atomic.cc (4115:cc1d6df13c7d) atomic.cc (4182:5b2c0d266107)
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 * Authors: Steve Reinhardt
29 */
30
31#include "arch/locked_mem.hh"
32#include "arch/mmaped_ipr.hh"
33#include "arch/utility.hh"
34#include "base/bigint.hh"
35#include "cpu/exetrace.hh"
36#include "cpu/simple/atomic.hh"
37#include "mem/packet.hh"
38#include "mem/packet_access.hh"
39#include "sim/builder.hh"
40#include "sim/system.hh"
41
42using namespace std;
43using namespace TheISA;
44
45AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
46 : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
47{
48}
49
50
51void
52AtomicSimpleCPU::TickEvent::process()
53{
54 cpu->tick();
55}
56
57const char *
58AtomicSimpleCPU::TickEvent::description()
59{
60 return "AtomicSimpleCPU tick event";
61}
62
63Port *
64AtomicSimpleCPU::getPort(const std::string &if_name, int idx)
65{
66 if (if_name == "dcache_port")
67 return &dcachePort;
68 else if (if_name == "icache_port")
69 return &icachePort;
70 else
71 panic("No Such Port\n");
72}
73
74void
75AtomicSimpleCPU::init()
76{
77 BaseCPU::init();
78#if FULL_SYSTEM
79 for (int i = 0; i < threadContexts.size(); ++i) {
80 ThreadContext *tc = threadContexts[i];
81
82 // initialize CPU, including PC
83 TheISA::initCPU(tc, tc->readCpuId());
84 }
85#endif
86}
87
88bool
89AtomicSimpleCPU::CpuPort::recvTiming(PacketPtr pkt)
90{
91 panic("AtomicSimpleCPU doesn't expect recvTiming callback!");
92 return true;
93}
94
95Tick
96AtomicSimpleCPU::CpuPort::recvAtomic(PacketPtr pkt)
97{
98 //Snooping a coherence request, just return
99 return 0;
100}
101
102void
103AtomicSimpleCPU::CpuPort::recvFunctional(PacketPtr pkt)
104{
105 //No internal storage to update, just return
106 return;
107}
108
109void
110AtomicSimpleCPU::CpuPort::recvStatusChange(Status status)
111{
112 if (status == RangeChange) {
113 if (!snoopRangeSent) {
114 snoopRangeSent = true;
115 sendStatusChange(Port::RangeChange);
116 }
117 return;
118 }
119
120 panic("AtomicSimpleCPU doesn't expect recvStatusChange callback!");
121}
122
123void
124AtomicSimpleCPU::CpuPort::recvRetry()
125{
126 panic("AtomicSimpleCPU doesn't expect recvRetry callback!");
127}
128
129
130AtomicSimpleCPU::AtomicSimpleCPU(Params *p)
131 : BaseSimpleCPU(p), tickEvent(this),
132 width(p->width), simulate_stalls(p->simulate_stalls),
133 icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
134{
135 _status = Idle;
136
137 icachePort.snoopRangeSent = false;
138 dcachePort.snoopRangeSent = false;
139
140 ifetch_req = new Request();
141 ifetch_req->setThreadContext(p->cpu_id, 0); // Add thread ID if we add MT
142 ifetch_pkt = new Packet(ifetch_req, MemCmd::ReadReq, Packet::Broadcast);
143 ifetch_pkt->dataStatic(&inst);
144
145 data_read_req = new Request();
146 data_read_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
147 data_read_pkt = new Packet(data_read_req, MemCmd::ReadReq,
148 Packet::Broadcast);
149 data_read_pkt->dataStatic(&dataReg);
150
151 data_write_req = new Request();
152 data_write_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
153 data_write_pkt = new Packet(data_write_req, MemCmd::WriteReq,
154 Packet::Broadcast);
155 data_swap_pkt = new Packet(data_write_req, MemCmd::SwapReq,
156 Packet::Broadcast);
157}
158
159
160AtomicSimpleCPU::~AtomicSimpleCPU()
161{
162}
163
164void
165AtomicSimpleCPU::serialize(ostream &os)
166{
167 SimObject::State so_state = SimObject::getState();
168 SERIALIZE_ENUM(so_state);
169 Status _status = status();
170 SERIALIZE_ENUM(_status);
171 BaseSimpleCPU::serialize(os);
172 nameOut(os, csprintf("%s.tickEvent", name()));
173 tickEvent.serialize(os);
174}
175
176void
177AtomicSimpleCPU::unserialize(Checkpoint *cp, const string &section)
178{
179 SimObject::State so_state;
180 UNSERIALIZE_ENUM(so_state);
181 UNSERIALIZE_ENUM(_status);
182 BaseSimpleCPU::unserialize(cp, section);
183 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
184}
185
186void
187AtomicSimpleCPU::resume()
188{
189 if (_status != SwitchedOut && _status != Idle) {
190 assert(system->getMemoryMode() == System::Atomic);
191
192 changeState(SimObject::Running);
193 if (thread->status() == ThreadContext::Active) {
194 if (!tickEvent.scheduled()) {
195 tickEvent.schedule(nextCycle());
196 }
197 }
198 }
199}
200
201void
202AtomicSimpleCPU::switchOut()
203{
204 assert(status() == Running || status() == Idle);
205 _status = SwitchedOut;
206
207 tickEvent.squash();
208}
209
210
211void
212AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
213{
214 BaseCPU::takeOverFrom(oldCPU);
215
216 assert(!tickEvent.scheduled());
217
218 // if any of this CPU's ThreadContexts are active, mark the CPU as
219 // running and schedule its tick event.
220 for (int i = 0; i < threadContexts.size(); ++i) {
221 ThreadContext *tc = threadContexts[i];
222 if (tc->status() == ThreadContext::Active && _status != Running) {
223 _status = Running;
224 tickEvent.schedule(nextCycle());
225 break;
226 }
227 }
228 if (_status != Running) {
229 _status = Idle;
230 }
231}
232
233
234void
235AtomicSimpleCPU::activateContext(int thread_num, int delay)
236{
237 assert(thread_num == 0);
238 assert(thread);
239
240 assert(_status == Idle);
241 assert(!tickEvent.scheduled());
242
243 notIdleFraction++;
244
245#if FULL_SYSTEM
246 // Connect the ThreadContext's memory ports (Functional/Virtual
247 // Ports)
248 tc->connectMemPorts();
249#endif
250
251 //Make sure ticks are still on multiples of cycles
252 tickEvent.schedule(nextCycle(curTick + cycles(delay)));
253 _status = Running;
254}
255
256
257void
258AtomicSimpleCPU::suspendContext(int thread_num)
259{
260 assert(thread_num == 0);
261 assert(thread);
262
263 assert(_status == Running);
264
265 // tick event may not be scheduled if this gets called from inside
266 // an instruction's execution, e.g. "quiesce"
267 if (tickEvent.scheduled())
268 tickEvent.deschedule();
269
270 notIdleFraction--;
271 _status = Idle;
272}
273
274
275template <class T>
276Fault
277AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
278{
279 // use the CPU's statically allocated read request and packet objects
280 Request *req = data_read_req;
281 PacketPtr pkt = data_read_pkt;
282
283 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
284
285 if (traceData) {
286 traceData->setAddr(addr);
287 }
288
289 // translate to physical address
290 Fault fault = thread->translateDataReadReq(req);
291
292 // Now do the access.
293 if (fault == NoFault) {
294 pkt->reinitFromRequest();
295
296 if (req->isMmapedIpr())
297 dcache_latency = TheISA::handleIprRead(thread->getTC(),pkt);
298 else
299 dcache_latency = dcachePort.sendAtomic(pkt);
300 dcache_access = true;
301#if !defined(NDEBUG)
302 if (pkt->result != Packet::Success)
303 panic("Unable to find responder for address pa = %#X va = %#X\n",
304 pkt->req->getPaddr(), pkt->req->getVaddr());
305#endif
306 data = pkt->get<T>();
307
308 if (req->isLocked()) {
309 TheISA::handleLockedRead(thread, req);
310 }
311 }
312
313 // This will need a new way to tell if it has a dcache attached.
314 if (req->isUncacheable())
315 recordEvent("Uncached Read");
316
317 return fault;
318}
319
320#ifndef DOXYGEN_SHOULD_SKIP_THIS
321
322template
323Fault
324AtomicSimpleCPU::read(Addr addr, Twin32_t &data, unsigned flags);
325
326template
327Fault
328AtomicSimpleCPU::read(Addr addr, Twin64_t &data, unsigned flags);
329
330template
331Fault
332AtomicSimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
333
334template
335Fault
336AtomicSimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
337
338template
339Fault
340AtomicSimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
341
342template
343Fault
344AtomicSimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
345
346#endif //DOXYGEN_SHOULD_SKIP_THIS
347
348template<>
349Fault
350AtomicSimpleCPU::read(Addr addr, double &data, unsigned flags)
351{
352 return read(addr, *(uint64_t*)&data, flags);
353}
354
355template<>
356Fault
357AtomicSimpleCPU::read(Addr addr, float &data, unsigned flags)
358{
359 return read(addr, *(uint32_t*)&data, flags);
360}
361
362
363template<>
364Fault
365AtomicSimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
366{
367 return read(addr, (uint32_t&)data, flags);
368}
369
370
371template <class T>
372Fault
373AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
374{
375 // use the CPU's statically allocated write request and packet objects
376 Request *req = data_write_req;
377 PacketPtr pkt;
378
379 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
380
381 if (req->isSwap())
382 pkt = data_swap_pkt;
383 else
384 pkt = data_write_pkt;
385
386 if (traceData) {
387 traceData->setAddr(addr);
388 }
389
390 // translate to physical address
391 Fault fault = thread->translateDataWriteReq(req);
392
393 // Now do the access.
394 if (fault == NoFault) {
395 bool do_access = true; // flag to suppress cache access
396
397 if (req->isLocked()) {
398 do_access = TheISA::handleLockedWrite(thread, req);
399 }
400 if (req->isCondSwap()) {
401 assert(res);
402 req->setExtraData(*res);
403 }
404
405
406 if (do_access) {
407 pkt->reinitFromRequest();
408 pkt->dataStatic(&data);
409
410 if (req->isMmapedIpr()) {
411 dcache_latency = TheISA::handleIprWrite(thread->getTC(), pkt);
412 } else {
413 data = htog(data);
414 dcache_latency = dcachePort.sendAtomic(pkt);
415 }
416 dcache_access = true;
417
418#if !defined(NDEBUG)
419 if (pkt->result != Packet::Success)
420 panic("Unable to find responder for address pa = %#X va = %#X\n",
421 pkt->req->getPaddr(), pkt->req->getVaddr());
422#endif
423 }
424
425 if (req->isSwap()) {
426 assert(res);
427 *res = pkt->get<T>();
428 } else if (res) {
429 *res = req->getExtraData();
430 }
431 }
432
433 // This will need a new way to tell if it's hooked up to a cache or not.
434 if (req->isUncacheable())
435 recordEvent("Uncached Write");
436
437 // If the write needs to have a fault on the access, consider calling
438 // changeStatus() and changing it to "bad addr write" or something.
439 return fault;
440}
441
442
443#ifndef DOXYGEN_SHOULD_SKIP_THIS
444template
445Fault
446AtomicSimpleCPU::write(uint64_t data, Addr addr,
447 unsigned flags, uint64_t *res);
448
449template
450Fault
451AtomicSimpleCPU::write(uint32_t data, Addr addr,
452 unsigned flags, uint64_t *res);
453
454template
455Fault
456AtomicSimpleCPU::write(uint16_t data, Addr addr,
457 unsigned flags, uint64_t *res);
458
459template
460Fault
461AtomicSimpleCPU::write(uint8_t data, Addr addr,
462 unsigned flags, uint64_t *res);
463
464#endif //DOXYGEN_SHOULD_SKIP_THIS
465
466template<>
467Fault
468AtomicSimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
469{
470 return write(*(uint64_t*)&data, addr, flags, res);
471}
472
473template<>
474Fault
475AtomicSimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
476{
477 return write(*(uint32_t*)&data, addr, flags, res);
478}
479
480
481template<>
482Fault
483AtomicSimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
484{
485 return write((uint32_t)data, addr, flags, res);
486}
487
488
489void
490AtomicSimpleCPU::tick()
491{
492 Tick latency = cycles(1); // instruction takes one cycle by default
493
494 for (int i = 0; i < width; ++i) {
495 numCycles++;
496
497 if (!curStaticInst || !curStaticInst->isDelayedCommit())
498 checkForInterrupts();
499
500 Fault fault = setupFetchRequest(ifetch_req);
501
502 if (fault == NoFault) {
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 * Authors: Steve Reinhardt
29 */
30
31#include "arch/locked_mem.hh"
32#include "arch/mmaped_ipr.hh"
33#include "arch/utility.hh"
34#include "base/bigint.hh"
35#include "cpu/exetrace.hh"
36#include "cpu/simple/atomic.hh"
37#include "mem/packet.hh"
38#include "mem/packet_access.hh"
39#include "sim/builder.hh"
40#include "sim/system.hh"
41
42using namespace std;
43using namespace TheISA;
44
45AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
46 : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
47{
48}
49
50
51void
52AtomicSimpleCPU::TickEvent::process()
53{
54 cpu->tick();
55}
56
57const char *
58AtomicSimpleCPU::TickEvent::description()
59{
60 return "AtomicSimpleCPU tick event";
61}
62
63Port *
64AtomicSimpleCPU::getPort(const std::string &if_name, int idx)
65{
66 if (if_name == "dcache_port")
67 return &dcachePort;
68 else if (if_name == "icache_port")
69 return &icachePort;
70 else
71 panic("No Such Port\n");
72}
73
74void
75AtomicSimpleCPU::init()
76{
77 BaseCPU::init();
78#if FULL_SYSTEM
79 for (int i = 0; i < threadContexts.size(); ++i) {
80 ThreadContext *tc = threadContexts[i];
81
82 // initialize CPU, including PC
83 TheISA::initCPU(tc, tc->readCpuId());
84 }
85#endif
86}
87
88bool
89AtomicSimpleCPU::CpuPort::recvTiming(PacketPtr pkt)
90{
91 panic("AtomicSimpleCPU doesn't expect recvTiming callback!");
92 return true;
93}
94
95Tick
96AtomicSimpleCPU::CpuPort::recvAtomic(PacketPtr pkt)
97{
98 //Snooping a coherence request, just return
99 return 0;
100}
101
102void
103AtomicSimpleCPU::CpuPort::recvFunctional(PacketPtr pkt)
104{
105 //No internal storage to update, just return
106 return;
107}
108
109void
110AtomicSimpleCPU::CpuPort::recvStatusChange(Status status)
111{
112 if (status == RangeChange) {
113 if (!snoopRangeSent) {
114 snoopRangeSent = true;
115 sendStatusChange(Port::RangeChange);
116 }
117 return;
118 }
119
120 panic("AtomicSimpleCPU doesn't expect recvStatusChange callback!");
121}
122
123void
124AtomicSimpleCPU::CpuPort::recvRetry()
125{
126 panic("AtomicSimpleCPU doesn't expect recvRetry callback!");
127}
128
129
130AtomicSimpleCPU::AtomicSimpleCPU(Params *p)
131 : BaseSimpleCPU(p), tickEvent(this),
132 width(p->width), simulate_stalls(p->simulate_stalls),
133 icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
134{
135 _status = Idle;
136
137 icachePort.snoopRangeSent = false;
138 dcachePort.snoopRangeSent = false;
139
140 ifetch_req = new Request();
141 ifetch_req->setThreadContext(p->cpu_id, 0); // Add thread ID if we add MT
142 ifetch_pkt = new Packet(ifetch_req, MemCmd::ReadReq, Packet::Broadcast);
143 ifetch_pkt->dataStatic(&inst);
144
145 data_read_req = new Request();
146 data_read_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
147 data_read_pkt = new Packet(data_read_req, MemCmd::ReadReq,
148 Packet::Broadcast);
149 data_read_pkt->dataStatic(&dataReg);
150
151 data_write_req = new Request();
152 data_write_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
153 data_write_pkt = new Packet(data_write_req, MemCmd::WriteReq,
154 Packet::Broadcast);
155 data_swap_pkt = new Packet(data_write_req, MemCmd::SwapReq,
156 Packet::Broadcast);
157}
158
159
160AtomicSimpleCPU::~AtomicSimpleCPU()
161{
162}
163
164void
165AtomicSimpleCPU::serialize(ostream &os)
166{
167 SimObject::State so_state = SimObject::getState();
168 SERIALIZE_ENUM(so_state);
169 Status _status = status();
170 SERIALIZE_ENUM(_status);
171 BaseSimpleCPU::serialize(os);
172 nameOut(os, csprintf("%s.tickEvent", name()));
173 tickEvent.serialize(os);
174}
175
176void
177AtomicSimpleCPU::unserialize(Checkpoint *cp, const string &section)
178{
179 SimObject::State so_state;
180 UNSERIALIZE_ENUM(so_state);
181 UNSERIALIZE_ENUM(_status);
182 BaseSimpleCPU::unserialize(cp, section);
183 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
184}
185
186void
187AtomicSimpleCPU::resume()
188{
189 if (_status != SwitchedOut && _status != Idle) {
190 assert(system->getMemoryMode() == System::Atomic);
191
192 changeState(SimObject::Running);
193 if (thread->status() == ThreadContext::Active) {
194 if (!tickEvent.scheduled()) {
195 tickEvent.schedule(nextCycle());
196 }
197 }
198 }
199}
200
201void
202AtomicSimpleCPU::switchOut()
203{
204 assert(status() == Running || status() == Idle);
205 _status = SwitchedOut;
206
207 tickEvent.squash();
208}
209
210
211void
212AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
213{
214 BaseCPU::takeOverFrom(oldCPU);
215
216 assert(!tickEvent.scheduled());
217
218 // if any of this CPU's ThreadContexts are active, mark the CPU as
219 // running and schedule its tick event.
220 for (int i = 0; i < threadContexts.size(); ++i) {
221 ThreadContext *tc = threadContexts[i];
222 if (tc->status() == ThreadContext::Active && _status != Running) {
223 _status = Running;
224 tickEvent.schedule(nextCycle());
225 break;
226 }
227 }
228 if (_status != Running) {
229 _status = Idle;
230 }
231}
232
233
234void
235AtomicSimpleCPU::activateContext(int thread_num, int delay)
236{
237 assert(thread_num == 0);
238 assert(thread);
239
240 assert(_status == Idle);
241 assert(!tickEvent.scheduled());
242
243 notIdleFraction++;
244
245#if FULL_SYSTEM
246 // Connect the ThreadContext's memory ports (Functional/Virtual
247 // Ports)
248 tc->connectMemPorts();
249#endif
250
251 //Make sure ticks are still on multiples of cycles
252 tickEvent.schedule(nextCycle(curTick + cycles(delay)));
253 _status = Running;
254}
255
256
257void
258AtomicSimpleCPU::suspendContext(int thread_num)
259{
260 assert(thread_num == 0);
261 assert(thread);
262
263 assert(_status == Running);
264
265 // tick event may not be scheduled if this gets called from inside
266 // an instruction's execution, e.g. "quiesce"
267 if (tickEvent.scheduled())
268 tickEvent.deschedule();
269
270 notIdleFraction--;
271 _status = Idle;
272}
273
274
275template <class T>
276Fault
277AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
278{
279 // use the CPU's statically allocated read request and packet objects
280 Request *req = data_read_req;
281 PacketPtr pkt = data_read_pkt;
282
283 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
284
285 if (traceData) {
286 traceData->setAddr(addr);
287 }
288
289 // translate to physical address
290 Fault fault = thread->translateDataReadReq(req);
291
292 // Now do the access.
293 if (fault == NoFault) {
294 pkt->reinitFromRequest();
295
296 if (req->isMmapedIpr())
297 dcache_latency = TheISA::handleIprRead(thread->getTC(),pkt);
298 else
299 dcache_latency = dcachePort.sendAtomic(pkt);
300 dcache_access = true;
301#if !defined(NDEBUG)
302 if (pkt->result != Packet::Success)
303 panic("Unable to find responder for address pa = %#X va = %#X\n",
304 pkt->req->getPaddr(), pkt->req->getVaddr());
305#endif
306 data = pkt->get<T>();
307
308 if (req->isLocked()) {
309 TheISA::handleLockedRead(thread, req);
310 }
311 }
312
313 // This will need a new way to tell if it has a dcache attached.
314 if (req->isUncacheable())
315 recordEvent("Uncached Read");
316
317 return fault;
318}
319
320#ifndef DOXYGEN_SHOULD_SKIP_THIS
321
322template
323Fault
324AtomicSimpleCPU::read(Addr addr, Twin32_t &data, unsigned flags);
325
326template
327Fault
328AtomicSimpleCPU::read(Addr addr, Twin64_t &data, unsigned flags);
329
330template
331Fault
332AtomicSimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
333
334template
335Fault
336AtomicSimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
337
338template
339Fault
340AtomicSimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
341
342template
343Fault
344AtomicSimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
345
346#endif //DOXYGEN_SHOULD_SKIP_THIS
347
348template<>
349Fault
350AtomicSimpleCPU::read(Addr addr, double &data, unsigned flags)
351{
352 return read(addr, *(uint64_t*)&data, flags);
353}
354
355template<>
356Fault
357AtomicSimpleCPU::read(Addr addr, float &data, unsigned flags)
358{
359 return read(addr, *(uint32_t*)&data, flags);
360}
361
362
363template<>
364Fault
365AtomicSimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
366{
367 return read(addr, (uint32_t&)data, flags);
368}
369
370
371template <class T>
372Fault
373AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
374{
375 // use the CPU's statically allocated write request and packet objects
376 Request *req = data_write_req;
377 PacketPtr pkt;
378
379 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
380
381 if (req->isSwap())
382 pkt = data_swap_pkt;
383 else
384 pkt = data_write_pkt;
385
386 if (traceData) {
387 traceData->setAddr(addr);
388 }
389
390 // translate to physical address
391 Fault fault = thread->translateDataWriteReq(req);
392
393 // Now do the access.
394 if (fault == NoFault) {
395 bool do_access = true; // flag to suppress cache access
396
397 if (req->isLocked()) {
398 do_access = TheISA::handleLockedWrite(thread, req);
399 }
400 if (req->isCondSwap()) {
401 assert(res);
402 req->setExtraData(*res);
403 }
404
405
406 if (do_access) {
407 pkt->reinitFromRequest();
408 pkt->dataStatic(&data);
409
410 if (req->isMmapedIpr()) {
411 dcache_latency = TheISA::handleIprWrite(thread->getTC(), pkt);
412 } else {
413 data = htog(data);
414 dcache_latency = dcachePort.sendAtomic(pkt);
415 }
416 dcache_access = true;
417
418#if !defined(NDEBUG)
419 if (pkt->result != Packet::Success)
420 panic("Unable to find responder for address pa = %#X va = %#X\n",
421 pkt->req->getPaddr(), pkt->req->getVaddr());
422#endif
423 }
424
425 if (req->isSwap()) {
426 assert(res);
427 *res = pkt->get<T>();
428 } else if (res) {
429 *res = req->getExtraData();
430 }
431 }
432
433 // This will need a new way to tell if it's hooked up to a cache or not.
434 if (req->isUncacheable())
435 recordEvent("Uncached Write");
436
437 // If the write needs to have a fault on the access, consider calling
438 // changeStatus() and changing it to "bad addr write" or something.
439 return fault;
440}
441
442
443#ifndef DOXYGEN_SHOULD_SKIP_THIS
444template
445Fault
446AtomicSimpleCPU::write(uint64_t data, Addr addr,
447 unsigned flags, uint64_t *res);
448
449template
450Fault
451AtomicSimpleCPU::write(uint32_t data, Addr addr,
452 unsigned flags, uint64_t *res);
453
454template
455Fault
456AtomicSimpleCPU::write(uint16_t data, Addr addr,
457 unsigned flags, uint64_t *res);
458
459template
460Fault
461AtomicSimpleCPU::write(uint8_t data, Addr addr,
462 unsigned flags, uint64_t *res);
463
464#endif //DOXYGEN_SHOULD_SKIP_THIS
465
466template<>
467Fault
468AtomicSimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
469{
470 return write(*(uint64_t*)&data, addr, flags, res);
471}
472
473template<>
474Fault
475AtomicSimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
476{
477 return write(*(uint32_t*)&data, addr, flags, res);
478}
479
480
481template<>
482Fault
483AtomicSimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
484{
485 return write((uint32_t)data, addr, flags, res);
486}
487
488
489void
490AtomicSimpleCPU::tick()
491{
492 Tick latency = cycles(1); // instruction takes one cycle by default
493
494 for (int i = 0; i < width; ++i) {
495 numCycles++;
496
497 if (!curStaticInst || !curStaticInst->isDelayedCommit())
498 checkForInterrupts();
499
500 Fault fault = setupFetchRequest(ifetch_req);
501
502 if (fault == NoFault) {
503 ifetch_pkt->reinitFromRequest();
503 Tick icache_latency = 0;
504 bool icache_access = false;
505 dcache_access = false; // assume no dcache access
504
506
505 Tick icache_latency = icachePort.sendAtomic(ifetch_pkt);
506 // ifetch_req is initialized to read the instruction directly
507 // into the CPU object's inst field.
507 //Fetch more instruction memory if necessary
508 if(predecoder.needMoreBytes())
509 {
510 icache_access = true;
511 ifetch_pkt->reinitFromRequest();
508
512
509 dcache_access = false; // assume no dcache access
513 icache_latency = icachePort.sendAtomic(ifetch_pkt);
514 // ifetch_req is initialized to read the instruction directly
515 // into the CPU object's inst field.
516 }
517
510 preExecute();
511
518 preExecute();
519
512 fault = curStaticInst->execute(this, traceData);
513 postExecute();
520 if(curStaticInst)
521 {
522 fault = curStaticInst->execute(this, traceData);
523 postExecute();
524 }
514
515 // @todo remove me after debugging with legion done
516 if (curStaticInst && (!curStaticInst->isMicroOp() ||
517 curStaticInst->isFirstMicroOp()))
518 instCnt++;
519
520 if (simulate_stalls) {
525
526 // @todo remove me after debugging with legion done
527 if (curStaticInst && (!curStaticInst->isMicroOp() ||
528 curStaticInst->isFirstMicroOp()))
529 instCnt++;
530
531 if (simulate_stalls) {
521 Tick icache_stall = icache_latency - cycles(1);
532 Tick icache_stall =
533 icache_access ? icache_latency - cycles(1) : 0;
522 Tick dcache_stall =
523 dcache_access ? dcache_latency - cycles(1) : 0;
524 Tick stall_cycles = (icache_stall + dcache_stall) / cycles(1);
525 if (cycles(stall_cycles) < (icache_stall + dcache_stall))
526 latency += cycles(stall_cycles+1);
527 else
528 latency += cycles(stall_cycles);
529 }
530
531 }
534 Tick dcache_stall =
535 dcache_access ? dcache_latency - cycles(1) : 0;
536 Tick stall_cycles = (icache_stall + dcache_stall) / cycles(1);
537 if (cycles(stall_cycles) < (icache_stall + dcache_stall))
538 latency += cycles(stall_cycles+1);
539 else
540 latency += cycles(stall_cycles);
541 }
542
543 }
532
533 advancePC(fault);
544 if(predecoder.needMoreBytes())
545 advancePC(fault);
534 }
535
536 if (_status != Idle)
537 tickEvent.schedule(curTick + latency);
538}
539
540
541////////////////////////////////////////////////////////////////////////
542//
543// AtomicSimpleCPU Simulation Object
544//
545BEGIN_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
546
547 Param<Counter> max_insts_any_thread;
548 Param<Counter> max_insts_all_threads;
549 Param<Counter> max_loads_any_thread;
550 Param<Counter> max_loads_all_threads;
551 Param<Tick> progress_interval;
552 SimObjectParam<System *> system;
553 Param<int> cpu_id;
554
555#if FULL_SYSTEM
556 SimObjectParam<TheISA::ITB *> itb;
557 SimObjectParam<TheISA::DTB *> dtb;
558 Param<Tick> profile;
559
560 Param<bool> do_quiesce;
561 Param<bool> do_checkpoint_insts;
562 Param<bool> do_statistics_insts;
563#else
564 SimObjectParam<Process *> workload;
565#endif // FULL_SYSTEM
566
567 Param<int> clock;
568 Param<int> phase;
569
570 Param<bool> defer_registration;
571 Param<int> width;
572 Param<bool> function_trace;
573 Param<Tick> function_trace_start;
574 Param<bool> simulate_stalls;
575
576END_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
577
578BEGIN_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
579
580 INIT_PARAM(max_insts_any_thread,
581 "terminate when any thread reaches this inst count"),
582 INIT_PARAM(max_insts_all_threads,
583 "terminate when all threads have reached this inst count"),
584 INIT_PARAM(max_loads_any_thread,
585 "terminate when any thread reaches this load count"),
586 INIT_PARAM(max_loads_all_threads,
587 "terminate when all threads have reached this load count"),
588 INIT_PARAM(progress_interval, "Progress interval"),
589 INIT_PARAM(system, "system object"),
590 INIT_PARAM(cpu_id, "processor ID"),
591
592#if FULL_SYSTEM
593 INIT_PARAM(itb, "Instruction TLB"),
594 INIT_PARAM(dtb, "Data TLB"),
595 INIT_PARAM(profile, ""),
596 INIT_PARAM(do_quiesce, ""),
597 INIT_PARAM(do_checkpoint_insts, ""),
598 INIT_PARAM(do_statistics_insts, ""),
599#else
600 INIT_PARAM(workload, "processes to run"),
601#endif // FULL_SYSTEM
602
603 INIT_PARAM(clock, "clock speed"),
604 INIT_PARAM_DFLT(phase, "clock phase", 0),
605 INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
606 INIT_PARAM(width, "cpu width"),
607 INIT_PARAM(function_trace, "Enable function trace"),
608 INIT_PARAM(function_trace_start, "Cycle to start function trace"),
609 INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
610
611END_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
612
613
614CREATE_SIM_OBJECT(AtomicSimpleCPU)
615{
616 AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
617 params->name = getInstanceName();
618 params->numberOfThreads = 1;
619 params->max_insts_any_thread = max_insts_any_thread;
620 params->max_insts_all_threads = max_insts_all_threads;
621 params->max_loads_any_thread = max_loads_any_thread;
622 params->max_loads_all_threads = max_loads_all_threads;
623 params->progress_interval = progress_interval;
624 params->deferRegistration = defer_registration;
625 params->phase = phase;
626 params->clock = clock;
627 params->functionTrace = function_trace;
628 params->functionTraceStart = function_trace_start;
629 params->width = width;
630 params->simulate_stalls = simulate_stalls;
631 params->system = system;
632 params->cpu_id = cpu_id;
633
634#if FULL_SYSTEM
635 params->itb = itb;
636 params->dtb = dtb;
637 params->profile = profile;
638 params->do_quiesce = do_quiesce;
639 params->do_checkpoint_insts = do_checkpoint_insts;
640 params->do_statistics_insts = do_statistics_insts;
641#else
642 params->process = workload;
643#endif
644
645 AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
646 return cpu;
647}
648
649REGISTER_SIM_OBJECT("AtomicSimpleCPU", AtomicSimpleCPU)
650
546 }
547
548 if (_status != Idle)
549 tickEvent.schedule(curTick + latency);
550}
551
552
553////////////////////////////////////////////////////////////////////////
554//
555// AtomicSimpleCPU Simulation Object
556//
557BEGIN_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
558
559 Param<Counter> max_insts_any_thread;
560 Param<Counter> max_insts_all_threads;
561 Param<Counter> max_loads_any_thread;
562 Param<Counter> max_loads_all_threads;
563 Param<Tick> progress_interval;
564 SimObjectParam<System *> system;
565 Param<int> cpu_id;
566
567#if FULL_SYSTEM
568 SimObjectParam<TheISA::ITB *> itb;
569 SimObjectParam<TheISA::DTB *> dtb;
570 Param<Tick> profile;
571
572 Param<bool> do_quiesce;
573 Param<bool> do_checkpoint_insts;
574 Param<bool> do_statistics_insts;
575#else
576 SimObjectParam<Process *> workload;
577#endif // FULL_SYSTEM
578
579 Param<int> clock;
580 Param<int> phase;
581
582 Param<bool> defer_registration;
583 Param<int> width;
584 Param<bool> function_trace;
585 Param<Tick> function_trace_start;
586 Param<bool> simulate_stalls;
587
588END_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
589
590BEGIN_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
591
592 INIT_PARAM(max_insts_any_thread,
593 "terminate when any thread reaches this inst count"),
594 INIT_PARAM(max_insts_all_threads,
595 "terminate when all threads have reached this inst count"),
596 INIT_PARAM(max_loads_any_thread,
597 "terminate when any thread reaches this load count"),
598 INIT_PARAM(max_loads_all_threads,
599 "terminate when all threads have reached this load count"),
600 INIT_PARAM(progress_interval, "Progress interval"),
601 INIT_PARAM(system, "system object"),
602 INIT_PARAM(cpu_id, "processor ID"),
603
604#if FULL_SYSTEM
605 INIT_PARAM(itb, "Instruction TLB"),
606 INIT_PARAM(dtb, "Data TLB"),
607 INIT_PARAM(profile, ""),
608 INIT_PARAM(do_quiesce, ""),
609 INIT_PARAM(do_checkpoint_insts, ""),
610 INIT_PARAM(do_statistics_insts, ""),
611#else
612 INIT_PARAM(workload, "processes to run"),
613#endif // FULL_SYSTEM
614
615 INIT_PARAM(clock, "clock speed"),
616 INIT_PARAM_DFLT(phase, "clock phase", 0),
617 INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
618 INIT_PARAM(width, "cpu width"),
619 INIT_PARAM(function_trace, "Enable function trace"),
620 INIT_PARAM(function_trace_start, "Cycle to start function trace"),
621 INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
622
623END_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
624
625
626CREATE_SIM_OBJECT(AtomicSimpleCPU)
627{
628 AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
629 params->name = getInstanceName();
630 params->numberOfThreads = 1;
631 params->max_insts_any_thread = max_insts_any_thread;
632 params->max_insts_all_threads = max_insts_all_threads;
633 params->max_loads_any_thread = max_loads_any_thread;
634 params->max_loads_all_threads = max_loads_all_threads;
635 params->progress_interval = progress_interval;
636 params->deferRegistration = defer_registration;
637 params->phase = phase;
638 params->clock = clock;
639 params->functionTrace = function_trace;
640 params->functionTraceStart = function_trace_start;
641 params->width = width;
642 params->simulate_stalls = simulate_stalls;
643 params->system = system;
644 params->cpu_id = cpu_id;
645
646#if FULL_SYSTEM
647 params->itb = itb;
648 params->dtb = dtb;
649 params->profile = profile;
650 params->do_quiesce = do_quiesce;
651 params->do_checkpoint_insts = do_checkpoint_insts;
652 params->do_statistics_insts = do_statistics_insts;
653#else
654 params->process = workload;
655#endif
656
657 AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
658 return cpu;
659}
660
661REGISTER_SIM_OBJECT("AtomicSimpleCPU", AtomicSimpleCPU)
662