atomic.cc revision 2663
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
35using namespace std;
36using namespace TheISA;
37
38AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
39    : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
40{
41}
42
43
44void
45AtomicSimpleCPU::TickEvent::process()
46{
47    cpu->tick();
48}
49
50const char *
51AtomicSimpleCPU::TickEvent::description()
52{
53    return "AtomicSimpleCPU tick event";
54}
55
56
57void
58AtomicSimpleCPU::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
80bool
81AtomicSimpleCPU::CpuPort::recvTiming(Packet *pkt)
82{
83    panic("AtomicSimpleCPU doesn't expect recvAtomic callback!");
84    return true;
85}
86
87Tick
88AtomicSimpleCPU::CpuPort::recvAtomic(Packet *pkt)
89{
90    panic("AtomicSimpleCPU doesn't expect recvAtomic callback!");
91    return curTick;
92}
93
94void
95AtomicSimpleCPU::CpuPort::recvFunctional(Packet *pkt)
96{
97    panic("AtomicSimpleCPU doesn't expect recvFunctional callback!");
98}
99
100void
101AtomicSimpleCPU::CpuPort::recvStatusChange(Status status)
102{
103    if (status == RangeChange)
104        return;
105
106    panic("AtomicSimpleCPU doesn't expect recvStatusChange callback!");
107}
108
109void
110AtomicSimpleCPU::CpuPort::recvRetry()
111{
112    panic("AtomicSimpleCPU doesn't expect recvRetry callback!");
113}
114
115
116AtomicSimpleCPU::AtomicSimpleCPU(Params *p)
117    : BaseSimpleCPU(p), tickEvent(this),
118      width(p->width), simulate_stalls(p->simulate_stalls),
119      icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
120{
121    _status = Idle;
122
123    // @todo fix me and get the real cpu id & thread number!!!
124    ifetch_req = new Request();
125    ifetch_pkt = new Packet(ifetch_req, Packet::ReadReq, Packet::Broadcast);
126    ifetch_pkt->dataStatic(&inst);
127
128    data_read_req = new Request();
129    data_read_pkt = new Packet(data_read_req, Packet::ReadReq,
130                               Packet::Broadcast);
131    data_read_pkt->dataStatic(&dataReg);
132
133    data_write_req = new Request();
134    data_write_pkt = new Packet(data_write_req, Packet::WriteReq,
135                                Packet::Broadcast);
136}
137
138
139AtomicSimpleCPU::~AtomicSimpleCPU()
140{
141}
142
143void
144AtomicSimpleCPU::serialize(ostream &os)
145{
146    BaseSimpleCPU::serialize(os);
147    SERIALIZE_ENUM(_status);
148    nameOut(os, csprintf("%s.tickEvent", name()));
149    tickEvent.serialize(os);
150}
151
152void
153AtomicSimpleCPU::unserialize(Checkpoint *cp, const string &section)
154{
155    BaseSimpleCPU::unserialize(cp, section);
156    UNSERIALIZE_ENUM(_status);
157    tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
158}
159
160void
161AtomicSimpleCPU::switchOut(Sampler *s)
162{
163    sampler = s;
164    if (status() == Running) {
165        _status = SwitchedOut;
166
167        tickEvent.squash();
168    }
169    sampler->signalSwitched();
170}
171
172
173void
174AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
175{
176    BaseCPU::takeOverFrom(oldCPU);
177
178    assert(!tickEvent.scheduled());
179
180    // if any of this CPU's ExecContexts are active, mark the CPU as
181    // running and schedule its tick event.
182    for (int i = 0; i < execContexts.size(); ++i) {
183        ExecContext *xc = execContexts[i];
184        if (xc->status() == ExecContext::Active && _status != Running) {
185            _status = Running;
186            tickEvent.schedule(curTick);
187            break;
188        }
189    }
190}
191
192
193void
194AtomicSimpleCPU::activateContext(int thread_num, int delay)
195{
196    assert(thread_num == 0);
197    assert(cpuXC);
198
199    assert(_status == Idle);
200    assert(!tickEvent.scheduled());
201
202    notIdleFraction++;
203    tickEvent.schedule(curTick + cycles(delay));
204    _status = Running;
205}
206
207
208void
209AtomicSimpleCPU::suspendContext(int thread_num)
210{
211    assert(thread_num == 0);
212    assert(cpuXC);
213
214    assert(_status == Running);
215
216    // tick event may not be scheduled if this gets called from inside
217    // an instruction's execution, e.g. "quiesce"
218    if (tickEvent.scheduled())
219        tickEvent.deschedule();
220
221    notIdleFraction--;
222    _status = Idle;
223}
224
225
226template <class T>
227Fault
228AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
229{
230    data_read_req->setVirt(0, addr, sizeof(T), flags, cpuXC->readPC());
231
232    if (traceData) {
233        traceData->setAddr(addr);
234    }
235
236    // translate to physical address
237    Fault fault = cpuXC->translateDataReadReq(data_read_req);
238
239    // Now do the access.
240    if (fault == NoFault) {
241        data_read_pkt->reinitFromRequest();
242
243        dcache_latency = dcachePort.sendAtomic(data_read_pkt);
244        dcache_access = true;
245
246        assert(data_read_pkt->result == Packet::Success);
247        data = data_read_pkt->get<T>();
248
249    }
250
251    // This will need a new way to tell if it has a dcache attached.
252    if (data_read_req->getFlags() & UNCACHEABLE)
253        recordEvent("Uncached Read");
254
255    return fault;
256}
257
258#ifndef DOXYGEN_SHOULD_SKIP_THIS
259
260template
261Fault
262AtomicSimpleCPU::read(Addr addr, uint64_t &data, unsigned flags);
263
264template
265Fault
266AtomicSimpleCPU::read(Addr addr, uint32_t &data, unsigned flags);
267
268template
269Fault
270AtomicSimpleCPU::read(Addr addr, uint16_t &data, unsigned flags);
271
272template
273Fault
274AtomicSimpleCPU::read(Addr addr, uint8_t &data, unsigned flags);
275
276#endif //DOXYGEN_SHOULD_SKIP_THIS
277
278template<>
279Fault
280AtomicSimpleCPU::read(Addr addr, double &data, unsigned flags)
281{
282    return read(addr, *(uint64_t*)&data, flags);
283}
284
285template<>
286Fault
287AtomicSimpleCPU::read(Addr addr, float &data, unsigned flags)
288{
289    return read(addr, *(uint32_t*)&data, flags);
290}
291
292
293template<>
294Fault
295AtomicSimpleCPU::read(Addr addr, int32_t &data, unsigned flags)
296{
297    return read(addr, (uint32_t&)data, flags);
298}
299
300
301template <class T>
302Fault
303AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
304{
305    data_write_req->setVirt(0, addr, sizeof(T), flags, cpuXC->readPC());
306
307    if (traceData) {
308        traceData->setAddr(addr);
309    }
310
311    // translate to physical address
312    Fault fault = cpuXC->translateDataWriteReq(data_write_req);
313
314    // Now do the access.
315    if (fault == NoFault) {
316        data = htog(data);
317        data_write_pkt->reinitFromRequest();
318        data_write_pkt->dataStatic(&data);
319
320        dcache_latency = dcachePort.sendAtomic(data_write_pkt);
321        dcache_access = true;
322
323        assert(data_write_pkt->result == Packet::Success);
324
325        if (res && data_write_req->getFlags() & LOCKED) {
326            *res = data_write_req->getScResult();
327        }
328    }
329
330    // This will need a new way to tell if it's hooked up to a cache or not.
331    if (data_write_req->getFlags() & UNCACHEABLE)
332        recordEvent("Uncached Write");
333
334    // If the write needs to have a fault on the access, consider calling
335    // changeStatus() and changing it to "bad addr write" or something.
336    return fault;
337}
338
339
340#ifndef DOXYGEN_SHOULD_SKIP_THIS
341template
342Fault
343AtomicSimpleCPU::write(uint64_t data, Addr addr,
344                       unsigned flags, uint64_t *res);
345
346template
347Fault
348AtomicSimpleCPU::write(uint32_t data, Addr addr,
349                       unsigned flags, uint64_t *res);
350
351template
352Fault
353AtomicSimpleCPU::write(uint16_t data, Addr addr,
354                       unsigned flags, uint64_t *res);
355
356template
357Fault
358AtomicSimpleCPU::write(uint8_t data, Addr addr,
359                       unsigned flags, uint64_t *res);
360
361#endif //DOXYGEN_SHOULD_SKIP_THIS
362
363template<>
364Fault
365AtomicSimpleCPU::write(double data, Addr addr, unsigned flags, uint64_t *res)
366{
367    return write(*(uint64_t*)&data, addr, flags, res);
368}
369
370template<>
371Fault
372AtomicSimpleCPU::write(float data, Addr addr, unsigned flags, uint64_t *res)
373{
374    return write(*(uint32_t*)&data, addr, flags, res);
375}
376
377
378template<>
379Fault
380AtomicSimpleCPU::write(int32_t data, Addr addr, unsigned flags, uint64_t *res)
381{
382    return write((uint32_t)data, addr, flags, res);
383}
384
385
386void
387AtomicSimpleCPU::tick()
388{
389    Tick latency = cycles(1); // instruction takes one cycle by default
390
391    for (int i = 0; i < width; ++i) {
392        numCycles++;
393
394        checkForInterrupts();
395
396        Fault fault = setupFetchRequest(ifetch_req);
397
398        if (fault == NoFault) {
399            ifetch_pkt->reinitFromRequest();
400
401            Tick icache_latency = icachePort.sendAtomic(ifetch_pkt);
402            // ifetch_req is initialized to read the instruction directly
403            // into the CPU object's inst field.
404
405            dcache_access = false; // assume no dcache access
406            preExecute();
407            fault = curStaticInst->execute(this, traceData);
408            postExecute();
409
410            if (simulate_stalls) {
411                // This calculation assumes that the icache and dcache
412                // access latencies are always a multiple of the CPU's
413                // cycle time.  If not, the next tick event may get
414                // scheduled at a non-integer multiple of the CPU
415                // cycle time.
416                Tick icache_stall = icache_latency - cycles(1);
417                Tick dcache_stall =
418                    dcache_access ? dcache_latency - cycles(1) : 0;
419                latency += icache_stall + dcache_stall;
420            }
421
422        }
423
424        advancePC(fault);
425    }
426
427    if (_status != Idle)
428        tickEvent.schedule(curTick + latency);
429}
430
431
432////////////////////////////////////////////////////////////////////////
433//
434//  AtomicSimpleCPU Simulation Object
435//
436BEGIN_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
437
438    Param<Counter> max_insts_any_thread;
439    Param<Counter> max_insts_all_threads;
440    Param<Counter> max_loads_any_thread;
441    Param<Counter> max_loads_all_threads;
442    SimObjectParam<MemObject *> mem;
443
444#if FULL_SYSTEM
445    SimObjectParam<AlphaITB *> itb;
446    SimObjectParam<AlphaDTB *> dtb;
447    SimObjectParam<System *> system;
448    Param<int> cpu_id;
449    Param<Tick> profile;
450#else
451    SimObjectParam<Process *> workload;
452#endif // FULL_SYSTEM
453
454    Param<int> clock;
455
456    Param<bool> defer_registration;
457    Param<int> width;
458    Param<bool> function_trace;
459    Param<Tick> function_trace_start;
460    Param<bool> simulate_stalls;
461
462END_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
463
464BEGIN_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
465
466    INIT_PARAM(max_insts_any_thread,
467               "terminate when any thread reaches this inst count"),
468    INIT_PARAM(max_insts_all_threads,
469               "terminate when all threads have reached this inst count"),
470    INIT_PARAM(max_loads_any_thread,
471               "terminate when any thread reaches this load count"),
472    INIT_PARAM(max_loads_all_threads,
473               "terminate when all threads have reached this load count"),
474    INIT_PARAM(mem, "memory"),
475
476#if FULL_SYSTEM
477    INIT_PARAM(itb, "Instruction TLB"),
478    INIT_PARAM(dtb, "Data TLB"),
479    INIT_PARAM(system, "system object"),
480    INIT_PARAM(cpu_id, "processor ID"),
481    INIT_PARAM(profile, ""),
482#else
483    INIT_PARAM(workload, "processes to run"),
484#endif // FULL_SYSTEM
485
486    INIT_PARAM(clock, "clock speed"),
487    INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
488    INIT_PARAM(width, "cpu width"),
489    INIT_PARAM(function_trace, "Enable function trace"),
490    INIT_PARAM(function_trace_start, "Cycle to start function trace"),
491    INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
492
493END_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
494
495
496CREATE_SIM_OBJECT(AtomicSimpleCPU)
497{
498    AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
499    params->name = getInstanceName();
500    params->numberOfThreads = 1;
501    params->max_insts_any_thread = max_insts_any_thread;
502    params->max_insts_all_threads = max_insts_all_threads;
503    params->max_loads_any_thread = max_loads_any_thread;
504    params->max_loads_all_threads = max_loads_all_threads;
505    params->deferRegistration = defer_registration;
506    params->clock = clock;
507    params->functionTrace = function_trace;
508    params->functionTraceStart = function_trace_start;
509    params->width = width;
510    params->simulate_stalls = simulate_stalls;
511    params->mem = mem;
512
513#if FULL_SYSTEM
514    params->itb = itb;
515    params->dtb = dtb;
516    params->system = system;
517    params->cpu_id = cpu_id;
518    params->profile = profile;
519#else
520    params->process = workload;
521#endif
522
523    AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
524    return cpu;
525}
526
527REGISTER_SIM_OBJECT("AtomicSimpleCPU", AtomicSimpleCPU)
528
529