GOSTSec/sgminer
1
0
Fork 0
mirror of https://github.com/GOSTSec/sgminer synced 2025-02-02 10:04:33 +00:00
Code Issues Releases Wiki Activity

Better HW error count; disable permanently those cores which fail often

Browse Source
This commit is contained in:
Vitalii Demianets 2013-10-18 15:16:11 +02:00 committed by Henrik Nordstrom
parent 53570ae8a0
commit 6d4a03ad81
1 changed files with 20 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
driver-knc-spi-fpga.c
View File

@ -24,6 +24,7 @@
/* How many hardware errors in a row before disabling the core */ /* How many hardware errors in a row before disabling the core */
#define HW_ERR_LIMIT 10 #define HW_ERR_LIMIT 10
#define DISA_ERR_LIMIT 3
#define MAX_ACTIVE_WORKS (192 * 2 * 6 * 2) #define MAX_ACTIVE_WORKS (192 * 2 * 6 * 2)
@ -148,6 +149,8 @@ struct knc_state {
struct active_work active_fifo[KNC_ACTIVE_BUFFER_SIZE]; struct active_work active_fifo[KNC_ACTIVE_BUFFER_SIZE];
uint8_t hwerrs[MAX_ASICS * 256]; uint8_t hwerrs[MAX_ASICS * 256];
uint8_t disa_cnt[MAX_ASICS * 256];
uint32_t hwerr_work_id[MAX_ASICS * 256];
int read_d, write_d; int read_d, write_d;
#define KNC_DISA_CORES_SIZE (MAX_ASICS * 256) #define KNC_DISA_CORES_SIZE (MAX_ASICS * 256)
struct core_disa_data disa_cores_fifo[KNC_DISA_CORES_SIZE]; struct core_disa_data disa_cores_fifo[KNC_DISA_CORES_SIZE];
@ -461,10 +464,15 @@ static int64_t knc_process_response(struct thr_info *thr, struct cgpu_info *cgpu
rxbuf->responses[i].core; rxbuf->responses[i].core;
if (submit_nonce(thr, work, if (submit_nonce(thr, work,
rxbuf->responses[i].nonce)) { rxbuf->responses[i].nonce)) {
if (cidx < sizeof(knc->hwerrs))
knc->hwerrs[cidx] = 0;
} else {
if (cidx < sizeof(knc->hwerrs)) { if (cidx < sizeof(knc->hwerrs)) {
knc->hwerrs[cidx] = 0;
knc->disa_cnt[cidx] = 0;
knc->hwerr_work_id[cidx] = 0xFFFFFFFF;
}
} else {
if ((cidx < sizeof(knc->hwerrs)) &&
(knc->hwerr_work_id[cidx] != rxbuf->responses[i].work_id)) {
knc->hwerr_work_id[cidx] = rxbuf->responses[i].work_id;
if (++(knc->hwerrs[cidx]) >= HW_ERR_LIMIT) { if (++(knc->hwerrs[cidx]) >= HW_ERR_LIMIT) {
struct core_disa_data *core; struct core_disa_data *core;
core = &knc->disa_cores_fifo[knc->write_d]; core = &knc->disa_cores_fifo[knc->write_d];
@ -472,10 +480,15 @@ static int64_t knc_process_response(struct thr_info *thr, struct cgpu_info *cgpu
core->asic = rxbuf->responses[i].asic; core->asic = rxbuf->responses[i].asic;
core->core = rxbuf->responses[i].core; core->core = rxbuf->responses[i].core;
disable_core(core->asic, core->core); disable_core(core->asic, core->core);
applog(LOG_WARNING, if (++(knc->disa_cnt[cidx]) >= DISA_ERR_LIMIT) {
applog(LOG_WARNING,
"KnC: core %u-%u was disabled permanently", core->asic, core->core);
} else {
applog(LOG_WARNING,
"KnC: core %u-%u was disabled due to %u HW errors in a row", "KnC: core %u-%u was disabled due to %u HW errors in a row",
core->asic, core->core, HW_ERR_LIMIT); core->asic, core->core, HW_ERR_LIMIT);
knc_disa_cores_fifo_inc_idx(&knc->write_d); knc_disa_cores_fifo_inc_idx(&knc->write_d);
}
} }
} }
}; };
@ -546,6 +559,7 @@ static bool knc_detect_one(struct spidev_context *ctx)
knc->read_d = 0; knc->read_d = 0;
knc->write_d = 1; knc->write_d = 1;
knc->salt = rand(); knc->salt = rand();
memset(knc->hwerr_work_id, 0xFF, sizeof(knc->hwerr_work_id));
_internal_knc_flush_fpga(knc); _internal_knc_flush_fpga(knc);