summaryrefslogtreecommitdiff
path: root/src/cpu/simple/atomic.cc
blob: 4b243e8626994829be546a1b700598d8f63ce1cd (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
/*
 * Copyright (c) 2002-2005 The Regents of The University of Michigan
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are
 * met: redistributions of source code must retain the above copyright
 * notice, this list of conditions and the following disclaimer;
 * redistributions in binary form must reproduce the above copyright
 * notice, this list of conditions and the following disclaimer in the
 * documentation and/or other materials provided with the distribution;
 * neither the name of the copyright holders nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Authors: Steve Reinhardt
 */

#include "arch/locked_mem.hh"
#include "arch/mmapped_ipr.hh"
#include "arch/utility.hh"
#include "base/bigint.hh"
#include "config/the_isa.hh"
#include "cpu/simple/atomic.hh"
#include "cpu/exetrace.hh"
#include "debug/ExecFaulting.hh"
#include "debug/SimpleCPU.hh"
#include "mem/packet.hh"
#include "mem/packet_access.hh"
#include "params/AtomicSimpleCPU.hh"
#include "sim/faults.hh"
#include "sim/system.hh"
#include "sim/full_system.hh"

using namespace std;
using namespace TheISA;

AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
    : Event(CPU_Tick_Pri), cpu(c)
{
}


void
AtomicSimpleCPU::TickEvent::process()
{
    cpu->tick();
}

const char *
AtomicSimpleCPU::TickEvent::description() const
{
    return "AtomicSimpleCPU tick";
}

Port *
AtomicSimpleCPU::getPort(const string &if_name, int idx)
{
    if (if_name == "dcache_port")
        return &dcachePort;
    else if (if_name == "icache_port")
        return &icachePort;
    else if (if_name == "physmem_port") {
        hasPhysMemPort = true;
        return &physmemPort;
    }
    else
        panic("No Such Port\n");
}

void
AtomicSimpleCPU::init()
{
    BaseCPU::init();
    if (FullSystem) {
        ThreadID size = threadContexts.size();
        for (ThreadID i = 0; i < size; ++i) {
            ThreadContext *tc = threadContexts[i];
            // initialize CPU, including PC
            TheISA::initCPU(tc, tc->contextId());
        }
    }

    // Initialise the ThreadContext's memory proxies
    tcBase()->initMemProxies(tcBase());

    if (hasPhysMemPort) {
        AddrRangeList pmAddrList = physmemPort.getPeer()->getAddrRanges();
        physMemAddr = *pmAddrList.begin();
    }
    // Atomic doesn't do MT right now, so contextId == threadId
    ifetch_req.setThreadContext(_cpuId, 0); // Add thread ID if we add MT
    data_read_req.setThreadContext(_cpuId, 0); // Add thread ID here too
    data_write_req.setThreadContext(_cpuId, 0); // Add thread ID here too
}

AtomicSimpleCPU::AtomicSimpleCPU(AtomicSimpleCPUParams *p)
    : BaseSimpleCPU(p), tickEvent(this), width(p->width), locked(false),
      simulate_data_stalls(p->simulate_data_stalls),
      simulate_inst_stalls(p->simulate_inst_stalls),
      icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this),
      physmemPort(name() + "-iport", this), hasPhysMemPort(false)
{
    _status = Idle;
}


AtomicSimpleCPU::~AtomicSimpleCPU()
{
    if (tickEvent.scheduled()) {
        deschedule(tickEvent);
    }
}

void
AtomicSimpleCPU::serialize(ostream &os)
{
    SimObject::State so_state = SimObject::getState();
    SERIALIZE_ENUM(so_state);
    SERIALIZE_SCALAR(locked);
    BaseSimpleCPU::serialize(os);
    nameOut(os, csprintf("%s.tickEvent", name()));
    tickEvent.serialize(os);
}

void
AtomicSimpleCPU::unserialize(Checkpoint *cp, const string &section)
{
    SimObject::State so_state;
    UNSERIALIZE_ENUM(so_state);
    UNSERIALIZE_SCALAR(locked);
    BaseSimpleCPU::unserialize(cp, section);
    tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
}

void
AtomicSimpleCPU::resume()
{
    if (_status == Idle || _status == SwitchedOut)
        return;

    DPRINTF(SimpleCPU, "Resume\n");
    assert(system->getMemoryMode() == Enums::atomic);

    changeState(SimObject::Running);
    if (thread->status() == ThreadContext::Active) {
        if (!tickEvent.scheduled())
            schedule(tickEvent, nextCycle());
    }
    system->totalNumInsts = 0;
}

void
AtomicSimpleCPU::switchOut()
{
    assert(_status == Running || _status == Idle);
    _status = SwitchedOut;

    tickEvent.squash();
}


void
AtomicSimpleCPU::takeOverFrom(BaseCPU *oldCPU)
{
    BaseCPU::takeOverFrom(oldCPU);

    assert(!tickEvent.scheduled());

    // if any of this CPU's ThreadContexts are active, mark the CPU as
    // running and schedule its tick event.
    ThreadID size = threadContexts.size();
    for (ThreadID i = 0; i < size; ++i) {
        ThreadContext *tc = threadContexts[i];
        if (tc->status() == ThreadContext::Active && _status != Running) {
            _status = Running;
            schedule(tickEvent, nextCycle());
            break;
        }
    }
    if (_status != Running) {
        _status = Idle;
    }
    assert(threadContexts.size() == 1);
    ifetch_req.setThreadContext(_cpuId, 0); // Add thread ID if we add MT
    data_read_req.setThreadContext(_cpuId, 0); // Add thread ID here too
    data_write_req.setThreadContext(_cpuId, 0); // Add thread ID here too
}


void
AtomicSimpleCPU::activateContext(ThreadID thread_num, int delay)
{
    DPRINTF(SimpleCPU, "ActivateContext %d (%d cycles)\n", thread_num, delay);

    assert(thread_num == 0);
    assert(thread);

    assert(_status == Idle);
    assert(!tickEvent.scheduled());

    notIdleFraction++;
    numCycles += tickToCycles(thread->lastActivate - thread->lastSuspend);

    //Make sure ticks are still on multiples of cycles
    schedule(tickEvent, nextCycle(curTick() + ticks(delay)));
    _status = Running;
}


void
AtomicSimpleCPU::suspendContext(ThreadID thread_num)
{
    DPRINTF(SimpleCPU, "SuspendContext %d\n", thread_num);

    assert(thread_num == 0);
    assert(thread);

    if (_status == Idle)
        return;

    assert(_status == Running);

    // tick event may not be scheduled if this gets called from inside
    // an instruction's execution, e.g. "quiesce"
    if (tickEvent.scheduled())
        deschedule(tickEvent);

    notIdleFraction--;
    _status = Idle;
}


Fault
AtomicSimpleCPU::readMem(Addr addr, uint8_t * data,
                         unsigned size, unsigned flags)
{
    // use the CPU's statically allocated read request and packet objects
    Request *req = &data_read_req;

    if (traceData) {
        traceData->setAddr(addr);
    }

    //The block size of our peer.
    unsigned blockSize = dcachePort.peerBlockSize();
    //The size of the data we're trying to read.
    int fullSize = size;

    //The address of the second part of this access if it needs to be split
    //across a cache line boundary.
    Addr secondAddr = roundDown(addr + size - 1, blockSize);

    if (secondAddr > addr)
        size = secondAddr - addr;

    dcache_latency = 0;

    while (1) {
        req->setVirt(0, addr, size, flags, dataMasterId(), thread->pcState().instAddr());

        // translate to physical address
        Fault fault = thread->dtb->translateAtomic(req, tc, BaseTLB::Read);

        // Now do the access.
        if (fault == NoFault && !req->getFlags().isSet(Request::NO_ACCESS)) {
            Packet pkt = Packet(req,
                    req->isLLSC() ? MemCmd::LoadLockedReq : MemCmd::ReadReq,
                    Packet::Broadcast);
            pkt.dataStatic(data);

            if (req->isMmappedIpr())
                dcache_latency += TheISA::handleIprRead(thread->getTC(), &pkt);
            else {
                if (hasPhysMemPort && pkt.getAddr() == physMemAddr)
                    dcache_latency += physmemPort.sendAtomic(&pkt);
                else
                    dcache_latency += dcachePort.sendAtomic(&pkt);
            }
            dcache_access = true;

            assert(!pkt.isError());

            if (req->isLLSC()) {
                TheISA::handleLockedRead(thread, req);
            }
        }

        //If there's a fault, return it
        if (fault != NoFault) {
            if (req->isPrefetch()) {
                return NoFault;
            } else {
                return fault;
            }
        }

        //If we don't need to access a second cache line, stop now.
        if (secondAddr <= addr)
        {
            if (req->isLocked() && fault == NoFault) {
                assert(!locked);
                locked = true;
            }
            return fault;
        }

        /*
         * Set up for accessing the second cache line.
         */

        //Move the pointer we're reading into to the correct location.
        data += size;
        //Adjust the size to get the remaining bytes.
        size = addr + fullSize - secondAddr;
        //And access the right address.
        addr = secondAddr;
    }
}


Fault
AtomicSimpleCPU::writeMem(uint8_t *data, unsigned size,
                          Addr addr, unsigned flags, uint64_t *res)
{
    // use the CPU's statically allocated write request and packet objects
    Request *req = &data_write_req;

    if (traceData) {
        traceData->setAddr(addr);
    }

    //The block size of our peer.
    unsigned blockSize = dcachePort.peerBlockSize();
    //The size of the data we're trying to read.
    int fullSize = size;

    //The address of the second part of this access if it needs to be split
    //across a cache line boundary.
    Addr secondAddr = roundDown(addr + size - 1, blockSize);

    if(secondAddr > addr)
        size = secondAddr - addr;

    dcache_latency = 0;

    while(1) {
        req->setVirt(0, addr, size, flags, dataMasterId(), thread->pcState().instAddr());

        // translate to physical address
        Fault fault = thread->dtb->translateAtomic(req, tc, BaseTLB::Write);

        // Now do the access.
        if (fault == NoFault) {
            MemCmd cmd = MemCmd::WriteReq; // default
            bool do_access = true;  // flag to suppress cache access

            if (req->isLLSC()) {
                cmd = MemCmd::StoreCondReq;
                do_access = TheISA::handleLockedWrite(thread, req);
            } else if (req->isSwap()) {
                cmd = MemCmd::SwapReq;
                if (req->isCondSwap()) {
                    assert(res);
                    req->setExtraData(*res);
                }
            }

            if (do_access && !req->getFlags().isSet(Request::NO_ACCESS)) {
                Packet pkt = Packet(req, cmd, Packet::Broadcast);
                pkt.dataStatic(data);

                if (req->isMmappedIpr()) {
                    dcache_latency +=
                        TheISA::handleIprWrite(thread->getTC(), &pkt);
                } else {
                    if (hasPhysMemPort && pkt.getAddr() == physMemAddr)
                        dcache_latency += physmemPort.sendAtomic(&pkt);
                    else
                        dcache_latency += dcachePort.sendAtomic(&pkt);
                }
                dcache_access = true;
                assert(!pkt.isError());

                if (req->isSwap()) {
                    assert(res);
                    memcpy(res, pkt.getPtr<uint8_t>(), fullSize);
                }
            }

            if (res && !req->isSwap()) {
                *res = req->getExtraData();
            }
        }

        //If there's a fault or we don't need to access a second cache line,
        //stop now.
        if (fault != NoFault || secondAddr <= addr)
        {
            if (req->isLocked() && fault == NoFault) {
                assert(locked);
                locked = false;
            }
            if (fault != NoFault && req->isPrefetch()) {
                return NoFault;
            } else {
                return fault;
            }
        }

        /*
         * Set up for accessing the second cache line.
         */

        //Move the pointer we're reading into to the correct location.
        data += size;
        //Adjust the size to get the remaining bytes.
        size = addr + fullSize - secondAddr;
        //And access the right address.
        addr = secondAddr;
    }
}


void
AtomicSimpleCPU::tick()
{
    DPRINTF(SimpleCPU, "Tick\n");

    Tick latency = 0;

    for (int i = 0; i < width || locked; ++i) {
        numCycles++;

        if (!curStaticInst || !curStaticInst->isDelayedCommit())
            checkForInterrupts();

        checkPcEventQueue();
        // We must have just got suspended by a PC event
        if (_status == Idle)
            return;

        Fault fault = NoFault;

        TheISA::PCState pcState = thread->pcState();

        bool needToFetch = !isRomMicroPC(pcState.microPC()) &&
                           !curMacroStaticInst;
        if (needToFetch) {
            setupFetchRequest(&ifetch_req);
            fault = thread->itb->translateAtomic(&ifetch_req, tc,
                                                 BaseTLB::Execute);
        }

        if (fault == NoFault) {
            Tick icache_latency = 0;
            bool icache_access = false;
            dcache_access = false; // assume no dcache access

            if (needToFetch) {
                // This is commented out because the predecoder would act like
                // a tiny cache otherwise. It wouldn't be flushed when needed
                // like the I cache. It should be flushed, and when that works
                // this code should be uncommented.
                //Fetch more instruction memory if necessary
                //if(predecoder.needMoreBytes())
                //{
                    icache_access = true;
                    Packet ifetch_pkt = Packet(&ifetch_req, MemCmd::ReadReq,
                                               Packet::Broadcast);
                    ifetch_pkt.dataStatic(&inst);

                    if (hasPhysMemPort && ifetch_pkt.getAddr() == physMemAddr)
                        icache_latency = physmemPort.sendAtomic(&ifetch_pkt);
                    else
                        icache_latency = icachePort.sendAtomic(&ifetch_pkt);

                    assert(!ifetch_pkt.isError());

                    // ifetch_req is initialized to read the instruction directly
                    // into the CPU object's inst field.
                //}
            }

            preExecute();

            if (curStaticInst) {
                fault = curStaticInst->execute(this, traceData);

                // keep an instruction count
                if (fault == NoFault)
                    countInst();
                else if (traceData && !DTRACE(ExecFaulting)) {
                    delete traceData;
                    traceData = NULL;
                }

                postExecute();
            }

            // @todo remove me after debugging with legion done
            if (curStaticInst && (!curStaticInst->isMicroop() ||
                        curStaticInst->isFirstMicroop()))
                instCnt++;

            Tick stall_ticks = 0;
            if (simulate_inst_stalls && icache_access)
                stall_ticks += icache_latency;

            if (simulate_data_stalls && dcache_access)
                stall_ticks += dcache_latency;

            if (stall_ticks) {
                Tick stall_cycles = stall_ticks / ticks(1);
                Tick aligned_stall_ticks = ticks(stall_cycles);

                if (aligned_stall_ticks < stall_ticks)
                    aligned_stall_ticks += 1;

                latency += aligned_stall_ticks;
            }

        }
        if(fault != NoFault || !stayAtPC)
            advancePC(fault);
    }

    // instruction takes at least one cycle
    if (latency < ticks(1))
        latency = ticks(1);

    if (_status != Idle)
        schedule(tickEvent, curTick() + latency);
}


void
AtomicSimpleCPU::printAddr(Addr a)
{
    dcachePort.printAddr(a);
}


////////////////////////////////////////////////////////////////////////
//
//  AtomicSimpleCPU Simulation Object
//
AtomicSimpleCPU *
AtomicSimpleCPUParams::create()
{
    numThreads = 1;
    if (!FullSystem && workload.size() != 1)
        panic("only one workload allowed");
    return new AtomicSimpleCPU(this);
}