lsq.hh revision 5529:9ae69b9cd7fd
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: Korey Sewell 29 */ 30 31#ifndef __CPU_O3_LSQ_HH__ 32#define __CPU_O3_LSQ_HH__ 33 34#include <map> 35#include <queue> 36 37#include "config/full_system.hh" 38#include "cpu/inst_seq.hh" 39#include "cpu/o3/lsq_unit.hh" 40#include "mem/port.hh" 41#include "sim/sim_object.hh" 42 43class DerivO3CPUParams; 44 45template <class Impl> 46class LSQ { 47 public: 48 typedef typename Impl::O3CPU O3CPU; 49 typedef typename Impl::DynInstPtr DynInstPtr; 50 typedef typename Impl::CPUPol::IEW IEW; 51 typedef typename Impl::CPUPol::LSQUnit LSQUnit; 52 53 /** SMT policy. */ 54 enum LSQPolicy { 55 Dynamic, 56 Partitioned, 57 Threshold 58 }; 59 60 /** Constructs an LSQ with the given parameters. */ 61 LSQ(O3CPU *cpu_ptr, IEW *iew_ptr, DerivO3CPUParams *params); 62 63 /** Returns the name of the LSQ. */ 64 std::string name() const; 65 66 /** Registers statistics of each LSQ unit. */ 67 void regStats(); 68 69 /** Returns dcache port. 70 * @todo: Dcache port needs to be moved up to this level for SMT 71 * to work. For now it just returns the port from one of the 72 * threads. 73 */ 74 Port *getDcachePort() { return &dcachePort; } 75 76 /** Sets the pointer to the list of active threads. */ 77 void setActiveThreads(std::list<unsigned> *at_ptr); 78 /** Switches out the LSQ. */ 79 void switchOut(); 80 /** Takes over execution from another CPU's thread. */ 81 void takeOverFrom(); 82 83 /** Number of entries needed for the given amount of threads.*/ 84 int entryAmount(int num_threads); 85 void removeEntries(unsigned tid); 86 /** Reset the max entries for each thread. */ 87 void resetEntries(); 88 /** Resize the max entries for a thread. */ 89 void resizeEntries(unsigned size, unsigned tid); 90 91 /** Ticks the LSQ. */ 92 void tick(); 93 /** Ticks a specific LSQ Unit. */ 94 void tick(unsigned tid) 95 { thread[tid].tick(); } 96 97 /** Inserts a load into the LSQ. */ 98 void insertLoad(DynInstPtr &load_inst); 99 /** Inserts a store into the LSQ. */ 100 void insertStore(DynInstPtr &store_inst); 101 102 /** Executes a load. */ 103 Fault executeLoad(DynInstPtr &inst); 104 105 /** Executes a store. */ 106 Fault executeStore(DynInstPtr &inst); 107 108 /** 109 * Commits loads up until the given sequence number for a specific thread. 110 */ 111 void commitLoads(InstSeqNum &youngest_inst, unsigned tid) 112 { thread[tid].commitLoads(youngest_inst); } 113 114 /** 115 * Commits stores up until the given sequence number for a specific thread. 116 */ 117 void commitStores(InstSeqNum &youngest_inst, unsigned tid) 118 { thread[tid].commitStores(youngest_inst); } 119 120 /** 121 * Attempts to write back stores until all cache ports are used or the 122 * interface becomes blocked. 123 */ 124 void writebackStores(); 125 /** Same as above, but only for one thread. */ 126 void writebackStores(unsigned tid); 127 128 /** 129 * Squash instructions from a thread until the specified sequence number. 130 */ 131 void squash(const InstSeqNum &squashed_num, unsigned tid) 132 { thread[tid].squash(squashed_num); } 133 134 /** Returns whether or not there was a memory ordering violation. */ 135 bool violation(); 136 /** 137 * Returns whether or not there was a memory ordering violation for a 138 * specific thread. 139 */ 140 bool violation(unsigned tid) 141 { return thread[tid].violation(); } 142 143 /** Returns if a load is blocked due to the memory system for a specific 144 * thread. 145 */ 146 bool loadBlocked(unsigned tid) 147 { return thread[tid].loadBlocked(); } 148 149 bool isLoadBlockedHandled(unsigned tid) 150 { return thread[tid].isLoadBlockedHandled(); } 151 152 void setLoadBlockedHandled(unsigned tid) 153 { thread[tid].setLoadBlockedHandled(); } 154 155 /** Gets the instruction that caused the memory ordering violation. */ 156 DynInstPtr getMemDepViolator(unsigned tid) 157 { return thread[tid].getMemDepViolator(); } 158 159 /** Returns the head index of the load queue for a specific thread. */ 160 int getLoadHead(unsigned tid) 161 { return thread[tid].getLoadHead(); } 162 163 /** Returns the sequence number of the head of the load queue. */ 164 InstSeqNum getLoadHeadSeqNum(unsigned tid) 165 { 166 return thread[tid].getLoadHeadSeqNum(); 167 } 168 169 /** Returns the head index of the store queue. */ 170 int getStoreHead(unsigned tid) 171 { return thread[tid].getStoreHead(); } 172 173 /** Returns the sequence number of the head of the store queue. */ 174 InstSeqNum getStoreHeadSeqNum(unsigned tid) 175 { 176 return thread[tid].getStoreHeadSeqNum(); 177 } 178 179 /** Returns the number of instructions in all of the queues. */ 180 int getCount(); 181 /** Returns the number of instructions in the queues of one thread. */ 182 int getCount(unsigned tid) 183 { return thread[tid].getCount(); } 184 185 /** Returns the total number of loads in the load queue. */ 186 int numLoads(); 187 /** Returns the total number of loads for a single thread. */ 188 int numLoads(unsigned tid) 189 { return thread[tid].numLoads(); } 190 191 /** Returns the total number of stores in the store queue. */ 192 int numStores(); 193 /** Returns the total number of stores for a single thread. */ 194 int numStores(unsigned tid) 195 { return thread[tid].numStores(); } 196 197 /** Returns the total number of loads that are ready. */ 198 int numLoadsReady(); 199 /** Returns the number of loads that are ready for a single thread. */ 200 int numLoadsReady(unsigned tid) 201 { return thread[tid].numLoadsReady(); } 202 203 /** Returns the number of free entries. */ 204 unsigned numFreeEntries(); 205 /** Returns the number of free entries for a specific thread. */ 206 unsigned numFreeEntries(unsigned tid); 207 208 /** Returns if the LSQ is full (either LQ or SQ is full). */ 209 bool isFull(); 210 /** 211 * Returns if the LSQ is full for a specific thread (either LQ or SQ is 212 * full). 213 */ 214 bool isFull(unsigned tid); 215 216 /** Returns if any of the LQs are full. */ 217 bool lqFull(); 218 /** Returns if the LQ of a given thread is full. */ 219 bool lqFull(unsigned tid); 220 221 /** Returns if any of the SQs are full. */ 222 bool sqFull(); 223 /** Returns if the SQ of a given thread is full. */ 224 bool sqFull(unsigned tid); 225 226 /** 227 * Returns if the LSQ is stalled due to a memory operation that must be 228 * replayed. 229 */ 230 bool isStalled(); 231 /** 232 * Returns if the LSQ of a specific thread is stalled due to a memory 233 * operation that must be replayed. 234 */ 235 bool isStalled(unsigned tid); 236 237 /** Returns whether or not there are any stores to write back to memory. */ 238 bool hasStoresToWB(); 239 240 /** Returns whether or not a specific thread has any stores to write back 241 * to memory. 242 */ 243 bool hasStoresToWB(unsigned tid) 244 { return thread[tid].hasStoresToWB(); } 245 246 /** Returns the number of stores a specific thread has to write back. */ 247 int numStoresToWB(unsigned tid) 248 { return thread[tid].numStoresToWB(); } 249 250 /** Returns if the LSQ will write back to memory this cycle. */ 251 bool willWB(); 252 /** Returns if the LSQ of a specific thread will write back to memory this 253 * cycle. 254 */ 255 bool willWB(unsigned tid) 256 { return thread[tid].willWB(); } 257 258 /** Returns if the cache is currently blocked. */ 259 bool cacheBlocked() 260 { return retryTid != -1; } 261 262 /** Sets the retry thread id, indicating that one of the LSQUnits 263 * tried to access the cache but the cache was blocked. */ 264 void setRetryTid(int tid) 265 { retryTid = tid; } 266 267 /** Debugging function to print out all instructions. */ 268 void dumpInsts(); 269 /** Debugging function to print out instructions from a specific thread. */ 270 void dumpInsts(unsigned tid) 271 { thread[tid].dumpInsts(); } 272 273 /** Executes a read operation, using the load specified at the load index. */ 274 template <class T> 275 Fault read(RequestPtr req, T &data, int load_idx); 276 277 /** Executes a store operation, using the store specified at the store 278 * index. 279 */ 280 template <class T> 281 Fault write(RequestPtr req, T &data, int store_idx); 282 283 /** The CPU pointer. */ 284 O3CPU *cpu; 285 286 /** The IEW stage pointer. */ 287 IEW *iewStage; 288 289 /** DcachePort class for this LSQ. Handles doing the 290 * communication with the cache/memory. 291 */ 292 class DcachePort : public Port 293 { 294 protected: 295 /** Pointer to LSQ. */ 296 LSQ *lsq; 297 298 public: 299 /** Default constructor. */ 300 DcachePort(LSQ *_lsq) 301 : Port(_lsq->name() + "-dport"), lsq(_lsq) 302 { } 303 304 bool snoopRangeSent; 305 306 virtual void setPeer(Port *port); 307 308 protected: 309 /** Atomic version of receive. Panics. */ 310 virtual Tick recvAtomic(PacketPtr pkt); 311 312 /** Functional version of receive. Panics. */ 313 virtual void recvFunctional(PacketPtr pkt); 314 315 /** Receives status change. Other than range changing, panics. */ 316 virtual void recvStatusChange(Status status); 317 318 /** Returns the address ranges of this device. */ 319 virtual void getDeviceAddressRanges(AddrRangeList &resp, 320 bool &snoop) 321 { resp.clear(); snoop = true; } 322 323 /** Timing version of receive. Handles writing back and 324 * completing the load or store that has returned from 325 * memory. */ 326 virtual bool recvTiming(PacketPtr pkt); 327 328 /** Handles doing a retry of the previous send. */ 329 virtual void recvRetry(); 330 }; 331 332 /** D-cache port. */ 333 DcachePort dcachePort; 334 335#if FULL_SYSTEM 336 /** Tell the CPU to update the Phys and Virt ports. */ 337 void updateMemPorts() { cpu->updateMemPorts(); } 338#endif 339 340 protected: 341 /** The LSQ policy for SMT mode. */ 342 LSQPolicy lsqPolicy; 343 344 /** The LSQ units for individual threads. */ 345 LSQUnit thread[Impl::MaxThreads]; 346 347 /** List of Active Threads in System. */ 348 std::list<unsigned> *activeThreads; 349 350 /** Total Size of LQ Entries. */ 351 unsigned LQEntries; 352 /** Total Size of SQ Entries. */ 353 unsigned SQEntries; 354 355 /** Max LQ Size - Used to Enforce Sharing Policies. */ 356 unsigned maxLQEntries; 357 358 /** Max SQ Size - Used to Enforce Sharing Policies. */ 359 unsigned maxSQEntries; 360 361 /** Number of Threads. */ 362 unsigned numThreads; 363 364 /** The thread id of the LSQ Unit that is currently waiting for a 365 * retry. */ 366 int retryTid; 367}; 368 369template <class Impl> 370template <class T> 371Fault 372LSQ<Impl>::read(RequestPtr req, T &data, int load_idx) 373{ 374 unsigned tid = req->getThreadNum(); 375 376 return thread[tid].read(req, data, load_idx); 377} 378 379template <class Impl> 380template <class T> 381Fault 382LSQ<Impl>::write(RequestPtr req, T &data, int store_idx) 383{ 384 unsigned tid = req->getThreadNum(); 385 386 return thread[tid].write(req, data, store_idx); 387} 388 389#endif // __CPU_O3_LSQ_HH__ 390