forked from RPCS3/rpcs3
-
Notifications
You must be signed in to change notification settings - Fork 1
/
CPUThread.cpp
729 lines (606 loc) · 14.5 KB
/
CPUThread.cpp
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
#include "stdafx.h"
#include "CPUThread.h"
#include "Emu/System.h"
#include "Emu/system_config.h"
#include "Emu/Memory/vm_locking.h"
#include "Emu/IdManager.h"
#include "Emu/GDB.h"
#include "Emu/Cell/PPUThread.h"
#include "Emu/Cell/SPUThread.h"
#include <thread>
#include <unordered_map>
#include <map>
DECLARE(cpu_thread::g_threads_created){0};
DECLARE(cpu_thread::g_threads_deleted){0};
LOG_CHANNEL(profiler);
LOG_CHANNEL(sys_log, "SYS");
template <>
void fmt_class_string<cpu_flag>::format(std::string& out, u64 arg)
{
format_enum(out, arg, [](cpu_flag f)
{
switch (f)
{
case cpu_flag::stop: return "STOP";
case cpu_flag::exit: return "EXIT";
case cpu_flag::wait: return "w";
case cpu_flag::pause: return "p";
case cpu_flag::suspend: return "s";
case cpu_flag::ret: return "ret";
case cpu_flag::signal: return "sig";
case cpu_flag::memory: return "mem";
case cpu_flag::dbg_global_pause: return "G-PAUSE";
case cpu_flag::dbg_global_stop: return "G-EXIT";
case cpu_flag::dbg_pause: return "PAUSE";
case cpu_flag::dbg_step: return "STEP";
case cpu_flag::__bitset_enum_max: break;
}
return unknown;
});
}
template<>
void fmt_class_string<bs_t<cpu_flag>>::format(std::string& out, u64 arg)
{
format_bitset(out, arg, "[", "|", "]", &fmt_class_string<cpu_flag>::format);
}
// CPU profiler thread
struct cpu_prof
{
// PPU/SPU id enqueued for registration
lf_queue<u32> registered;
struct sample_info
{
// Weak pointer to the thread
std::weak_ptr<cpu_thread> wptr;
// Block occurences: name -> sample_count
std::unordered_map<u64, u64, value_hash<u64>> freq;
// Total number of samples
u64 samples = 0, idle = 0;
sample_info(const std::shared_ptr<cpu_thread>& ptr)
: wptr(ptr)
{
}
void reset()
{
freq.clear();
samples = 0;
idle = 0;
}
// Print info
void print(u32 id) const
{
// Make reversed map: sample_count -> name
std::multimap<u64, u64, std::greater<u64>> chart;
for (auto& [name, count] : freq)
{
chart.emplace(count, name);
}
// Print results
std::string results;
results.reserve(5100);
// Fraction of non-idle samples
const f64 busy = 1. * (samples - idle) / samples;
for (auto& [count, name] : chart)
{
const f64 _frac = count / busy / samples;
// Print only 7 hash characters out of 11 (which covers roughly 48 bits)
fmt::append(results, "\n\t[%s", fmt::base57(be_t<u64>{name}));
results.resize(results.size() - 4);
// Print chunk address from lowest 16 bits
fmt::append(results, "...chunk-0x%05x]: %.4f%% (%u)", (name & 0xffff) * 4, _frac * 100., count);
if (results.size() >= 5000)
{
// Stop printing after reaching some arbitrary limit in characters
break;
}
}
profiler.notice("Thread [0x%08x]: %u samples (%.4f%% idle):%s", id, samples, 100. * idle / samples, results);
}
};
void operator()()
{
std::unordered_map<u32, sample_info, value_hash<u64>> threads;
while (thread_ctrl::state() != thread_state::aborting)
{
bool flush = false;
// Handle registration channel
for (u32 id : registered.pop_all())
{
if (id == 0)
{
// Handle id zero as a command to flush results
flush = true;
continue;
}
std::shared_ptr<cpu_thread> ptr;
if (id >> 24 == 1)
{
ptr = idm::get<named_thread<ppu_thread>>(id);
}
else if (id >> 24 == 2)
{
ptr = idm::get<named_thread<spu_thread>>(id);
}
else
{
profiler.error("Invalid Thread ID: 0x%08x", id);
continue;
}
if (ptr)
{
auto [found, add] = threads.try_emplace(id, ptr);
if (!add)
{
// Overwritten: print previous data
found->second.print(id);
found->second.reset();
found->second.wptr = ptr;
}
}
}
if (threads.empty())
{
// Wait for messages if no work (don't waste CPU)
registered.wait();
continue;
}
// Sample active threads
for (auto& [id, info] : threads)
{
if (auto ptr = info.wptr.lock())
{
// Get short function hash
const u64 name = atomic_storage<u64>::load(ptr->block_hash);
// Append occurrence
info.samples++;
if (!(ptr->state.load() & (cpu_flag::wait + cpu_flag::stop + cpu_flag::dbg_global_pause)))
{
info.freq[name]++;
// Append verification time to fixed common name 0000000...chunk-0x3fffc
if ((name & 0xffff) == 0)
info.freq[0xffff]++;
}
else
{
info.idle++;
}
}
}
// Cleanup and print results for deleted threads
for (auto it = threads.begin(), end = threads.end(); it != end;)
{
if (it->second.wptr.expired())
it->second.print(it->first), it = threads.erase(it);
else
it++;
}
if (flush)
{
profiler.success("Flushing profiling results...");
// Print all results and cleanup
for (auto& [id, info] : threads)
{
info.print(id);
info.reset();
}
}
// Wait, roughly for 20µs
thread_ctrl::wait_for(20, false);
}
// Print all remaining results
for (auto& [id, info] : threads)
{
info.print(id);
}
}
static constexpr auto thread_name = "CPU Profiler"sv;
};
using cpu_profiler = named_thread<cpu_prof>;
thread_local cpu_thread* g_tls_current_cpu_thread = nullptr;
struct cpu_counter
{
// For synchronizing suspend_all operation
alignas(64) shared_mutex cpu_suspend_lock;
// Semaphore for global thread array (global counter)
alignas(64) atomic_t<u32> cpu_array_sema{0};
// Semaphore subdivision for each array slot (64 x N in total)
atomic_t<u64> cpu_array_bits[6]{};
// All registered threads
atomic_t<cpu_thread*> cpu_array[sizeof(cpu_array_bits) * 8]{};
u64 add(cpu_thread* _this)
{
if (!cpu_array_sema.try_inc(sizeof(cpu_counter::cpu_array_bits) * 8))
{
return -1;
}
u64 array_slot = -1;
for (u32 i = 0;; i = (i + 1) % ::size32(cpu_array_bits))
{
const auto [bits, ok] = cpu_array_bits[i].fetch_op([](u64& bits) -> u64
{
if (~bits) [[likely]]
{
// Set lowest clear bit
bits |= bits + 1;
return true;
}
return false;
});
if (ok) [[likely]]
{
// Get actual slot number
array_slot = i * 64 + utils::cnttz64(~bits, false);
break;
}
}
// Register and wait if necessary
verify("cpu_counter::add()" HERE), cpu_array[array_slot].exchange(_this) == nullptr;
_this->state += cpu_flag::wait;
cpu_suspend_lock.lock_unlock();
return array_slot;
}
void remove(cpu_thread* _this, u64 slot)
{
// Unregister and wait if necessary
_this->state += cpu_flag::wait;
if (cpu_array[slot].exchange(nullptr) != _this)
sys_log.fatal("Inconsistency for array slot %u", slot);
cpu_array_bits[slot / 64] &= ~(1ull << (slot % 64));
cpu_array_sema--;
cpu_suspend_lock.lock_unlock();
}
};
template <typename F>
void for_all_cpu(F&& func) noexcept
{
auto ctr = g_fxo->get<cpu_counter>();
for (u32 i = 0; i < ::size32(ctr->cpu_array_bits); i++)
{
for (u64 bits = ctr->cpu_array_bits[i]; bits; bits &= bits - 1)
{
const u64 index = i * 64 + utils::cnttz64(bits, true);
if (cpu_thread* cpu = ctr->cpu_array[index].load())
{
func(cpu);
}
}
}
}
void cpu_thread::operator()()
{
g_tls_current_cpu_thread = this;
if (g_cfg.core.thread_scheduler_enabled)
{
thread_ctrl::set_thread_affinity_mask(thread_ctrl::get_affinity_mask(id_type() == 1 ? thread_class::ppu : thread_class::spu));
}
if (g_cfg.core.lower_spu_priority && id_type() == 2)
{
thread_ctrl::set_native_priority(-1);
}
if (id_type() == 2)
{
// force input/output denormals to zero for SPU threads (FTZ/DAZ)
_mm_setcsr( _mm_getcsr() | 0x8040 );
const volatile int a = 0x1fc00000;
__m128 b = _mm_castsi128_ps(_mm_set1_epi32(a));
int c = _mm_cvtsi128_si32(_mm_castps_si128(_mm_mul_ps(b,b)));
if (c != 0)
{
sys_log.fatal("Could not disable denormals.");
}
}
while (!g_fxo->get<cpu_counter>() && !g_fxo->get<cpu_profiler>())
{
// Can we have a little race, right? First thread is started concurrently with g_fxo->init()
std::this_thread::sleep_for(1ms);
}
if (id_type() == 1 && false)
{
g_fxo->get<cpu_profiler>()->registered.push(id);
}
if (id_type() == 2 && g_cfg.core.spu_prof)
{
g_fxo->get<cpu_profiler>()->registered.push(id);
}
// Register thread in g_cpu_array
const u64 array_slot = g_fxo->get<cpu_counter>()->add(this);
if (array_slot == umax)
{
sys_log.fatal("Too many threads.");
return;
}
static thread_local struct thread_cleanup_t
{
cpu_thread* _this;
u64 slot;
std::string name;
thread_cleanup_t(cpu_thread* _this, u64 slot)
: _this(_this)
, slot(slot)
, name(thread_ctrl::get_name())
{
}
void cleanup()
{
if (_this == nullptr)
{
return;
}
if (auto ptr = vm::g_tls_locked)
{
ptr->compare_and_swap(_this, nullptr);
}
g_fxo->get<cpu_counter>()->remove(_this, slot);
_this = nullptr;
}
~thread_cleanup_t()
{
if (_this)
{
sys_log.warning("CPU Thread '%s' terminated abnormally:\n%s", name, _this->dump_all());
cleanup();
}
}
} cleanup{this, array_slot};
// Check thread status
while (!(state & (cpu_flag::exit + cpu_flag::dbg_global_stop)) && thread_ctrl::state() != thread_state::aborting)
{
// Check stop status
if (!(state & cpu_flag::stop))
{
cpu_task();
state -= cpu_flag::ret;
continue;
}
thread_ctrl::wait();
}
// Complete cleanup gracefully
cleanup.cleanup();
}
cpu_thread::~cpu_thread()
{
vm::cleanup_unlock(*this);
g_threads_deleted++;
}
cpu_thread::cpu_thread(u32 id)
: id(id)
{
g_threads_created++;
}
bool cpu_thread::check_state() noexcept
{
if (state & cpu_flag::dbg_pause)
{
g_fxo->get<gdb_server>()->pause_from(this);
}
bool cpu_sleep_called = false;
bool cpu_flag_memory = false;
if (!(state & cpu_flag::wait))
{
state += cpu_flag::wait;
}
while (true)
{
if (state & cpu_flag::memory)
{
if (auto& ptr = vm::g_tls_locked)
{
ptr->compare_and_swap(this, nullptr);
ptr = nullptr;
}
cpu_flag_memory = true;
state -= cpu_flag::memory;
}
if (state & (cpu_flag::exit + cpu_flag::dbg_global_stop))
{
state += cpu_flag::wait;
return true;
}
const auto [state0, escape] = state.fetch_op([&](bs_t<cpu_flag>& flags)
{
// Atomically clean wait flag and escape
if (!(flags & (cpu_flag::exit + cpu_flag::dbg_global_stop + cpu_flag::ret + cpu_flag::stop)))
{
// Check pause flags which hold thread inside check_state
if (flags & (cpu_flag::pause + cpu_flag::suspend + cpu_flag::dbg_global_pause + cpu_flag::dbg_pause))
{
return false;
}
flags -= cpu_flag::wait;
}
return true;
});
if (state & cpu_flag::signal && state.test_and_reset(cpu_flag::signal))
{
cpu_sleep_called = false;
}
if (escape)
{
if (cpu_flag_memory)
{
cpu_mem();
}
break;
}
else if (!cpu_sleep_called && state0 & cpu_flag::suspend)
{
cpu_sleep();
cpu_sleep_called = true;
continue;
}
if (state0 & (cpu_flag::suspend + cpu_flag::dbg_global_pause + cpu_flag::dbg_pause))
{
thread_ctrl::wait();
}
else
{
// If only cpu_flag::pause was set, notification won't arrive
g_fxo->get<cpu_counter>()->cpu_suspend_lock.lock_unlock();
}
}
const auto state_ = state.load();
if (state_ & (cpu_flag::ret + cpu_flag::stop))
{
return true;
}
if (state_ & cpu_flag::dbg_step)
{
state += cpu_flag::dbg_pause;
state -= cpu_flag::dbg_step;
}
return false;
}
void cpu_thread::notify()
{
// Downcast to correct type
if (id_type() == 1)
{
thread_ctrl::notify(*static_cast<named_thread<ppu_thread>*>(this));
}
else if (id_type() == 2)
{
thread_ctrl::notify(*static_cast<named_thread<spu_thread>*>(this));
}
else
{
fmt::throw_exception("Invalid cpu_thread type" HERE);
}
}
void cpu_thread::abort()
{
// Downcast to correct type
if (id_type() == 1)
{
*static_cast<named_thread<ppu_thread>*>(this) = thread_state::aborting;
}
else if (id_type() == 2)
{
*static_cast<named_thread<spu_thread>*>(this) = thread_state::aborting;
}
else
{
fmt::throw_exception("Invalid cpu_thread type" HERE);
}
}
std::string cpu_thread::get_name() const
{
// Downcast to correct type
if (id_type() == 1)
{
return thread_ctrl::get_name(*static_cast<const named_thread<ppu_thread>*>(this));
}
else if (id_type() == 2)
{
return thread_ctrl::get_name(*static_cast<const named_thread<spu_thread>*>(this));
}
else
{
fmt::throw_exception("Invalid cpu_thread type" HERE);
}
}
std::string cpu_thread::dump_all() const
{
return {};
}
std::string cpu_thread::dump_regs() const
{
return {};
}
std::string cpu_thread::dump_callstack() const
{
return {};
}
std::vector<u32> cpu_thread::dump_callstack_list() const
{
return {};
}
std::string cpu_thread::dump_misc() const
{
return fmt::format("Type: %s\n" "State: %s\n", typeid(*this).name(), state.load());
}
cpu_thread::suspend_all::suspend_all(cpu_thread* _this) noexcept
: m_this(_this)
{
if (m_this)
{
m_this->state += cpu_flag::wait;
}
g_fxo->get<cpu_counter>()->cpu_suspend_lock.lock_vip();
for_all_cpu([](cpu_thread* cpu)
{
cpu->state += cpu_flag::pause;
});
busy_wait(500);
while (true)
{
bool ok = true;
for_all_cpu([&](cpu_thread* cpu)
{
if (!(cpu->state & cpu_flag::wait))
{
ok = false;
}
});
if (ok) [[likely]]
{
break;
}
busy_wait(500);
}
}
cpu_thread::suspend_all::~suspend_all()
{
// Make sure the latest thread does the cleanup and notifies others
if (g_fxo->get<cpu_counter>()->cpu_suspend_lock.downgrade_unique_vip_lock_to_low_or_unlock())
{
for_all_cpu([&](cpu_thread* cpu)
{
cpu->state -= cpu_flag::pause;
});
g_fxo->get<cpu_counter>()->cpu_suspend_lock.unlock_low();
}
else
{
g_fxo->get<cpu_counter>()->cpu_suspend_lock.lock_unlock();
}
if (m_this)
{
m_this->check_state();
}
}
void cpu_thread::stop_all() noexcept
{
if (g_tls_current_cpu_thread)
{
// Report unsupported but unnecessary case
sys_log.fatal("cpu_thread::stop_all() has been called from a CPU thread.");
return;
}
else
{
::vip_lock lock(g_fxo->get<cpu_counter>()->cpu_suspend_lock);
for_all_cpu([](cpu_thread* cpu)
{
cpu->state += cpu_flag::dbg_global_stop;
cpu->abort();
});
}
sys_log.notice("All CPU threads have been signaled.");
while (g_fxo->get<cpu_counter>()->cpu_array_sema)
{
std::this_thread::sleep_for(10ms);
}
sys_log.notice("All CPU threads have been stopped.");
}
void cpu_thread::flush_profilers() noexcept
{
if (!g_fxo->get<cpu_profiler>())
{
profiler.fatal("cpu_thread::flush_profilers() has been called incorrectly." HERE);
return;
}
if (g_cfg.core.spu_prof || false)
{
g_fxo->get<cpu_profiler>()->registered.push(0);
}
}