| 1 | /* |
| 2 | * ========== Copyright Header Begin ========================================== |
| 3 | * |
| 4 | * OpenSPARC T2 Processor File: workerthread.h |
| 5 | * Copyright (c) 2006 Sun Microsystems, Inc. All Rights Reserved. |
| 6 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES. |
| 7 | * |
| 8 | * The above named program is free software; you can redistribute it and/or |
| 9 | * modify it under the terms of the GNU General Public |
| 10 | * License version 2 as published by the Free Software Foundation. |
| 11 | * |
| 12 | * The above named program is distributed in the hope that it will be |
| 13 | * useful, but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public |
| 18 | * License along with this work; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. |
| 20 | * |
| 21 | * ========== Copyright Header End ============================================ |
| 22 | */ |
| 23 | /* standard includes */ |
| 24 | #include <pthread.h> |
| 25 | #include <synch.h> |
| 26 | #include <signal.h> |
| 27 | #include <stdio.h> |
| 28 | #include <stdlib.h> |
| 29 | #include <errno.h> |
| 30 | #include <assert.h> |
| 31 | #include <list> |
| 32 | |
| 33 | /* project includes */ |
| 34 | #include "types.h" |
| 35 | #include "blaze_globals.h" |
| 36 | #include "ui.h" |
| 37 | #include "system.h" |
| 38 | #include "system_impl.h" |
| 39 | #include "atomic.h" |
| 40 | #include "tracemod.h" |
| 41 | #include "prioque.h" |
| 42 | #include "cpu_interface.h" |
| 43 | #include "vtracer_async_queue.h" |
| 44 | |
| 45 | |
| 46 | extern "C" { |
| 47 | static inline void * system_worker_thread (void * arg); |
| 48 | }; |
| 49 | |
| 50 | class WorkerThread { |
| 51 | |
| 52 | static Vcpu * first_cpu; |
| 53 | |
| 54 | static volatile uint64_t GlobalTimeUsecs; |
| 55 | // |
| 56 | // each workerthread maintains a local copy time variable, which is |
| 57 | // used for comparison in doEventQueueCallbacks() |
| 58 | // |
| 59 | volatile uint64_t simTime; |
| 60 | |
| 61 | |
| 62 | // approximate tick count per cpu/strand. only for "mips" ui-cmd, |
| 63 | // this is currently Broken ~!~ |
| 64 | static volatile int64_t GlobalTicks; |
| 65 | |
| 66 | |
| 67 | static WorkerThread * wrkthds; // worker thread array |
| 68 | static volatile int numThds; // total number of workerthreads |
| 69 | |
| 70 | pthread_t tid; // the real worker thread |
| 71 | int worker_id; // 0..numthds-1 |
| 72 | int num_cpus; // # of vcpu's assigned to this workerthread |
| 73 | Vcpu *cpus[MAX_MP]; // pointers to vcpu's assigned to this thread, |
| 74 | // index = [0 ... num_cpus - 1] |
| 75 | |
| 76 | EventQue * eq; // per worker thread event queue |
| 77 | |
| 78 | static pthread_key_t key; // to bind the eq to workerthread's private data |
| 79 | |
| 80 | int64_t done_x_1024[MAX_MP]; |
| 81 | |
| 82 | WorkerThread(){ |
| 83 | called_from_stept = false; |
| 84 | KillThread = false; |
| 85 | Nusecs = Ninstrs = Ncycles = 0; |
| 86 | sema_init (&KICK,0, USYNC_THREAD, NULL); |
| 87 | assert (pthread_create (&tid,NULL, &system_worker_thread, (void*)this) == 0); |
| 88 | eq = new EventQue(); |
| 89 | /////assert (pthread_setspecific(key,(void *) this) == 0); |
| 90 | simTime = GlobalTimeUsecs; |
| 91 | } |
| 92 | |
| 93 | ~WorkerThread(){ |
| 94 | sema_destroy (&KICK); |
| 95 | delete eq; |
| 96 | } |
| 97 | |
| 98 | |
| 99 | sema_t KICK; // <--- KICK semaphores |
| 100 | bool KillThread; |
| 101 | static int64_t Nusecs; /* stept microseconds */ |
| 102 | static int64_t Ninstrs; /* stepi instruction count */ |
| 103 | static int64_t Ncycles; /* for exec-driven mode */ |
| 104 | |
| 105 | static volatile int BarrierCount; |
| 106 | static volatile int BarrierLock; |
| 107 | static volatile int BarrierTemp; |
| 108 | static volatile uint64_t step_remainder; /* see note in .cc for explanation */ |
| 109 | |
| 110 | static volatile uint64_t stick_incr; |
| 111 | // to allow for stick frequencies that are not a multiple of 10^6 |
| 112 | static volatile uint64_t stick_remainder; |
| 113 | |
| 114 | void stepi(uint64_t n); |
| 115 | void stept(uint64_t usecs); |
| 116 | void stepc(int64_t ncycles); |
| 117 | int barrier(); |
| 118 | static void kill_worker_threads (); |
| 119 | |
| 120 | bool called_from_stept; |
| 121 | |
| 122 | void killThread(){ |
| 123 | KillThread = true; |
| 124 | sema_post (&KICK); |
| 125 | } |
| 126 | |
| 127 | void doEventqueCallbacks (){ |
| 128 | Event_t event; |
| 129 | |
| 130 | while (! eq->empty() |
| 131 | && eq->top_time() <= simTime ) { |
| 132 | |
| 133 | eq->get_top (&event); |
| 134 | (*(EventFunc_T*)event.pq_cbfunc) (event.pq_cbarg1, event.pq_cbarg2); |
| 135 | } |
| 136 | } |
| 137 | |
| 138 | |
| 139 | void doTrace() { |
| 140 | if (cpus[0] == first_cpu) { |
| 141 | while (trace_async_queue->getsize()) { |
| 142 | VCPU_AsyncData async_rec; |
| 143 | trace_async_queue->dequeue(&async_rec); |
| 144 | first_cpu->sys_intf.vtrace->async(&async_rec); |
| 145 | } // if this is the first cpu in the system |
| 146 | } |
| 147 | |
| 148 | // write sync records to all cpus. Use GlobalTimeUsecs as a possible seq # |
| 149 | vtracer_syncinfo syncinfo; |
| 150 | syncinfo.synctype = vtracer_syncinfo::synctype_LOCAL; |
| 151 | syncinfo.syncid = GlobalTimeUsecs; |
| 152 | syncinfo.data = 0; |
| 153 | int i; |
| 154 | for (i=0; i<num_cpus; i++) { |
| 155 | syncinfo.cpuid = cpus[i]->id(); |
| 156 | cpus[i]->sys_intf.vtrace->sync(&syncinfo); |
| 157 | } |
| 158 | |
| 159 | } |
| 160 | |
| 161 | |
| 162 | public: |
| 163 | // external interface exported by workerthreads |
| 164 | |
| 165 | void run_workerThread(){ |
| 166 | assert (pthread_setspecific(key,(void *) this) == 0); |
| 167 | for (;;) { |
| 168 | sema_wait (&KICK); // --------------------------------- STANDBY !!! |
| 169 | |
| 170 | if (KillThread) { |
| 171 | pthread_exit (0); |
| 172 | tid = 0; |
| 173 | } |
| 174 | if (Nusecs) |
| 175 | stept(Nusecs); |
| 176 | else if(Ninstrs) |
| 177 | stepi(Ninstrs); |
| 178 | else if (Ncycles) |
| 179 | stepc(Ncycles); |
| 180 | } |
| 181 | } |
| 182 | static void kick_stepi(uint64_t n){ |
| 183 | // initialize the stick related variabled |
| 184 | stick_incr = the_arch.stick_freq/1000000ull; |
| 185 | |
| 186 | if(!wrkthds) |
| 187 | create_worker_threads (SYSTEM_get_ncpu(),SYSTEM_get_cpus_per_thread(),\ |
| 188 | SYSTEM_get_numthreads()); |
| 189 | Ninstrs = n; Nusecs = 0; Ncycles = 0; |
| 190 | for (int i = 0; i < numThds; i++) |
| 191 | sema_post (&wrkthds[i].KICK); |
| 192 | } |
| 193 | |
| 194 | static void kick_stept(uint64_t usecs){ |
| 195 | // initialize the stick related variabled |
| 196 | stick_incr = the_arch.stick_freq/1000000ull; |
| 197 | |
| 198 | if(!wrkthds) |
| 199 | create_worker_threads (SYSTEM_get_ncpu(),SYSTEM_get_cpus_per_thread(), \ |
| 200 | SYSTEM_get_numthreads()); |
| 201 | Nusecs = usecs; Ninstrs = 0; Ncycles = 0; |
| 202 | for (int i = 0; i < numThds; i++) |
| 203 | sema_post (&wrkthds[i].KICK); |
| 204 | } |
| 205 | |
| 206 | static void kick_stepc(int64_t ncycles) { |
| 207 | stick_incr = the_arch.stick_freq/1000000ull; |
| 208 | if (!wrkthds) |
| 209 | create_worker_threads(SYSTEM_get_ncpu(), SYSTEM_get_ncpu(), 1); |
| 210 | |
| 211 | Ncycles = ncycles; Nusecs = Ninstrs = 0; |
| 212 | sema_post(&wrkthds[0].KICK); |
| 213 | } |
| 214 | |
| 215 | void info(){ |
| 216 | ui->verbose("sid = {"); |
| 217 | for (int j=0; j < num_cpus; j++) |
| 218 | { |
| 219 | if (j>0) ui->verbose (","); |
| 220 | ui->verbose ("%d", cpus[j]->id()); |
| 221 | } |
| 222 | ui->verbose ("} \n"); |
| 223 | } |
| 224 | |
| 225 | |
| 226 | |
| 227 | static void registerEvent (uint64_t stime, |
| 228 | EventFunc_T* callbackfunc, void* arg1, void* arg2, |
| 229 | UnloadFunc_T* unloadfunc, |
| 230 | const char * debugstring) |
| 231 | { |
| 232 | WorkerThread * wt = (WorkerThread*) pthread_getspecific (key); |
| 233 | EventQue * myeq = wt ? wt->eq : 0; |
| 234 | if (!myeq) { |
| 235 | |
| 236 | if (!wrkthds) { |
| 237 | |
| 238 | create_worker_threads(SYSTEM_get_ncpu(), |
| 239 | SYSTEM_get_cpus_per_thread(), |
| 240 | SYSTEM_get_numthreads()); |
| 241 | } |
| 242 | myeq = wrkthds[0].eq; |
| 243 | } |
| 244 | |
| 245 | myeq->insert_callback ( stime, (void*)callbackfunc, arg1, arg2, |
| 246 | (void*)unloadfunc, debugstring, |
| 247 | (wt ? wt->worker_id : -1));/*more debug*/ |
| 248 | } |
| 249 | |
| 250 | |
| 251 | static void doEventqueUnloads () |
| 252 | { |
| 253 | for (int i = 0; i < numThds; i++) |
| 254 | wrkthds[i].eq->unload(); |
| 255 | } |
| 256 | |
| 257 | |
| 258 | static void doEventquePrint () |
| 259 | { |
| 260 | for (int i = 0; i < numThds; i++) |
| 261 | wrkthds[i].eq->print (); |
| 262 | } |
| 263 | |
| 264 | |
| 265 | |
| 266 | |
| 267 | static void create_worker_threads (int NumCpus, int cpusPerThread, int numThreads); |
| 268 | |
| 269 | // store any needed parameter in the form of NAME VALUE pair in fp |
| 270 | static void dump(FILE * fp); |
| 271 | // restore parameters stored in the form of NAME VALUE pair from fp |
| 272 | static int restore(char * line); |
| 273 | |
| 274 | |
| 275 | static uint64_t get_time(){ |
| 276 | WorkerThread * wt = (WorkerThread*) pthread_getspecific(key); |
| 277 | |
| 278 | return (wt ? wt->simTime : GlobalTimeUsecs); |
| 279 | } |
| 280 | |
| 281 | |
| 282 | static int64_t get_ticks(){ return GlobalTicks;} |
| 283 | |
| 284 | |
| 285 | static volatile int64_t u_instrs; // for "mips" ui command |
| 286 | static volatile int64_t k_instrs; |
| 287 | |
| 288 | static volatile int64_t u_intervals; // ditto |
| 289 | static volatile int64_t k_intervals; |
| 290 | }; |
| 291 | |
| 292 | extern "C" { |
| 293 | static inline void * system_worker_thread (void * arg){ |
| 294 | WorkerThread * sp = (WorkerThread*) arg; |
| 295 | sp->run_workerThread(); |
| 296 | return 0; |
| 297 | } |
| 298 | }; |
| 299 | |