2 * Copyright (c) 2010 ARM Limited
5 * The license below extends only to copyright in the software and shall
6 * not be construed as granting a license to any other intellectual
7 * property including but not limited to intellectual property relating
8 * to a hardware implementation of the functionality of the software
9 * licensed hereunder. You may use the software subject to the license
10 * terms below provided that you ensure that this notice is replicated
11 * unmodified and in its entirety in all distributions of the software,
12 * modified or unmodified, in source code or in binary form.
14 * Copyright (c) 2004-2006 The Regents of The University of Michigan
15 * All rights reserved.
17 * Redistribution and use in source and binary forms, with or without
18 * modification, are permitted provided that the following conditions are
19 * met: redistributions of source code must retain the above copyright
20 * notice, this list of conditions and the following disclaimer;
21 * redistributions in binary form must reproduce the above copyright
22 * notice, this list of conditions and the following disclaimer in the
23 * documentation and/or other materials provided with the distribution;
24 * neither the name of the copyright holders nor the names of its
25 * contributors may be used to endorse or promote products derived from
26 * this software without specific prior written permission.
28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
44 #ifndef __CPU_O3_FETCH_HH__
45 #define __CPU_O3_FETCH_HH__
47 #include "arch/predecoder.hh"
48 #include "arch/utility.hh"
49 #include "base/statistics.hh"
50 #include "config/the_isa.hh"
51 #include "cpu/pc_event.hh"
52 #include "cpu/timebuf.hh"
53 #include "cpu/translation.hh"
54 #include "mem/packet.hh"
55 #include "mem/port.hh"
56 #include "sim/eventq.hh"
58 class DerivO3CPUParams;
61 * DefaultFetch class handles both single threaded and SMT fetch. Its
62 * width is specified by the parameters; each cycle it tries to fetch
63 * that many instructions. It supports using a branch predictor to
64 * predict direction and targets.
65 * It supports the idling functionality of the CPU by indicating to
66 * the CPU when it is active and inactive.
72 /** Typedefs from Impl. */
73 typedef typename Impl::CPUPol CPUPol;
74 typedef typename Impl::DynInst DynInst;
75 typedef typename Impl::DynInstPtr DynInstPtr;
76 typedef typename Impl::O3CPU O3CPU;
78 /** Typedefs from the CPU policy. */
79 typedef typename CPUPol::BPredUnit BPredUnit;
80 typedef typename CPUPol::FetchStruct FetchStruct;
81 typedef typename CPUPol::TimeStruct TimeStruct;
83 /** Typedefs from ISA. */
84 typedef TheISA::MachInst MachInst;
85 typedef TheISA::ExtMachInst ExtMachInst;
87 /** IcachePort class for DefaultFetch. Handles doing the
88 * communication with the cache/memory.
90 class IcachePort : public Port
93 /** Pointer to fetch. */
94 DefaultFetch<Impl> *fetch;
97 /** Default constructor. */
98 IcachePort(DefaultFetch<Impl> *_fetch)
99 : Port(_fetch->name() + "-iport", _fetch->cpu), fetch(_fetch)
104 virtual void setPeer(Port *port);
107 /** Atomic version of receive. Panics. */
108 virtual Tick recvAtomic(PacketPtr pkt);
110 /** Functional version of receive. Panics. */
111 virtual void recvFunctional(PacketPtr pkt);
113 /** Receives status change. Other than range changing, panics. */
114 virtual void recvStatusChange(Status status);
116 /** Returns the address ranges of this device. */
117 virtual void getDeviceAddressRanges(AddrRangeList &resp,
119 { resp.clear(); snoop = true; }
121 /** Timing version of receive. Handles setting fetch to the
122 * proper status to start fetching. */
123 virtual bool recvTiming(PacketPtr pkt);
125 /** Handles doing a retry of a failed fetch. */
126 virtual void recvRetry();
129 class FetchTranslation : public BaseTLB::Translation
132 DefaultFetch<Impl> *fetch;
135 FetchTranslation(DefaultFetch<Impl> *_fetch)
144 finish(Fault fault, RequestPtr req, ThreadContext *tc,
147 assert(mode == BaseTLB::Execute);
148 fetch->finishTranslation(fault, req);
154 /* Event to delay delivery of a fetch translation result in case of
155 * a fault and the nop to carry the fault cannot be generated
157 class FinishTranslationEvent : public Event
160 DefaultFetch<Impl> *fetch;
165 FinishTranslationEvent(DefaultFetch<Impl> *_fetch)
169 void setFault(Fault _fault)
174 void setReq(RequestPtr _req)
179 /** Process the delayed finish translation */
182 assert(fetch->numInst < fetch->fetchWidth);
183 fetch->finishTranslation(fault, req);
186 const char *description() const
188 return "FullO3CPU FetchFinishTranslation";
193 /** Overall fetch status. Used to determine if the CPU can
194 * deschedule itsef due to a lack of activity.
201 /** Individual thread status. */
214 IcacheAccessComplete,
218 /** Fetching Policy, Add new policies here.*/
231 /** Per-thread status. */
232 ThreadStatus fetchStatus[Impl::MaxThreads];
235 FetchPriority fetchPolicy;
237 /** List that has the threads organized by priority. */
238 std::list<ThreadID> priorityList;
241 /** DefaultFetch constructor. */
242 DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params);
244 /** Returns the name of fetch. */
245 std::string name() const;
247 /** Registers statistics. */
250 /** Returns the icache port. */
251 Port *getIcachePort() { return icachePort; }
253 /** Sets the main backwards communication time buffer pointer. */
254 void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer);
256 /** Sets pointer to list of active threads. */
257 void setActiveThreads(std::list<ThreadID> *at_ptr);
259 /** Sets pointer to time buffer used to communicate to the next stage. */
260 void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr);
262 /** Initialize stage. */
265 /** Tells the fetch stage that the Icache is set. */
268 /** Processes cache completion event. */
269 void processCacheCompletion(PacketPtr pkt);
271 /** Begins the drain of the fetch stage. */
274 /** Resumes execution after a drain. */
277 /** Tells fetch stage to prepare to be switched out. */
280 /** Takes over from another CPU's thread. */
283 /** Checks if the fetch stage is switched out. */
284 bool isSwitchedOut() { return switchedOut; }
286 /** Tells fetch to wake up from a quiesce instruction. */
287 void wakeFromQuiesce();
290 /** Changes the status of this stage to active, and indicates this
293 inline void switchToActive();
295 /** Changes the status of this stage to inactive, and indicates
298 inline void switchToInactive();
301 * Looks up in the branch predictor to see if the next PC should be
302 * either next PC+=MachInst or a branch target.
303 * @param next_PC Next PC variable passed in by reference. It is
304 * expected to be set to the current PC; it will be updated with what
305 * the next PC will be.
306 * @param next_NPC Used for ISAs which use delay slots.
307 * @return Whether or not a branch was predicted as taken.
309 bool lookupAndUpdateNextPC(DynInstPtr &inst, TheISA::PCState &pc);
312 * Fetches the cache line that contains fetch_PC. Returns any
313 * fault that happened. Puts the data into the class variable
315 * @param vaddr The memory address that is being fetched from.
316 * @param ret_fault The fault reference that will be set to the result of
318 * @param tid Thread id.
319 * @param pc The actual PC of the current instruction.
320 * @return Any fault that occured.
322 bool fetchCacheLine(Addr vaddr, ThreadID tid, Addr pc);
323 void finishTranslation(Fault fault, RequestPtr mem_req);
326 /** Check if an interrupt is pending and that we need to handle
329 checkInterrupt(Addr pc)
331 return (interruptPending && (THE_ISA != ALPHA_ISA || !(pc & 0x3)));
334 /** Squashes a specific thread and resets the PC. */
335 inline void doSquash(const TheISA::PCState &newPC,
336 const DynInstPtr squashInst, ThreadID tid);
338 /** Squashes a specific thread and resets the PC. Also tells the CPU to
339 * remove any instructions between fetch and decode that should be sqaushed.
341 void squashFromDecode(const TheISA::PCState &newPC,
342 const DynInstPtr squashInst,
343 const InstSeqNum seq_num, ThreadID tid);
345 /** Checks if a thread is stalled. */
346 bool checkStall(ThreadID tid) const;
348 /** Updates overall fetch stage status; to be called at the end of each
350 FetchStatus updateFetchStatus();
353 /** Squashes a specific thread and resets the PC. Also tells the CPU to
354 * remove any instructions that are not in the ROB. The source of this
355 * squash should be the commit stage.
357 void squash(const TheISA::PCState &newPC, const InstSeqNum seq_num,
358 DynInstPtr squashInst, ThreadID tid);
360 /** Ticks the fetch stage, processing all inputs signals and fetching
361 * as many instructions as possible.
365 /** Checks all input signals and updates the status as necessary.
366 * @return: Returns if the status has changed due to input signals.
368 bool checkSignalsAndUpdate(ThreadID tid);
370 /** Does the actual fetching of instructions and passing them on to the
372 * @param status_change fetch() sets this variable if there was a status
373 * change (ie switching to IcacheMissStall).
375 void fetch(bool &status_change);
377 /** Align a PC to the start of an I-cache block. */
378 Addr icacheBlockAlignPC(Addr addr)
380 return (addr & ~(cacheBlkMask));
384 DynInstPtr buildInst(ThreadID tid, StaticInstPtr staticInst,
385 StaticInstPtr curMacroop, TheISA::PCState thisPC,
386 TheISA::PCState nextPC, bool trace);
388 /** Handles retrying the fetch access. */
391 /** Returns the appropriate thread to fetch, given the fetch policy. */
392 ThreadID getFetchingThread(FetchPriority &fetch_priority);
394 /** Returns the appropriate thread to fetch using a round robin policy. */
395 ThreadID roundRobin();
397 /** Returns the appropriate thread to fetch using the IQ count policy. */
400 /** Returns the appropriate thread to fetch using the LSQ count policy. */
403 /** Returns the appropriate thread to fetch using the branch count
405 ThreadID branchCount();
407 /** Pipeline the next I-cache access to the current one. */
408 void pipelineIcacheAccesses(ThreadID tid);
410 /** Profile the reasons of fetch stall. */
411 void profileStall(ThreadID tid);
414 /** Pointer to the O3CPU. */
417 /** Time buffer interface. */
418 TimeBuffer<TimeStruct> *timeBuffer;
420 /** Wire to get decode's information from backwards time buffer. */
421 typename TimeBuffer<TimeStruct>::wire fromDecode;
423 /** Wire to get rename's information from backwards time buffer. */
424 typename TimeBuffer<TimeStruct>::wire fromRename;
426 /** Wire to get iew's information from backwards time buffer. */
427 typename TimeBuffer<TimeStruct>::wire fromIEW;
429 /** Wire to get commit's information from backwards time buffer. */
430 typename TimeBuffer<TimeStruct>::wire fromCommit;
432 /** Internal fetch instruction queue. */
433 TimeBuffer<FetchStruct> *fetchQueue;
435 //Might be annoying how this name is different than the queue.
436 /** Wire used to write any information heading to decode. */
437 typename TimeBuffer<FetchStruct>::wire toDecode;
439 /** Icache interface. */
440 IcachePort *icachePort;
443 BPredUnit branchPred;
446 TheISA::Predecoder predecoder;
448 TheISA::PCState pc[Impl::MaxThreads];
450 Addr fetchOffset[Impl::MaxThreads];
452 StaticInstPtr macroop[Impl::MaxThreads];
454 /** Can the fetch stage redirect from an interrupt on this instruction? */
455 bool delayedCommit[Impl::MaxThreads];
457 /** Memory request used to access cache. */
458 RequestPtr memReq[Impl::MaxThreads];
460 /** Variable that tracks if fetch has written to the time buffer this
461 * cycle. Used to tell CPU if there is activity this cycle.
463 bool wroteToTimeBuffer;
465 /** Tracks how many instructions has been fetched this cycle. */
468 /** Source of possible stalls. */
476 /** Tracks which stages are telling fetch to stall. */
477 Stalls stalls[Impl::MaxThreads];
479 /** Decode to fetch delay, in ticks. */
480 unsigned decodeToFetchDelay;
482 /** Rename to fetch delay, in ticks. */
483 unsigned renameToFetchDelay;
485 /** IEW to fetch delay, in ticks. */
486 unsigned iewToFetchDelay;
488 /** Commit to fetch delay, in ticks. */
489 unsigned commitToFetchDelay;
491 /** The width of fetch in instructions. */
494 /** Is the cache blocked? If so no threads can access it. */
497 /** The packet that is waiting to be retried. */
500 /** The thread that is waiting on the cache to tell fetch to retry. */
503 /** Cache block size. */
506 /** Mask to get a cache block's address. */
509 /** The cache line being fetched. */
510 uint8_t *cacheData[Impl::MaxThreads];
512 /** The PC of the cacheline that has been loaded. */
513 Addr cacheDataPC[Impl::MaxThreads];
515 /** Whether or not the cache data is valid. */
516 bool cacheDataValid[Impl::MaxThreads];
518 /** Size of instructions. */
521 /** Icache stall statistics. */
522 Counter lastIcacheStall[Impl::MaxThreads];
524 /** List of Active Threads */
525 std::list<ThreadID> *activeThreads;
527 /** Number of threads. */
530 /** Number of threads that are actively fetching. */
531 ThreadID numFetchingThreads;
533 /** Thread ID being fetched. */
534 ThreadID threadFetched;
536 /** Checks if there is an interrupt pending. If there is, fetch
537 * must stop once it is not fetching PAL instructions.
539 bool interruptPending;
541 /** Is there a drain pending. */
544 /** Records if fetch is switched out. */
547 /** Set to true if a pipelined I-cache request should be issued. */
548 bool issuePipelinedIfetch[Impl::MaxThreads];
550 /** Event used to delay fault generation of translation faults */
551 FinishTranslationEvent finishTranslationEvent;
553 // @todo: Consider making these vectors and tracking on a per thread basis.
554 /** Stat for total number of cycles stalled due to an icache miss. */
555 Stats::Scalar icacheStallCycles;
556 /** Stat for total number of fetched instructions. */
557 Stats::Scalar fetchedInsts;
558 /** Total number of fetched branches. */
559 Stats::Scalar fetchedBranches;
560 /** Stat for total number of predicted branches. */
561 Stats::Scalar predictedBranches;
562 /** Stat for total number of cycles spent fetching. */
563 Stats::Scalar fetchCycles;
564 /** Stat for total number of cycles spent squashing. */
565 Stats::Scalar fetchSquashCycles;
566 /** Stat for total number of cycles spent waiting for translation */
567 Stats::Scalar fetchTlbCycles;
568 /** Stat for total number of cycles spent blocked due to other stages in
571 Stats::Scalar fetchIdleCycles;
572 /** Total number of cycles spent blocked. */
573 Stats::Scalar fetchBlockedCycles;
574 /** Total number of cycles spent in any other state. */
575 Stats::Scalar fetchMiscStallCycles;
576 /** Total number of cycles spent in waiting for drains. */
577 Stats::Scalar fetchPendingDrainCycles;
578 /** Total number of stall cycles caused by no active threads to run. */
579 Stats::Scalar fetchNoActiveThreadStallCycles;
580 /** Total number of stall cycles caused by pending traps. */
581 Stats::Scalar fetchPendingTrapStallCycles;
582 /** Total number of stall cycles caused by pending quiesce instructions. */
583 Stats::Scalar fetchPendingQuiesceStallCycles;
584 /** Total number of stall cycles caused by I-cache wait retrys. */
585 Stats::Scalar fetchIcacheWaitRetryStallCycles;
586 /** Stat for total number of fetched cache lines. */
587 Stats::Scalar fetchedCacheLines;
588 /** Total number of outstanding icache accesses that were dropped
591 Stats::Scalar fetchIcacheSquashes;
592 /** Total number of outstanding tlb accesses that were dropped
595 Stats::Scalar fetchTlbSquashes;
596 /** Distribution of number of instructions fetched each cycle. */
597 Stats::Distribution fetchNisnDist;
598 /** Rate of how often fetch was idle. */
599 Stats::Formula idleRate;
600 /** Number of branch fetches per cycle. */
601 Stats::Formula branchRate;
602 /** Number of instruction fetched per cycle. */
603 Stats::Formula fetchRate;
606 #endif //__CPU_O3_FETCH_HH__