fetch.hh revision 7720:65d338a8dba4
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 *          Korey Sewell
30 */
31
32#ifndef __CPU_O3_FETCH_HH__
33#define __CPU_O3_FETCH_HH__
34
35#include "arch/utility.hh"
36#include "arch/predecoder.hh"
37#include "base/statistics.hh"
38#include "base/timebuf.hh"
39#include "config/the_isa.hh"
40#include "cpu/pc_event.hh"
41#include "mem/packet.hh"
42#include "mem/port.hh"
43#include "sim/eventq.hh"
44
45class DerivO3CPUParams;
46
47/**
48 * DefaultFetch class handles both single threaded and SMT fetch. Its
49 * width is specified by the parameters; each cycle it tries to fetch
50 * that many instructions. It supports using a branch predictor to
51 * predict direction and targets.
52 * It supports the idling functionality of the CPU by indicating to
53 * the CPU when it is active and inactive.
54 */
55template <class Impl>
56class DefaultFetch
57{
58  public:
59    /** Typedefs from Impl. */
60    typedef typename Impl::CPUPol CPUPol;
61    typedef typename Impl::DynInst DynInst;
62    typedef typename Impl::DynInstPtr DynInstPtr;
63    typedef typename Impl::O3CPU O3CPU;
64
65    /** Typedefs from the CPU policy. */
66    typedef typename CPUPol::BPredUnit BPredUnit;
67    typedef typename CPUPol::FetchStruct FetchStruct;
68    typedef typename CPUPol::TimeStruct TimeStruct;
69
70    /** Typedefs from ISA. */
71    typedef TheISA::MachInst MachInst;
72    typedef TheISA::ExtMachInst ExtMachInst;
73
74    /** IcachePort class for DefaultFetch.  Handles doing the
75     * communication with the cache/memory.
76     */
77    class IcachePort : public Port
78    {
79      protected:
80        /** Pointer to fetch. */
81        DefaultFetch<Impl> *fetch;
82
83      public:
84        /** Default constructor. */
85        IcachePort(DefaultFetch<Impl> *_fetch)
86            : Port(_fetch->name() + "-iport", _fetch->cpu), fetch(_fetch)
87        { }
88
89        bool snoopRangeSent;
90
91        virtual void setPeer(Port *port);
92
93      protected:
94        /** Atomic version of receive.  Panics. */
95        virtual Tick recvAtomic(PacketPtr pkt);
96
97        /** Functional version of receive.  Panics. */
98        virtual void recvFunctional(PacketPtr pkt);
99
100        /** Receives status change.  Other than range changing, panics. */
101        virtual void recvStatusChange(Status status);
102
103        /** Returns the address ranges of this device. */
104        virtual void getDeviceAddressRanges(AddrRangeList &resp,
105                                            bool &snoop)
106        { resp.clear(); snoop = true; }
107
108        /** Timing version of receive.  Handles setting fetch to the
109         * proper status to start fetching. */
110        virtual bool recvTiming(PacketPtr pkt);
111
112        /** Handles doing a retry of a failed fetch. */
113        virtual void recvRetry();
114    };
115
116
117  public:
118    /** Overall fetch status. Used to determine if the CPU can
119     * deschedule itsef due to a lack of activity.
120     */
121    enum FetchStatus {
122        Active,
123        Inactive
124    };
125
126    /** Individual thread status. */
127    enum ThreadStatus {
128        Running,
129        Idle,
130        Squashing,
131        Blocked,
132        Fetching,
133        TrapPending,
134        QuiescePending,
135        SwitchOut,
136        IcacheWaitResponse,
137        IcacheWaitRetry,
138        IcacheAccessComplete
139    };
140
141    /** Fetching Policy, Add new policies here.*/
142    enum FetchPriority {
143        SingleThread,
144        RoundRobin,
145        Branch,
146        IQ,
147        LSQ
148    };
149
150  private:
151    /** Fetch status. */
152    FetchStatus _status;
153
154    /** Per-thread status. */
155    ThreadStatus fetchStatus[Impl::MaxThreads];
156
157    /** Fetch policy. */
158    FetchPriority fetchPolicy;
159
160    /** List that has the threads organized by priority. */
161    std::list<ThreadID> priorityList;
162
163  public:
164    /** DefaultFetch constructor. */
165    DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params);
166
167    /** Returns the name of fetch. */
168    std::string name() const;
169
170    /** Registers statistics. */
171    void regStats();
172
173    /** Returns the icache port. */
174    Port *getIcachePort() { return icachePort; }
175
176    /** Sets the main backwards communication time buffer pointer. */
177    void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer);
178
179    /** Sets pointer to list of active threads. */
180    void setActiveThreads(std::list<ThreadID> *at_ptr);
181
182    /** Sets pointer to time buffer used to communicate to the next stage. */
183    void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr);
184
185    /** Initialize stage. */
186    void initStage();
187
188    /** Tells the fetch stage that the Icache is set. */
189    void setIcache();
190
191    /** Processes cache completion event. */
192    void processCacheCompletion(PacketPtr pkt);
193
194    /** Begins the drain of the fetch stage. */
195    bool drain();
196
197    /** Resumes execution after a drain. */
198    void resume();
199
200    /** Tells fetch stage to prepare to be switched out. */
201    void switchOut();
202
203    /** Takes over from another CPU's thread. */
204    void takeOverFrom();
205
206    /** Checks if the fetch stage is switched out. */
207    bool isSwitchedOut() { return switchedOut; }
208
209    /** Tells fetch to wake up from a quiesce instruction. */
210    void wakeFromQuiesce();
211
212  private:
213    /** Changes the status of this stage to active, and indicates this
214     * to the CPU.
215     */
216    inline void switchToActive();
217
218    /** Changes the status of this stage to inactive, and indicates
219     * this to the CPU.
220     */
221    inline void switchToInactive();
222
223    /**
224     * Looks up in the branch predictor to see if the next PC should be
225     * either next PC+=MachInst or a branch target.
226     * @param next_PC Next PC variable passed in by reference.  It is
227     * expected to be set to the current PC; it will be updated with what
228     * the next PC will be.
229     * @param next_NPC Used for ISAs which use delay slots.
230     * @return Whether or not a branch was predicted as taken.
231     */
232    bool lookupAndUpdateNextPC(DynInstPtr &inst, TheISA::PCState &pc);
233
234    /**
235     * Fetches the cache line that contains fetch_PC.  Returns any
236     * fault that happened.  Puts the data into the class variable
237     * cacheData.
238     * @param fetch_PC The PC address that is being fetched from.
239     * @param ret_fault The fault reference that will be set to the result of
240     * the icache access.
241     * @param tid Thread id.
242     * @return Any fault that occured.
243     */
244    bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, ThreadID tid);
245
246    /** Squashes a specific thread and resets the PC. */
247    inline void doSquash(const TheISA::PCState &newPC, ThreadID tid);
248
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.
251     */
252    void squashFromDecode(const TheISA::PCState &newPC,
253                          const InstSeqNum &seq_num, ThreadID tid);
254
255    /** Checks if a thread is stalled. */
256    bool checkStall(ThreadID tid) const;
257
258    /** Updates overall fetch stage status; to be called at the end of each
259     * cycle. */
260    FetchStatus updateFetchStatus();
261
262  public:
263    /** Squashes a specific thread and resets the PC. Also tells the CPU to
264     * remove any instructions that are not in the ROB. The source of this
265     * squash should be the commit stage.
266     */
267    void squash(const TheISA::PCState &newPC,
268                const InstSeqNum &seq_num, ThreadID tid);
269
270    /** Ticks the fetch stage, processing all inputs signals and fetching
271     * as many instructions as possible.
272     */
273    void tick();
274
275    /** Checks all input signals and updates the status as necessary.
276     *  @return: Returns if the status has changed due to input signals.
277     */
278    bool checkSignalsAndUpdate(ThreadID tid);
279
280    /** Does the actual fetching of instructions and passing them on to the
281     * next stage.
282     * @param status_change fetch() sets this variable if there was a status
283     * change (ie switching to IcacheMissStall).
284     */
285    void fetch(bool &status_change);
286
287    /** Align a PC to the start of an I-cache block. */
288    Addr icacheBlockAlignPC(Addr addr)
289    {
290        return (addr & ~(cacheBlkMask));
291    }
292
293  private:
294    /** Handles retrying the fetch access. */
295    void recvRetry();
296
297    /** Returns the appropriate thread to fetch, given the fetch policy. */
298    ThreadID getFetchingThread(FetchPriority &fetch_priority);
299
300    /** Returns the appropriate thread to fetch using a round robin policy. */
301    ThreadID roundRobin();
302
303    /** Returns the appropriate thread to fetch using the IQ count policy. */
304    ThreadID iqCount();
305
306    /** Returns the appropriate thread to fetch using the LSQ count policy. */
307    ThreadID lsqCount();
308
309    /** Returns the appropriate thread to fetch using the branch count
310     * policy. */
311    ThreadID branchCount();
312
313  private:
314    /** Pointer to the O3CPU. */
315    O3CPU *cpu;
316
317    /** Time buffer interface. */
318    TimeBuffer<TimeStruct> *timeBuffer;
319
320    /** Wire to get decode's information from backwards time buffer. */
321    typename TimeBuffer<TimeStruct>::wire fromDecode;
322
323    /** Wire to get rename's information from backwards time buffer. */
324    typename TimeBuffer<TimeStruct>::wire fromRename;
325
326    /** Wire to get iew's information from backwards time buffer. */
327    typename TimeBuffer<TimeStruct>::wire fromIEW;
328
329    /** Wire to get commit's information from backwards time buffer. */
330    typename TimeBuffer<TimeStruct>::wire fromCommit;
331
332    /** Internal fetch instruction queue. */
333    TimeBuffer<FetchStruct> *fetchQueue;
334
335    //Might be annoying how this name is different than the queue.
336    /** Wire used to write any information heading to decode. */
337    typename TimeBuffer<FetchStruct>::wire toDecode;
338
339    /** Icache interface. */
340    IcachePort *icachePort;
341
342    /** BPredUnit. */
343    BPredUnit branchPred;
344
345    /** Predecoder. */
346    TheISA::Predecoder predecoder;
347
348    TheISA::PCState pc[Impl::MaxThreads];
349
350    /** Memory request used to access cache. */
351    RequestPtr memReq[Impl::MaxThreads];
352
353    /** Variable that tracks if fetch has written to the time buffer this
354     * cycle. Used to tell CPU if there is activity this cycle.
355     */
356    bool wroteToTimeBuffer;
357
358    /** Tracks how many instructions has been fetched this cycle. */
359    int numInst;
360
361    /** Source of possible stalls. */
362    struct Stalls {
363        bool decode;
364        bool rename;
365        bool iew;
366        bool commit;
367    };
368
369    /** Tracks which stages are telling fetch to stall. */
370    Stalls stalls[Impl::MaxThreads];
371
372    /** Decode to fetch delay, in ticks. */
373    unsigned decodeToFetchDelay;
374
375    /** Rename to fetch delay, in ticks. */
376    unsigned renameToFetchDelay;
377
378    /** IEW to fetch delay, in ticks. */
379    unsigned iewToFetchDelay;
380
381    /** Commit to fetch delay, in ticks. */
382    unsigned commitToFetchDelay;
383
384    /** The width of fetch in instructions. */
385    unsigned fetchWidth;
386
387    /** Is the cache blocked?  If so no threads can access it. */
388    bool cacheBlocked;
389
390    /** The packet that is waiting to be retried. */
391    PacketPtr retryPkt;
392
393    /** The thread that is waiting on the cache to tell fetch to retry. */
394    ThreadID retryTid;
395
396    /** Cache block size. */
397    int cacheBlkSize;
398
399    /** Mask to get a cache block's address. */
400    Addr cacheBlkMask;
401
402    /** The cache line being fetched. */
403    uint8_t *cacheData[Impl::MaxThreads];
404
405    /** The PC of the cacheline that has been loaded. */
406    Addr cacheDataPC[Impl::MaxThreads];
407
408    /** Whether or not the cache data is valid. */
409    bool cacheDataValid[Impl::MaxThreads];
410
411    /** Size of instructions. */
412    int instSize;
413
414    /** Icache stall statistics. */
415    Counter lastIcacheStall[Impl::MaxThreads];
416
417    /** List of Active Threads */
418    std::list<ThreadID> *activeThreads;
419
420    /** Number of threads. */
421    ThreadID numThreads;
422
423    /** Number of threads that are actively fetching. */
424    ThreadID numFetchingThreads;
425
426    /** Thread ID being fetched. */
427    ThreadID threadFetched;
428
429    /** Checks if there is an interrupt pending.  If there is, fetch
430     * must stop once it is not fetching PAL instructions.
431     */
432    bool interruptPending;
433
434    /** Is there a drain pending. */
435    bool drainPending;
436
437    /** Records if fetch is switched out. */
438    bool switchedOut;
439
440    // @todo: Consider making these vectors and tracking on a per thread basis.
441    /** Stat for total number of cycles stalled due to an icache miss. */
442    Stats::Scalar icacheStallCycles;
443    /** Stat for total number of fetched instructions. */
444    Stats::Scalar fetchedInsts;
445    /** Total number of fetched branches. */
446    Stats::Scalar fetchedBranches;
447    /** Stat for total number of predicted branches. */
448    Stats::Scalar predictedBranches;
449    /** Stat for total number of cycles spent fetching. */
450    Stats::Scalar fetchCycles;
451    /** Stat for total number of cycles spent squashing. */
452    Stats::Scalar fetchSquashCycles;
453    /** Stat for total number of cycles spent blocked due to other stages in
454     * the pipeline.
455     */
456    Stats::Scalar fetchIdleCycles;
457    /** Total number of cycles spent blocked. */
458    Stats::Scalar fetchBlockedCycles;
459    /** Total number of cycles spent in any other state. */
460    Stats::Scalar fetchMiscStallCycles;
461    /** Stat for total number of fetched cache lines. */
462    Stats::Scalar fetchedCacheLines;
463    /** Total number of outstanding icache accesses that were dropped
464     * due to a squash.
465     */
466    Stats::Scalar fetchIcacheSquashes;
467    /** Distribution of number of instructions fetched each cycle. */
468    Stats::Distribution fetchNisnDist;
469    /** Rate of how often fetch was idle. */
470    Stats::Formula idleRate;
471    /** Number of branch fetches per cycle. */
472    Stats::Formula branchRate;
473    /** Number of instruction fetched per cycle. */
474    Stats::Formula fetchRate;
475};
476
477#endif //__CPU_O3_FETCH_HH__
478