atomic.cc (4776:8c8407243a2c) atomic.cc (4870:fcc39d001154)
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;

--- 22 unchanged lines hidden (view full) ---

31#include "arch/locked_mem.hh"
32#include "arch/mmaped_ipr.hh"
33#include "arch/utility.hh"
34#include "base/bigint.hh"
35#include "cpu/exetrace.hh"
36#include "cpu/simple/atomic.hh"
37#include "mem/packet.hh"
38#include "mem/packet_access.hh"
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;

--- 22 unchanged lines hidden (view full) ---

31#include "arch/locked_mem.hh"
32#include "arch/mmaped_ipr.hh"
33#include "arch/utility.hh"
34#include "base/bigint.hh"
35#include "cpu/exetrace.hh"
36#include "cpu/simple/atomic.hh"
37#include "mem/packet.hh"
38#include "mem/packet_access.hh"
39#include "params/AtomicSimpleCPU.hh"
39#include "sim/builder.hh"
40#include "sim/system.hh"
41
42using namespace std;
43using namespace TheISA;
44
45AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
46 : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
47{

--- 95 unchanged lines hidden (view full) ---

143 width(p->width), simulate_stalls(p->simulate_stalls),
144 icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
145{
146 _status = Idle;
147
148 icachePort.snoopRangeSent = false;
149 dcachePort.snoopRangeSent = false;
150
40#include "sim/system.hh"
41
42using namespace std;
43using namespace TheISA;
44
45AtomicSimpleCPU::TickEvent::TickEvent(AtomicSimpleCPU *c)
46 : Event(&mainEventQueue, CPU_Tick_Pri), cpu(c)
47{

--- 95 unchanged lines hidden (view full) ---

143 width(p->width), simulate_stalls(p->simulate_stalls),
144 icachePort(name() + "-iport", this), dcachePort(name() + "-iport", this)
145{
146 _status = Idle;
147
148 icachePort.snoopRangeSent = false;
149 dcachePort.snoopRangeSent = false;
150
151 ifetch_req = new Request();
152 ifetch_req->setThreadContext(p->cpu_id, 0); // Add thread ID if we add MT
153 ifetch_pkt = new Packet(ifetch_req, MemCmd::ReadReq, Packet::Broadcast);
154 ifetch_pkt->dataStatic(&inst);
155
156 data_read_req = new Request();
157 data_read_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
158 data_read_pkt = new Packet(data_read_req, MemCmd::ReadReq,
159 Packet::Broadcast);
160 data_read_pkt->dataStatic(&dataReg);
161
162 data_write_req = new Request();
163 data_write_req->setThreadContext(p->cpu_id, 0); // Add thread ID here too
164 data_write_pkt = new Packet(data_write_req, MemCmd::WriteReq,
165 Packet::Broadcast);
166 data_swap_pkt = new Packet(data_write_req, MemCmd::SwapReq,
167 Packet::Broadcast);
151 ifetch_req.setThreadContext(p->cpu_id, 0); // Add thread ID if we add MT
152 data_read_req.setThreadContext(p->cpu_id, 0); // Add thread ID here too
153 data_write_req.setThreadContext(p->cpu_id, 0); // Add thread ID here too
168}
169
170
171AtomicSimpleCPU::~AtomicSimpleCPU()
172{
173}
174
175void

--- 17 unchanged lines hidden (view full) ---

193 BaseSimpleCPU::unserialize(cp, section);
194 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
195}
196
197void
198AtomicSimpleCPU::resume()
199{
200 if (_status != SwitchedOut && _status != Idle) {
154}
155
156
157AtomicSimpleCPU::~AtomicSimpleCPU()
158{
159}
160
161void

--- 17 unchanged lines hidden (view full) ---

179 BaseSimpleCPU::unserialize(cp, section);
180 tickEvent.unserialize(cp, csprintf("%s.tickEvent", section));
181}
182
183void
184AtomicSimpleCPU::resume()
185{
186 if (_status != SwitchedOut && _status != Idle) {
201 assert(system->getMemoryMode() == Enums::atomic);
187 assert(system->getMemoryMode() == System::Atomic);
202
203 changeState(SimObject::Running);
204 if (thread->status() == ThreadContext::Active) {
205 if (!tickEvent.scheduled()) {
206 tickEvent.schedule(nextCycle());
207 }
208 }
209 }

--- 67 unchanged lines hidden (view full) ---

277}
278
279
280template <class T>
281Fault
282AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
283{
284 // use the CPU's statically allocated read request and packet objects
188
189 changeState(SimObject::Running);
190 if (thread->status() == ThreadContext::Active) {
191 if (!tickEvent.scheduled()) {
192 tickEvent.schedule(nextCycle());
193 }
194 }
195 }

--- 67 unchanged lines hidden (view full) ---

263}
264
265
266template <class T>
267Fault
268AtomicSimpleCPU::read(Addr addr, T &data, unsigned flags)
269{
270 // use the CPU's statically allocated read request and packet objects
285 Request *req = data_read_req;
286 PacketPtr pkt = data_read_pkt;
287
271 Request *req = &data_read_req;
288 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
289
290 if (traceData) {
291 traceData->setAddr(addr);
292 }
293
294 // translate to physical address
295 Fault fault = thread->translateDataReadReq(req);
296
297 // Now do the access.
298 if (fault == NoFault) {
272 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
273
274 if (traceData) {
275 traceData->setAddr(addr);
276 }
277
278 // translate to physical address
279 Fault fault = thread->translateDataReadReq(req);
280
281 // Now do the access.
282 if (fault == NoFault) {
299 pkt->reinitFromRequest();
283 Packet pkt = Packet(req, MemCmd::ReadReq, Packet::Broadcast);
284 pkt.dataStatic(&data);
300
301 if (req->isMmapedIpr())
285
286 if (req->isMmapedIpr())
302 dcache_latency = TheISA::handleIprRead(thread->getTC(),pkt);
287 dcache_latency = TheISA::handleIprRead(thread->getTC(), &pkt);
303 else
288 else
304 dcache_latency = dcachePort.sendAtomic(pkt);
289 dcache_latency = dcachePort.sendAtomic(&pkt);
305 dcache_access = true;
290 dcache_access = true;
306#if !defined(NDEBUG)
307 if (pkt->result != Packet::Success)
308 panic("Unable to find responder for address pa = %#X va = %#X\n",
309 pkt->req->getPaddr(), pkt->req->getVaddr());
310#endif
311 data = pkt->get<T>();
291 assert(!pkt.isError());
312
313 if (req->isLocked()) {
314 TheISA::handleLockedRead(thread, req);
315 }
316 }
317
318 // This will need a new way to tell if it has a dcache attached.
319 if (req->isUncacheable())

--- 53 unchanged lines hidden (view full) ---

373}
374
375
376template <class T>
377Fault
378AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
379{
380 // use the CPU's statically allocated write request and packet objects
292
293 if (req->isLocked()) {
294 TheISA::handleLockedRead(thread, req);
295 }
296 }
297
298 // This will need a new way to tell if it has a dcache attached.
299 if (req->isUncacheable())

--- 53 unchanged lines hidden (view full) ---

353}
354
355
356template <class T>
357Fault
358AtomicSimpleCPU::write(T data, Addr addr, unsigned flags, uint64_t *res)
359{
360 // use the CPU's statically allocated write request and packet objects
381 Request *req = data_write_req;
382 PacketPtr pkt;
383
361 Request *req = &data_write_req;
384 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
385
362 req->setVirt(0, addr, sizeof(T), flags, thread->readPC());
363
386 if (req->isSwap())
387 pkt = data_swap_pkt;
388 else
389 pkt = data_write_pkt;
390
391 if (traceData) {
392 traceData->setAddr(addr);
393 }
394
395 // translate to physical address
396 Fault fault = thread->translateDataWriteReq(req);
397
398 // Now do the access.
399 if (fault == NoFault) {
364 if (traceData) {
365 traceData->setAddr(addr);
366 }
367
368 // translate to physical address
369 Fault fault = thread->translateDataWriteReq(req);
370
371 // Now do the access.
372 if (fault == NoFault) {
373 Packet pkt =
374 Packet(req, req->isSwap() ? MemCmd::SwapReq : MemCmd::WriteReq,
375 Packet::Broadcast);
376 pkt.dataStatic(&data);
377
400 bool do_access = true; // flag to suppress cache access
401
402 if (req->isLocked()) {
403 do_access = TheISA::handleLockedWrite(thread, req);
404 }
405 if (req->isCondSwap()) {
406 assert(res);
407 req->setExtraData(*res);
408 }
409
410
411 if (do_access) {
378 bool do_access = true; // flag to suppress cache access
379
380 if (req->isLocked()) {
381 do_access = TheISA::handleLockedWrite(thread, req);
382 }
383 if (req->isCondSwap()) {
384 assert(res);
385 req->setExtraData(*res);
386 }
387
388
389 if (do_access) {
412 pkt->reinitFromRequest();
413 pkt->dataStatic(&data);
414
415 if (req->isMmapedIpr()) {
390 if (req->isMmapedIpr()) {
416 dcache_latency = TheISA::handleIprWrite(thread->getTC(), pkt);
391 dcache_latency = TheISA::handleIprWrite(thread->getTC(), &pkt);
417 } else {
418 data = htog(data);
392 } else {
393 data = htog(data);
419 dcache_latency = dcachePort.sendAtomic(pkt);
394 dcache_latency = dcachePort.sendAtomic(&pkt);
420 }
421 dcache_access = true;
395 }
396 dcache_access = true;
422
423#if !defined(NDEBUG)
424 if (pkt->result != Packet::Success)
425 panic("Unable to find responder for address pa = %#X va = %#X\n",
426 pkt->req->getPaddr(), pkt->req->getVaddr());
427#endif
397 assert(!pkt.isError());
428 }
429
430 if (req->isSwap()) {
431 assert(res);
398 }
399
400 if (req->isSwap()) {
401 assert(res);
432 *res = pkt->get<T>();
402 *res = pkt.get<T>();
433 } else if (res) {
434 *res = req->getExtraData();
435 }
436 }
437
438 // This will need a new way to tell if it's hooked up to a cache or not.
439 if (req->isUncacheable())
440 recordEvent("Uncached Write");

--- 67 unchanged lines hidden (view full) ---

508 Tick latency = cycles(1); // instruction takes one cycle by default
509
510 for (int i = 0; i < width; ++i) {
511 numCycles++;
512
513 if (!curStaticInst || !curStaticInst->isDelayedCommit())
514 checkForInterrupts();
515
403 } else if (res) {
404 *res = req->getExtraData();
405 }
406 }
407
408 // This will need a new way to tell if it's hooked up to a cache or not.
409 if (req->isUncacheable())
410 recordEvent("Uncached Write");

--- 67 unchanged lines hidden (view full) ---

478 Tick latency = cycles(1); // instruction takes one cycle by default
479
480 for (int i = 0; i < width; ++i) {
481 numCycles++;
482
483 if (!curStaticInst || !curStaticInst->isDelayedCommit())
484 checkForInterrupts();
485
516 Fault fault = setupFetchRequest(ifetch_req);
486 Fault fault = setupFetchRequest(&ifetch_req);
517
518 if (fault == NoFault) {
519 Tick icache_latency = 0;
520 bool icache_access = false;
521 dcache_access = false; // assume no dcache access
522
523 //Fetch more instruction memory if necessary
524 //if(predecoder.needMoreBytes())
525 //{
526 icache_access = true;
487
488 if (fault == NoFault) {
489 Tick icache_latency = 0;
490 bool icache_access = false;
491 dcache_access = false; // assume no dcache access
492
493 //Fetch more instruction memory if necessary
494 //if(predecoder.needMoreBytes())
495 //{
496 icache_access = true;
527 ifetch_pkt->reinitFromRequest();
497 Packet ifetch_pkt = Packet(&ifetch_req, MemCmd::ReadReq,
498 Packet::Broadcast);
499 ifetch_pkt.dataStatic(&inst);
528
500
529 icache_latency = icachePort.sendAtomic(ifetch_pkt);
501 icache_latency = icachePort.sendAtomic(&ifetch_pkt);
530 // ifetch_req is initialized to read the instruction directly
531 // into the CPU object's inst field.
532 //}
533
534 preExecute();
535
536 if(curStaticInst)
537 {

--- 27 unchanged lines hidden (view full) ---

565 tickEvent.schedule(curTick + latency);
566}
567
568
569////////////////////////////////////////////////////////////////////////
570//
571// AtomicSimpleCPU Simulation Object
572//
502 // ifetch_req is initialized to read the instruction directly
503 // into the CPU object's inst field.
504 //}
505
506 preExecute();
507
508 if(curStaticInst)
509 {

--- 27 unchanged lines hidden (view full) ---

537 tickEvent.schedule(curTick + latency);
538}
539
540
541////////////////////////////////////////////////////////////////////////
542//
543// AtomicSimpleCPU Simulation Object
544//
573AtomicSimpleCPU *
574AtomicSimpleCPUParams::create()
545BEGIN_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
546
547 Param<Counter> max_insts_any_thread;
548 Param<Counter> max_insts_all_threads;
549 Param<Counter> max_loads_any_thread;
550 Param<Counter> max_loads_all_threads;
551 Param<Tick> progress_interval;
552 SimObjectParam<System *> system;
553 Param<int> cpu_id;
554
555#if FULL_SYSTEM
556 SimObjectParam<TheISA::ITB *> itb;
557 SimObjectParam<TheISA::DTB *> dtb;
558 Param<Tick> profile;
559
560 Param<bool> do_quiesce;
561 Param<bool> do_checkpoint_insts;
562 Param<bool> do_statistics_insts;
563#else
564 SimObjectParam<Process *> workload;
565#endif // FULL_SYSTEM
566
567 Param<int> clock;
568 Param<int> phase;
569
570 Param<bool> defer_registration;
571 Param<int> width;
572 Param<bool> function_trace;
573 Param<Tick> function_trace_start;
574 Param<bool> simulate_stalls;
575
576END_DECLARE_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
577
578BEGIN_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
579
580 INIT_PARAM(max_insts_any_thread,
581 "terminate when any thread reaches this inst count"),
582 INIT_PARAM(max_insts_all_threads,
583 "terminate when all threads have reached this inst count"),
584 INIT_PARAM(max_loads_any_thread,
585 "terminate when any thread reaches this load count"),
586 INIT_PARAM(max_loads_all_threads,
587 "terminate when all threads have reached this load count"),
588 INIT_PARAM(progress_interval, "Progress interval"),
589 INIT_PARAM(system, "system object"),
590 INIT_PARAM(cpu_id, "processor ID"),
591
592#if FULL_SYSTEM
593 INIT_PARAM(itb, "Instruction TLB"),
594 INIT_PARAM(dtb, "Data TLB"),
595 INIT_PARAM(profile, ""),
596 INIT_PARAM(do_quiesce, ""),
597 INIT_PARAM(do_checkpoint_insts, ""),
598 INIT_PARAM(do_statistics_insts, ""),
599#else
600 INIT_PARAM(workload, "processes to run"),
601#endif // FULL_SYSTEM
602
603 INIT_PARAM(clock, "clock speed"),
604 INIT_PARAM_DFLT(phase, "clock phase", 0),
605 INIT_PARAM(defer_registration, "defer system registration (for sampling)"),
606 INIT_PARAM(width, "cpu width"),
607 INIT_PARAM(function_trace, "Enable function trace"),
608 INIT_PARAM(function_trace_start, "Cycle to start function trace"),
609 INIT_PARAM(simulate_stalls, "Simulate cache stall cycles")
610
611END_INIT_SIM_OBJECT_PARAMS(AtomicSimpleCPU)
612
613
614CREATE_SIM_OBJECT(AtomicSimpleCPU)
575{
576 AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
615{
616 AtomicSimpleCPU::Params *params = new AtomicSimpleCPU::Params();
577 params->name = name;
617 params->name = getInstanceName();
578 params->numberOfThreads = 1;
579 params->max_insts_any_thread = max_insts_any_thread;
580 params->max_insts_all_threads = max_insts_all_threads;
581 params->max_loads_any_thread = max_loads_any_thread;
582 params->max_loads_all_threads = max_loads_all_threads;
583 params->progress_interval = progress_interval;
584 params->deferRegistration = defer_registration;
585 params->phase = phase;
586 params->clock = clock;
587 params->functionTrace = function_trace;
588 params->functionTraceStart = function_trace_start;
589 params->width = width;
590 params->simulate_stalls = simulate_stalls;
591 params->system = system;
592 params->cpu_id = cpu_id;
618 params->numberOfThreads = 1;
619 params->max_insts_any_thread = max_insts_any_thread;
620 params->max_insts_all_threads = max_insts_all_threads;
621 params->max_loads_any_thread = max_loads_any_thread;
622 params->max_loads_all_threads = max_loads_all_threads;
623 params->progress_interval = progress_interval;
624 params->deferRegistration = defer_registration;
625 params->phase = phase;
626 params->clock = clock;
627 params->functionTrace = function_trace;
628 params->functionTraceStart = function_trace_start;
629 params->width = width;
630 params->simulate_stalls = simulate_stalls;
631 params->system = system;
632 params->cpu_id = cpu_id;
593 params->tracer = tracer;
594
595#if FULL_SYSTEM
596 params->itb = itb;
597 params->dtb = dtb;
598 params->profile = profile;
599 params->do_quiesce = do_quiesce;
600 params->do_checkpoint_insts = do_checkpoint_insts;
601 params->do_statistics_insts = do_statistics_insts;
602#else
633
634#if FULL_SYSTEM
635 params->itb = itb;
636 params->dtb = dtb;
637 params->profile = profile;
638 params->do_quiesce = do_quiesce;
639 params->do_checkpoint_insts = do_checkpoint_insts;
640 params->do_statistics_insts = do_statistics_insts;
641#else
603 if (workload.size() != 1)
604 panic("only one workload allowed");
605 params->process = workload[0];
642 params->process = workload;
606#endif
607
608 AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
609 return cpu;
610}
643#endif
644
645 AtomicSimpleCPU *cpu = new AtomicSimpleCPU(params);
646 return cpu;
647}
648
649REGISTER_SIM_OBJECT("AtomicSimpleCPU", AtomicSimpleCPU)
650