fetch.hh revision 2670:9107b8bd08cd
1/* 2 * Copyright (c) 2004-2006 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 * Authors: Kevin Lim 29 */ 30 31#ifndef __CPU_O3_FETCH_HH__ 32#define __CPU_O3_FETCH_HH__ 33 34#include "arch/utility.hh" 35#include "base/statistics.hh" 36#include "base/timebuf.hh" 37#include "cpu/pc_event.hh" 38#include "mem/packet.hh" 39#include "mem/port.hh" 40#include "sim/eventq.hh" 41 42class Sampler; 43 44/** 45 * DefaultFetch class handles both single threaded and SMT fetch. Its 46 * width is specified by the parameters; each cycle it tries to fetch 47 * that many instructions. It supports using a branch predictor to 48 * predict direction and targets. 49 * It supports the idling functionalitiy of the CPU by indicating to 50 * the CPU when it is active and inactive. 51 */ 52template <class Impl> 53class DefaultFetch 54{ 55 public: 56 /** Typedefs from Impl. */ 57 typedef typename Impl::CPUPol CPUPol; 58 typedef typename Impl::DynInst DynInst; 59 typedef typename Impl::DynInstPtr DynInstPtr; 60 typedef typename Impl::FullCPU FullCPU; 61 typedef typename Impl::Params Params; 62 63 /** Typedefs from the CPU policy. */ 64 typedef typename CPUPol::BPredUnit BPredUnit; 65 typedef typename CPUPol::FetchStruct FetchStruct; 66 typedef typename CPUPol::TimeStruct TimeStruct; 67 68 /** Typedefs from ISA. */ 69 typedef TheISA::MachInst MachInst; 70 typedef TheISA::ExtMachInst ExtMachInst; 71 72 class IcachePort : public Port 73 { 74 protected: 75 DefaultFetch<Impl> *fetch; 76 77 public: 78 IcachePort(DefaultFetch<Impl> *_fetch) 79 : Port(_fetch->name() + "-iport"), fetch(_fetch) 80 { } 81 82 protected: 83 virtual Tick recvAtomic(PacketPtr pkt); 84 85 virtual void recvFunctional(PacketPtr pkt); 86 87 virtual void recvStatusChange(Status status); 88 89 virtual void getDeviceAddressRanges(AddrRangeList &resp, 90 AddrRangeList &snoop) 91 { resp.clear(); snoop.clear(); } 92 93 virtual bool recvTiming(PacketPtr pkt); 94 95 virtual void recvRetry(); 96 }; 97 98 public: 99 /** Overall fetch status. Used to determine if the CPU can 100 * deschedule itsef due to a lack of activity. 101 */ 102 enum FetchStatus { 103 Active, 104 Inactive 105 }; 106 107 /** Individual thread status. */ 108 enum ThreadStatus { 109 Running, 110 Idle, 111 Squashing, 112 Blocked, 113 Fetching, 114 TrapPending, 115 QuiescePending, 116 SwitchOut, 117 IcacheWaitResponse, 118 IcacheRetry, 119 IcacheAccessComplete 120 }; 121 122 /** Fetching Policy, Add new policies here.*/ 123 enum FetchPriority { 124 SingleThread, 125 RoundRobin, 126 Branch, 127 IQ, 128 LSQ 129 }; 130 131 private: 132 /** Fetch status. */ 133 FetchStatus _status; 134 135 /** Per-thread status. */ 136 ThreadStatus fetchStatus[Impl::MaxThreads]; 137 138 /** Fetch policy. */ 139 FetchPriority fetchPolicy; 140 141 /** List that has the threads organized by priority. */ 142 std::list<unsigned> priorityList; 143 144 public: 145 /** DefaultFetch constructor. */ 146 DefaultFetch(Params *params); 147 148 /** Returns the name of fetch. */ 149 std::string name() const; 150 151 /** Registers statistics. */ 152 void regStats(); 153 154 /** Sets CPU pointer. */ 155 void setCPU(FullCPU *cpu_ptr); 156 157 /** Sets the main backwards communication time buffer pointer. */ 158 void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer); 159 160 /** Sets pointer to list of active threads. */ 161 void setActiveThreads(std::list<unsigned> *at_ptr); 162 163 /** Sets pointer to time buffer used to communicate to the next stage. */ 164 void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr); 165 166 /** Sets pointer to page table. */ 167// void setPageTable(PageTable *pt_ptr); 168 169 /** Initialize stage. */ 170 void initStage(); 171 172 /** Processes cache completion event. */ 173 void processCacheCompletion(PacketPtr pkt); 174 175 void switchOut(); 176 177 void doSwitchOut(); 178 179 void takeOverFrom(); 180 181 bool isSwitchedOut() { return switchedOut; } 182 183 void wakeFromQuiesce(); 184 185 private: 186 /** Changes the status of this stage to active, and indicates this 187 * to the CPU. 188 */ 189 inline void switchToActive(); 190 191 /** Changes the status of this stage to inactive, and indicates 192 * this to the CPU. 193 */ 194 inline void switchToInactive(); 195 196 /** 197 * Looks up in the branch predictor to see if the next PC should be 198 * either next PC+=MachInst or a branch target. 199 * @param next_PC Next PC variable passed in by reference. It is 200 * expected to be set to the current PC; it will be updated with what 201 * the next PC will be. 202 * @return Whether or not a branch was predicted as taken. 203 */ 204 bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC); 205 206 /** 207 * Fetches the cache line that contains fetch_PC. Returns any 208 * fault that happened. Puts the data into the class variable 209 * cacheData. 210 * @param fetch_PC The PC address that is being fetched from. 211 * @param ret_fault The fault reference that will be set to the result of 212 * the icache access. 213 * @param tid Thread id. 214 * @return Any fault that occured. 215 */ 216 bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid); 217 218 /** Squashes a specific thread and resets the PC. */ 219 inline void doSquash(const Addr &new_PC, unsigned tid); 220 221 /** Squashes a specific thread and resets the PC. Also tells the CPU to 222 * remove any instructions between fetch and decode that should be sqaushed. 223 */ 224 void squashFromDecode(const Addr &new_PC, const InstSeqNum &seq_num, 225 unsigned tid); 226 227 /** Checks if a thread is stalled. */ 228 bool checkStall(unsigned tid) const; 229 230 /** Updates overall fetch stage status; to be called at the end of each 231 * cycle. */ 232 FetchStatus updateFetchStatus(); 233 234 public: 235 /** Squashes a specific thread and resets the PC. Also tells the CPU to 236 * remove any instructions that are not in the ROB. The source of this 237 * squash should be the commit stage. 238 */ 239 void squash(const Addr &new_PC, unsigned tid); 240 241 /** Ticks the fetch stage, processing all inputs signals and fetching 242 * as many instructions as possible. 243 */ 244 void tick(); 245 246 /** Checks all input signals and updates the status as necessary. 247 * @return: Returns if the status has changed due to input signals. 248 */ 249 bool checkSignalsAndUpdate(unsigned tid); 250 251 /** Does the actual fetching of instructions and passing them on to the 252 * next stage. 253 * @param status_change fetch() sets this variable if there was a status 254 * change (ie switching to IcacheMissStall). 255 */ 256 void fetch(bool &status_change); 257 258 /** Align a PC to the start of an I-cache block. */ 259 Addr icacheBlockAlignPC(Addr addr) 260 { 261 addr = TheISA::realPCToFetchPC(addr); 262 return (addr & ~(cacheBlkMask)); 263 } 264 265 private: 266 /** Returns the appropriate thread to fetch, given the fetch policy. */ 267 int getFetchingThread(FetchPriority &fetch_priority); 268 269 /** Returns the appropriate thread to fetch using a round robin policy. */ 270 int roundRobin(); 271 272 /** Returns the appropriate thread to fetch using the IQ count policy. */ 273 int iqCount(); 274 275 /** Returns the appropriate thread to fetch using the LSQ count policy. */ 276 int lsqCount(); 277 278 /** Returns the appropriate thread to fetch using the branch count policy. */ 279 int branchCount(); 280 281 private: 282 /** Pointer to the FullCPU. */ 283 FullCPU *cpu; 284 285 /** Time buffer interface. */ 286 TimeBuffer<TimeStruct> *timeBuffer; 287 288 /** Wire to get decode's information from backwards time buffer. */ 289 typename TimeBuffer<TimeStruct>::wire fromDecode; 290 291 /** Wire to get rename's information from backwards time buffer. */ 292 typename TimeBuffer<TimeStruct>::wire fromRename; 293 294 /** Wire to get iew's information from backwards time buffer. */ 295 typename TimeBuffer<TimeStruct>::wire fromIEW; 296 297 /** Wire to get commit's information from backwards time buffer. */ 298 typename TimeBuffer<TimeStruct>::wire fromCommit; 299 300 /** Internal fetch instruction queue. */ 301 TimeBuffer<FetchStruct> *fetchQueue; 302 303 //Might be annoying how this name is different than the queue. 304 /** Wire used to write any information heading to decode. */ 305 typename TimeBuffer<FetchStruct>::wire toDecode; 306 307 MemObject *mem; 308 309 /** Icache interface. */ 310 IcachePort *icachePort; 311 312 /** BPredUnit. */ 313 BPredUnit branchPred; 314 315 Addr PC[Impl::MaxThreads]; 316 317 Addr nextPC[Impl::MaxThreads]; 318 319 /** Memory packet used to access cache. */ 320 PacketPtr memPkt[Impl::MaxThreads]; 321 322 /** Variable that tracks if fetch has written to the time buffer this 323 * cycle. Used to tell CPU if there is activity this cycle. 324 */ 325 bool wroteToTimeBuffer; 326 327 /** Tracks how many instructions has been fetched this cycle. */ 328 int numInst; 329 330 /** Source of possible stalls. */ 331 struct Stalls { 332 bool decode; 333 bool rename; 334 bool iew; 335 bool commit; 336 }; 337 338 /** Tracks which stages are telling fetch to stall. */ 339 Stalls stalls[Impl::MaxThreads]; 340 341 /** Decode to fetch delay, in ticks. */ 342 unsigned decodeToFetchDelay; 343 344 /** Rename to fetch delay, in ticks. */ 345 unsigned renameToFetchDelay; 346 347 /** IEW to fetch delay, in ticks. */ 348 unsigned iewToFetchDelay; 349 350 /** Commit to fetch delay, in ticks. */ 351 unsigned commitToFetchDelay; 352 353 /** The width of fetch in instructions. */ 354 unsigned fetchWidth; 355 356 /** Cache block size. */ 357 int cacheBlkSize; 358 359 /** Mask to get a cache block's address. */ 360 Addr cacheBlkMask; 361 362 /** The cache line being fetched. */ 363 uint8_t *cacheData[Impl::MaxThreads]; 364 365 /** Size of instructions. */ 366 int instSize; 367 368 /** Icache stall statistics. */ 369 Counter lastIcacheStall[Impl::MaxThreads]; 370 371 /** List of Active Threads */ 372 std::list<unsigned> *activeThreads; 373 374 /** Number of threads. */ 375 unsigned numThreads; 376 377 /** Number of threads that are actively fetching. */ 378 unsigned numFetchingThreads; 379 380 /** Thread ID being fetched. */ 381 int threadFetched; 382 383 bool interruptPending; 384 385 bool switchedOut; 386 387#if !FULL_SYSTEM 388 /** Page table pointer. */ 389// PageTable *pTable; 390#endif 391 392 // @todo: Consider making these vectors and tracking on a per thread basis. 393 /** Stat for total number of cycles stalled due to an icache miss. */ 394 Stats::Scalar<> icacheStallCycles; 395 /** Stat for total number of fetched instructions. */ 396 Stats::Scalar<> fetchedInsts; 397 Stats::Scalar<> fetchedBranches; 398 /** Stat for total number of predicted branches. */ 399 Stats::Scalar<> predictedBranches; 400 /** Stat for total number of cycles spent fetching. */ 401 Stats::Scalar<> fetchCycles; 402 /** Stat for total number of cycles spent squashing. */ 403 Stats::Scalar<> fetchSquashCycles; 404 /** Stat for total number of cycles spent blocked due to other stages in 405 * the pipeline. 406 */ 407 Stats::Scalar<> fetchIdleCycles; 408 Stats::Scalar<> fetchBlockedCycles; 409 410 Stats::Scalar<> fetchMiscStallCycles; 411 /** Stat for total number of fetched cache lines. */ 412 Stats::Scalar<> fetchedCacheLines; 413 414 Stats::Scalar<> fetchIcacheSquashes; 415 /** Distribution of number of instructions fetched each cycle. */ 416 Stats::Distribution<> fetchNisnDist; 417 Stats::Formula idleRate; 418 Stats::Formula branchRate; 419 Stats::Formula fetchRate; 420}; 421 422#endif //__CPU_O3_FETCH_HH__ 423