PPCScheduleE5500.td revision 263508
1//===-- PPCScheduleE500mc.td - e5500 Scheduling Defs -------*- 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 Freescale e5500 64-bit 11// Power processor. 12// 13// All information is derived from the "e5500 Core Reference Manual", 14// Freescale Document Number e5500RM, Rev. 1, 03/2012. 15// 16//===----------------------------------------------------------------------===// 17// Relevant functional units in the Freescale e5500 core 18// (These are the same as for the e500mc) 19// 20// * Decode & Dispatch 21// Can dispatch up to 2 instructions per clock cycle to either the GPR Issue 22// queues (GIQx), FP Issue Queue (FIQ), or Branch issue queue (BIQ). 23// def DIS0 : FuncUnit; 24// def DIS1 : FuncUnit; 25 26// * Execute 27// 6 pipelined execution units: SFX0, SFX1, BU, FPU, LSU, CFX. 28// The CFX has a bypass path, allowing non-divide instructions to execute 29// while a divide instruction is being executed. 30// def SFX0 : FuncUnit; // Simple unit 0 31// def SFX1 : FuncUnit; // Simple unit 1 32// def BU : FuncUnit; // Branch unit 33// def CFX_DivBypass 34// : FuncUnit; // CFX divide bypass path 35// def CFX_0 : FuncUnit; // CFX pipeline stage 0 36 37def CFX_1 : FuncUnit; // CFX pipeline stage 1 38 39// def LSU_0 : FuncUnit; // LSU pipeline 40// def FPU_0 : FuncUnit; // FPU pipeline 41 42// def CR_Bypass : Bypass; 43 44def PPCE5500Itineraries : ProcessorItineraries< 45 [DIS0, DIS1, SFX0, SFX1, BU, CFX_DivBypass, CFX_0, CFX_1, 46 LSU_0, FPU_0], 47 [CR_Bypass, GPR_Bypass, FPR_Bypass], [ 48 InstrItinData<IntSimple , [InstrStage<1, [DIS0, DIS1], 0>, 49 InstrStage<1, [SFX0, SFX1]>], 50 [5, 2, 2], // Latency = 1 51 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 52 InstrItinData<IntGeneral , [InstrStage<1, [DIS0, DIS1], 0>, 53 InstrStage<1, [SFX0, SFX1]>], 54 [5, 2, 2], // Latency = 1 55 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 56 InstrItinData<IntCompare , [InstrStage<1, [DIS0, DIS1], 0>, 57 InstrStage<1, [SFX0, SFX1]>], 58 [6, 2, 2], // Latency = 1 or 2 59 [CR_Bypass, GPR_Bypass, GPR_Bypass]>, 60 InstrItinData<IntDivD , [InstrStage<1, [DIS0, DIS1], 0>, 61 InstrStage<1, [CFX_0], 0>, 62 InstrStage<26, [CFX_DivBypass]>], 63 [30, 2, 2], // Latency= 4..26, Repeat rate= 4..26 64 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 65 InstrItinData<IntDivW , [InstrStage<1, [DIS0, DIS1], 0>, 66 InstrStage<1, [CFX_0], 0>, 67 InstrStage<16, [CFX_DivBypass]>], 68 [20, 2, 2], // Latency= 4..16, Repeat rate= 4..16 69 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 70 InstrItinData<IntMFFS , [InstrStage<1, [DIS0, DIS1], 0>, 71 InstrStage<1, [FPU_0]>], 72 [11], // Latency = 7, Repeat rate = 1 73 [FPR_Bypass]>, 74 InstrItinData<IntMTFSB0 , [InstrStage<1, [DIS0, DIS1], 0>, 75 InstrStage<7, [FPU_0]>], 76 [11, 2, 2], // Latency = 7, Repeat rate = 7 77 [NoBypass, NoBypass, NoBypass]>, 78 InstrItinData<IntMulHD , [InstrStage<1, [DIS0, DIS1], 0>, 79 InstrStage<1, [CFX_0], 0>, 80 InstrStage<2, [CFX_1]>], 81 [9, 2, 2], // Latency = 4..7, Repeat rate = 2..4 82 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 83 InstrItinData<IntMulHW , [InstrStage<1, [DIS0, DIS1], 0>, 84 InstrStage<1, [CFX_0], 0>, 85 InstrStage<1, [CFX_1]>], 86 [8, 2, 2], // Latency = 4, Repeat rate = 1 87 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 88 InstrItinData<IntMulHWU , [InstrStage<1, [DIS0, DIS1], 0>, 89 InstrStage<1, [CFX_0], 0>, 90 InstrStage<1, [CFX_1]>], 91 [8, 2, 2], // Latency = 4, Repeat rate = 1 92 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 93 InstrItinData<IntMulLI , [InstrStage<1, [DIS0, DIS1], 0>, 94 InstrStage<1, [CFX_0], 0>, 95 InstrStage<2, [CFX_1]>], 96 [8, 2, 2], // Latency = 4 or 5, Repeat = 2 97 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 98 InstrItinData<IntRotate , [InstrStage<1, [DIS0, DIS1], 0>, 99 InstrStage<1, [SFX0, SFX1]>], 100 [5, 2, 2], // Latency = 1 101 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 102 InstrItinData<IntRotateD , [InstrStage<1, [DIS0, DIS1], 0>, 103 InstrStage<2, [SFX0, SFX1]>], 104 [6, 2, 2], // Latency = 2, Repeat rate = 2 105 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 106 InstrItinData<IntRotateDI , [InstrStage<1, [DIS0, DIS1], 0>, 107 InstrStage<1, [SFX0, SFX1]>], 108 [5, 2, 2], // Latency = 1, Repeat rate = 1 109 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 110 InstrItinData<IntShift , [InstrStage<1, [DIS0, DIS1], 0>, 111 InstrStage<2, [SFX0, SFX1]>], 112 [6, 2, 2], // Latency = 2, Repeat rate = 2 113 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 114 InstrItinData<IntTrapW , [InstrStage<1, [DIS0, DIS1], 0>, 115 InstrStage<2, [SFX0]>], 116 [6, 2], // Latency = 2, Repeat rate = 2 117 [GPR_Bypass, GPR_Bypass]>, 118 InstrItinData<BrB , [InstrStage<1, [DIS0, DIS1], 0>, 119 InstrStage<1, [BU]>], 120 [5, 2], // Latency = 1 121 [NoBypass, GPR_Bypass]>, 122 InstrItinData<BrCR , [InstrStage<1, [DIS0, DIS1], 0>, 123 InstrStage<1, [BU]>], 124 [5, 2, 2], // Latency = 1 125 [CR_Bypass, CR_Bypass, CR_Bypass]>, 126 InstrItinData<BrMCR , [InstrStage<1, [DIS0, DIS1], 0>, 127 InstrStage<1, [BU]>], 128 [5, 2], // Latency = 1 129 [CR_Bypass, CR_Bypass]>, 130 InstrItinData<BrMCRX , [InstrStage<1, [DIS0, DIS1], 0>, 131 InstrStage<1, [CFX_0]>], 132 [5, 2, 2], // Latency = 1 133 [CR_Bypass, GPR_Bypass]>, 134 InstrItinData<LdStDCBA , [InstrStage<1, [DIS0, DIS1], 0>, 135 InstrStage<1, [LSU_0]>], 136 [7, 2], // Latency = 3, Repeat rate = 1 137 [GPR_Bypass, GPR_Bypass]>, 138 InstrItinData<LdStDCBF , [InstrStage<1, [DIS0, DIS1], 0>, 139 InstrStage<1, [LSU_0]>], 140 [7, 2], // Latency = 3, Repeat rate = 1 141 [GPR_Bypass, GPR_Bypass]>, 142 InstrItinData<LdStDCBI , [InstrStage<1, [DIS0, DIS1], 0>, 143 InstrStage<1, [LSU_0]>], 144 [7, 2], // Latency = 3, Repeat rate = 1 145 [GPR_Bypass, GPR_Bypass]>, 146 InstrItinData<LdStLoad , [InstrStage<1, [DIS0, DIS1], 0>, 147 InstrStage<1, [LSU_0]>], 148 [7, 2], // Latency = 3 149 [GPR_Bypass, GPR_Bypass]>, 150 InstrItinData<LdStLoadUpd , [InstrStage<1, [DIS0, DIS1], 0>, 151 InstrStage<1, [SFX0, SFX1], 0>, 152 InstrStage<1, [LSU_0]>], 153 [7, 2], // Latency = 3, Repeat rate = 1 154 [GPR_Bypass, GPR_Bypass], 155 2>, // 2 micro-ops 156 InstrItinData<LdStLD , [InstrStage<1, [DIS0, DIS1], 0>, 157 InstrStage<1, [LSU_0]>], 158 [7, 2], // Latency = 3, Repeat rate = 1 159 [GPR_Bypass, GPR_Bypass]>, 160 InstrItinData<LdStLDARX , [InstrStage<1, [DIS0, DIS1], 0>, 161 InstrStage<3, [LSU_0]>], 162 [7, 2], // Latency = 3, Repeat rate = 3 163 [GPR_Bypass, GPR_Bypass]>, 164 InstrItinData<LdStLDU , [InstrStage<1, [DIS0, DIS1], 0>, 165 InstrStage<1, [SFX0, SFX1], 0>, 166 InstrStage<1, [LSU_0]>], 167 [7, 2], // Latency = 3, Repeat rate = 1 168 [GPR_Bypass, GPR_Bypass], 169 2>, // 2 micro-ops 170 InstrItinData<LdStStore , [InstrStage<1, [DIS0, DIS1], 0>, 171 InstrStage<1, [LSU_0]>], 172 [7, 2], // Latency = 3, Repeat rate = 1 173 [NoBypass, GPR_Bypass]>, 174 InstrItinData<LdStStoreUpd, [InstrStage<1, [DIS0, DIS1], 0>, 175 InstrStage<1, [SFX0, SFX1], 0>, 176 InstrStage<1, [LSU_0]>], 177 [7, 2], // Latency = 3, Repeat rate = 1 178 [NoBypass, GPR_Bypass], 179 2>, // 2 micro-ops 180 InstrItinData<LdStICBI , [InstrStage<1, [DIS0, DIS1], 0>, 181 InstrStage<1, [LSU_0]>], 182 [7, 2], // Latency = 3, Repeat rate = 1 183 [NoBypass, GPR_Bypass]>, 184 InstrItinData<LdStSTFD , [InstrStage<1, [DIS0, DIS1], 0>, 185 InstrStage<1, [LSU_0]>], 186 [7, 2, 2], // Latency = 3, Repeat rate = 1 187 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 188 InstrItinData<LdStSTFDU , [InstrStage<1, [DIS0, DIS1], 0>, 189 InstrStage<1, [SFX0, SFX1], 0>, 190 InstrStage<1, [LSU_0]>], 191 [7, 2, 2], // Latency = 3, Repeat rate = 1 192 [GPR_Bypass, GPR_Bypass, GPR_Bypass], 193 2>, // 2 micro-ops 194 InstrItinData<LdStLFD , [InstrStage<1, [DIS0, DIS1], 0>, 195 InstrStage<1, [LSU_0]>], 196 [8, 2, 2], // Latency = 4, Repeat rate = 1 197 [FPR_Bypass, GPR_Bypass, GPR_Bypass], 198 2>, // 2 micro-ops 199 InstrItinData<LdStLFDU , [InstrStage<1, [DIS0, DIS1], 0>, 200 InstrStage<1, [SFX0, SFX1], 0>, 201 InstrStage<1, [LSU_0]>], 202 [8, 2, 2], // Latency = 4, Repeat rate = 1 203 [FPR_Bypass, GPR_Bypass, GPR_Bypass], 204 2>, // 2 micro-ops 205 InstrItinData<LdStLHA , [InstrStage<1, [DIS0, DIS1], 0>, 206 InstrStage<1, [LSU_0]>], 207 [7, 2], // Latency = 3 208 [GPR_Bypass, GPR_Bypass]>, 209 InstrItinData<LdStLHAU , [InstrStage<1, [DIS0, DIS1], 0>, 210 InstrStage<1, [SFX0, SFX1], 0>, 211 InstrStage<1, [LSU_0]>], 212 [7, 2], // Latency = 3, Repeat rate = 1 213 [GPR_Bypass, GPR_Bypass], 214 2>, // 2 micro-ops 215 InstrItinData<LdStLMW , [InstrStage<1, [DIS0, DIS1], 0>, 216 InstrStage<4, [LSU_0]>], 217 [8, 2], // Latency = r+3, Repeat rate = r+3 218 [NoBypass, GPR_Bypass]>, 219 InstrItinData<LdStLWARX , [InstrStage<1, [DIS0, DIS1], 0>, 220 InstrStage<3, [LSU_0]>], 221 [7, 2, 2], // Latency = 3, Repeat rate = 3 222 [GPR_Bypass, GPR_Bypass, GPR_Bypass]>, 223 InstrItinData<LdStSTD , [InstrStage<1, [DIS0, DIS1], 0>, 224 InstrStage<1, [LSU_0]>], 225 [7, 2], // Latency = 3, Repeat rate = 1 226 [NoBypass, GPR_Bypass]>, 227 InstrItinData<LdStSTDCX , [InstrStage<1, [DIS0, DIS1], 0>, 228 InstrStage<1, [LSU_0]>], 229 [7, 2], // Latency = 3, Repeat rate = 1 230 [NoBypass, GPR_Bypass]>, 231 InstrItinData<LdStSTDU , [InstrStage<1, [DIS0, DIS1], 0>, 232 InstrStage<1, [SFX0, SFX1], 0>, 233 InstrStage<1, [LSU_0]>], 234 [7, 2], // Latency = 3, Repeat rate = 1 235 [NoBypass, GPR_Bypass], 236 2>, // 2 micro-ops 237 InstrItinData<LdStSTWCX , [InstrStage<1, [DIS0, DIS1], 0>, 238 InstrStage<1, [LSU_0]>], 239 [7, 2], // Latency = 3, Repeat rate = 1 240 [NoBypass, GPR_Bypass]>, 241 InstrItinData<LdStSync , [InstrStage<1, [DIS0, DIS1], 0>, 242 InstrStage<1, [LSU_0]>]>, 243 InstrItinData<SprMTMSR , [InstrStage<1, [DIS0, DIS1], 0>, 244 InstrStage<2, [CFX_0]>], 245 [6, 2], // Latency = 2, Repeat rate = 4 246 [GPR_Bypass, GPR_Bypass]>, 247 InstrItinData<SprTLBSYNC , [InstrStage<1, [DIS0, DIS1], 0>, 248 InstrStage<1, [LSU_0], 0>]>, 249 InstrItinData<SprMFCR , [InstrStage<1, [DIS0, DIS1], 0>, 250 InstrStage<5, [CFX_0]>], 251 [9, 2], // Latency = 5, Repeat rate = 5 252 [GPR_Bypass, CR_Bypass]>, 253 InstrItinData<SprMFMSR , [InstrStage<1, [DIS0, DIS1], 0>, 254 InstrStage<4, [SFX0]>], 255 [8, 2], // Latency = 4, Repeat rate = 4 256 [GPR_Bypass, GPR_Bypass]>, 257 InstrItinData<SprMFSPR , [InstrStage<1, [DIS0, DIS1], 0>, 258 InstrStage<1, [CFX_0]>], 259 [5], // Latency = 1, Repeat rate = 1 260 [GPR_Bypass]>, 261 InstrItinData<SprMFTB , [InstrStage<1, [DIS0, DIS1], 0>, 262 InstrStage<4, [CFX_0]>], 263 [8, 2], // Latency = 4, Repeat rate = 4 264 [NoBypass, GPR_Bypass]>, 265 InstrItinData<SprMTSPR , [InstrStage<1, [DIS0, DIS1], 0>, 266 InstrStage<1, [SFX0, SFX1]>], 267 [5], // Latency = 1, Repeat rate = 1 268 [GPR_Bypass]>, 269 InstrItinData<FPGeneral , [InstrStage<1, [DIS0, DIS1], 0>, 270 InstrStage<1, [FPU_0]>], 271 [11, 2, 2], // Latency = 7, Repeat rate = 1 272 [FPR_Bypass, FPR_Bypass, FPR_Bypass]>, 273 InstrItinData<FPAddSub , [InstrStage<1, [DIS0, DIS1], 0>, 274 InstrStage<1, [FPU_0]>], 275 [11, 2, 2], // Latency = 7, Repeat rate = 1 276 [FPR_Bypass, FPR_Bypass, FPR_Bypass]>, 277 InstrItinData<FPCompare , [InstrStage<1, [DIS0, DIS1], 0>, 278 InstrStage<1, [FPU_0]>], 279 [11, 2, 2], // Latency = 7, Repeat rate = 1 280 [CR_Bypass, FPR_Bypass, FPR_Bypass]>, 281 InstrItinData<FPDivD , [InstrStage<1, [DIS0, DIS1], 0>, 282 InstrStage<31, [FPU_0]>], 283 [39, 2, 2], // Latency = 35, Repeat rate = 31 284 [FPR_Bypass, FPR_Bypass, FPR_Bypass]>, 285 InstrItinData<FPDivS , [InstrStage<1, [DIS0, DIS1], 0>, 286 InstrStage<16, [FPU_0]>], 287 [24, 2, 2], // Latency = 20, Repeat rate = 16 288 [FPR_Bypass, FPR_Bypass, FPR_Bypass]>, 289 InstrItinData<FPFused , [InstrStage<1, [DIS0, DIS1], 0>, 290 InstrStage<1, [FPU_0]>], 291 [11, 2, 2, 2], // Latency = 7, Repeat rate = 1 292 [FPR_Bypass, FPR_Bypass, FPR_Bypass, FPR_Bypass]>, 293 InstrItinData<FPRes , [InstrStage<1, [DIS0, DIS1], 0>, 294 InstrStage<2, [FPU_0]>], 295 [12, 2], // Latency = 8, Repeat rate = 2 296 [FPR_Bypass, FPR_Bypass]> 297]>; 298 299// ===---------------------------------------------------------------------===// 300// e5500 machine model for scheduling and other instruction cost heuristics. 301 302def PPCE5500Model : SchedMachineModel { 303 let IssueWidth = 2; // 2 micro-ops are dispatched per cycle. 304 let MinLatency = -1; // OperandCycles are interpreted as MinLatency. 305 let LoadLatency = 6; // Optimistic load latency assuming bypass. 306 // This is overriden by OperandCycles if the 307 // Itineraries are queried instead. 308 309 let Itineraries = PPCE5500Itineraries; 310} 311