base_dyn_inst.hh revision 2090
1/* 2 * Copyright (c) 2004-2005 The Regents of The University of Michigan 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are 7 * met: redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer; 9 * redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution; 12 * neither the name of the copyright holders nor the names of its 13 * contributors may be used to endorse or promote products derived from 14 * this software without specific prior written permission. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29#ifndef __CPU_BASE_DYN_INST_HH__ 30#define __CPU_BASE_DYN_INST_HH__ 31 32#include <string> 33#include <vector> 34 35#include "base/fast_alloc.hh" 36#include "base/trace.hh" 37#include "config/full_system.hh" 38#include "cpu/exetrace.hh" 39#include "cpu/inst_seq.hh" 40#include "cpu/o3/comm.hh" 41#include "cpu/static_inst.hh" 42#include "encumbered/cpu/full/bpred_update.hh" 43#include "encumbered/cpu/full/op_class.hh" 44#include "encumbered/cpu/full/spec_memory.hh" 45#include "encumbered/cpu/full/spec_state.hh" 46#include "encumbered/mem/functional/main.hh" 47 48/** 49 * @file 50 * Defines a dynamic instruction context. 51 */ 52 53// Forward declaration. 54template <class ISA> 55class StaticInstPtr; 56 57template <class Impl> 58class BaseDynInst : public FastAlloc, public RefCounted 59{ 60 public: 61 // Typedef for the CPU. 62 typedef typename Impl::FullCPU FullCPU; 63 64 //Typedef to get the ISA. 65 typedef typename Impl::ISA ISA; 66 67 /// Binary machine instruction type. 68 typedef typename ISA::MachInst MachInst; 69 /// Memory address type. 70 typedef typename ISA::Addr Addr; 71 /// Logical register index type. 72 typedef typename ISA::RegIndex RegIndex; 73 /// Integer register index type. 74 typedef typename ISA::IntReg IntReg; 75 76 enum { 77 MaxInstSrcRegs = ISA::MaxInstSrcRegs, //< Max source regs 78 MaxInstDestRegs = ISA::MaxInstDestRegs, //< Max dest regs 79 }; 80 81 /** The static inst used by this dyn inst. */ 82 StaticInstPtr<ISA> staticInst; 83 84 //////////////////////////////////////////// 85 // 86 // INSTRUCTION EXECUTION 87 // 88 //////////////////////////////////////////// 89 Trace::InstRecord *traceData; 90 91 template <class T> 92 Fault * read(Addr addr, T &data, unsigned flags); 93 94 template <class T> 95 Fault * write(T data, Addr addr, unsigned flags, 96 uint64_t *res); 97 98 void prefetch(Addr addr, unsigned flags); 99 void writeHint(Addr addr, int size, unsigned flags); 100 Fault * copySrcTranslate(Addr src); 101 Fault * copy(Addr dest); 102 103 /** @todo: Consider making this private. */ 104 public: 105 /** Is this instruction valid. */ 106 bool valid; 107 108 /** The sequence number of the instruction. */ 109 InstSeqNum seqNum; 110 111 /** How many source registers are ready. */ 112 unsigned readyRegs; 113 114 /** Is the instruction completed. */ 115 bool completed; 116 117 /** Can this instruction issue. */ 118 bool canIssue; 119 120 /** Has this instruction issued. */ 121 bool issued; 122 123 /** Has this instruction executed (or made it through execute) yet. */ 124 bool executed; 125 126 /** Can this instruction commit. */ 127 bool canCommit; 128 129 /** Is this instruction squashed. */ 130 bool squashed; 131 132 /** Is this instruction squashed in the instruction queue. */ 133 bool squashedInIQ; 134 135 /** Is this a recover instruction. */ 136 bool recoverInst; 137 138 /** Is this a thread blocking instruction. */ 139 bool blockingInst; /* this inst has called thread_block() */ 140 141 /** Is this a thread syncrhonization instruction. */ 142 bool threadsyncWait; 143 144 /** The thread this instruction is from. */ 145 short threadNumber; 146 147 /** data address space ID, for loads & stores. */ 148 short asid; 149 150 /** Pointer to the FullCPU object. */ 151 FullCPU *cpu; 152 153 /** Pointer to the exec context. Will not exist in the final version. */ 154 ExecContext *xc; 155 156 /** The kind of fault this instruction has generated. */ 157 Fault * fault; 158 159 /** The effective virtual address (lds & stores only). */ 160 Addr effAddr; 161 162 /** The effective physical address. */ 163 Addr physEffAddr; 164 165 /** Effective virtual address for a copy source. */ 166 Addr copySrcEffAddr; 167 168 /** Effective physical address for a copy source. */ 169 Addr copySrcPhysEffAddr; 170 171 /** The memory request flags (from translation). */ 172 unsigned memReqFlags; 173 174 /** The size of the data to be stored. */ 175 int storeSize; 176 177 /** The data to be stored. */ 178 IntReg storeData; 179 180 union Result { 181 uint64_t integer; 182 float fp; 183 double dbl; 184 }; 185 186 /** The result of the instruction; assumes for now that there's only one 187 * destination register. 188 */ 189 Result instResult; 190 191 /** PC of this instruction. */ 192 Addr PC; 193 194 /** Next non-speculative PC. It is not filled in at fetch, but rather 195 * once the target of the branch is truly known (either decode or 196 * execute). 197 */ 198 Addr nextPC; 199 200 /** Predicted next PC. */ 201 Addr predPC; 202 203 /** Count of total number of dynamic instructions. */ 204 static int instcount; 205 206 /** Whether or not the source register is ready. Not sure this should be 207 * here vs. the derived class. 208 */ 209 bool _readySrcRegIdx[MaxInstSrcRegs]; 210 211 public: 212 /** BaseDynInst constructor given a binary instruction. */ 213 BaseDynInst(MachInst inst, Addr PC, Addr Pred_PC, InstSeqNum seq_num, 214 FullCPU *cpu); 215 216 /** BaseDynInst constructor given a static inst pointer. */ 217 BaseDynInst(StaticInstPtr<ISA> &_staticInst); 218 219 /** BaseDynInst destructor. */ 220 ~BaseDynInst(); 221 222 private: 223 /** Function to initialize variables in the constructors. */ 224 void initVars(); 225 226 public: 227 void 228 trace_mem(Fault * fault, // last fault 229 MemCmd cmd, // last command 230 Addr addr, // virtual address of access 231 void *p, // memory accessed 232 int nbytes); // access size 233 234 /** Dumps out contents of this BaseDynInst. */ 235 void dump(); 236 237 /** Dumps out contents of this BaseDynInst into given string. */ 238 void dump(std::string &outstring); 239 240 /** Returns the fault type. */ 241 Fault * getFault() { return fault; } 242 243 /** Checks whether or not this instruction has had its branch target 244 * calculated yet. For now it is not utilized and is hacked to be 245 * always false. 246 */ 247 bool doneTargCalc() { return false; } 248 249 /** Returns the next PC. This could be the speculative next PC if it is 250 * called prior to the actual branch target being calculated. 251 */ 252 Addr readNextPC() { return nextPC; } 253 254 /** Set the predicted target of this current instruction. */ 255 void setPredTarg(Addr predicted_PC) { predPC = predicted_PC; } 256 257 /** Returns the predicted target of the branch. */ 258 Addr readPredTarg() { return predPC; } 259 260 /** Returns whether the instruction was predicted taken or not. */ 261 bool predTaken() { 262 return( predPC != (PC + sizeof(MachInst) ) ); 263 } 264 265 /** Returns whether the instruction mispredicted. */ 266 bool mispredicted() { return (predPC != nextPC); } 267 268 // 269 // Instruction types. Forward checks to StaticInst object. 270 // 271 bool isNop() const { return staticInst->isNop(); } 272 bool isMemRef() const { return staticInst->isMemRef(); } 273 bool isLoad() const { return staticInst->isLoad(); } 274 bool isStore() const { return staticInst->isStore(); } 275 bool isInstPrefetch() const { return staticInst->isInstPrefetch(); } 276 bool isDataPrefetch() const { return staticInst->isDataPrefetch(); } 277 bool isCopy() const { return staticInst->isCopy(); } 278 bool isInteger() const { return staticInst->isInteger(); } 279 bool isFloating() const { return staticInst->isFloating(); } 280 bool isControl() const { return staticInst->isControl(); } 281 bool isCall() const { return staticInst->isCall(); } 282 bool isReturn() const { return staticInst->isReturn(); } 283 bool isDirectCtrl() const { return staticInst->isDirectCtrl(); } 284 bool isIndirectCtrl() const { return staticInst->isIndirectCtrl(); } 285 bool isCondCtrl() const { return staticInst->isCondCtrl(); } 286 bool isUncondCtrl() const { return staticInst->isUncondCtrl(); } 287 bool isThreadSync() const { return staticInst->isThreadSync(); } 288 bool isSerializing() const { return staticInst->isSerializing(); } 289 bool isMemBarrier() const { return staticInst->isMemBarrier(); } 290 bool isWriteBarrier() const { return staticInst->isWriteBarrier(); } 291 bool isNonSpeculative() const { return staticInst->isNonSpeculative(); } 292 293 /** Returns the opclass of this instruction. */ 294 OpClass opClass() const { return staticInst->opClass(); } 295 296 /** Returns the branch target address. */ 297 Addr branchTarget() const { return staticInst->branchTarget(PC); } 298 299 /** Number of source registers. */ 300 int8_t numSrcRegs() const { return staticInst->numSrcRegs(); } 301 302 /** Number of destination registers. */ 303 int8_t numDestRegs() const { return staticInst->numDestRegs(); } 304 305 // the following are used to track physical register usage 306 // for machines with separate int & FP reg files 307 int8_t numFPDestRegs() const { return staticInst->numFPDestRegs(); } 308 int8_t numIntDestRegs() const { return staticInst->numIntDestRegs(); } 309 310 /** Returns the logical register index of the i'th destination register. */ 311 RegIndex destRegIdx(int i) const 312 { 313 return staticInst->destRegIdx(i); 314 } 315 316 /** Returns the logical register index of the i'th source register. */ 317 RegIndex srcRegIdx(int i) const 318 { 319 return staticInst->srcRegIdx(i); 320 } 321 322 /** Returns the result of an integer instruction. */ 323 uint64_t readIntResult() { return instResult.integer; } 324 325 /** Returns the result of a floating point instruction. */ 326 float readFloatResult() { return instResult.fp; } 327 328 /** Returns the result of a floating point (double) instruction. */ 329 double readDoubleResult() { return instResult.dbl; } 330 331 //Push to .cc file. 332 /** Records that one of the source registers is ready. */ 333 void markSrcRegReady() 334 { 335 ++readyRegs; 336 if(readyRegs == numSrcRegs()) { 337 canIssue = true; 338 } 339 } 340 341 /** Marks a specific register as ready. 342 * @todo: Move this to .cc file. 343 */ 344 void markSrcRegReady(RegIndex src_idx) 345 { 346 ++readyRegs; 347 348 _readySrcRegIdx[src_idx] = 1; 349 350 if(readyRegs == numSrcRegs()) { 351 canIssue = true; 352 } 353 } 354 355 /** Returns if a source register is ready. */ 356 bool isReadySrcRegIdx(int idx) const 357 { 358 return this->_readySrcRegIdx[idx]; 359 } 360 361 /** Sets this instruction as completed. */ 362 void setCompleted() { completed = true; } 363 364 /** Returns whethe or not this instruction is completed. */ 365 bool isCompleted() const { return completed; } 366 367 /** Sets this instruction as ready to issue. */ 368 void setCanIssue() { canIssue = true; } 369 370 /** Returns whether or not this instruction is ready to issue. */ 371 bool readyToIssue() const { return canIssue; } 372 373 /** Sets this instruction as issued from the IQ. */ 374 void setIssued() { issued = true; } 375 376 /** Returns whether or not this instruction has issued. */ 377 bool isIssued() const { return issued; } 378 379 /** Sets this instruction as executed. */ 380 void setExecuted() { executed = true; } 381 382 /** Returns whether or not this instruction has executed. */ 383 bool isExecuted() const { return executed; } 384 385 /** Sets this instruction as ready to commit. */ 386 void setCanCommit() { canCommit = true; } 387 388 /** Clears this instruction as being ready to commit. */ 389 void clearCanCommit() { canCommit = false; } 390 391 /** Returns whether or not this instruction is ready to commit. */ 392 bool readyToCommit() const { return canCommit; } 393 394 /** Sets this instruction as squashed. */ 395 void setSquashed() { squashed = true; } 396 397 /** Returns whether or not this instruction is squashed. */ 398 bool isSquashed() const { return squashed; } 399 400 /** Sets this instruction as squashed in the IQ. */ 401 void setSquashedInIQ() { squashedInIQ = true; } 402 403 /** Returns whether or not this instruction is squashed in the IQ. */ 404 bool isSquashedInIQ() const { return squashedInIQ; } 405 406 /** Read the PC of this instruction. */ 407 const Addr readPC() const { return PC; } 408 409 /** Set the next PC of this instruction (its actual target). */ 410 void setNextPC(uint64_t val) { nextPC = val; } 411 412 /** Returns the exec context. 413 * @todo: Remove this once the ExecContext is no longer used. 414 */ 415 ExecContext *xcBase() { return xc; } 416 417 private: 418 /** Instruction effective address. 419 * @todo: Consider if this is necessary or not. 420 */ 421 Addr instEffAddr; 422 /** Whether or not the effective address calculation is completed. 423 * @todo: Consider if this is necessary or not. 424 */ 425 bool eaCalcDone; 426 427 public: 428 /** Sets the effective address. */ 429 void setEA(Addr &ea) { instEffAddr = ea; eaCalcDone = true; } 430 431 /** Returns the effective address. */ 432 const Addr &getEA() const { return instEffAddr; } 433 434 /** Returns whether or not the eff. addr. calculation has been completed. */ 435 bool doneEACalc() { return eaCalcDone; } 436 437 /** Returns whether or not the eff. addr. source registers are ready. */ 438 bool eaSrcsReady(); 439 440 public: 441 /** Load queue index. */ 442 int16_t lqIdx; 443 444 /** Store queue index. */ 445 int16_t sqIdx; 446}; 447 448template<class Impl> 449template<class T> 450inline Fault * 451BaseDynInst<Impl>::read(Addr addr, T &data, unsigned flags) 452{ 453 MemReqPtr req = new MemReq(addr, xc, sizeof(T), flags); 454 req->asid = asid; 455 456 fault = cpu->translateDataReadReq(req); 457 458 // Record key MemReq parameters so we can generate another one 459 // just like it for the timing access without calling translate() 460 // again (which might mess up the TLB). 461 // Do I ever really need this? -KTL 3/05 462 effAddr = req->vaddr; 463 physEffAddr = req->paddr; 464 memReqFlags = req->flags; 465 466 /** 467 * @todo 468 * Replace the disjoint functional memory with a unified one and remove 469 * this hack. 470 */ 471#if !FULL_SYSTEM 472 req->paddr = req->vaddr; 473#endif 474 475 if (fault == NoFault) { 476 fault = cpu->read(req, data, lqIdx); 477 } else { 478 // Return a fixed value to keep simulation deterministic even 479 // along misspeculated paths. 480 data = (T)-1; 481 } 482 483 if (traceData) { 484 traceData->setAddr(addr); 485 traceData->setData(data); 486 } 487 488 return fault; 489} 490 491template<class Impl> 492template<class T> 493inline Fault * 494BaseDynInst<Impl>::write(T data, Addr addr, unsigned flags, uint64_t *res) 495{ 496 if (traceData) { 497 traceData->setAddr(addr); 498 traceData->setData(data); 499 } 500 501 MemReqPtr req = new MemReq(addr, xc, sizeof(T), flags); 502 503 req->asid = asid; 504 505 fault = cpu->translateDataWriteReq(req); 506 507 // Record key MemReq parameters so we can generate another one 508 // just like it for the timing access without calling translate() 509 // again (which might mess up the TLB). 510 effAddr = req->vaddr; 511 physEffAddr = req->paddr; 512 memReqFlags = req->flags; 513 514 /** 515 * @todo 516 * Replace the disjoint functional memory with a unified one and remove 517 * this hack. 518 */ 519#if !FULL_SYSTEM 520 req->paddr = req->vaddr; 521#endif 522 523 if (fault == NoFault) { 524 fault = cpu->write(req, data, sqIdx); 525 } 526 527 if (res) { 528 // always return some result to keep misspeculated paths 529 // (which will ignore faults) deterministic 530 *res = (fault == NoFault) ? req->result : 0; 531 } 532 533 return fault; 534} 535 536#endif // __CPU_BASE_DYN_INST_HH__ 537