cpu: fixed how O3 CPU executes an exit system call
[gem5.git] / src / cpu / o3 / fetch.hh
1 /*
2 * Copyright (c) 2010-2012, 2014 ARM Limited
3 * All rights reserved
4 *
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.
13 *
14 * Copyright (c) 2004-2006 The Regents of The University of Michigan
15 * All rights reserved.
16 *
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.
27 *
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.
39 *
40 * Authors: Kevin Lim
41 * Korey Sewell
42 */
43
44 #ifndef __CPU_O3_FETCH_HH__
45 #define __CPU_O3_FETCH_HH__
46
47 #include "arch/decoder.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/pred/bpred_unit.hh"
53 #include "cpu/timebuf.hh"
54 #include "cpu/translation.hh"
55 #include "enums/FetchPolicy.hh"
56 #include "mem/packet.hh"
57 #include "mem/port.hh"
58 #include "sim/eventq.hh"
59 #include "sim/probe/probe.hh"
60
61 struct DerivO3CPUParams;
62
63 /**
64 * DefaultFetch class handles both single threaded and SMT fetch. Its
65 * width is specified by the parameters; each cycle it tries to fetch
66 * that many instructions. It supports using a branch predictor to
67 * predict direction and targets.
68 * It supports the idling functionality of the CPU by indicating to
69 * the CPU when it is active and inactive.
70 */
71 template <class Impl>
72 class DefaultFetch
73 {
74 public:
75 /** Typedefs from Impl. */
76 typedef typename Impl::CPUPol CPUPol;
77 typedef typename Impl::DynInst DynInst;
78 typedef typename Impl::DynInstPtr DynInstPtr;
79 typedef typename Impl::O3CPU O3CPU;
80
81 /** Typedefs from the CPU policy. */
82 typedef typename CPUPol::FetchStruct FetchStruct;
83 typedef typename CPUPol::TimeStruct TimeStruct;
84
85 /** Typedefs from ISA. */
86 typedef TheISA::MachInst MachInst;
87
88 class FetchTranslation : public BaseTLB::Translation
89 {
90 protected:
91 DefaultFetch<Impl> *fetch;
92
93 public:
94 FetchTranslation(DefaultFetch<Impl> *_fetch)
95 : fetch(_fetch)
96 {}
97
98 void
99 markDelayed()
100 {}
101
102 void
103 finish(const Fault &fault, const RequestPtr &req, ThreadContext *tc,
104 BaseTLB::Mode mode)
105 {
106 assert(mode == BaseTLB::Execute);
107 fetch->finishTranslation(fault, req);
108 delete this;
109 }
110 };
111
112 private:
113 /* Event to delay delivery of a fetch translation result in case of
114 * a fault and the nop to carry the fault cannot be generated
115 * immediately */
116 class FinishTranslationEvent : public Event
117 {
118 private:
119 DefaultFetch<Impl> *fetch;
120 Fault fault;
121 RequestPtr req;
122
123 public:
124 FinishTranslationEvent(DefaultFetch<Impl> *_fetch)
125 : fetch(_fetch), req(nullptr)
126 {}
127
128 void setFault(Fault _fault)
129 {
130 fault = _fault;
131 }
132
133 void setReq(const RequestPtr &_req)
134 {
135 req = _req;
136 }
137
138 /** Process the delayed finish translation */
139 void process()
140 {
141 assert(fetch->numInst < fetch->fetchWidth);
142 fetch->finishTranslation(fault, req);
143 }
144
145 const char *description() const
146 {
147 return "FullO3CPU FetchFinishTranslation";
148 }
149 };
150
151 public:
152 /** Overall fetch status. Used to determine if the CPU can
153 * deschedule itsef due to a lack of activity.
154 */
155 enum FetchStatus {
156 Active,
157 Inactive
158 };
159
160 /** Individual thread status. */
161 enum ThreadStatus {
162 Running,
163 Idle,
164 Squashing,
165 Blocked,
166 Fetching,
167 TrapPending,
168 QuiescePending,
169 ItlbWait,
170 IcacheWaitResponse,
171 IcacheWaitRetry,
172 IcacheAccessComplete,
173 NoGoodAddr
174 };
175
176 private:
177 /** Fetch status. */
178 FetchStatus _status;
179
180 /** Per-thread status. */
181 ThreadStatus fetchStatus[Impl::MaxThreads];
182
183 /** Fetch policy. */
184 FetchPolicy fetchPolicy;
185
186 /** List that has the threads organized by priority. */
187 std::list<ThreadID> priorityList;
188
189 /** Probe points. */
190 ProbePointArg<DynInstPtr> *ppFetch;
191 /** To probe when a fetch request is successfully sent. */
192 ProbePointArg<RequestPtr> *ppFetchRequestSent;
193
194 public:
195 /** DefaultFetch constructor. */
196 DefaultFetch(O3CPU *_cpu, DerivO3CPUParams *params);
197
198 /** Returns the name of fetch. */
199 std::string name() const;
200
201 /** Registers statistics. */
202 void regStats();
203
204 /** Registers probes. */
205 void regProbePoints();
206
207 /** Sets the main backwards communication time buffer pointer. */
208 void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer);
209
210 /** Sets pointer to list of active threads. */
211 void setActiveThreads(std::list<ThreadID> *at_ptr);
212
213 /** Sets pointer to time buffer used to communicate to the next stage. */
214 void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr);
215
216 /** Initialize stage. */
217 void startupStage();
218
219 /** Clear all thread-specific states*/
220 void clearStates(ThreadID tid);
221
222 /** Handles retrying the fetch access. */
223 void recvReqRetry();
224
225 /** Processes cache completion event. */
226 void processCacheCompletion(PacketPtr pkt);
227
228 /** Resume after a drain. */
229 void drainResume();
230
231 /** Perform sanity checks after a drain. */
232 void drainSanityCheck() const;
233
234 /** Has the stage drained? */
235 bool isDrained() const;
236
237 /** Takes over from another CPU's thread. */
238 void takeOverFrom();
239
240 /**
241 * Stall the fetch stage after reaching a safe drain point.
242 *
243 * The CPU uses this method to stop fetching instructions from a
244 * thread that has been drained. The drain stall is different from
245 * all other stalls in that it is signaled instantly from the
246 * commit stage (without the normal communication delay) when it
247 * has reached a safe point to drain from.
248 */
249 void drainStall(ThreadID tid);
250
251 /** Tells fetch to wake up from a quiesce instruction. */
252 void wakeFromQuiesce();
253
254 /** For priority-based fetch policies, need to keep update priorityList */
255 void deactivateThread(ThreadID tid);
256 private:
257 /** Reset this pipeline stage */
258 void resetStage();
259
260 /** Changes the status of this stage to active, and indicates this
261 * to the CPU.
262 */
263 inline void switchToActive();
264
265 /** Changes the status of this stage to inactive, and indicates
266 * this to the CPU.
267 */
268 inline void switchToInactive();
269
270 /**
271 * Looks up in the branch predictor to see if the next PC should be
272 * either next PC+=MachInst or a branch target.
273 * @param next_PC Next PC variable passed in by reference. It is
274 * expected to be set to the current PC; it will be updated with what
275 * the next PC will be.
276 * @param next_NPC Used for ISAs which use delay slots.
277 * @return Whether or not a branch was predicted as taken.
278 */
279 bool lookupAndUpdateNextPC(const DynInstPtr &inst, TheISA::PCState &pc);
280
281 /**
282 * Fetches the cache line that contains the fetch PC. Returns any
283 * fault that happened. Puts the data into the class variable
284 * fetchBuffer, which may not hold the entire fetched cache line.
285 * @param vaddr The memory address that is being fetched from.
286 * @param ret_fault The fault reference that will be set to the result of
287 * the icache access.
288 * @param tid Thread id.
289 * @param pc The actual PC of the current instruction.
290 * @return Any fault that occured.
291 */
292 bool fetchCacheLine(Addr vaddr, ThreadID tid, Addr pc);
293 void finishTranslation(const Fault &fault, const RequestPtr &mem_req);
294
295
296 /** Check if an interrupt is pending and that we need to handle
297 */
298 bool
299 checkInterrupt(Addr pc)
300 {
301 return (interruptPending && (THE_ISA != ALPHA_ISA || !(pc & 0x3)));
302 }
303
304 /** Squashes a specific thread and resets the PC. */
305 inline void doSquash(const TheISA::PCState &newPC,
306 const DynInstPtr squashInst, ThreadID tid);
307
308 /** Squashes a specific thread and resets the PC. Also tells the CPU to
309 * remove any instructions between fetch and decode that should be sqaushed.
310 */
311 void squashFromDecode(const TheISA::PCState &newPC,
312 const DynInstPtr squashInst,
313 const InstSeqNum seq_num, ThreadID tid);
314
315 /** Checks if a thread is stalled. */
316 bool checkStall(ThreadID tid) const;
317
318 /** Updates overall fetch stage status; to be called at the end of each
319 * cycle. */
320 FetchStatus updateFetchStatus();
321
322 public:
323 /** Squashes a specific thread and resets the PC. Also tells the CPU to
324 * remove any instructions that are not in the ROB. The source of this
325 * squash should be the commit stage.
326 */
327 void squash(const TheISA::PCState &newPC, const InstSeqNum seq_num,
328 DynInstPtr squashInst, ThreadID tid);
329
330 /** Ticks the fetch stage, processing all inputs signals and fetching
331 * as many instructions as possible.
332 */
333 void tick();
334
335 /** Checks all input signals and updates the status as necessary.
336 * @return: Returns if the status has changed due to input signals.
337 */
338 bool checkSignalsAndUpdate(ThreadID tid);
339
340 /** Does the actual fetching of instructions and passing them on to the
341 * next stage.
342 * @param status_change fetch() sets this variable if there was a status
343 * change (ie switching to IcacheMissStall).
344 */
345 void fetch(bool &status_change);
346
347 /** Align a PC to the start of a fetch buffer block. */
348 Addr fetchBufferAlignPC(Addr addr)
349 {
350 return (addr & ~(fetchBufferMask));
351 }
352
353 /** The decoder. */
354 TheISA::Decoder *decoder[Impl::MaxThreads];
355
356 private:
357 DynInstPtr buildInst(ThreadID tid, StaticInstPtr staticInst,
358 StaticInstPtr curMacroop, TheISA::PCState thisPC,
359 TheISA::PCState nextPC, bool trace);
360
361 /** Returns the appropriate thread to fetch, given the fetch policy. */
362 ThreadID getFetchingThread();
363
364 /** Returns the appropriate thread to fetch using a round robin policy. */
365 ThreadID roundRobin();
366
367 /** Returns the appropriate thread to fetch using the IQ count policy. */
368 ThreadID iqCount();
369
370 /** Returns the appropriate thread to fetch using the LSQ count policy. */
371 ThreadID lsqCount();
372
373 /** Returns the appropriate thread to fetch using the branch count
374 * policy. */
375 ThreadID branchCount();
376
377 /** Pipeline the next I-cache access to the current one. */
378 void pipelineIcacheAccesses(ThreadID tid);
379
380 /** Profile the reasons of fetch stall. */
381 void profileStall(ThreadID tid);
382
383 private:
384 /** Pointer to the O3CPU. */
385 O3CPU *cpu;
386
387 /** Time buffer interface. */
388 TimeBuffer<TimeStruct> *timeBuffer;
389
390 /** Wire to get decode's information from backwards time buffer. */
391 typename TimeBuffer<TimeStruct>::wire fromDecode;
392
393 /** Wire to get rename's information from backwards time buffer. */
394 typename TimeBuffer<TimeStruct>::wire fromRename;
395
396 /** Wire to get iew's information from backwards time buffer. */
397 typename TimeBuffer<TimeStruct>::wire fromIEW;
398
399 /** Wire to get commit's information from backwards time buffer. */
400 typename TimeBuffer<TimeStruct>::wire fromCommit;
401
402 //Might be annoying how this name is different than the queue.
403 /** Wire used to write any information heading to decode. */
404 typename TimeBuffer<FetchStruct>::wire toDecode;
405
406 /** BPredUnit. */
407 BPredUnit *branchPred;
408
409 TheISA::PCState pc[Impl::MaxThreads];
410
411 Addr fetchOffset[Impl::MaxThreads];
412
413 StaticInstPtr macroop[Impl::MaxThreads];
414
415 /** Can the fetch stage redirect from an interrupt on this instruction? */
416 bool delayedCommit[Impl::MaxThreads];
417
418 /** Memory request used to access cache. */
419 RequestPtr memReq[Impl::MaxThreads];
420
421 /** Variable that tracks if fetch has written to the time buffer this
422 * cycle. Used to tell CPU if there is activity this cycle.
423 */
424 bool wroteToTimeBuffer;
425
426 /** Tracks how many instructions has been fetched this cycle. */
427 int numInst;
428
429 /** Source of possible stalls. */
430 struct Stalls {
431 bool decode;
432 bool drain;
433 };
434
435 /** Tracks which stages are telling fetch to stall. */
436 Stalls stalls[Impl::MaxThreads];
437
438 /** Decode to fetch delay. */
439 Cycles decodeToFetchDelay;
440
441 /** Rename to fetch delay. */
442 Cycles renameToFetchDelay;
443
444 /** IEW to fetch delay. */
445 Cycles iewToFetchDelay;
446
447 /** Commit to fetch delay. */
448 Cycles commitToFetchDelay;
449
450 /** The width of fetch in instructions. */
451 unsigned fetchWidth;
452
453 /** The width of decode in instructions. */
454 unsigned decodeWidth;
455
456 /** Is the cache blocked? If so no threads can access it. */
457 bool cacheBlocked;
458
459 /** The packet that is waiting to be retried. */
460 PacketPtr retryPkt;
461
462 /** The thread that is waiting on the cache to tell fetch to retry. */
463 ThreadID retryTid;
464
465 /** Cache block size. */
466 unsigned int cacheBlkSize;
467
468 /** The size of the fetch buffer in bytes. The fetch buffer
469 * itself may be smaller than a cache line.
470 */
471 unsigned fetchBufferSize;
472
473 /** Mask to align a fetch address to a fetch buffer boundary. */
474 Addr fetchBufferMask;
475
476 /** The fetch data that is being fetched and buffered. */
477 uint8_t *fetchBuffer[Impl::MaxThreads];
478
479 /** The PC of the first instruction loaded into the fetch buffer. */
480 Addr fetchBufferPC[Impl::MaxThreads];
481
482 /** The size of the fetch queue in micro-ops */
483 unsigned fetchQueueSize;
484
485 /** Queue of fetched instructions. Per-thread to prevent HoL blocking. */
486 std::deque<DynInstPtr> fetchQueue[Impl::MaxThreads];
487
488 /** Whether or not the fetch buffer data is valid. */
489 bool fetchBufferValid[Impl::MaxThreads];
490
491 /** Size of instructions. */
492 int instSize;
493
494 /** Icache stall statistics. */
495 Counter lastIcacheStall[Impl::MaxThreads];
496
497 /** List of Active Threads */
498 std::list<ThreadID> *activeThreads;
499
500 /** Number of threads. */
501 ThreadID numThreads;
502
503 /** Number of threads that are actively fetching. */
504 ThreadID numFetchingThreads;
505
506 /** Thread ID being fetched. */
507 ThreadID threadFetched;
508
509 /** Checks if there is an interrupt pending. If there is, fetch
510 * must stop once it is not fetching PAL instructions.
511 */
512 bool interruptPending;
513
514 /** Set to true if a pipelined I-cache request should be issued. */
515 bool issuePipelinedIfetch[Impl::MaxThreads];
516
517 /** Event used to delay fault generation of translation faults */
518 FinishTranslationEvent finishTranslationEvent;
519
520 // @todo: Consider making these vectors and tracking on a per thread basis.
521 /** Stat for total number of cycles stalled due to an icache miss. */
522 Stats::Scalar icacheStallCycles;
523 /** Stat for total number of fetched instructions. */
524 Stats::Scalar fetchedInsts;
525 /** Total number of fetched branches. */
526 Stats::Scalar fetchedBranches;
527 /** Stat for total number of predicted branches. */
528 Stats::Scalar predictedBranches;
529 /** Stat for total number of cycles spent fetching. */
530 Stats::Scalar fetchCycles;
531 /** Stat for total number of cycles spent squashing. */
532 Stats::Scalar fetchSquashCycles;
533 /** Stat for total number of cycles spent waiting for translation */
534 Stats::Scalar fetchTlbCycles;
535 /** Stat for total number of cycles spent blocked due to other stages in
536 * the pipeline.
537 */
538 Stats::Scalar fetchIdleCycles;
539 /** Total number of cycles spent blocked. */
540 Stats::Scalar fetchBlockedCycles;
541 /** Total number of cycles spent in any other state. */
542 Stats::Scalar fetchMiscStallCycles;
543 /** Total number of cycles spent in waiting for drains. */
544 Stats::Scalar fetchPendingDrainCycles;
545 /** Total number of stall cycles caused by no active threads to run. */
546 Stats::Scalar fetchNoActiveThreadStallCycles;
547 /** Total number of stall cycles caused by pending traps. */
548 Stats::Scalar fetchPendingTrapStallCycles;
549 /** Total number of stall cycles caused by pending quiesce instructions. */
550 Stats::Scalar fetchPendingQuiesceStallCycles;
551 /** Total number of stall cycles caused by I-cache wait retrys. */
552 Stats::Scalar fetchIcacheWaitRetryStallCycles;
553 /** Stat for total number of fetched cache lines. */
554 Stats::Scalar fetchedCacheLines;
555 /** Total number of outstanding icache accesses that were dropped
556 * due to a squash.
557 */
558 Stats::Scalar fetchIcacheSquashes;
559 /** Total number of outstanding tlb accesses that were dropped
560 * due to a squash.
561 */
562 Stats::Scalar fetchTlbSquashes;
563 /** Distribution of number of instructions fetched each cycle. */
564 Stats::Distribution fetchNisnDist;
565 /** Rate of how often fetch was idle. */
566 Stats::Formula idleRate;
567 /** Number of branch fetches per cycle. */
568 Stats::Formula branchRate;
569 /** Number of instruction fetched per cycle. */
570 Stats::Formula fetchRate;
571 };
572
573 #endif //__CPU_O3_FETCH_HH__