Merge ktlim@zizzer:/bk/m5
[gem5.git] / src / cpu / simple / atomic.cc
1 /*
2 * Copyright (c) 2002-2005 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
29 #include "arch/utility.hh"
30 #include "cpu/exetrace.hh"
31 #include "cpu/simple/atomic.hh"
32 #include "mem/packet_impl.hh"
33 #include "sim/builder.hh"
34
35 using namespace std;
36 using namespace TheISA;
37
38 AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
39 : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
40 {
41 }
42
43
44 void
45 AtomicSimpleCPU::TickEvent::process()
46 {
47 cpu->tick();
48 }
49
50 const char *
51 AtomicSimpleCPU::TickEvent::description()
52 {
53 return "AtomicSimpleCPU tick event";
54 }
55
56
57 void
58 AtomicSimpleCPU::init()
59 {
60 //Create Memory Ports (conect them up)
61 Port *mem_dport = mem->getPort("");
62 dcachePort.setPeer(mem_dport);
63 mem_dport->setPeer(&dcachePort);
64
65 Port *mem_iport = mem->getPort("");
66 icachePort.setPeer(mem_iport);
67 mem_iport->setPeer(&icachePort);
68
69 BaseCPU::init();
70 #if FULL_SYSTEM
71 for (int i = 0; i < execContexts.size(); ++i) {
72 ExecContext *xc = execContexts[i];
73
74 // initialize CPU, including PC
75 TheISA::initCPU(xc, xc->readCpuId());
76 }
77 #endif
78 }
79
80 bool
81 AtomicSimpleCPU::CpuPort::recvTiming(Packet *pkt)
82 {
83 panic("AtomicSimpleCPU doesn't expect recvAtomic callback!");
84 return true;
85 }
86
87 Tick
88 AtomicSimpleCPU::CpuPort::recvAtomic(Packet *pkt)
89 {
90 panic("AtomicSimpleCPU doesn't expect recvAtomic callback!");
91 return curTick;
92 }
93
94 void
95 AtomicSimpleCPU::CpuPort::recvFunctional(Packet *pkt)
96 {
97 panic("AtomicSimpleCPU doesn't expect recvFunctional callback!");
98 }
99
100 void
101 AtomicSimpleCPU::CpuPort::recvStatusChange(Status status)
102 {
103 if (status == RangeChange)
104 return;
105
106 panic("AtomicSimpleCPU doesn't expect recvStatusChange callback!");
107 }
108
109 Packet *
110 AtomicSimpleCPU::CpuPort::recvRetry()
111 {
112 panic("AtomicSimpleCPU doesn't expect recvRetry callback!");
113 return NULL;
114 }
115
116
117 AtomicSimpleCPU::AtomicSimpleCPU(Params *p)
118 : BaseSimpleCPU(p), tickEvent(this),
119 width(p->width), simulate_stalls(p->simulate_stalls),
120 icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
121 {
122 _status = Idle;
123
124 ifetch_req = new Request(true);
125 ifetch_req->setAsid(0);
126 // @todo fix me and get the real cpu iD!!!
127 ifetch_req->setCpuNum(0);
128 ifetch_req->setSize(sizeof(MachInst));
129 ifetch_pkt = new Packet(ifetch_req, Packet::ReadReq, Packet::Broadcast);
130 ifetch_pkt->dataStatic(&inst);
131
132 data_read_req = new Request(true);
133 // @todo fix me and get the real cpu iD!!!
134 data_read_req->setCpuNum(0);
135 data_read_req->setAsid(0);
136 data_read_pkt = new Packet(data_read_req, Packet::ReadReq,
137 Packet::Broadcast);
138 data_read_pkt->dataStatic(&dataReg);
139
140 data_write_req = new Request(true);
141 // @todo fix me and get the real cpu iD!!!
142 data_write_req->setCpuNum(0);
143 data_write_req->setAsid(0);
144 data_write_pkt = new Packet(data_write_req, Packet::WriteReq,
145 Packet::Broadcast);
146 }
147
148
149 AtomicSimpleCPU::~AtomicSimpleCPU()
150 {
151 }
152
153 void
154 AtomicSimpleCPU::serialize(ostream &os)
155 {
156 BaseSimpleCPU::serialize(os);
157 SERIALIZE_ENUM(_status);
158 nameOut(os, csprintf("%s.tickEvent", name()));
159 tickEvent.serialize(os);
160 }
161
162 void
163 AtomicSimpleCPU::unserialize(Checkpoint *cp, const string &section)
164 {
165 BaseSimpleCPU::unserialize(cp, section);
166 UNSERIALIZE_ENUM(_status);
167 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
168 }
169
170 void
171 AtomicSimpleCPU::switchOut(Sampler *s)
172 {
173 sampler = s;
174 if (status() == Running) {
175 _status = SwitchedOut;
176
177 tickEvent.squash();
178 }
179 sampler->signalSwitched();
180 }
181
182
183 void
184 AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
185 {
186 BaseCPU::takeOverFrom(oldCPU);
187
188 assert(!tickEvent.scheduled());
189
190 // if any of this CPU's ExecContexts are active, mark the CPU as
191 // running and schedule its tick event.
192 for (int i = 0; i < execContexts.size(); ++i) {
193 ExecContext *xc = execContexts[i];
194 if (xc->status() == ExecContext::Active && _status != Running) {
195 _status = Running;
196 tickEvent.schedule(curTick);
197 break;
198 }
199 }
200 }
201
202
203 void
204 AtomicSimpleCPU::activateContext(int thread_num, int delay)
205 {
206 assert(thread_num == 0);
207 assert(cpuXC);
208
209 assert(_status == Idle);
210 assert(!tickEvent.scheduled());
211
212 notIdleFraction++;
213 tickEvent.schedule(curTick + cycles(delay));
214 _status = Running;
215 }
216
217
218 void
219 AtomicSimpleCPU::suspendContext(int thread_num)
220 {
221 assert(thread_num == 0);
222 assert(cpuXC);
223
224 assert(_status == Running);
225
226 // tick event may not be scheduled if this gets called from inside
227 // an instruction's execution, e.g. "quiesce"
228 if (tickEvent.scheduled())
229 tickEvent.deschedule();
230
231 notIdleFraction--;
232 _status = Idle;
233 }
234
235
236 template <class T>
237 Fault
238 AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
239 {
240 data_read_req->setVaddr(addr);
241 data_read_req->setSize(sizeof(T));
242 data_read_req->setFlags(flags);
243 data_read_req->setTime(curTick);
244
245 if (traceData) {
246 traceData->setAddr(addr);
247 }
248
249 // translate to physical address
250 Fault fault = cpuXC->translateDataReadReq(data_read_req);
251
252 // Now do the access.
253 if (fault == NoFault) {
254 data_read_pkt->reset();
255 data_read_pkt->reinitFromRequest();
256
257 dcache_complete = dcachePort.sendAtomic(data_read_pkt);
258 dcache_access = true;
259
260 assert(data_read_pkt->result == Packet::Success);
261 data = data_read_pkt->get<T>();
262
263 }
264
265 // This will need a new way to tell if it has a dcache attached.
266 if (data_read_req->getFlags() & UNCACHEABLE)
267 recordEvent("Uncached Read");
268
269 return fault;
270 }
271
272 #ifndef DOXYGEN_SHOULD_SKIP_THIS
273
274 template
275 Fault
276 AtomicSimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
277
278 template
279 Fault
280 AtomicSimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
281
282 template
283 Fault
284 AtomicSimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
285
286 template
287 Fault
288 AtomicSimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
289
290 #endif //DOXYGEN_SHOULD_SKIP_THIS
291
292 template<>
293 Fault
294 AtomicSimpleCPU::read(Addr addr, double &data, unsigned flags)
295 {
296 return read(addr, *(uint64_t*)&data, flags);
297 }
298
299 template<>
300 Fault
301 AtomicSimpleCPU::read(Addr addr, float &data, unsigned flags)
302 {
303 return read(addr, *(uint32_t*)&data, flags);
304 }
305
306
307 template<>
308 Fault
309 AtomicSimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
310 {
311 return read(addr, (uint32_t&)data, flags);
312 }
313
314
315 template <class T>
316 Fault
317 AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
318 {
319 data_write_req->setVaddr(addr);
320 data_write_req->setTime(curTick);
321 data_write_req->setSize(sizeof(T));
322 data_write_req->setFlags(flags);
323
324 if (traceData) {
325 traceData->setAddr(addr);
326 }
327
328 // translate to physical address
329 Fault fault = cpuXC->translateDataWriteReq(data_write_req);
330
331 // Now do the access.
332 if (fault == NoFault) {
333 data_write_pkt->reset();
334 data = htog(data);
335 data_write_pkt->dataStatic(&data);
336 data_write_pkt->reinitFromRequest();
337
338 dcache_complete = dcachePort.sendAtomic(data_write_pkt);
339 dcache_access = true;
340
341 assert(data_write_pkt->result == Packet::Success);
342
343 if (res && data_write_req->getFlags() & LOCKED) {
344 *res = data_write_req->getScResult();
345 }
346 }
347
348 // This will need a new way to tell if it's hooked up to a cache or not.
349 if (data_write_req->getFlags() & UNCACHEABLE)
350 recordEvent("Uncached Write");
351
352 // If the write needs to have a fault on the access, consider calling
353 // changeStatus() and changing it to "bad addr write" or something.
354 return fault;
355 }
356
357
358 #ifndef DOXYGEN_SHOULD_SKIP_THIS
359 template
360 Fault
361 AtomicSimpleCPU::write(uint64_t data, Addr addr,
362 unsigned flags, uint64_t *res);
363
364 template
365 Fault
366 AtomicSimpleCPU::write(uint32_t data, Addr addr,
367 unsigned flags, uint64_t *res);
368
369 template
370 Fault
371 AtomicSimpleCPU::write(uint16_t data, Addr addr,
372 unsigned flags, uint64_t *res);
373
374 template
375 Fault
376 AtomicSimpleCPU::write(uint8_t data, Addr addr,
377 unsigned flags, uint64_t *res);
378
379 #endif //DOXYGEN_SHOULD_SKIP_THIS
380
381 template<>
382 Fault
383 AtomicSimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
384 {
385 return write(*(uint64_t*)&data, addr, flags, res);
386 }
387
388 template<>
389 Fault
390 AtomicSimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
391 {
392 return write(*(uint32_t*)&data, addr, flags, res);
393 }
394
395
396 template<>
397 Fault
398 AtomicSimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
399 {
400 return write((uint32_t)data, addr, flags, res);
401 }
402
403
404 void
405 AtomicSimpleCPU::tick()
406 {
407 Tick latency = cycles(1); // instruction takes one cycle by default
408
409 for (int i = 0; i < width; ++i) {
410 numCycles++;
411
412 checkForInterrupts();
413
414 ifetch_req->resetMin();
415 ifetch_pkt->reset();
416 Fault fault = setupFetchPacket(ifetch_pkt);
417
418 if (fault == NoFault) {
419 Tick icache_complete = icachePort.sendAtomic(ifetch_pkt);
420 // ifetch_req is initialized to read the instruction directly
421 // into the CPU object's inst field.
422
423 dcache_access = false; // assume no dcache access
424 preExecute();
425 fault = curStaticInst->execute(this, traceData);
426 postExecute();
427
428 if (simulate_stalls) {
429 // This calculation assumes that the icache and dcache
430 // access latencies are always a multiple of the CPU's
431 // cycle time. If not, the next tick event may get
432 // scheduled at a non-integer multiple of the CPU
433 // cycle time.
434 Tick icache_stall = icache_complete - curTick - cycles(1);
435 Tick dcache_stall =
436 dcache_access ? dcache_complete - curTick - cycles(1) : 0;
437 latency += icache_stall + dcache_stall;
438 }
439
440 }
441
442 advancePC(fault);
443 }
444
445 if (_status != Idle)
446 tickEvent.schedule(curTick + latency);
447 }
448
449
450 ////////////////////////////////////////////////////////////////////////
451 //
452 // AtomicSimpleCPU Simulation Object
453 //
454 BEGIN_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
455
456 Param<Counter> max_insts_any_thread;
457 Param<Counter> max_insts_all_threads;
458 Param<Counter> max_loads_any_thread;
459 Param<Counter> max_loads_all_threads;
460 SimObjectParam<MemObject *> mem;
461
462 #if FULL_SYSTEM
463 SimObjectParam<AlphaITB *> itb;
464 SimObjectParam<AlphaDTB *> dtb;
465 SimObjectParam<System *> system;
466 Param<int> cpu_id;
467 Param<Tick> profile;
468 #else
469 SimObjectParam<Process *> workload;
470 #endif // FULL_SYSTEM
471
472 Param<int> clock;
473
474 Param<bool> defer_registration;
475 Param<int> width;
476 Param<bool> function_trace;
477 Param<Tick> function_trace_start;
478 Param<bool> simulate_stalls;
479
480 END_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
481
482 BEGIN_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
483
484 INIT_PARAM(max_insts_any_thread,
485 "terminate when any thread reaches this inst count"),
486 INIT_PARAM(max_insts_all_threads,
487 "terminate when all threads have reached this inst count"),
488 INIT_PARAM(max_loads_any_thread,
489 "terminate when any thread reaches this load count"),
490 INIT_PARAM(max_loads_all_threads,
491 "terminate when all threads have reached this load count"),
492 INIT_PARAM(mem, "memory"),
493
494 #if FULL_SYSTEM
495 INIT_PARAM(itb, "Instruction TLB"),
496 INIT_PARAM(dtb, "Data TLB"),
497 INIT_PARAM(system, "system object"),
498 INIT_PARAM(cpu_id, "processor ID"),
499 INIT_PARAM(profile, ""),
500 #else
501 INIT_PARAM(workload, "processes to run"),
502 #endif // FULL_SYSTEM
503
504 INIT_PARAM(clock, "clock speed"),
505 INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
506 INIT_PARAM(width, "cpu width"),
507 INIT_PARAM(function_trace, "Enable function trace"),
508 INIT_PARAM(function_trace_start, "Cycle to start function trace"),
509 INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
510
511 END_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
512
513
514 CREATE_SIM_OBJECT(AtomicSimpleCPU)
515 {
516 AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
517 params->name = getInstanceName();
518 params->numberOfThreads = 1;
519 params->max_insts_any_thread = max_insts_any_thread;
520 params->max_insts_all_threads = max_insts_all_threads;
521 params->max_loads_any_thread = max_loads_any_thread;
522 params->max_loads_all_threads = max_loads_all_threads;
523 params->deferRegistration = defer_registration;
524 params->clock = clock;
525 params->functionTrace = function_trace;
526 params->functionTraceStart = function_trace_start;
527 params->width = width;
528 params->simulate_stalls = simulate_stalls;
529 params->mem = mem;
530
531 #if FULL_SYSTEM
532 params->itb = itb;
533 params->dtb = dtb;
534 params->system = system;
535 params->cpu_id = cpu_id;
536 params->profile = profile;
537 #else
538 params->process = workload;
539 #endif
540
541 AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
542 return cpu;
543 }
544
545 REGISTER_SIM_OBJECT("AtomicSimpleCPU", AtomicSimpleCPU)
546