe]>],
 602                               [4, 4, 5, 5, 2, 1, 1, 1, 1, 1, 2, 2]>,
 603   //
 604   // VLD4dup
 605   InstrItinData<IIC_VLD4dup,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 606                                InstrStage<3, [A8_NLSPipe], 0>,
 607                                InstrStage<3, [A8_LSPipe]>],
 608                               [2, 2, 3, 3, 1]>,
 609   //
 610   // VLD4dupu
 611   InstrItinData<IIC_VLD4dupu, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 612                                InstrStage<3, [A8_NLSPipe], 0>,
 613                                InstrStage<3, [A8_LSPipe]>],
 614                               [2, 2, 3, 3, 2, 1, 1]>,
 615   //
 616   // VST1
 617   InstrItinData<IIC_VST1,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 618                                InstrStage<2, [A8_NLSPipe], 0>,
 619                                InstrStage<2, [A8_LSPipe]>],
 620                               [1, 1, 1]>,
 621   //
 622   // VST1x2
 623   InstrItinData<IIC_VST1x2,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 624                                InstrStage<2, [A8_NLSPipe], 0>,
 625                                InstrStage<2, [A8_LSPipe]>],
 626                               [1, 1, 1, 1]>,
 627   //
 628   // VST1x3
 629   InstrItinData<IIC_VST1x3,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 630                                InstrStage<3, [A8_NLSPipe], 0>,
 631                                InstrStage<3, [A8_LSPipe]>],
 632                               [1, 1, 1, 1, 2]>,
 633   //
 634   // VST1x4
 635   InstrItinData<IIC_VST1x4,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 636                                InstrStage<3, [A8_NLSPipe], 0>,
 637                                InstrStage<3, [A8_LSPipe]>],
 638                               [1, 1, 1, 1, 2, 2]>,
 639   //
 640   // VST1u
 641   InstrItinData<IIC_VST1u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 642                                InstrStage<2, [A8_NLSPipe], 0>,
 643                                InstrStage<2, [A8_LSPipe]>],
 644                               [2, 1, 1, 1, 1]>,
 645   //
 646   // VST1x2u
 647   InstrItinData<IIC_VST1x2u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 648                                InstrStage<2, [A8_NLSPipe], 0>,
 649                                InstrStage<2, [A8_LSPipe]>],
 650                               [2, 1, 1, 1, 1, 1]>,
 651   //
 652   // VST1x3u
 653   InstrItinData<IIC_VST1x3u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 654                                InstrStage<3, [A8_NLSPipe], 0>,
 655                                InstrStage<3, [A8_LSPipe]>],
 656                               [2, 1, 1, 1, 1, 1, 2]>,
 657   //
 658   // VST1x4u
 659   InstrItinData<IIC_VST1x4u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 660                                InstrStage<3, [A8_NLSPipe], 0>,
 661                                InstrStage<3, [A8_LSPipe]>],
 662                               [2, 1, 1, 1, 1, 1, 2, 2]>,
 663   //
 664   // VST1ln
 665   InstrItinData<IIC_VST1ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 666                                InstrStage<2, [A8_NLSPipe], 0>,
 667                                InstrStage<2, [A8_LSPipe]>],
 668                               [1, 1, 1]>,
 669   //
 670   // VST1lnu
 671   InstrItinData<IIC_VST1lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
 672                                InstrStage<2, [A8_NLSPipe], 0>,
 673                                InstrStage<2, [A8_LSPipe]>],
 674                               [2, 1, 1, 1, 1]>,
 675   //
 676   // VST2
 677   InstrItinData<IIC_VST2,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 678                                InstrStage<2, [A8_NLSPipe], 0>,
 679                                InstrStage<2, [A8_LSPipe]>],
 680                               [1, 1, 1, 1]>,
 681   //
 682   // VST2x2
 683   InstrItinData<IIC_VST2x2,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 684                                InstrStage<4, [A8_NLSPipe], 0>,
 685                                InstrStage<4, [A8_LSPipe]>],
 686                               [1, 1, 1, 1, 2, 2]>,
 687   //
 688   // VST2u
 689   InstrItinData<IIC_VST2u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 690                                InstrStage<2, [A8_NLSPipe], 0>,
 691                                InstrStage<2, [A8_LSPipe]>],
 692                               [2, 1, 1, 1, 1, 1]>,
 693   //
 694   // VST2x2u
 695   InstrItinData<IIC_VST2x2u,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 696                                InstrStage<4, [A8_NLSPipe], 0>,
 697                                InstrStage<4, [A8_LSPipe]>],
 698                               [2, 1, 1, 1, 1, 1, 2, 2]>,
 699   //
 700   // VST2ln
 701   InstrItinData<IIC_VST2ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 702                                InstrStage<2, [A8_NLSPipe], 0>,
 703                                InstrStage<2, [A8_LSPipe]>],
 704                               [1, 1, 1, 1]>,
 705   //
 706   // VST2lnu
 707   InstrItinData<IIC_VST2lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 708                                InstrStage<2, [A8_NLSPipe], 0>,
 709                                InstrStage<2, [A8_LSPipe]>],
 710                               [2, 1, 1, 1, 1, 1]>,
 711   //
 712   // VST3
 713   InstrItinData<IIC_VST3,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 714                                InstrStage<3, [A8_NLSPipe], 0>,
 715                                InstrStage<3, [A8_LSPipe]>],
 716                               [1, 1, 1, 1, 2]>,
 717   //
 718   // VST3u
 719   InstrItinData<IIC_VST3u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 720                                InstrStage<3, [A8_NLSPipe], 0>,
 721                                InstrStage<3, [A8_LSPipe]>],
 722                               [2, 1, 1, 1, 1, 1, 2]>,
 723   //
 724   // VST3ln
 725   InstrItinData<IIC_VST3ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 726                                InstrStage<3, [A8_NLSPipe], 0>,
 727                                InstrStage<3, [A8_LSPipe]>],
 728                               [1, 1, 1, 1, 2]>,
 729   //
 730   // VST3lnu
 731   InstrItinData<IIC_VST3lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 732                                InstrStage<3, [A8_NLSPipe], 0>,
 733                                InstrStage<3, [A8_LSPipe]>],
 734                               [2, 1, 1, 1, 1, 1, 2]>,
 735   //
 736   // VST4
 737   InstrItinData<IIC_VST4,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 738                                InstrStage<4, [A8_NLSPipe], 0>,
 739                                InstrStage<4, [A8_LSPipe]>],
 740                               [1, 1, 1, 1, 2, 2]>,
 741   //
 742   // VST4u
 743   InstrItinData<IIC_VST4u,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 744                                InstrStage<4, [A8_NLSPipe], 0>,
 745                                InstrStage<4, [A8_LSPipe]>],
 746                               [2, 1, 1, 1, 1, 1, 2, 2]>,
 747   //
 748   // VST4ln
 749   InstrItinData<IIC_VST4ln,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 750                                InstrStage<4, [A8_NLSPipe], 0>,
 751                                InstrStage<4, [A8_LSPipe]>],
 752                               [1, 1, 1, 1, 2, 2]>,
 753   //
 754   // VST4lnu
 755   InstrItinData<IIC_VST4lnu,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 756                                InstrStage<4, [A8_NLSPipe], 0>,
 757                                InstrStage<4, [A8_LSPipe]>],
 758                               [2, 1, 1, 1, 1, 1, 2, 2]>,
 759   //
 760   // Double-register FP Unary
 761   InstrItinData<IIC_VUNAD,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 762                                InstrStage<1, [A8_NPipe]>], [5, 2]>,
 763   //
 764   // Quad-register FP Unary
 765   // Result written in N5, but that is relative to the last cycle of multicycle,
 766   // so we use 6 for those cases
 767   InstrItinData<IIC_VUNAQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 768                                InstrStage<2, [A8_NPipe]>], [6, 2]>,
 769   //
 770   // Double-register FP Binary
 771   InstrItinData<IIC_VBIND,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 772                                InstrStage<1, [A8_NPipe]>], [5, 2, 2]>,
 773   //
 774   // VPADD, etc.
 775   InstrItinData<IIC_VPBIND,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 776                                InstrStage<1, [A8_NPipe]>], [5, 2, 2]>,
 777   //
 778   // Double-register FP VMUL
 779   InstrItinData<IIC_VFMULD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 780                                InstrStage<1, [A8_NPipe]>], [5, 2, 1]>,
 781
 782   //
 783   // Quad-register FP Binary
 784   // Result written in N5, but that is relative to the last cycle of multicycle,
 785   // so we use 6 for those cases
 786   InstrItinData<IIC_VBINQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 787                                InstrStage<2, [A8_NPipe]>], [6, 2, 2]>,
 788   //
 789   // Quad-register FP VMUL
 790   InstrItinData<IIC_VFMULQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 791                                InstrStage<1, [A8_NPipe]>], [6, 2, 1]>,
 792   //
 793   // Move
 794   InstrItinData<IIC_VMOV,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 795                                InstrStage<1, [A8_NPipe]>], [1, 1]>,
 796   //
 797   // Move Immediate
 798   InstrItinData<IIC_VMOVImm,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 799                                InstrStage<1, [A8_NPipe]>], [3]>,
 800   //
 801   // Double-register Permute Move
 802   InstrItinData<IIC_VMOVD,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 803                                InstrStage<1, [A8_NLSPipe]>], [2, 1]>,
 804   //
 805   // Quad-register Permute Move
 806   // Result written in N2, but that is relative to the last cycle of multicycle,
 807   // so we use 3 for those cases
 808   InstrItinData<IIC_VMOVQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 809                                InstrStage<2, [A8_NLSPipe]>], [3, 1]>,
 810   //
 811   // Integer to Single-precision Move
 812   InstrItinData<IIC_VMOVIS ,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 813                                InstrStage<1, [A8_NLSPipe]>], [2, 1]>,
 814   //
 815   // Integer to Double-precision Move
 816   InstrItinData<IIC_VMOVID ,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 817                                InstrStage<1, [A8_NLSPipe]>], [2, 1, 1]>,
 818   //
 819   // Single-precision to Integer Move
 820   InstrItinData<IIC_VMOVSI ,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 821                                InstrStage<1, [A8_NLSPipe]>], [20, 1]>,
 822   //
 823   // Double-precision to Integer Move
 824   InstrItinData<IIC_VMOVDI ,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 825                                InstrStage<1, [A8_NLSPipe]>], [20, 20, 1]>,
 826   //
 827   // Integer to Lane Move
 828   InstrItinData<IIC_VMOVISL , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 829                                InstrStage<2, [A8_NLSPipe]>], [3, 1, 1]>,
 830   //
 831   // Vector narrow move
 832   InstrItinData<IIC_VMOVN   , [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 833                                InstrStage<1, [A8_NPipe]>], [2, 1]>,
 834   //
 835   // Double-register Permute
 836   InstrItinData<IIC_VPERMD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 837                                InstrStage<1, [A8_NLSPipe]>], [2, 2, 1, 1]>,
 838   //
 839   // Quad-register Permute
 840   // Result written in N2, but that is relative to the last cycle of multicycle,
 841   // so we use 3 for those cases
 842   InstrItinData<IIC_VPERMQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 843                                InstrStage<2, [A8_NLSPipe]>], [3, 3, 1, 1]>,
 844   //
 845   // Quad-register Permute (3 cycle issue)
 846   // Result written in N2, but that is relative to the last cycle of multicycle,
 847   // so we use 4 for those cases
 848   InstrItinData<IIC_VPERMQ3,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 849                                InstrStage<1, [A8_NLSPipe]>,
 850                                InstrStage<1, [A8_NPipe], 0>,
 851                                InstrStage<2, [A8_NLSPipe]>], [4, 4, 1, 1]>,
 852   //
 853   // Double-register FP Multiple-Accumulate
 854   InstrItinData<IIC_VMACD,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 855                                InstrStage<1, [A8_NPipe]>], [9, 3, 2, 2]>,
 856   //
 857   // Quad-register FP Multiple-Accumulate
 858   // Result written in N9, but that is relative to the last cycle of multicycle,
 859   // so we use 10 for those cases
 860   InstrItinData<IIC_VMACQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 861                                InstrStage<2, [A8_NPipe]>], [10, 3, 2, 2]>,
 862   //
 863   // Double-register Reciprical Step
 864   InstrItinData<IIC_VRECSD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 865                                InstrStage<1, [A8_NPipe]>], [9, 2, 2]>,
 866   //
 867   // Quad-register Reciprical Step
 868   InstrItinData<IIC_VRECSQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 869                                InstrStage<2, [A8_NPipe]>], [10, 2, 2]>,
 870   //
 871   // Double-register Integer Count
 872   InstrItinData<IIC_VCNTiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 873                                InstrStage<1, [A8_NPipe]>], [3, 2, 2]>,
 874   //
 875   // Quad-register Integer Count
 876   // Result written in N3, but that is relative to the last cycle of multicycle,
 877   // so we use 4 for those cases
 878   InstrItinData<IIC_VCNTiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 879                                InstrStage<2, [A8_NPipe]>], [4, 2, 2]>,
 880   //
 881   // Double-register Integer Unary
 882   InstrItinData<IIC_VUNAiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 883                                InstrStage<1, [A8_NPipe]>], [4, 2]>,
 884   //
 885   // Quad-register Integer Unary
 886   InstrItinData<IIC_VUNAiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 887                                InstrStage<1, [A8_NPipe]>], [4, 2]>,
 888   //
 889   // Double-register Integer Q-Unary
 890   InstrItinData<IIC_VQUNAiD,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 891                                InstrStage<1, [A8_NPipe]>], [4, 1]>,
 892   //
 893   // Quad-register Integer CountQ-Unary
 894   InstrItinData<IIC_VQUNAiQ,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 895                                InstrStage<1, [A8_NPipe]>], [4, 1]>,
 896   //
 897   // Double-register Integer Binary
 898   InstrItinData<IIC_VBINiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 899                                InstrStage<1, [A8_NPipe]>], [3, 2, 2]>,
 900   //
 901   // Quad-register Integer Binary
 902   InstrItinData<IIC_VBINiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 903                                InstrStage<1, [A8_NPipe]>], [3, 2, 2]>,
 904   //
 905   // Double-register Integer Binary (4 cycle)
 906   InstrItinData<IIC_VBINi4D,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 907                                InstrStage<1, [A8_NPipe]>], [4, 2, 1]>,
 908   //
 909   // Quad-register Integer Binary (4 cycle)
 910   InstrItinData<IIC_VBINi4Q,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 911                                InstrStage<1, [A8_NPipe]>], [4, 2, 1]>,
 912
 913   //
 914   // Double-register Integer Subtract
 915   InstrItinData<IIC_VSUBiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 916                                InstrStage<1, [A8_NPipe]>], [3, 2, 1]>,
 917   //
 918   // Quad-register Integer Subtract
 919   InstrItinData<IIC_VSUBiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 920                                InstrStage<1, [A8_NPipe]>], [3, 2, 1]>,
 921   //
 922   // Double-register Integer Subtract
 923   InstrItinData<IIC_VSUBi4D,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 924                                InstrStage<1, [A8_NPipe]>], [4, 2, 1]>,
 925   //
 926   // Quad-register Integer Subtract
 927   InstrItinData<IIC_VSUBi4Q,  [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 928                                InstrStage<1, [A8_NPipe]>], [4, 2, 1]>,
 929   //
 930   // Double-register Integer Shift
 931   InstrItinData<IIC_VSHLiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 932                                InstrStage<1, [A8_NPipe]>], [3, 1, 1]>,
 933   //
 934   // Quad-register Integer Shift
 935   InstrItinData<IIC_VSHLiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 936                                InstrStage<2, [A8_NPipe]>], [4, 1, 1]>,
 937   //
 938   // Double-register Integer Shift (4 cycle)
 939   InstrItinData<IIC_VSHLi4D,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 940                                InstrStage<1, [A8_NPipe]>], [4, 1, 1]>,
 941   //
 942   // Quad-register Integer Shift (4 cycle)
 943   InstrItinData<IIC_VSHLi4Q,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 944                                InstrStage<2, [A8_NPipe]>], [5, 1, 1]>,
 945   //
 946   // Double-register Integer Pair Add Long
 947   InstrItinData<IIC_VPALiD,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 948                                InstrStage<1, [A8_NPipe]>], [6, 3, 1]>,
 949   //
 950   // Quad-register Integer Pair Add Long
 951   InstrItinData<IIC_VPALiQ,   [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 952                                InstrStage<2, [A8_NPipe]>], [7, 3, 1]>,
 953   //
 954   // Double-register Absolute Difference and Accumulate
 955   InstrItinData<IIC_VABAD,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 956                                InstrStage<1, [A8_NPipe]>], [6, 3, 2, 1]>,
 957   //
 958   // Quad-register Absolute Difference and Accumulate
 959   InstrItinData<IIC_VABAQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 960                                InstrStage<2, [A8_NPipe]>], [6, 3, 2, 1]>,
 961
 962   //
 963   // Double-register Integer Multiply (.8, .16)
 964   InstrItinData<IIC_VMULi16D, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 965                                InstrStage<1, [A8_NPipe]>], [6, 2, 2]>,
 966   //
 967   // Double-register Integer Multiply (.32)
 968   InstrItinData<IIC_VMULi32D, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 969                                InstrStage<2, [A8_NPipe]>], [7, 2, 1]>,
 970   //
 971   // Quad-register Integer Multiply (.8, .16)
 972   InstrItinData<IIC_VMULi16Q, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 973                                InstrStage<2, [A8_NPipe]>], [7, 2, 2]>,
 974   //
 975   // Quad-register Integer Multiply (.32)
 976   InstrItinData<IIC_VMULi32Q, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 977                                InstrStage<1, [A8_NPipe]>,
 978                                InstrStage<2, [A8_NLSPipe], 0>,
 979                                InstrStage<3, [A8_NPipe]>], [9, 2, 1]>,
 980   //
 981   // Double-register Integer Multiply-Accumulate (.8, .16)
 982   InstrItinData<IIC_VMACi16D, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 983                                InstrStage<1, [A8_NPipe]>], [6, 3, 2, 2]>,
 984   //
 985   // Double-register Integer Multiply-Accumulate (.32)
 986   InstrItinData<IIC_VMACi32D, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 987                                InstrStage<2, [A8_NPipe]>], [7, 3, 2, 1]>,
 988   //
 989   // Quad-register Integer Multiply-Accumulate (.8, .16)
 990   InstrItinData<IIC_VMACi16Q, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 991                                InstrStage<2, [A8_NPipe]>], [7, 3, 2, 2]>,
 992   //
 993   // Quad-register Integer Multiply-Accumulate (.32)
 994   InstrItinData<IIC_VMACi32Q, [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
 995                                InstrStage<1, [A8_NPipe]>,
 996                                InstrStage<2, [A8_NLSPipe], 0>,
 997                                InstrStage<3, [A8_NPipe]>], [9, 3, 2, 1]>,
 998   //
 999   // Double-register VEXT
1000   InstrItinData<IIC_VEXTD,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1001                                InstrStage<1, [A8_NLSPipe]>], [2, 1, 1]>,
1002   //
1003   // Quad-register VEXT
1004   InstrItinData<IIC_VEXTQ,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1005                                InstrStage<2, [A8_NLSPipe]>], [3, 1, 1]>,
1006   //
1007   // VTB
1008   InstrItinData<IIC_VTB1,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1009                                InstrStage<2, [A8_NLSPipe]>], [3, 2, 1]>,
1010   InstrItinData<IIC_VTB2,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1011                                InstrStage<2, [A8_NLSPipe]>], [3, 2, 2, 1]>,
1012   InstrItinData<IIC_VTB3,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1013                                InstrStage<1, [A8_NLSPipe]>,
1014                                InstrStage<1, [A8_NPipe], 0>,
1015                                InstrStage<2, [A8_NLSPipe]>], [4, 2, 2, 3, 1]>,
1016   InstrItinData<IIC_VTB4,     [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1017                                InstrStage<1, [A8_NLSPipe]>,
1018                                InstrStage<1, [A8_NPipe], 0>,
1019                                InstrStage<2, [A8_NLSPipe]>],[4, 2, 2, 3, 3, 1]>,
1020   //
1021   // VTBX
1022   InstrItinData<IIC_VTBX1,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1023                                InstrStage<2, [A8_NLSPipe]>], [3, 1, 2, 1]>,
1024   InstrItinData<IIC_VTBX2,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1025                                InstrStage<2, [A8_NLSPipe]>], [3, 1, 2, 2, 1]>,
1026   InstrItinData<IIC_VTBX3,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1027                                InstrStage<1, [A8_NLSPipe]>,
1028                                InstrStage<1, [A8_NPipe], 0>,
1029                                InstrStage<2, [A8_NLSPipe]>],[4, 1, 2, 2, 3, 1]>,
1030   InstrItinData<IIC_VTBX4,    [InstrStage<1, [A8_Pipe0, A8_Pipe1], 0>,
1031                                InstrStage<1, [A8_NLSPipe]>,
1032                                InstrStage<1, [A8_NPipe], 0>,
1033                             InstrStage<2, [A8_NLSPipe]>], [4, 1, 2, 2, 3, 3, 1]>
1034 ]>;