2 * Copyright (c) 2004-2006 The Regents of The University of Michigan
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.
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.
32 #ifndef __CPU_O3_FETCH_HH__
33 #define __CPU_O3_FETCH_HH__
35 #include "arch/utility.hh"
36 #include "arch/predecoder.hh"
37 #include "base/statistics.hh"
38 #include "base/timebuf.hh"
39 #include "cpu/pc_event.hh"
40 #include "mem/packet.hh"
41 #include "mem/port.hh"
42 #include "sim/eventq.hh"
44 class DerivO3CPUParams;
47 * DefaultFetch class handles both single threaded and SMT fetch. Its
48 * width is specified by the parameters; each cycle it tries to fetch
49 * that many instructions. It supports using a branch predictor to
50 * predict direction and targets.
51 * It supports the idling functionality of the CPU by indicating to
52 * the CPU when it is active and inactive.
58 /** Typedefs from Impl. */
59 typedef typename Impl::CPUPol CPUPol;
60 typedef typename Impl::DynInst DynInst;
61 typedef typename Impl::DynInstPtr DynInstPtr;
62 typedef typename Impl::O3CPU O3CPU;
64 /** Typedefs from the CPU policy. */
65 typedef typename CPUPol::BPredUnit BPredUnit;
66 typedef typename CPUPol::FetchStruct FetchStruct;
67 typedef typename CPUPol::TimeStruct TimeStruct;
69 /** Typedefs from ISA. */
70 typedef TheISA::MachInst MachInst;
71 typedef TheISA::ExtMachInst ExtMachInst;
73 /** IcachePort class for DefaultFetch. Handles doing the
74 * communication with the cache/memory.
76 class IcachePort : public Port
79 /** Pointer to fetch. */
80 DefaultFetch<Impl> *fetch;
83 /** Default constructor. */
84 IcachePort(DefaultFetch<Impl> *_fetch)
85 : Port(_fetch->name() + "-iport", _fetch->cpu), fetch(_fetch)
90 virtual void setPeer(Port *port);
93 /** Atomic version of receive. Panics. */
94 virtual Tick recvAtomic(PacketPtr pkt);
96 /** Functional version of receive. Panics. */
97 virtual void recvFunctional(PacketPtr pkt);
99 /** Receives status change. Other than range changing, panics. */
100 virtual void recvStatusChange(Status status);
102 /** Returns the address ranges of this device. */
103 virtual void getDeviceAddressRanges(AddrRangeList &resp,
105 { resp.clear(); snoop = true; }
107 /** Timing version of receive. Handles setting fetch to the
108 * proper status to start fetching. */
109 virtual bool recvTiming(PacketPtr pkt);
111 /** Handles doing a retry of a failed fetch. */
112 virtual void recvRetry();
117 /** Overall fetch status. Used to determine if the CPU can
118 * deschedule itsef due to a lack of activity.
125 /** Individual thread status. */
140 /** Fetching Policy, Add new policies here.*/
153 /** Per-thread status. */
154 ThreadStatus fetchStatus[Impl::MaxThreads];
157 FetchPriority fetchPolicy;
159 /** List that has the threads organized by priority. */
160 std::list<unsigned> priorityList;
163 /** DefaultFetch constructor. */
164 DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params);
166 /** Returns the name of fetch. */
167 std::string name() const;
169 /** Registers statistics. */
172 /** Returns the icache port. */
173 Port *getIcachePort() { return icachePort; }
175 /** Sets the main backwards communication time buffer pointer. */
176 void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer);
178 /** Sets pointer to list of active threads. */
179 void setActiveThreads(std::list<unsigned> *at_ptr);
181 /** Sets pointer to time buffer used to communicate to the next stage. */
182 void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr);
184 /** Initialize stage. */
187 /** Tells the fetch stage that the Icache is set. */
190 /** Processes cache completion event. */
191 void processCacheCompletion(PacketPtr pkt);
193 /** Begins the drain of the fetch stage. */
196 /** Resumes execution after a drain. */
199 /** Tells fetch stage to prepare to be switched out. */
202 /** Takes over from another CPU's thread. */
205 /** Checks if the fetch stage is switched out. */
206 bool isSwitchedOut() { return switchedOut; }
208 /** Tells fetch to wake up from a quiesce instruction. */
209 void wakeFromQuiesce();
212 /** Changes the status of this stage to active, and indicates this
215 inline void switchToActive();
217 /** Changes the status of this stage to inactive, and indicates
220 inline void switchToInactive();
223 * Looks up in the branch predictor to see if the next PC should be
224 * either next PC+=MachInst or a branch target.
225 * @param next_PC Next PC variable passed in by reference. It is
226 * expected to be set to the current PC; it will be updated with what
227 * the next PC will be.
228 * @param next_NPC Used for ISAs which use delay slots.
229 * @return Whether or not a branch was predicted as taken.
231 bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, Addr &next_NPC, Addr &next_MicroPC);
234 * Fetches the cache line that contains fetch_PC. Returns any
235 * fault that happened. Puts the data into the class variable
237 * @param fetch_PC The PC address that is being fetched from.
238 * @param ret_fault The fault reference that will be set to the result of
240 * @param tid Thread id.
241 * @return Any fault that occured.
243 bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid);
245 /** Squashes a specific thread and resets the PC. */
246 inline void doSquash(const Addr &new_PC, const Addr &new_NPC,
247 const Addr &new_MicroPC, unsigned tid);
249 /** Squashes a specific thread and resets the PC. Also tells the CPU to
250 * remove any instructions between fetch and decode that should be sqaushed.
252 void squashFromDecode(const Addr &new_PC, const Addr &new_NPC,
253 const Addr &new_MicroPC,
254 const InstSeqNum &seq_num, unsigned tid);
256 /** Checks if a thread is stalled. */
257 bool checkStall(unsigned tid) const;
259 /** Updates overall fetch stage status; to be called at the end of each
261 FetchStatus updateFetchStatus();
264 /** Squashes a specific thread and resets the PC. Also tells the CPU to
265 * remove any instructions that are not in the ROB. The source of this
266 * squash should be the commit stage.
268 void squash(const Addr &new_PC, const Addr &new_NPC,
269 const Addr &new_MicroPC,
270 const InstSeqNum &seq_num, unsigned tid);
272 /** Ticks the fetch stage, processing all inputs signals and fetching
273 * as many instructions as possible.
277 /** Checks all input signals and updates the status as necessary.
278 * @return: Returns if the status has changed due to input signals.
280 bool checkSignalsAndUpdate(unsigned tid);
282 /** Does the actual fetching of instructions and passing them on to the
284 * @param status_change fetch() sets this variable if there was a status
285 * change (ie switching to IcacheMissStall).
287 void fetch(bool &status_change);
289 /** Align a PC to the start of an I-cache block. */
290 Addr icacheBlockAlignPC(Addr addr)
292 addr = TheISA::realPCToFetchPC(addr);
293 return (addr & ~(cacheBlkMask));
297 /** Handles retrying the fetch access. */
300 /** Returns the appropriate thread to fetch, given the fetch policy. */
301 int getFetchingThread(FetchPriority &fetch_priority);
303 /** Returns the appropriate thread to fetch using a round robin policy. */
306 /** Returns the appropriate thread to fetch using the IQ count policy. */
309 /** Returns the appropriate thread to fetch using the LSQ count policy. */
312 /** Returns the appropriate thread to fetch using the branch count policy. */
316 /** Pointer to the O3CPU. */
319 /** Time buffer interface. */
320 TimeBuffer<TimeStruct> *timeBuffer;
322 /** Wire to get decode's information from backwards time buffer. */
323 typename TimeBuffer<TimeStruct>::wire fromDecode;
325 /** Wire to get rename's information from backwards time buffer. */
326 typename TimeBuffer<TimeStruct>::wire fromRename;
328 /** Wire to get iew's information from backwards time buffer. */
329 typename TimeBuffer<TimeStruct>::wire fromIEW;
331 /** Wire to get commit's information from backwards time buffer. */
332 typename TimeBuffer<TimeStruct>::wire fromCommit;
334 /** Internal fetch instruction queue. */
335 TimeBuffer<FetchStruct> *fetchQueue;
337 //Might be annoying how this name is different than the queue.
338 /** Wire used to write any information heading to decode. */
339 typename TimeBuffer<FetchStruct>::wire toDecode;
341 /** Icache interface. */
342 IcachePort *icachePort;
345 BPredUnit branchPred;
348 TheISA::Predecoder predecoder;
350 /** Per-thread fetch PC. */
351 Addr PC[Impl::MaxThreads];
353 /** Per-thread fetch micro PC. */
354 Addr microPC[Impl::MaxThreads];
356 /** Per-thread next PC. */
357 Addr nextPC[Impl::MaxThreads];
359 /** Memory request used to access cache. */
360 RequestPtr memReq[Impl::MaxThreads];
362 /** Variable that tracks if fetch has written to the time buffer this
363 * cycle. Used to tell CPU if there is activity this cycle.
365 bool wroteToTimeBuffer;
367 /** Tracks how many instructions has been fetched this cycle. */
370 /** Source of possible stalls. */
378 /** Tracks which stages are telling fetch to stall. */
379 Stalls stalls[Impl::MaxThreads];
381 /** Decode to fetch delay, in ticks. */
382 unsigned decodeToFetchDelay;
384 /** Rename to fetch delay, in ticks. */
385 unsigned renameToFetchDelay;
387 /** IEW to fetch delay, in ticks. */
388 unsigned iewToFetchDelay;
390 /** Commit to fetch delay, in ticks. */
391 unsigned commitToFetchDelay;
393 /** The width of fetch in instructions. */
396 /** Is the cache blocked? If so no threads can access it. */
399 /** The packet that is waiting to be retried. */
402 /** The thread that is waiting on the cache to tell fetch to retry. */
405 /** Cache block size. */
408 /** Mask to get a cache block's address. */
411 /** The cache line being fetched. */
412 uint8_t *cacheData[Impl::MaxThreads];
414 /** The PC of the cacheline that has been loaded. */
415 Addr cacheDataPC[Impl::MaxThreads];
417 /** Whether or not the cache data is valid. */
418 bool cacheDataValid[Impl::MaxThreads];
420 /** Size of instructions. */
423 /** Icache stall statistics. */
424 Counter lastIcacheStall[Impl::MaxThreads];
426 /** List of Active Threads */
427 std::list<unsigned> *activeThreads;
429 /** Number of threads. */
432 /** Number of threads that are actively fetching. */
433 unsigned numFetchingThreads;
435 /** Thread ID being fetched. */
438 /** Checks if there is an interrupt pending. If there is, fetch
439 * must stop once it is not fetching PAL instructions.
441 bool interruptPending;
443 /** Is there a drain pending. */
446 /** Records if fetch is switched out. */
449 // @todo: Consider making these vectors and tracking on a per thread basis.
450 /** Stat for total number of cycles stalled due to an icache miss. */
451 Stats::Scalar icacheStallCycles;
452 /** Stat for total number of fetched instructions. */
453 Stats::Scalar fetchedInsts;
454 /** Total number of fetched branches. */
455 Stats::Scalar fetchedBranches;
456 /** Stat for total number of predicted branches. */
457 Stats::Scalar predictedBranches;
458 /** Stat for total number of cycles spent fetching. */
459 Stats::Scalar fetchCycles;
460 /** Stat for total number of cycles spent squashing. */
461 Stats::Scalar fetchSquashCycles;
462 /** Stat for total number of cycles spent blocked due to other stages in
465 Stats::Scalar fetchIdleCycles;
466 /** Total number of cycles spent blocked. */
467 Stats::Scalar fetchBlockedCycles;
468 /** Total number of cycles spent in any other state. */
469 Stats::Scalar fetchMiscStallCycles;
470 /** Stat for total number of fetched cache lines. */
471 Stats::Scalar fetchedCacheLines;
472 /** Total number of outstanding icache accesses that were dropped
475 Stats::Scalar fetchIcacheSquashes;
476 /** Distribution of number of instructions fetched each cycle. */
477 Stats::Distribution fetchNisnDist;
478 /** Rate of how often fetch was idle. */
479 Stats::Formula idleRate;
480 /** Number of branch fetches per cycle. */
481 Stats::Formula branchRate;
482 /** Number of instruction fetched per cycle. */
483 Stats::Formula fetchRate;
486 #endif //__CPU_O3_FETCH_HH__