Browse Source

Change sick/dead processing to use device pointer, not gpu array.

Change BFL timing to adjust only when hashing complete (not error/idle etc.).
nfactor-troky
Paul Sheppard 13 years ago
parent
commit
d3e2b62c54
  1. 44
      cgminer.c
  2. 2
      driver-bitforce.c

44
cgminer.c

@ -4484,14 +4484,14 @@ static void *watchdog_thread(void __maybe_unused *userdata)
if (cgpu->api->get_stats) if (cgpu->api->get_stats)
cgpu->api->get_stats(cgpu); cgpu->api->get_stats(cgpu);
gpu = thr->cgpu->device_id; gpu = cgpu->device_id;
denable = &cgpu->deven; denable = &cgpu->deven;
sprintf(dev_str, "%s%d", cgpu->api->name, gpu); sprintf(dev_str, "%s%d", cgpu->api->name, gpu);
#ifdef HAVE_ADL #ifdef HAVE_ADL
if (adl_active && gpus[gpu].has_adl) if (adl_active && cgpu->has_adl)
gpu_autotune(gpu, denable); gpu_autotune(gpu, denable);
if (opt_debug && gpus[gpu].has_adl) { if (opt_debug && cgpu->has_adl) {
int engineclock = 0, memclock = 0, activity = 0, fanspeed = 0, fanpercent = 0, powertune = 0; int engineclock = 0, memclock = 0, activity = 0, fanspeed = 0, fanpercent = 0, powertune = 0;
float temp = 0, vddc = 0; float temp = 0, vddc = 0;
@ -4505,48 +4505,48 @@ static void *watchdog_thread(void __maybe_unused *userdata)
if (thr->getwork || *denable == DEV_DISABLED) if (thr->getwork || *denable == DEV_DISABLED)
continue; continue;
if (gpus[gpu].status != LIFE_WELL && now.tv_sec - thr->last.tv_sec < 60) { if (cgpu->status != LIFE_WELL && now.tv_sec - thr->last.tv_sec < 60) {
applog(LOG_ERR, "%s: Recovered, declaring WELL!", dev_str); applog(LOG_ERR, "%s: Recovered, declaring WELL!", dev_str);
gpus[gpu].status = LIFE_WELL; cgpu->status = LIFE_WELL;
gpus[gpu].device_last_well = time(NULL); cgpu->device_last_well = time(NULL);
} else if (now.tv_sec - thr->last.tv_sec > 60 && gpus[gpu].status == LIFE_WELL) { } else if (now.tv_sec - thr->last.tv_sec > 60 && cgpu->status == LIFE_WELL) {
thr->rolling = thr->cgpu->rolling = 0; thr->rolling = cgpu->rolling = 0;
gpus[gpu].status = LIFE_SICK; cgpu->status = LIFE_SICK;
applog(LOG_ERR, "%s: Idle for more than 60 seconds, declaring SICK!", dev_str); applog(LOG_ERR, "%s: Idle for more than 60 seconds, declaring SICK!", dev_str);
gettimeofday(&thr->sick, NULL); gettimeofday(&thr->sick, NULL);
gpus[gpu].device_last_not_well = time(NULL); cgpu->device_last_not_well = time(NULL);
gpus[gpu].device_not_well_reason = REASON_DEV_SICK_IDLE_60; cgpu->device_not_well_reason = REASON_DEV_SICK_IDLE_60;
gpus[gpu].dev_sick_idle_60_count++; cgpu->dev_sick_idle_60_count++;
#ifdef HAVE_ADL #ifdef HAVE_ADL
if (adl_active && gpus[gpu].has_adl && gpu_activity(gpu) > 50) { if (adl_active && cgpu->has_adl && gpu_activity(gpu) > 50) {
applog(LOG_ERR, "GPU still showing activity suggesting a hard hang."); applog(LOG_ERR, "GPU still showing activity suggesting a hard hang.");
applog(LOG_ERR, "Will not attempt to auto-restart it."); applog(LOG_ERR, "Will not attempt to auto-restart it.");
} else } else
#endif #endif
if (opt_restart) { if (opt_restart) {
applog(LOG_ERR, "%s: Attempting to restart", dev_str); applog(LOG_ERR, "%s: Attempting to restart", dev_str);
reinit_device(thr->cgpu); reinit_device(cgpu);
} }
} else if (now.tv_sec - thr->last.tv_sec > 600 && gpus[i].status == LIFE_SICK) { } else if (now.tv_sec - thr->last.tv_sec > 600 && cgpu->status == LIFE_SICK) {
gpus[gpu].status = LIFE_DEAD; cgpu->status = LIFE_DEAD;
applog(LOG_ERR, "%s: Not responded for more than 10 minutes, declaring DEAD!", dev_str); applog(LOG_ERR, "%s: Not responded for more than 10 minutes, declaring DEAD!", dev_str);
gettimeofday(&thr->sick, NULL); gettimeofday(&thr->sick, NULL);
gpus[gpu].device_last_not_well = time(NULL); cgpu->device_last_not_well = time(NULL);
gpus[gpu].device_not_well_reason = REASON_DEV_DEAD_IDLE_600; cgpu->device_not_well_reason = REASON_DEV_DEAD_IDLE_600;
gpus[gpu].dev_dead_idle_600_count++; cgpu->dev_dead_idle_600_count++;
} else if (now.tv_sec - thr->sick.tv_sec > 60 && } else if (now.tv_sec - thr->sick.tv_sec > 60 &&
(gpus[i].status == LIFE_SICK || gpus[i].status == LIFE_DEAD)) { (cgpu->status == LIFE_SICK || cgpu->status == LIFE_DEAD)) {
/* Attempt to restart a GPU that's sick or dead once every minute */ /* Attempt to restart a GPU that's sick or dead once every minute */
gettimeofday(&thr->sick, NULL); gettimeofday(&thr->sick, NULL);
#ifdef HAVE_ADL #ifdef HAVE_ADL
if (adl_active && gpus[gpu].has_adl && gpu_activity(gpu) > 50) { if (adl_active && cgpu->has_adl && gpu_activity(gpu) > 50) {
/* Again do not attempt to restart a device that may have hard hung */ /* Again do not attempt to restart a device that may have hard hung */
} else } else
#endif #endif
if (opt_restart) if (opt_restart)
reinit_device(thr->cgpu); reinit_device(cgpu);
} }
} }
} }

2
driver-bitforce.c

@ -292,7 +292,7 @@ static uint64_t bitforce_get_result(struct thr_info *thr, struct work *work)
bitforce->device_not_well_reason = REASON_DEV_OVER_HEAT; bitforce->device_not_well_reason = REASON_DEV_OVER_HEAT;
bitforce->dev_over_heat_count++; bitforce->dev_over_heat_count++;
return 1; return 1;
} else { } else if (pdevbuf[0] == 'N') {/* Hashing complete (NONCE-FOUND or NO-NONCE) */
/* Simple timing adjustment */ /* Simple timing adjustment */
if (bitforce->wait_ms > (bitforce->sleep_ms + WORK_CHECK_INTERVAL_MS)) if (bitforce->wait_ms > (bitforce->sleep_ms + WORK_CHECK_INTERVAL_MS))
bitforce->sleep_ms += WORK_CHECK_INTERVAL_MS; bitforce->sleep_ms += WORK_CHECK_INTERVAL_MS;

Loading…
Cancel
Save