Browse Source

bflsc increase flush count to handle parallel work

nfactor-troky
Kano 12 years ago
parent
commit
89d0062ac8
  1. 7
      driver-bflsc.c

7
driver-bflsc.c

@ -150,6 +150,7 @@ struct bflsc_info { @@ -150,6 +150,7 @@ struct bflsc_info {
int que_noncecount;
int que_fld_min;
int que_fld_max;
int flush_size;
};
#define BFLSC_XLINKHDR '@'
@ -1131,6 +1132,8 @@ reinit: @@ -1131,6 +1132,8 @@ reinit:
sc_info->que_noncecount = QUE_NONCECOUNT_V1;
sc_info->que_fld_min = QUE_FLD_MIN_V1;
sc_info->que_fld_max = QUE_FLD_MAX_V1;
// Only Jalapeno uses 1.0.0
sc_info->flush_size = 1;
break;
case BFLSC_DRV2:
case BFLSC_DRVUNDEF:
@ -1144,6 +1147,8 @@ reinit: @@ -1144,6 +1147,8 @@ reinit:
sc_info->que_noncecount = QUE_NONCECOUNT_V2;
sc_info->que_fld_min = QUE_FLD_MIN_V2;
sc_info->que_fld_max = QUE_FLD_MAX_V2;
// TODO: this can be reduced to total chip count
sc_info->flush_size = 16 * sc_info->sc_count;
break;
}
@ -2043,7 +2048,7 @@ static int64_t bflsc_scanwork(struct thr_info *thr) @@ -2043,7 +2048,7 @@ static int64_t bflsc_scanwork(struct thr_info *thr)
// Is there any flushed work that can be removed?
rd_lock(&(sc_info->stat_lock));
if (sc_info->sc_devs[dev].flushed) {
if (sc_info->sc_devs[dev].result_id > (sc_info->sc_devs[dev].flush_id + 1))
if (sc_info->sc_devs[dev].result_id > (sc_info->sc_devs[dev].flush_id + sc_info->flush_size))
cleanup = true;
}
rd_unlock(&(sc_info->stat_lock));

Loading…
Cancel
Save