Merge pull request #93 from luke-jr/bitforce_temp

Generalize --temp-cutoff and implement support for reading temperature from BitFORCE FPGAs

README

@@ -153,6 +153,7 @@ Options for both config file and command line:
 --socks-proxy <arg> Set socks4 proxy (host:port)
 --submit-stale      Submit shares even if they would normally be considered stale
 --syslog            Use system log for output messages (default: standard error)
+--temp-cutoff <arg> Temperature where a device will be automatically disabled, one value or comma separated list (default: 95)
 --text-only|-T      Disable ncurses formatted screen output
 --url|-o <arg>      URL for bitcoin JSON-RPC server
 --user|-u <arg>     Username for bitcoin JSON-RPC server
@@ -183,7 +184,6 @@ GPU only options:
 --kernel-path|-K <arg> Specify a path to where the kernel .cl files are (default: "/usr/local/bin")
 --kernel|-k <arg>   Select kernel to use (poclbm or phatk - default: auto)
 --no-restart        Do not attempt to restart GPUs that hang
---temp-cutoff <arg> Temperature where a GPU device will be automatically disabled, one value or comma separated list (default: 95)
 --temp-hysteresis <arg> Set how much the temperature can fluctuate outside limits when automanaging speeds (default: 3)
 --temp-overheat <arg> Overheat temperature when automatically managing fan and GPU speeds (default: 85)
 --temp-target <arg> Target temperature when automatically managing fan and GPU speeds (default: 75)

adl.c

@@ -26,7 +26,6 @@ bool opt_reorder = false;
 int opt_hysteresis = 3;
 const int opt_targettemp = 75;
 const int opt_overheattemp = 85;
-const int opt_cutofftemp = 95;
 static pthread_mutex_t adl_lock;
 
 struct gpu_adapters {
@@ -443,8 +442,8 @@ void init_adl(int nDevs)
 			ga->targettemp = opt_targettemp;
 		if (!ga->overtemp)
 			ga->overtemp = opt_overheattemp;
-		if (!ga->cutofftemp)
+		if (!gpus[gpu].cutofftemp)
-			ga->cutofftemp = opt_cutofftemp;
+			gpus[gpu].cutofftemp = opt_cutofftemp;
 		if (opt_autofan) {
 			ga->autofan = true;
 			/* Set a safe starting default if we're automanaging fan speeds */
@@ -503,6 +502,7 @@ float gpu_temp(int gpu)
 	lock_adl();
 	ret = __gpu_temp(ga);
 	unlock_adl();
+	gpus[gpu].temp = ret;
 	return ret;
 }
 
@@ -1055,7 +1055,7 @@ void gpu_autotune(int gpu, bool *enable)
 	}
 
 	if (engine && ga->autoengine) {
-		if (temp > ga->cutofftemp) {
+		if (temp > cgpu->cutofftemp) {
 			applog(LOG_WARNING, "Hit thermal cutoff limit on GPU %d, disabling!", gpu);
 			*enable = false;
 			newengine = ga->minspeed;
@@ -1122,7 +1122,7 @@ void change_autosettings(int gpu)
 
 	wlogprint("Target temperature: %d\n", ga->targettemp);
 	wlogprint("Overheat temperature: %d\n", ga->overtemp);
-	wlogprint("Cutoff temperature: %d\n", ga->cutofftemp);
+	wlogprint("Cutoff temperature: %d\n", gpus[gpu].cutofftemp);
 	wlogprint("Toggle [F]an auto [G]PU auto\nChange [T]arget [O]verheat [C]utoff\n");
 	wlogprint("Or press any other key to continue\n");
 	input = getch();
@@ -1159,7 +1159,7 @@ void change_autosettings(int gpu)
 		if (val <= ga->overtemp || val > 200)
 			wlogprint("Invalid temperature");
 		else
-			ga->cutofftemp = val;
+			gpus[gpu].cutofftemp = val;
 	}
 }
 

adl.h

@@ -6,7 +6,6 @@ bool opt_reorder;
 int opt_hysteresis;
 const int opt_targettemp;
 const int opt_overheattemp;
-const int opt_cutofftemp;
 void init_adl(int nDevs);
 float gpu_temp(int gpu);
 int gpu_engineclock(int gpu);

bitforce.c

@@ -10,6 +10,7 @@
 #include <limits.h>
 #include <pthread.h>
 #include <stdio.h>
+#include <strings.h>
 #include <sys/time.h>
 #include <sys/types.h>
 #include <dirent.h>
@@ -146,6 +147,16 @@ static void bitforce_detect()
 	bitforce_detect_auto();
 }
 
+static void get_bitforce_statline_before(char *buf, struct cgpu_info *bitforce)
+{
+	float gt = bitforce->temp;
+	if (gt > 0)
+		tailsprintf(buf, "%5.1fC ", gt);
+	else
+		tailsprintf(buf, "       ", gt);
+	tailsprintf(buf, "        | ");
+}
+
 static bool bitforce_thread_prepare(struct thr_info *thr)
 {
 	struct cgpu_info *bitforce = thr->cgpu;
@@ -221,6 +232,23 @@ static uint64_t bitforce_scanhash(struct thr_info *thr, struct work *work, uint6
 		return 0;
 	}
 
+	BFwrite(fdDev, "ZKX", 3);
+	BFgets(pdevbuf, sizeof(pdevbuf), fdDev);
+	if (unlikely(!pdevbuf[0])) {
+		applog(LOG_ERR, "Error reading from BitForce (ZKX)");
+		return 0;
+	}
+	if (!strncasecmp(pdevbuf, "TEMP:", 5)) {
+		float temp = strtof(pdevbuf + 5, NULL);
+		if (temp > 0) {
+			bitforce->temp = temp;
+			if (temp > bitforce->cutofftemp) {
+				applog(LOG_WARNING, "Hit thermal cutoff limit on %s %d, disabling!", bitforce->api->name, bitforce->device_id);
+				bitforce->enabled = false;
+			}
+		}
+	}
+
 	usleep(4500000);
 	i = 4500;
 	while (1) {
@@ -267,6 +295,7 @@ struct device_api bitforce_api = {
 	.name = "BFL",
 	.api_detect = bitforce_detect,
 	// .reinit_device = TODO
+	.get_statline_before = get_bitforce_statline_before,
 	.thread_prepare = bitforce_thread_prepare,
 	.scanhash = bitforce_scanhash,
 };

cgminer.c

@@ -90,6 +90,7 @@ bool use_syslog = false;
 static bool opt_quiet = false;
 static bool opt_realquiet = false;
 bool opt_loginput = false;
+const int opt_cutofftemp = 95;
 static int opt_retries = -1;
 static int opt_fail_pause = 5;
 static int fail_pause = 5;
@@ -537,6 +538,46 @@ static char *set_schedtime(const char *arg, struct schedtime *st)
 	return NULL;
 }
 
+static char *temp_cutoff_str = NULL;
+
+char *set_temp_cutoff(char *arg)
+{
+	int val;
+
+	if (!(arg && arg[0]))
+		return "Invalid parameters for set temp cutoff";
+	val = atoi(arg);
+	if (val < 0 || val > 200)
+		return "Invalid value passed to set temp cutoff";
+	temp_cutoff_str = arg;
+
+	return NULL;
+}
+
+static void load_temp_cutoffs()
+{
+	int i, val = 0, device = 0;
+	char *nextptr;
+
+	if (temp_cutoff_str) {
+		for (device = 0, nextptr = strtok(temp_cutoff_str, ","); nextptr; ++device, nextptr = strtok(NULL, ",")) {
+			if (device >= total_devices)
+				quit(1, "Too many values passed to set temp cutoff");
+			val = atoi(nextptr);
+			if (val < 0 || val > 200)
+				quit(1, "Invalid value passed to set temp cutoff");
+
+			devices[device]->cutofftemp = val;
+		}
+	}
+	else
+		val = opt_cutofftemp;
+	if (device <= 1) {
+		for (i = device; i < total_devices; ++i)
+			devices[i]->cutofftemp = val;
+	}
+}
+
 static char *set_api_allow(const char *arg)
 {
 	opt_set_charp(arg, &opt_api_allow);
@@ -764,10 +805,12 @@ static struct opt_table opt_config_table[] = {
 			opt_set_bool, &use_syslog,
 			"Use system log for output messages (default: standard error)"),
 #endif
-#ifdef HAVE_ADL
+#if defined(HAVE_ADL) || defined(USE_BITFORCE)
 	OPT_WITH_ARG("--temp-cutoff",
 		     set_temp_cutoff, opt_show_intval, &opt_cutofftemp,
-		     "Temperature where a GPU device will be automatically disabled, one value or comma separated list"),
+		     "Temperature where a device will be automatically disabled, one value or comma separated list"),
+#endif
+#ifdef HAVE_ADL
 	OPT_WITH_ARG("--temp-hysteresis",
 		     set_int_1_to_10, opt_show_intval, &opt_hysteresis,
 		     "Set how much the temperature can fluctuate outside limits when automanaging speeds"),
@@ -2237,7 +2280,7 @@ void write_config(FILE *fcfg)
 			fprintf(fcfg, "%s%1.3f", i > 0 ? "," : "", gpus[i].gpu_vddc);
 		fputs("\",\n\"temp-cutoff\" : \"", fcfg);
 		for(i = 0; i < nDevs; i++)
-			fprintf(fcfg, "%s%d", i > 0 ? "," : "", gpus[i].adl.cutofftemp);
+			fprintf(fcfg, "%s%d", i > 0 ? "," : "", gpus[i].cutofftemp);
 		fputs("\",\n\"temp-overheat\" : \"", fcfg);
 		for(i = 0; i < nDevs; i++)
 			fprintf(fcfg, "%s%d", i > 0 ? "," : "", gpus[i].adl.overtemp);
@@ -4237,6 +4280,8 @@ int main (int argc, char *argv[])
 	if (!total_devices)
 		quit(1, "All devices disabled, cannot mine!");
 
+	load_temp_cutoffs();
+
 	devcursor = 8;
 	logstart = devcursor + total_devices + 1;
 	logcursor = logstart + 1;

device-gpu.c

@@ -269,39 +269,6 @@ char *set_gpu_vddc(char *arg)
 	return NULL;
 }
 
-char *set_temp_cutoff(char *arg)
-{
-	int i, val = 0, device = 0, *tco;
-	char *nextptr;
-
-	nextptr = strtok(arg, ",");
-	if (nextptr == NULL)
-		return "Invalid parameters for set temp cutoff";
-	val = atoi(nextptr);
-	if (val < 0 || val > 200)
-		return "Invalid value passed to set temp cutoff";
-
-	tco = &gpus[device++].adl.cutofftemp;
-	*tco = val;
-
-	while ((nextptr = strtok(NULL, ",")) != NULL) {
-		val = atoi(nextptr);
-		if (val < 0 || val > 200)
-			return "Invalid value passed to set temp cutoff";
-
-		tco = &gpus[device++].adl.cutofftemp;
-		*tco = val;
-	}
-	if (device == 1) {
-		for (i = device; i < MAX_GPUDEVICES; i++) {
-			tco = &gpus[i].adl.cutofftemp;
-			*tco = val;
-		}
-	}
-
-	return NULL;
-}
-
 char *set_temp_overheat(char *arg)
 {
 	int i, val = 0, device = 0, *to;

device-gpu.h

@@ -12,7 +12,6 @@ extern char *set_gpu_memclock(char *arg);
 extern char *set_gpu_memdiff(char *arg);
 extern char *set_gpu_powertune(char *arg);
 extern char *set_gpu_vddc(char *arg);
-extern char *set_temp_cutoff(char *arg);
 extern char *set_temp_overheat(char *arg);
 extern char *set_temp_target(char *arg);
 extern char *set_intensity(char *arg);

miner.h

@@ -221,7 +221,6 @@ struct gpu_adl {
 	int targetfan;
 	int targettemp;
 	int overtemp;
-	int cutofftemp;
 	int minspeed;
 	int maxspeed;
 
@@ -280,6 +279,10 @@ struct cgpu_info {
 	int virtual_gpu;
 	bool dynamic;
 	int intensity;
+
+	float temp;
+	int cutofftemp;
+
 #ifdef HAVE_ADL
 	bool has_adl;
 	struct gpu_adl adl;
@@ -590,6 +593,7 @@ extern int total_accepted, total_rejected;
 extern int total_getworks, total_stale, total_discarded;
 extern unsigned int local_work;
 extern unsigned int total_go, total_ro;
+extern const int opt_cutofftemp;
 extern int opt_log_interval;
 
 #ifdef HAVE_OPENCL
@@ -686,6 +690,7 @@ enum cl_kernels {
 
 extern void get_datestamp(char *, struct timeval *);
 bool submit_nonce(struct thr_info *thr, struct work *work, uint32_t nonce);
+extern void tailsprintf(char *f, const char *fmt, ...);
 extern void wlogprint(const char *f, ...);
 extern int curses_int(const char *query);
 extern char *curses_input(const char *query);