summaryrefslogtreecommitdiff
path: root/src/mem/cache/cache_impl.hh
blob: 76a99109d3bfbf952d93062c709cdfc3f0da83df (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
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
/*
 * Copyright (c) 2002-2005 The Regents of The University of Michigan
 * Copyright (c) 2010 Advanced Micro Devices, Inc.
 * 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: Erik Hallnor
 *          Dave Greene
 *          Nathan Binkert
 *          Steve Reinhardt
 *          Ron Dreslinski
 */

/**
 * @file
 * Cache definitions.
 */

#include "base/fast_alloc.hh"
#include "base/misc.hh"
#include "base/range.hh"
#include "base/types.hh"
#include "mem/cache/blk.hh"
#include "mem/cache/cache.hh"
#include "mem/cache/mshr.hh"
#include "mem/cache/prefetch/base.hh"
#include "sim/sim_exit.hh"

template<class TagStore>
Cache<TagStore>::Cache(const Params *p, TagStore *tags, BasePrefetcher *pf)
    : BaseCache(p),
      tags(tags),
      prefetcher(pf),
      doFastWrites(true),
      prefetchOnAccess(p->prefetch_on_access)
{
    tempBlock = new BlkType();
    tempBlock->data = new uint8_t[blkSize];

    cpuSidePort = new CpuSidePort(p->name + "-cpu_side_port", this,
                                  "CpuSidePort");
    memSidePort = new MemSidePort(p->name + "-mem_side_port", this,
                                  "MemSidePort");
    cpuSidePort->setOtherPort(memSidePort);
    memSidePort->setOtherPort(cpuSidePort);

    tags->setCache(this);
    if (prefetcher)
        prefetcher->setCache(this);
}

template<class TagStore>
void
Cache<TagStore>::regStats()
{
    BaseCache::regStats();
    tags->regStats(name());
    if (prefetcher)
        prefetcher->regStats(name());
}

template<class TagStore>
Port *
Cache<TagStore>::getPort(const std::string &if_name, int idx)
{
    if (if_name == "" || if_name == "cpu_side") {
        return cpuSidePort;
    } else if (if_name == "mem_side") {
        return memSidePort;
    } else if (if_name == "functional") {
        CpuSidePort *funcPort =
            new CpuSidePort(name() + "-cpu_side_funcport", this,
                            "CpuSideFuncPort");
        funcPort->setOtherPort(memSidePort);
        return funcPort;
    } else {
        panic("Port name %s unrecognized\n", if_name);
    }
}

template<class TagStore>
void
Cache<TagStore>::deletePortRefs(Port *p)
{
    if (cpuSidePort == p || memSidePort == p)
        panic("Can only delete functional ports\n");

    delete p;
}


template<class TagStore>
void
Cache<TagStore>::cmpAndSwap(BlkType *blk, PacketPtr pkt)
{
    uint64_t overwrite_val;
    bool overwrite_mem;
    uint64_t condition_val64;
    uint32_t condition_val32;

    int offset = tags->extractBlkOffset(pkt->getAddr());
    uint8_t *blk_data = blk->data + offset;

    assert(sizeof(uint64_t) >= pkt->getSize());

    overwrite_mem = true;
    // keep a copy of our possible write value, and copy what is at the
    // memory address into the packet
    pkt->writeData((uint8_t *)&overwrite_val);
    pkt->setData(blk_data);

    if (pkt->req->isCondSwap()) {
        if (pkt->getSize() == sizeof(uint64_t)) {
            condition_val64 = pkt->req->getExtraData();
            overwrite_mem = !std::memcmp(&condition_val64, blk_data,
                                         sizeof(uint64_t));
        } else if (pkt->getSize() == sizeof(uint32_t)) {
            condition_val32 = (uint32_t)pkt->req->getExtraData();
            overwrite_mem = !std::memcmp(&condition_val32, blk_data,
                                         sizeof(uint32_t));
        } else
            panic("Invalid size for conditional read/write\n");
    }

    if (overwrite_mem) {
        std::memcpy(blk_data, &overwrite_val, pkt->getSize());
        blk->status |= BlkDirty;
    }
}


template<class TagStore>
void
Cache<TagStore>::satisfyCpuSideRequest(PacketPtr pkt, BlkType *blk)
{
    assert(blk);
    // Occasionally this is not true... if we are a lower-level cache
    // satisfying a string of Read and ReadEx requests from
    // upper-level caches, a Read will mark the block as shared but we
    // can satisfy a following ReadEx anyway since we can rely on the
    // Read requester(s) to have buffered the ReadEx snoop and to
    // invalidate their blocks after receiving them.
    // assert(pkt->needsExclusive() ? blk->isWritable() : blk->isValid());
    assert(pkt->getOffset(blkSize) + pkt->getSize() <= blkSize);

    // Check RMW operations first since both isRead() and
    // isWrite() will be true for them
    if (pkt->cmd == MemCmd::SwapReq) {
        cmpAndSwap(blk, pkt);
    } else if (pkt->isWrite()) {
        if (blk->checkWrite(pkt)) {
            pkt->writeDataToBlock(blk->data, blkSize);
            blk->status |= BlkDirty;
        }
    } else if (pkt->isRead()) {
        if (pkt->isLLSC()) {
            blk->trackLoadLocked(pkt);
        }
        pkt->setDataFromBlock(blk->data, blkSize);
        if (pkt->getSize() == blkSize) {
            // special handling for coherent block requests from
            // upper-level caches
            if (pkt->needsExclusive()) {
                // on ReadExReq we give up our copy
                tags->invalidateBlk(blk);
            } else {
                // on ReadReq we create shareable copies here and in
                // the requester
                pkt->assertShared();
                blk->status &= ~BlkWritable;
            }
        }
    } else {
        // Not a read or write... must be an upgrade.  it's OK
        // to just ack those as long as we have an exclusive
        // copy at this level.
        assert(pkt->isUpgrade());
        tags->invalidateBlk(blk);
    }
}


/////////////////////////////////////////////////////
//
// MSHR helper functions
//
/////////////////////////////////////////////////////


template<class TagStore>
void
Cache<TagStore>::markInService(MSHR *mshr)
{
    markInServiceInternal(mshr);
#if 0
        if (mshr->originalCmd == MemCmd::HardPFReq) {
            DPRINTF(HWPrefetch, "%s:Marking a HW_PF in service\n",
                    name());
            //Also clear pending if need be
            if (!prefetcher->havePending())
            {
                deassertMemSideBusRequest(Request_PF);
            }
        }
#endif
}


template<class TagStore>
void
Cache<TagStore>::squash(int threadNum)
{
    bool unblock = false;
    BlockedCause cause = NUM_BLOCKED_CAUSES;

    if (noTargetMSHR && noTargetMSHR->threadNum == threadNum) {
        noTargetMSHR = NULL;
        unblock = true;
        cause = Blocked_NoTargets;
    }
    if (mshrQueue.isFull()) {
        unblock = true;
        cause = Blocked_NoMSHRs;
    }
    mshrQueue.squash(threadNum);
    if (unblock && !mshrQueue.isFull()) {
        clearBlocked(cause);
    }
}

/////////////////////////////////////////////////////
//
// Access path: requests coming in from the CPU side
//
/////////////////////////////////////////////////////

template<class TagStore>
bool
Cache<TagStore>::access(PacketPtr pkt, BlkType *&blk,
                        int &lat, PacketList &writebacks)
{
    if (pkt->req->isUncacheable())  {
        blk = NULL;
        lat = hitLatency;
        return false;
    }

    int id = pkt->req->hasContextId() ? pkt->req->contextId() : -1;
    blk = tags->accessBlock(pkt->getAddr(), lat, id);

    DPRINTF(Cache, "%s%s %x %s\n", pkt->cmdString(),
            pkt->req->isInstFetch() ? " (ifetch)" : "",
            pkt->getAddr(), (blk) ? "hit" : "miss");

    if (blk != NULL) {

        if (pkt->needsExclusive() ? blk->isWritable() : blk->isReadable()) {
            // OK to satisfy access
            incHitCount(pkt, id);
            satisfyCpuSideRequest(pkt, blk);
            return true;
        }
    }

    // Can't satisfy access normally... either no block (blk == NULL)
    // or have block but need exclusive & only have shared.

    // Writeback handling is special case.  We can write the block
    // into the cache without having a writeable copy (or any copy at
    // all).
    if (pkt->cmd == MemCmd::Writeback) {
        assert(blkSize == pkt->getSize());
        if (blk == NULL) {
            // need to do a replacement
            blk = allocateBlock(pkt->getAddr(), writebacks);
            if (blk == NULL) {
                // no replaceable block available, give up.
                // writeback will be forwarded to next level.
                incMissCount(pkt, id);
                return false;
            }
            int id = pkt->req->hasContextId() ? pkt->req->contextId() : -1;
            tags->insertBlock(pkt->getAddr(), blk, id);
            blk->status = BlkValid | BlkReadable;
        }
        std::memcpy(blk->data, pkt->getPtr<uint8_t>(), blkSize);
        blk->status |= BlkDirty;
        // nothing else to do; writeback doesn't expect response
        assert(!pkt->needsResponse());
        incHitCount(pkt, id);
        return true;
    }

    incMissCount(pkt, id);

    if (blk == NULL && pkt->isLLSC() && pkt->isWrite()) {
        // complete miss on store conditional... just give up now
        pkt->req->setExtraData(0);
        return true;
    }

    return false;
}


class ForwardResponseRecord : public Packet::SenderState, public FastAlloc
{
    Packet::SenderState *prevSenderState;
    int prevSrc;
#ifndef NDEBUG
    BaseCache *cache;
#endif
  public:
    ForwardResponseRecord(Packet *pkt, BaseCache *_cache)
        : prevSenderState(pkt->senderState), prevSrc(pkt->getSrc())
#ifndef NDEBUG
          , cache(_cache)
#endif
    {}
    void restore(Packet *pkt, BaseCache *_cache)
    {
        assert(_cache == cache);
        pkt->senderState = prevSenderState;
        pkt->setDest(prevSrc);
    }
};


template<class TagStore>
bool
Cache<TagStore>::timingAccess(PacketPtr pkt)
{
//@todo Add back in MemDebug Calls
//    MemDebug::cacheAccess(pkt);

    // we charge hitLatency for doing just about anything here
    Tick time =  curTick + hitLatency;

    if (pkt->isResponse()) {
        // must be cache-to-cache response from upper to lower level
        ForwardResponseRecord *rec =
            dynamic_cast<ForwardResponseRecord *>(pkt->senderState);
        assert(rec != NULL);
        rec->restore(pkt, this);
        delete rec;
        memSidePort->respond(pkt, time);
        return true;
    }

    assert(pkt->isRequest());

    if (pkt->memInhibitAsserted()) {
        DPRINTF(Cache, "mem inhibited on 0x%x: not responding\n",
                pkt->getAddr());
        assert(!pkt->req->isUncacheable());
        // Special tweak for multilevel coherence: snoop downward here
        // on invalidates since there may be other caches below here
        // that have shared copies.  Not necessary if we know that
        // supplier had exclusive copy to begin with.
        if (pkt->needsExclusive() && !pkt->isSupplyExclusive()) {
            Packet *snoopPkt = new Packet(pkt, true);  // clear flags
            snoopPkt->setExpressSnoop();
            snoopPkt->assertMemInhibit();
            memSidePort->sendTiming(snoopPkt);
            // main memory will delete snoopPkt
        }
        // since we're the official target but we aren't responding,
        // delete the packet now.
        delete pkt;
        return true;
    }

    if (pkt->req->isUncacheable()) {
        // writes go in write buffer, reads use MSHR
        if (pkt->isWrite() && !pkt->isRead()) {
            allocateWriteBuffer(pkt, time, true);
        } else {
            allocateUncachedReadBuffer(pkt, time, true);
        }
        assert(pkt->needsResponse()); // else we should delete it here??
        return true;
    }

    int lat = hitLatency;
    BlkType *blk = NULL;
    PacketList writebacks;

    bool satisfied = access(pkt, blk, lat, writebacks);

#if 0
    /** @todo make the fast write alloc (wh64) work with coherence. */

    // If this is a block size write/hint (WH64) allocate the block here
    // if the coherence protocol allows it.
    if (!blk && pkt->getSize() >= blkSize && coherence->allowFastWrites() &&
        (pkt->cmd == MemCmd::WriteReq
         || pkt->cmd == MemCmd::WriteInvalidateReq) ) {
        // not outstanding misses, can do this
        MSHR *outstanding_miss = mshrQueue.findMatch(pkt->getAddr());
        if (pkt->cmd == MemCmd::WriteInvalidateReq || !outstanding_miss) {
            if (outstanding_miss) {
                warn("WriteInv doing a fastallocate"
                     "with an outstanding miss to the same address\n");
            }
            blk = handleFill(NULL, pkt, BlkValid | BlkWritable,
                                   writebacks);
            ++fastWrites;
        }
    }
#endif

    // track time of availability of next prefetch, if any
    Tick next_pf_time = 0;

    bool needsResponse = pkt->needsResponse();

    if (satisfied) {
        if (needsResponse) {
            pkt->makeTimingResponse();
            cpuSidePort->respond(pkt, curTick+lat);
        } else {
            delete pkt;
        }

        if (prefetcher && (prefetchOnAccess || (blk && blk->wasPrefetched()))) {
            if (blk)
                blk->status &= ~BlkHWPrefetched;
            next_pf_time = prefetcher->notify(pkt, time);
        }
    } else {
        // miss

        Addr blk_addr = blockAlign(pkt->getAddr());
        MSHR *mshr = mshrQueue.findMatch(blk_addr);

        if (mshr) {
            // MSHR hit
            //@todo remove hw_pf here
            mshr_hits[pkt->cmdToIndex()][0/*pkt->req->threadId()*/]++;
            if (mshr->threadNum != 0/*pkt->req->threadId()*/) {
                mshr->threadNum = -1;
            }
            mshr->allocateTarget(pkt, time, order++);
            if (mshr->getNumTargets() == numTarget) {
                noTargetMSHR = mshr;
                setBlocked(Blocked_NoTargets);
                // need to be careful with this... if this mshr isn't
                // ready yet (i.e. time > curTick_, we don't want to
                // move it ahead of mshrs that are ready
                // mshrQueue.moveToFront(mshr);
            }
        } else {
            // no MSHR
            mshr_misses[pkt->cmdToIndex()][0/*pkt->req->threadId()*/]++;
            // always mark as cache fill for now... if we implement
            // no-write-allocate or bypass accesses this will have to
            // be changed.
            if (pkt->cmd == MemCmd::Writeback) {
                allocateWriteBuffer(pkt, time, true);
            } else {
                if (blk && blk->isValid()) {
                    // If we have a write miss to a valid block, we
                    // need to mark the block non-readable.  Otherwise
                    // if we allow reads while there's an outstanding
                    // write miss, the read could return stale data
                    // out of the cache block... a more aggressive
                    // system could detect the overlap (if any) and
                    // forward data out of the MSHRs, but we don't do
                    // that yet.  Note that we do need to leave the
                    // block valid so that it stays in the cache, in
                    // case we get an upgrade response (and hence no
                    // new data) when the write miss completes.
                    // As long as CPUs do proper store/load forwarding
                    // internally, and have a sufficiently weak memory
                    // model, this is probably unnecessary, but at some
                    // point it must have seemed like we needed it...
                    assert(pkt->needsExclusive() && !blk->isWritable());
                    blk->status &= ~BlkReadable;
                }

                allocateMissBuffer(pkt, time, true);
            }

            if (prefetcher) {
                next_pf_time = prefetcher->notify(pkt, time);
            }
        }
    }

    if (next_pf_time != 0)
        requestMemSideBus(Request_PF, std::max(time, next_pf_time));

    // copy writebacks to write buffer
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        allocateWriteBuffer(wbPkt, time, true);
        writebacks.pop_front();
    }

    return true;
}


// See comment in cache.hh.
template<class TagStore>
PacketPtr
Cache<TagStore>::getBusPacket(PacketPtr cpu_pkt, BlkType *blk,
                              bool needsExclusive)
{
    bool blkValid = blk && blk->isValid();

    if (cpu_pkt->req->isUncacheable()) {
        assert(blk == NULL);
        return NULL;
    }

    if (!blkValid &&
        (cpu_pkt->cmd == MemCmd::Writeback || cpu_pkt->isUpgrade())) {
        // Writebacks that weren't allocated in access() and upgrades
        // from upper-level caches that missed completely just go
        // through.
        return NULL;
    }

    assert(cpu_pkt->needsResponse());

    MemCmd cmd;
    // @TODO make useUpgrades a parameter.
    // Note that ownership protocols require upgrade, otherwise a
    // write miss on a shared owned block will generate a ReadExcl,
    // which will clobber the owned copy.
    const bool useUpgrades = true;
    if (blkValid && useUpgrades) {
        // only reason to be here is that blk is shared
        // (read-only) and we need exclusive
        assert(needsExclusive && !blk->isWritable());
        cmd = cpu_pkt->isLLSC() ? MemCmd::SCUpgradeReq : MemCmd::UpgradeReq;
    } else {
        // block is invalid
        cmd = needsExclusive ? MemCmd::ReadExReq : MemCmd::ReadReq;
    }
    PacketPtr pkt = new Packet(cpu_pkt->req, cmd, Packet::Broadcast, blkSize);

    pkt->allocate();
    return pkt;
}


template<class TagStore>
Tick
Cache<TagStore>::atomicAccess(PacketPtr pkt)
{
    int lat = hitLatency;

    // @TODO: make this a parameter
    bool last_level_cache = false;

    if (pkt->memInhibitAsserted()) {
        assert(!pkt->req->isUncacheable());
        // have to invalidate ourselves and any lower caches even if
        // upper cache will be responding
        if (pkt->isInvalidate()) {
            BlkType *blk = tags->findBlock(pkt->getAddr());
            if (blk && blk->isValid()) {
                tags->invalidateBlk(blk);
                DPRINTF(Cache, "rcvd mem-inhibited %s on 0x%x: invalidating\n",
                        pkt->cmdString(), pkt->getAddr());
            }
            if (!last_level_cache) {
                DPRINTF(Cache, "forwarding mem-inhibited %s on 0x%x\n",
                        pkt->cmdString(), pkt->getAddr());
                lat += memSidePort->sendAtomic(pkt);
            }
        } else {
            DPRINTF(Cache, "rcvd mem-inhibited %s on 0x%x: not responding\n",
                    pkt->cmdString(), pkt->getAddr());
        }

        return lat;
    }

    // should assert here that there are no outstanding MSHRs or
    // writebacks... that would mean that someone used an atomic
    // access in timing mode

    BlkType *blk = NULL;
    PacketList writebacks;

    if (!access(pkt, blk, lat, writebacks)) {
        // MISS
        PacketPtr bus_pkt = getBusPacket(pkt, blk, pkt->needsExclusive());

        bool is_forward = (bus_pkt == NULL);

        if (is_forward) {
            // just forwarding the same request to the next level
            // no local cache operation involved
            bus_pkt = pkt;
        }

        DPRINTF(Cache, "Sending an atomic %s for %x\n",
                bus_pkt->cmdString(), bus_pkt->getAddr());

#if TRACING_ON
        CacheBlk::State old_state = blk ? blk->status : 0;
#endif

        lat += memSidePort->sendAtomic(bus_pkt);

        DPRINTF(Cache, "Receive response: %s for addr %x in state %i\n",
                bus_pkt->cmdString(), bus_pkt->getAddr(), old_state);

        assert(!bus_pkt->wasNacked());

        // If packet was a forward, the response (if any) is already
        // in place in the bus_pkt == pkt structure, so we don't need
        // to do anything.  Otherwise, use the separate bus_pkt to
        // generate response to pkt and then delete it.
        if (!is_forward) {
            if (pkt->needsResponse()) {
                assert(bus_pkt->isResponse());
                if (bus_pkt->isError()) {
                    pkt->makeAtomicResponse();
                    pkt->copyError(bus_pkt);
                } else if (bus_pkt->isRead() ||
                           bus_pkt->cmd == MemCmd::UpgradeResp) {
                    // we're updating cache state to allow us to
                    // satisfy the upstream request from the cache
                    blk = handleFill(bus_pkt, blk, writebacks);
                    satisfyCpuSideRequest(pkt, blk);
                } else {
                    // we're satisfying the upstream request without
                    // modifying cache state, e.g., a write-through
                    pkt->makeAtomicResponse();
                }
            }
            delete bus_pkt;
        }
    }

    // Note that we don't invoke the prefetcher at all in atomic mode.
    // It's not clear how to do it properly, particularly for
    // prefetchers that aggressively generate prefetch candidates and
    // rely on bandwidth contention to throttle them; these will tend
    // to pollute the cache in atomic mode since there is no bandwidth
    // contention.  If we ever do want to enable prefetching in atomic
    // mode, though, this is the place to do it... see timingAccess()
    // for an example (though we'd want to issue the prefetch(es)
    // immediately rather than calling requestMemSideBus() as we do
    // there).

    // Handle writebacks if needed
    while (!writebacks.empty()){
        PacketPtr wbPkt = writebacks.front();
        memSidePort->sendAtomic(wbPkt);
        writebacks.pop_front();
        delete wbPkt;
    }

    // We now have the block one way or another (hit or completed miss)

    if (pkt->needsResponse()) {
        pkt->makeAtomicResponse();
    }

    return lat;
}


template<class TagStore>
void
Cache<TagStore>::functionalAccess(PacketPtr pkt,
                                  CachePort *incomingPort,
                                  CachePort *otherSidePort)
{
    Addr blk_addr = blockAlign(pkt->getAddr());
    BlkType *blk = tags->findBlock(pkt->getAddr());

    pkt->pushLabel(name());

    CacheBlkPrintWrapper cbpw(blk);
    bool done =
        (blk && pkt->checkFunctional(&cbpw, blk_addr, blkSize, blk->data))
        || incomingPort->checkFunctional(pkt)
        || mshrQueue.checkFunctional(pkt, blk_addr)
        || writeBuffer.checkFunctional(pkt, blk_addr)
        || otherSidePort->checkFunctional(pkt);

    // We're leaving the cache, so pop cache->name() label
    pkt->popLabel();

    if (!done) {
        otherSidePort->sendFunctional(pkt);
    }
}


/////////////////////////////////////////////////////
//
// Response handling: responses from the memory side
//
/////////////////////////////////////////////////////


template<class TagStore>
void
Cache<TagStore>::handleResponse(PacketPtr pkt)
{
    Tick time = curTick + hitLatency;
    MSHR *mshr = dynamic_cast<MSHR*>(pkt->senderState);
    bool is_error = pkt->isError();

    assert(mshr);

    if (pkt->wasNacked()) {
        //pkt->reinitFromRequest();
        warn("NACKs from devices not connected to the same bus "
             "not implemented\n");
        return;
    }
    if (is_error) {
        DPRINTF(Cache, "Cache received packet with error for address %x, "
                "cmd: %s\n", pkt->getAddr(), pkt->cmdString());
    }

    DPRINTF(Cache, "Handling response to %x\n", pkt->getAddr());

    MSHRQueue *mq = mshr->queue;
    bool wasFull = mq->isFull();

    if (mshr == noTargetMSHR) {
        // we always clear at least one target
        clearBlocked(Blocked_NoTargets);
        noTargetMSHR = NULL;
    }

    // Initial target is used just for stats
    MSHR::Target *initial_tgt = mshr->getTarget();
    BlkType *blk = tags->findBlock(pkt->getAddr());
    int stats_cmd_idx = initial_tgt->pkt->cmdToIndex();
    Tick miss_latency = curTick - initial_tgt->recvTime;
    PacketList writebacks;

    if (pkt->req->isUncacheable()) {
        mshr_uncacheable_lat[stats_cmd_idx][0/*pkt->req->threadId()*/] +=
            miss_latency;
    } else {
        mshr_miss_latency[stats_cmd_idx][0/*pkt->req->threadId()*/] +=
            miss_latency;
    }

    bool is_fill = !mshr->isForward &&
        (pkt->isRead() || pkt->cmd == MemCmd::UpgradeResp);

    if (is_fill && !is_error) {
        DPRINTF(Cache, "Block for addr %x being updated in Cache\n",
                pkt->getAddr());

        // give mshr a chance to do some dirty work
        mshr->handleFill(pkt, blk);

        blk = handleFill(pkt, blk, writebacks);
        assert(blk != NULL);
    }

    // First offset for critical word first calculations
    int initial_offset = 0;

    if (mshr->hasTargets()) {
        initial_offset = mshr->getTarget()->pkt->getOffset(blkSize);
    }

    while (mshr->hasTargets()) {
        MSHR::Target *target = mshr->getTarget();

        switch (target->source) {
          case MSHR::Target::FromCPU:
            Tick completion_time;
            if (is_fill) {
                satisfyCpuSideRequest(target->pkt, blk);
                // How many bytes past the first request is this one
                int transfer_offset =
                    target->pkt->getOffset(blkSize) - initial_offset;
                if (transfer_offset < 0) {
                    transfer_offset += blkSize;
                }

                // If critical word (no offset) return first word time
                completion_time = tags->getHitLatency() +
                    (transfer_offset ? pkt->finishTime : pkt->firstWordTime);

                assert(!target->pkt->req->isUncacheable());
                missLatency[target->pkt->cmdToIndex()][0/*pkt->req->threadId()*/] +=
                    completion_time - target->recvTime;
            } else if (target->pkt->cmd == MemCmd::StoreCondReq &&
                       pkt->cmd == MemCmd::UpgradeFailResp) {
                // failed StoreCond upgrade
                completion_time = tags->getHitLatency() + pkt->finishTime;
                target->pkt->req->setExtraData(0);
            } else {
                // not a cache fill, just forwarding response
                completion_time = tags->getHitLatency() + pkt->finishTime;
                if (pkt->isRead() && !is_error) {
                    target->pkt->setData(pkt->getPtr<uint8_t>());
                }
            }
            target->pkt->makeTimingResponse();
            // if this packet is an error copy that to the new packet
            if (is_error)
                target->pkt->copyError(pkt);
            if (pkt->isInvalidate()) {
                // If intermediate cache got ReadRespWithInvalidate,
                // propagate that.  Response should not have
                // isInvalidate() set otherwise.
                assert(target->pkt->cmd == MemCmd::ReadResp);
                assert(pkt->cmd == MemCmd::ReadRespWithInvalidate);
                target->pkt->cmd = MemCmd::ReadRespWithInvalidate;
            }
            cpuSidePort->respond(target->pkt, completion_time);
            break;

          case MSHR::Target::FromPrefetcher:
            assert(target->pkt->cmd == MemCmd::HardPFReq);
            if (blk)
                blk->status |= BlkHWPrefetched;
            delete target->pkt->req;
            delete target->pkt;
            break;

          case MSHR::Target::FromSnoop:
            // I don't believe that a snoop can be in an error state
            assert(!is_error);
            // response to snoop request
            DPRINTF(Cache, "processing deferred snoop...\n");
            handleSnoop(target->pkt, blk, true, true,
                        mshr->pendingInvalidate || pkt->isInvalidate());
            break;

          default:
            panic("Illegal target->source enum %d\n", target->source);
        }

        mshr->popTarget();
    }

    if (pkt->isInvalidate()) {
        tags->invalidateBlk(blk);
    }

    if (mshr->promoteDeferredTargets()) {
        // avoid later read getting stale data while write miss is
        // outstanding.. see comment in timingAccess()
        blk->status &= ~BlkReadable;
        MSHRQueue *mq = mshr->queue;
        mq->markPending(mshr);
        requestMemSideBus((RequestCause)mq->index, pkt->finishTime);
    } else {
        mq->deallocate(mshr);
        if (wasFull && !mq->isFull()) {
            clearBlocked((BlockedCause)mq->index);
        }
    }

    // copy writebacks to write buffer
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        allocateWriteBuffer(wbPkt, time, true);
        writebacks.pop_front();
    }
    // if we used temp block, clear it out
    if (blk == tempBlock) {
        if (blk->isDirty()) {
            allocateWriteBuffer(writebackBlk(blk), time, true);
        }
        tags->invalidateBlk(blk);
    }

    delete pkt;
}




template<class TagStore>
PacketPtr
Cache<TagStore>::writebackBlk(BlkType *blk)
{
    assert(blk && blk->isValid() && blk->isDirty());

    writebacks[0/*pkt->req->threadId()*/]++;

    Request *writebackReq =
        new Request(tags->regenerateBlkAddr(blk->tag, blk->set), blkSize, 0);
    PacketPtr writeback = new Packet(writebackReq, MemCmd::Writeback, -1);
    writeback->allocate();
    std::memcpy(writeback->getPtr<uint8_t>(), blk->data, blkSize);

    blk->status &= ~BlkDirty;
    return writeback;
}


template<class TagStore>
typename Cache<TagStore>::BlkType*
Cache<TagStore>::allocateBlock(Addr addr, PacketList &writebacks)
{
    BlkType *blk = tags->findVictim(addr, writebacks);

    if (blk->isValid()) {
        Addr repl_addr = tags->regenerateBlkAddr(blk->tag, blk->set);
        MSHR *repl_mshr = mshrQueue.findMatch(repl_addr);
        if (repl_mshr) {
            // must be an outstanding upgrade request on block
            // we're about to replace...
            assert(!blk->isWritable());
            assert(repl_mshr->needsExclusive());
            // too hard to replace block with transient state
            // allocation failed, block not inserted
            return NULL;
        } else {
            DPRINTF(Cache, "replacement: replacing %x with %x: %s\n",
                    repl_addr, addr,
                    blk->isDirty() ? "writeback" : "clean");

            if (blk->isDirty()) {
                // Save writeback packet for handling by caller
                writebacks.push_back(writebackBlk(blk));
            }
        }
    }

    return blk;
}


// Note that the reason we return a list of writebacks rather than
// inserting them directly in the write buffer is that this function
// is called by both atomic and timing-mode accesses, and in atomic
// mode we don't mess with the write buffer (we just perform the
// writebacks atomically once the original request is complete).
template<class TagStore>
typename Cache<TagStore>::BlkType*
Cache<TagStore>::handleFill(PacketPtr pkt, BlkType *blk,
                            PacketList &writebacks)
{
    Addr addr = pkt->getAddr();
#if TRACING_ON
    CacheBlk::State old_state = blk ? blk->status : 0;
#endif

    if (blk == NULL) {
        // better have read new data...
        assert(pkt->hasData());
        // need to do a replacement
        blk = allocateBlock(addr, writebacks);
        if (blk == NULL) {
            // No replaceable block... just use temporary storage to
            // complete the current request and then get rid of it
            assert(!tempBlock->isValid());
            blk = tempBlock;
            tempBlock->set = tags->extractSet(addr);
            tempBlock->tag = tags->extractTag(addr);
            DPRINTF(Cache, "using temp block for %x\n", addr);
        } else {
            int id = pkt->req->hasContextId() ? pkt->req->contextId() : -1;
            tags->insertBlock(pkt->getAddr(), blk, id);
        }
    } else {
        // existing block... probably an upgrade
        assert(blk->tag == tags->extractTag(addr));
        // either we're getting new data or the block should already be valid
        assert(pkt->hasData() || blk->isValid());
    }

    blk->status = BlkValid | BlkReadable;

    if (!pkt->sharedAsserted()) {
        blk->status |= BlkWritable;
        // If we got this via cache-to-cache transfer (i.e., from a
        // cache that was an owner) and took away that owner's copy,
        // then we need to write it back.  Normally this happens
        // anyway as a side effect of getting a copy to write it, but
        // there are cases (such as failed store conditionals or
        // compare-and-swaps) where we'll demand an exclusive copy but
        // end up not writing it.
        if (pkt->memInhibitAsserted())
            blk->status |= BlkDirty;
    }

    DPRINTF(Cache, "Block addr %x moving from state %i to %i\n",
            addr, old_state, blk->status);

    // if we got new data, copy it in
    if (pkt->isRead()) {
        std::memcpy(blk->data, pkt->getPtr<uint8_t>(), blkSize);
    }

    blk->whenReady = pkt->finishTime;

    return blk;
}


/////////////////////////////////////////////////////
//
// Snoop path: requests coming in from the memory side
//
/////////////////////////////////////////////////////

template<class TagStore>
void
Cache<TagStore>::
doTimingSupplyResponse(PacketPtr req_pkt, uint8_t *blk_data,
                       bool already_copied, bool pending_inval)
{
    // timing-mode snoop responses require a new packet, unless we
    // already made a copy...
    PacketPtr pkt = already_copied ? req_pkt : new Packet(req_pkt);
    assert(req_pkt->isInvalidate() || pkt->sharedAsserted());
    pkt->allocate();
    pkt->makeTimingResponse();
    if (pkt->isRead()) {
        pkt->setDataFromBlock(blk_data, blkSize);
    }
    if (pkt->cmd == MemCmd::ReadResp && pending_inval) {
        // Assume we defer a response to a read from a far-away cache
        // A, then later defer a ReadExcl from a cache B on the same
        // bus as us.  We'll assert MemInhibit in both cases, but in
        // the latter case MemInhibit will keep the invalidation from
        // reaching cache A.  This special response tells cache A that
        // it gets the block to satisfy its read, but must immediately
        // invalidate it.
        pkt->cmd = MemCmd::ReadRespWithInvalidate;
    }
    memSidePort->respond(pkt, curTick + hitLatency);
}

template<class TagStore>
void
Cache<TagStore>::handleSnoop(PacketPtr pkt, BlkType *blk,
                             bool is_timing, bool is_deferred,
                             bool pending_inval)
{
    // deferred snoops can only happen in timing mode
    assert(!(is_deferred && !is_timing));
    // pending_inval only makes sense on deferred snoops
    assert(!(pending_inval && !is_deferred));
    assert(pkt->isRequest());

    if (forwardSnoops) {
        // first propagate snoop upward to see if anyone above us wants to
        // handle it.  save & restore packet src since it will get
        // rewritten to be relative to cpu-side bus (if any)
        bool alreadyResponded = pkt->memInhibitAsserted();
        if (is_timing) {
            Packet *snoopPkt = new Packet(pkt, true);  // clear flags
            snoopPkt->setExpressSnoop();
            snoopPkt->senderState = new ForwardResponseRecord(pkt, this);
            cpuSidePort->sendTiming(snoopPkt);
            if (snoopPkt->memInhibitAsserted()) {
                // cache-to-cache response from some upper cache
                assert(!alreadyResponded);
                pkt->assertMemInhibit();
            } else {
                delete snoopPkt->senderState;
            }
            if (snoopPkt->sharedAsserted()) {
                pkt->assertShared();
            }
            delete snoopPkt;
        } else {
            int origSrc = pkt->getSrc();
            cpuSidePort->sendAtomic(pkt);
            if (!alreadyResponded && pkt->memInhibitAsserted()) {
                // cache-to-cache response from some upper cache:
                // forward response to original requester
                assert(pkt->isResponse());
            }
            pkt->setSrc(origSrc);
        }
    }

    if (!blk || !blk->isValid()) {
        return;
    }

    // we may end up modifying both the block state and the packet (if
    // we respond in atomic mode), so just figure out what to do now
    // and then do it later
    bool respond = blk->isDirty() && pkt->needsResponse();
    bool have_exclusive = blk->isWritable();
    bool invalidate = pkt->isInvalidate();

    if (pkt->isRead() && !pkt->isInvalidate()) {
        assert(!pkt->needsExclusive());
        pkt->assertShared();
        int bits_to_clear = BlkWritable;
        const bool haveOwnershipState = true; // for now
        if (!haveOwnershipState) {
            // if we don't support pure ownership (dirty && !writable),
            // have to clear dirty bit here, assume memory snarfs data
            // on cache-to-cache xfer
            bits_to_clear |= BlkDirty;
        }
        blk->status &= ~bits_to_clear;
    }

    DPRINTF(Cache, "snooped a %s request for addr %x, %snew state is %i\n",
            pkt->cmdString(), blockAlign(pkt->getAddr()),
            respond ? "responding, " : "", invalidate ? 0 : blk->status);

    if (respond) {
        assert(!pkt->memInhibitAsserted());
        pkt->assertMemInhibit();
        if (have_exclusive) {
            pkt->setSupplyExclusive();
        }
        if (is_timing) {
            doTimingSupplyResponse(pkt, blk->data, is_deferred, pending_inval);
        } else {
            pkt->makeAtomicResponse();
            pkt->setDataFromBlock(blk->data, blkSize);
        }
    } else if (is_timing && is_deferred) {
        // if it's a deferred timing snoop then we've made a copy of
        // the packet, and so if we're not using that copy to respond
        // then we need to delete it here.
        delete pkt;
    }

    // Do this last in case it deallocates block data or something
    // like that
    if (invalidate) {
        tags->invalidateBlk(blk);
    }
}


template<class TagStore>
void
Cache<TagStore>::snoopTiming(PacketPtr pkt)
{
    // Note that some deferred snoops don't have requests, since the
    // original access may have already completed
    if ((pkt->req && pkt->req->isUncacheable()) ||
        pkt->cmd == MemCmd::Writeback) {
        //Can't get a hit on an uncacheable address
        //Revisit this for multi level coherence
        return;
    }

    BlkType *blk = tags->findBlock(pkt->getAddr());

    Addr blk_addr = blockAlign(pkt->getAddr());
    MSHR *mshr = mshrQueue.findMatch(blk_addr);

    // Let the MSHR itself track the snoop and decide whether we want
    // to go ahead and do the regular cache snoop
    if (mshr && mshr->handleSnoop(pkt, order++)) {
        DPRINTF(Cache, "Deferring snoop on in-service MSHR to blk %x\n",
                blk_addr);
        if (mshr->getNumTargets() > numTarget)
            warn("allocating bonus target for snoop"); //handle later
        return;
    }

    //We also need to check the writeback buffers and handle those
    std::vector<MSHR *> writebacks;
    if (writeBuffer.findMatches(blk_addr, writebacks)) {
        DPRINTF(Cache, "Snoop hit in writeback to addr: %x\n",
                pkt->getAddr());

        //Look through writebacks for any non-uncachable writes, use that
        for (int i = 0; i < writebacks.size(); i++) {
            mshr = writebacks[i];
            assert(!mshr->isUncacheable());
            assert(mshr->getNumTargets() == 1);
            PacketPtr wb_pkt = mshr->getTarget()->pkt;
            assert(wb_pkt->cmd == MemCmd::Writeback);

            assert(!pkt->memInhibitAsserted());
            pkt->assertMemInhibit();
            if (!pkt->needsExclusive()) {
                pkt->assertShared();
            } else {
                // if we're not asserting the shared line, we need to
                // invalidate our copy.  we'll do that below as long as
                // the packet's invalidate flag is set...
                assert(pkt->isInvalidate());
            }
            doTimingSupplyResponse(pkt, wb_pkt->getPtr<uint8_t>(),
                                   false, false);

            if (pkt->isInvalidate()) {
                // Invalidation trumps our writeback... discard here
                markInService(mshr);
                delete wb_pkt;
            }

            // If this was a shared writeback, there may still be
            // other shared copies above that require invalidation.
            // We could be more selective and return here if the
            // request is non-exclusive or if the writeback is
            // exclusive.
            break;
        }
    }

    handleSnoop(pkt, blk, true, false, false);
}


template<class TagStore>
Tick
Cache<TagStore>::snoopAtomic(PacketPtr pkt)
{
    if (pkt->req->isUncacheable() || pkt->cmd == MemCmd::Writeback) {
        // Can't get a hit on an uncacheable address
        // Revisit this for multi level coherence
        return hitLatency;
    }

    BlkType *blk = tags->findBlock(pkt->getAddr());
    handleSnoop(pkt, blk, false, false, false);
    return hitLatency;
}


template<class TagStore>
MSHR *
Cache<TagStore>::getNextMSHR()
{
    // Check both MSHR queue and write buffer for potential requests
    MSHR *miss_mshr  = mshrQueue.getNextMSHR();
    MSHR *write_mshr = writeBuffer.getNextMSHR();

    // Now figure out which one to send... some cases are easy
    if (miss_mshr && !write_mshr) {
        return miss_mshr;
    }
    if (write_mshr && !miss_mshr) {
        return write_mshr;
    }

    if (miss_mshr && write_mshr) {
        // We have one of each... normally we favor the miss request
        // unless the write buffer is full
        if (writeBuffer.isFull() && writeBuffer.inServiceEntries == 0) {
            // Write buffer is full, so we'd like to issue a write;
            // need to search MSHR queue for conflicting earlier miss.
            MSHR *conflict_mshr =
                mshrQueue.findPending(write_mshr->addr, write_mshr->size);

            if (conflict_mshr && conflict_mshr->order < write_mshr->order) {
                // Service misses in order until conflict is cleared.
                return conflict_mshr;
            }

            // No conflicts; issue write
            return write_mshr;
        }

        // Write buffer isn't full, but need to check it for
        // conflicting earlier writeback
        MSHR *conflict_mshr =
            writeBuffer.findPending(miss_mshr->addr, miss_mshr->size);
        if (conflict_mshr) {
            // not sure why we don't check order here... it was in the
            // original code but commented out.

            // The only way this happens is if we are
            // doing a write and we didn't have permissions
            // then subsequently saw a writeback (owned got evicted)
            // We need to make sure to perform the writeback first
            // To preserve the dirty data, then we can issue the write

            // should we return write_mshr here instead?  I.e. do we
            // have to flush writes in order?  I don't think so... not
            // for Alpha anyway.  Maybe for x86?
            return conflict_mshr;
        }

        // No conflicts; issue read
        return miss_mshr;
    }

    // fall through... no pending requests.  Try a prefetch.
    assert(!miss_mshr && !write_mshr);
    if (prefetcher && !mshrQueue.isFull()) {
        // If we have a miss queue slot, we can try a prefetch
        PacketPtr pkt = prefetcher->getPacket();
        if (pkt) {
            Addr pf_addr = blockAlign(pkt->getAddr());
            if (!tags->findBlock(pf_addr) && !mshrQueue.findMatch(pf_addr)) {
                // Update statistic on number of prefetches issued
                // (hwpf_mshr_misses)
                mshr_misses[pkt->cmdToIndex()][0/*pkt->req->threadId()*/]++;
                // Don't request bus, since we already have it
                return allocateMissBuffer(pkt, curTick, false);
            }
        }
    }

    return NULL;
}


template<class TagStore>
PacketPtr
Cache<TagStore>::getTimingPacket()
{
    MSHR *mshr = getNextMSHR();

    if (mshr == NULL) {
        return NULL;
    }

    // use request from 1st target
    PacketPtr tgt_pkt = mshr->getTarget()->pkt;
    PacketPtr pkt = NULL;

    if (tgt_pkt->cmd == MemCmd::SCUpgradeFailReq) {
        // SCUpgradeReq saw invalidation while queued in MSHR, so now
        // that we are getting around to processing it, just treat it
        // as if we got a failure response
        pkt = new Packet(tgt_pkt);
        pkt->cmd = MemCmd::UpgradeFailResp;
        pkt->senderState = mshr;
        pkt->firstWordTime = pkt->finishTime = curTick;
        handleResponse(pkt);
        return NULL;
    } else if (mshr->isForwardNoResponse()) {
        // no response expected, just forward packet as it is
        assert(tags->findBlock(mshr->addr) == NULL);
        pkt = tgt_pkt;
    } else {
        BlkType *blk = tags->findBlock(mshr->addr);
        pkt = getBusPacket(tgt_pkt, blk, mshr->needsExclusive());

        mshr->isForward = (pkt == NULL);

        if (mshr->isForward) {
            // not a cache block request, but a response is expected
            // make copy of current packet to forward, keep current
            // copy for response handling
            pkt = new Packet(tgt_pkt);
            pkt->allocate();
            if (pkt->isWrite()) {
                pkt->setData(tgt_pkt->getPtr<uint8_t>());
            }
        }
    }

    assert(pkt != NULL);
    pkt->senderState = mshr;
    return pkt;
}


template<class TagStore>
Tick
Cache<TagStore>::nextMSHRReadyTime()
{
    Tick nextReady = std::min(mshrQueue.nextMSHRReadyTime(),
                              writeBuffer.nextMSHRReadyTime());

    if (prefetcher) {
        nextReady = std::min(nextReady,
                             prefetcher->nextPrefetchReadyTime());
    }

    return nextReady;
}


///////////////
//
// CpuSidePort
//
///////////////

template<class TagStore>
void
Cache<TagStore>::CpuSidePort::
getDeviceAddressRanges(AddrRangeList &resp, bool &snoop)
{
    // CPU side port doesn't snoop; it's a target only.  It can
    // potentially respond to any address.
    snoop = false;
    resp.push_back(myCache()->getAddrRange());
}


template<class TagStore>
bool
Cache<TagStore>::CpuSidePort::recvTiming(PacketPtr pkt)
{
    // illegal to block responses... can lead to deadlock
    if (pkt->isRequest() && !pkt->memInhibitAsserted() && blocked) {
        DPRINTF(Cache,"Scheduling a retry while blocked\n");
        mustSendRetry = true;
        return false;
    }

    myCache()->timingAccess(pkt);
    return true;
}


template<class TagStore>
Tick
Cache<TagStore>::CpuSidePort::recvAtomic(PacketPtr pkt)
{
    return myCache()->atomicAccess(pkt);
}


template<class TagStore>
void
Cache<TagStore>::CpuSidePort::recvFunctional(PacketPtr pkt)
{
    myCache()->functionalAccess(pkt, this, otherPort);
}


template<class TagStore>
Cache<TagStore>::
CpuSidePort::CpuSidePort(const std::string &_name, Cache<TagStore> *_cache,
                         const std::string &_label)
    : BaseCache::CachePort(_name, _cache, _label)
{
}

///////////////
//
// MemSidePort
//
///////////////

template<class TagStore>
void
Cache<TagStore>::MemSidePort::
getDeviceAddressRanges(AddrRangeList &resp, bool &snoop)
{
    // Memory-side port always snoops, but never passes requests
    // through to targets on the cpu side (so we don't add anything to
    // the address range list).
    snoop = true;
}


template<class TagStore>
bool
Cache<TagStore>::MemSidePort::recvTiming(PacketPtr pkt)
{
    // this needs to be fixed so that the cache updates the mshr and sends the
    // packet back out on the link, but it probably won't happen so until this
    // gets fixed, just panic when it does
    if (pkt->wasNacked())
        panic("Need to implement cache resending nacked packets!\n");

    if (pkt->isRequest() && blocked) {
        DPRINTF(Cache,"Scheduling a retry while blocked\n");
        mustSendRetry = true;
        return false;
    }

    if (pkt->isResponse()) {
        myCache()->handleResponse(pkt);
    } else {
        myCache()->snoopTiming(pkt);
    }
    return true;
}


template<class TagStore>
Tick
Cache<TagStore>::MemSidePort::recvAtomic(PacketPtr pkt)
{
    // in atomic mode, responses go back to the sender via the
    // function return from sendAtomic(), not via a separate
    // sendAtomic() from the responder.  Thus we should never see a
    // response packet in recvAtomic() (anywhere, not just here).
    assert(!pkt->isResponse());
    return myCache()->snoopAtomic(pkt);
}


template<class TagStore>
void
Cache<TagStore>::MemSidePort::recvFunctional(PacketPtr pkt)
{
    myCache()->functionalAccess(pkt, this, otherPort);
}



template<class TagStore>
void
Cache<TagStore>::MemSidePort::sendPacket()
{
    // if we have responses that are ready, they take precedence
    if (deferredPacketReady()) {
        bool success = sendTiming(transmitList.front().pkt);

        if (success) {
            //send successful, remove packet
            transmitList.pop_front();
        }

        waitingOnRetry = !success;
    } else {
        // check for non-response packets (requests & writebacks)
        PacketPtr pkt = myCache()->getTimingPacket();
        if (pkt == NULL) {
            // can happen if e.g. we attempt a writeback and fail, but
            // before the retry, the writeback is eliminated because
            // we snoop another cache's ReadEx.
            waitingOnRetry = false;
        } else {
            MSHR *mshr = dynamic_cast<MSHR*>(pkt->senderState);

            bool success = sendTiming(pkt);

            waitingOnRetry = !success;
            if (waitingOnRetry) {
                DPRINTF(CachePort, "now waiting on a retry\n");
                if (!mshr->isForwardNoResponse()) {
                    delete pkt;
                }
            } else {
                myCache()->markInService(mshr);
            }
        }
    }


    // tried to send packet... if it was successful (no retry), see if
    // we need to rerequest bus or not
    if (!waitingOnRetry) {
        Tick nextReady = std::min(deferredPacketReadyTime(),
                                  myCache()->nextMSHRReadyTime());
        // @TODO: need to facotr in prefetch requests here somehow
        if (nextReady != MaxTick) {
            DPRINTF(CachePort, "more packets to send @ %d\n", nextReady);
            schedule(sendEvent, std::max(nextReady, curTick + 1));
        } else {
            // no more to send right now: if we're draining, we may be done
            if (drainEvent) {
                drainEvent->process();
                drainEvent = NULL;
            }
        }
    }
}

template<class TagStore>
void
Cache<TagStore>::MemSidePort::recvRetry()
{
    assert(waitingOnRetry);
    sendPacket();
}


template<class TagStore>
void
Cache<TagStore>::MemSidePort::processSendEvent()
{
    assert(!waitingOnRetry);
    sendPacket();
}


template<class TagStore>
Cache<TagStore>::
MemSidePort::MemSidePort(const std::string &_name, Cache<TagStore> *_cache,
                         const std::string &_label)
    : BaseCache::CachePort(_name, _cache, _label)
{
    // override default send event from SimpleTimingPort
    delete sendEvent;
    sendEvent = new SendEvent(this);
}