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"
45 * DefaultFetch class handles both single threaded and SMT fetch. Its
46 * width is specified by the parameters; each cycle it tries to fetch
47 * that many instructions. It supports using a branch predictor to
48 * predict direction and targets.
49 * It supports the idling functionality of the CPU by indicating to
50 * the CPU when it is active and inactive.
56 /** Typedefs from Impl. */
57 typedef typename Impl::CPUPol CPUPol;
58 typedef typename Impl::DynInst DynInst;
59 typedef typename Impl::DynInstPtr DynInstPtr;
60 typedef typename Impl::O3CPU O3CPU;
61 typedef typename Impl::Params Params;
63 /** Typedefs from the CPU policy. */
64 typedef typename CPUPol::BPredUnit BPredUnit;
65 typedef typename CPUPol::FetchStruct FetchStruct;
66 typedef typename CPUPol::TimeStruct TimeStruct;
68 /** Typedefs from ISA. */
69 typedef TheISA::MachInst MachInst;
70 typedef TheISA::ExtMachInst ExtMachInst;
72 /** IcachePort class for DefaultFetch. Handles doing the
73 * communication with the cache/memory.
75 class IcachePort : public Port
78 /** Pointer to fetch. */
79 DefaultFetch<Impl> *fetch;
82 /** Default constructor. */
83 IcachePort(DefaultFetch<Impl> *_fetch)
84 : Port(_fetch->name() + "-iport"), fetch(_fetch)
89 virtual void setPeer(Port *port);
92 /** Atomic version of receive. Panics. */
93 virtual Tick recvAtomic(PacketPtr pkt);
95 /** Functional version of receive. Panics. */
96 virtual void recvFunctional(PacketPtr pkt);
98 /** Receives status change. Other than range changing, panics. */
99 virtual void recvStatusChange(Status status);
101 /** Returns the address ranges of this device. */
102 virtual void getDeviceAddressRanges(AddrRangeList &resp,
103 AddrRangeList &snoop)
104 { resp.clear(); snoop.clear(); snoop.push_back(RangeSize(0,0)); }
106 /** Timing version of receive. Handles setting fetch to the
107 * proper status to start fetching. */
108 virtual bool recvTiming(PacketPtr pkt);
110 /** Handles doing a retry of a failed fetch. */
111 virtual void recvRetry();
116 /** Overall fetch status. Used to determine if the CPU can
117 * deschedule itsef due to a lack of activity.
124 /** Individual thread status. */
139 /** Fetching Policy, Add new policies here.*/
152 /** Per-thread status. */
153 ThreadStatus fetchStatus[Impl::MaxThreads];
156 FetchPriority fetchPolicy;
158 /** List that has the threads organized by priority. */
159 std::list<unsigned> priorityList;
162 /** DefaultFetch constructor. */
163 DefaultFetch(Params *params);
165 /** Returns the name of fetch. */
166 std::string name() const;
168 /** Registers statistics. */
171 /** Returns the icache port. */
172 Port *getIcachePort() { return icachePort; }
174 /** Sets CPU pointer. */
175 void setCPU(O3CPU *cpu_ptr);
177 /** Sets the main backwards communication time buffer pointer. */
178 void setTimeBuffer(TimeBuffer<TimeStruct> *time_buffer);
180 /** Sets pointer to list of active threads. */
181 void setActiveThreads(std::list<unsigned> *at_ptr);
183 /** Sets pointer to time buffer used to communicate to the next stage. */
184 void setFetchQueue(TimeBuffer<FetchStruct> *fq_ptr);
186 /** Initialize stage. */
189 /** Tells the fetch stage that the Icache is set. */
192 /** Processes cache completion event. */
193 void processCacheCompletion(PacketPtr pkt);
195 /** Begins the drain of the fetch stage. */
198 /** Resumes execution after a drain. */
201 /** Tells fetch stage to prepare to be switched out. */
204 /** Takes over from another CPU's thread. */
207 /** Checks if the fetch stage is switched out. */
208 bool isSwitchedOut() { return switchedOut; }
210 /** Tells fetch to wake up from a quiesce instruction. */
211 void wakeFromQuiesce();
214 /** Changes the status of this stage to active, and indicates this
217 inline void switchToActive();
219 /** Changes the status of this stage to inactive, and indicates
222 inline void switchToInactive();
225 * Looks up in the branch predictor to see if the next PC should be
226 * either next PC+=MachInst or a branch target.
227 * @param next_PC Next PC variable passed in by reference. It is
228 * expected to be set to the current PC; it will be updated with what
229 * the next PC will be.
230 * @param next_NPC Used for ISAs which use delay slots.
231 * @return Whether or not a branch was predicted as taken.
233 bool lookupAndUpdateNextPC(DynInstPtr &inst, Addr &next_PC, Addr &next_NPC);
236 * Fetches the cache line that contains fetch_PC. Returns any
237 * fault that happened. Puts the data into the class variable
239 * @param fetch_PC The PC address that is being fetched from.
240 * @param ret_fault The fault reference that will be set to the result of
242 * @param tid Thread id.
243 * @return Any fault that occured.
245 bool fetchCacheLine(Addr fetch_PC, Fault &ret_fault, unsigned tid);
247 /** Squashes a specific thread and resets the PC. */
248 inline void doSquash(const Addr &new_PC, const Addr &new_NPC, unsigned tid);
250 /** Squashes a specific thread and resets the PC. Also tells the CPU to
251 * remove any instructions between fetch and decode that should be sqaushed.
253 void squashFromDecode(const Addr &new_PC, const Addr &new_NPC,
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 InstSeqNum &seq_num,
270 bool squash_delay_slot, 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 next PC. */
354 Addr nextPC[Impl::MaxThreads];
356 /** Per-thread next Next PC.
357 * This is not a real register but is used for
358 * architectures that use a branch-delay slot.
359 * (such as MIPS or Sparc)
361 Addr nextNPC[Impl::MaxThreads];
363 /** Memory request used to access cache. */
364 RequestPtr memReq[Impl::MaxThreads];
366 /** Variable that tracks if fetch has written to the time buffer this
367 * cycle. Used to tell CPU if there is activity this cycle.
369 bool wroteToTimeBuffer;
371 /** Tracks how many instructions has been fetched this cycle. */
374 /** Source of possible stalls. */
382 /** Tracks which stages are telling fetch to stall. */
383 Stalls stalls[Impl::MaxThreads];
385 /** Decode to fetch delay, in ticks. */
386 unsigned decodeToFetchDelay;
388 /** Rename to fetch delay, in ticks. */
389 unsigned renameToFetchDelay;
391 /** IEW to fetch delay, in ticks. */
392 unsigned iewToFetchDelay;
394 /** Commit to fetch delay, in ticks. */
395 unsigned commitToFetchDelay;
397 /** The width of fetch in instructions. */
400 /** Is the cache blocked? If so no threads can access it. */
403 /** The packet that is waiting to be retried. */
406 /** The thread that is waiting on the cache to tell fetch to retry. */
409 /** Cache block size. */
412 /** Mask to get a cache block's address. */
415 /** The cache line being fetched. */
416 uint8_t *cacheData[Impl::MaxThreads];
418 /** The PC of the cacheline that has been loaded. */
419 Addr cacheDataPC[Impl::MaxThreads];
421 /** Whether or not the cache data is valid. */
422 bool cacheDataValid[Impl::MaxThreads];
424 /** Size of instructions. */
427 /** Icache stall statistics. */
428 Counter lastIcacheStall[Impl::MaxThreads];
430 /** List of Active Threads */
431 std::list<unsigned> *activeThreads;
433 /** Number of threads. */
436 /** Number of threads that are actively fetching. */
437 unsigned numFetchingThreads;
439 /** Thread ID being fetched. */
442 /** Checks if there is an interrupt pending. If there is, fetch
443 * must stop once it is not fetching PAL instructions.
445 bool interruptPending;
447 /** Is there a drain pending. */
450 /** Records if fetch is switched out. */
453 // @todo: Consider making these vectors and tracking on a per thread basis.
454 /** Stat for total number of cycles stalled due to an icache miss. */
455 Stats::Scalar<> icacheStallCycles;
456 /** Stat for total number of fetched instructions. */
457 Stats::Scalar<> fetchedInsts;
458 /** Total number of fetched branches. */
459 Stats::Scalar<> fetchedBranches;
460 /** Stat for total number of predicted branches. */
461 Stats::Scalar<> predictedBranches;
462 /** Stat for total number of cycles spent fetching. */
463 Stats::Scalar<> fetchCycles;
464 /** Stat for total number of cycles spent squashing. */
465 Stats::Scalar<> fetchSquashCycles;
466 /** Stat for total number of cycles spent blocked due to other stages in
469 Stats::Scalar<> fetchIdleCycles;
470 /** Total number of cycles spent blocked. */
471 Stats::Scalar<> fetchBlockedCycles;
472 /** Total number of cycles spent in any other state. */
473 Stats::Scalar<> fetchMiscStallCycles;
474 /** Stat for total number of fetched cache lines. */
475 Stats::Scalar<> fetchedCacheLines;
476 /** Total number of outstanding icache accesses that were dropped
479 Stats::Scalar<> fetchIcacheSquashes;
480 /** Distribution of number of instructions fetched each cycle. */
481 Stats::Distribution<> fetchNisnDist;
482 /** Rate of how often fetch was idle. */
483 Stats::Formula idleRate;
484 /** Number of branch fetches per cycle. */
485 Stats::Formula branchRate;
486 /** Number of instruction fetched per cycle. */
487 Stats::Formula fetchRate;
490 #endif //__CPU_O3_FETCH_HH__