X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=steadystate.c;h=224995d1d26899a8b4fcf4f719a66924969dbd78;hp=c7a7e4c70195ac193ce7742adab8044f9cd745b9;hb=17df7023156b50ee7288eea0f118f1ac20b21ea2;hpb=f0c50c66a602fa964b2aa2de416deb515dd39ac9

diff --git a/steadystate.c b/steadystate.c
index c7a7e4c7..224995d1 100644
--- a/steadystate.c
+++ b/steadystate.c
@@ -10,11 +10,13 @@ static void steadystate_alloc(struct thread_data *td)
 {
 	int i;
 
-	td->ss.bw_data = malloc(td->ss.dur * sizeof(unsigned long));
-	td->ss.iops_data = malloc(td->ss.dur * sizeof(unsigned long));
+	td->ss.bw_data = malloc(td->ss.dur * sizeof(uint64_t));
+	td->ss.iops_data = malloc(td->ss.dur * sizeof(uint64_t));
 	/* initialize so that it is obvious if the cache is not full in the output */
 	for (i = 0; i < td->ss.dur; i++)
 		td->ss.iops_data[i] = td->ss.bw_data[i] = 0;
+
+	td->ss.state |= __FIO_SS_DATA;
 }
 
 void steadystate_setup(void)
@@ -33,7 +35,7 @@ void steadystate_setup(void)
 	prev_groupid = -1;
 	prev_td = NULL;
 	for_each_td(td, i) {
-		if (td->ts.ss == NULL)
+		if (!td->ss.dur)
 			continue;
 
 		if (!td->o.group_reporting) {
@@ -43,7 +45,6 @@ void steadystate_setup(void)
 
 		if (prev_groupid != td->groupid) {
 			if (prev_td != NULL) {
-				prev_td->ss.last_in_group = 1;
 				steadystate_alloc(prev_td);
 			}
 			prev_groupid = td->groupid;
@@ -52,7 +53,6 @@ void steadystate_setup(void)
 	}
 
 	if (prev_td != NULL && prev_td->o.group_reporting) {
-		prev_td->ss.last_in_group = 1;
 		steadystate_alloc(prev_td);
 	}
 }
@@ -68,33 +68,33 @@ static bool steadystate_slope(unsigned long iops, unsigned long bw,
 	ss->bw_data[ss->tail] = bw;
 	ss->iops_data[ss->tail] = iops;
 
-	if (ss->mode & __FIO_SS_IOPS)
+	if (ss->state & __FIO_SS_IOPS)
 		new_val = iops;
 	else
 		new_val = bw;
 
-	if (ss->tail < ss->head || (ss->tail - ss->head == ss->dur - 1)) {
-		if (ss->sum_y == 0) {	/* first time through */
-			for(i = 0; i < ss->dur; i++) {
-				if (ss->mode & __FIO_SS_IOPS)
+	if (ss->state & __FIO_SS_BUFFER_FULL || ss->tail - ss->head == ss->dur - 1) {
+		if (!(ss->state & __FIO_SS_BUFFER_FULL)) {
+			/* first time through */
+			for(i = 0, ss->sum_y = 0; i < ss->dur; i++) {
+				if (ss->state & __FIO_SS_IOPS)
 					ss->sum_y += ss->iops_data[i];
 				else
 					ss->sum_y += ss->bw_data[i];
-				j = ss->head + i;
-				if (j >= ss->dur)
-					j -= ss->dur;
-				if (ss->mode & __FIO_SS_IOPS)
-					ss->sum_xy += ss->iops_data[j];
+				j = (ss->head + i) % ss->dur;
+				if (ss->state & __FIO_SS_IOPS)
+					ss->sum_xy += i * ss->iops_data[j];
 				else
-					ss->sum_xy += ss->bw_data[j];
+					ss->sum_xy += i * ss->bw_data[j];
 			}
+			ss->state |= __FIO_SS_BUFFER_FULL;
 		} else {		/* easy to update the sums */
 			ss->sum_y -= ss->oldest_y;
 			ss->sum_y += new_val;
 			ss->sum_xy = ss->sum_xy - ss->sum_y + ss->dur * new_val;
 		}
 
-		if (ss->mode & __FIO_SS_IOPS)
+		if (ss->state & __FIO_SS_IOPS)
 			ss->oldest_y = ss->iops_data[ss->head];
 		else
 			ss->oldest_y = ss->bw_data[ss->head];
@@ -107,7 +107,7 @@ static bool steadystate_slope(unsigned long iops, unsigned long bw,
 		 */
 		ss->slope = (ss->sum_xy - (double) ss->sum_x * ss->sum_y / ss->dur) /
 				(ss->sum_x_sq - (double) ss->sum_x * ss->sum_x / ss->dur);
-		if (ss->pct)
+		if (ss->state & __FIO_SS_PCT)
 			ss->criterion = 100.0 * ss->slope / (ss->sum_y / ss->dur);
 		else
 			ss->criterion = ss->slope;
@@ -141,22 +141,24 @@ static bool steadystate_deviation(unsigned long iops, unsigned long bw,
 	ss->bw_data[ss->tail] = bw;
 	ss->iops_data[ss->tail] = iops;
 
-	if (ss->tail < ss->head || (ss->tail - ss->head == ss->dur - 1)) {
-		if (ss->sum_y == 0) {	/* first time through */
-			for(i = 0; i < ss->dur; i++)
-				if (ss->mode & __FIO_SS_IOPS)
+	if (ss->state & __FIO_SS_BUFFER_FULL || ss->tail - ss->head == ss->dur - 1) {
+		if (!(ss->state & __FIO_SS_BUFFER_FULL)) {
+			/* first time through */
+			for(i = 0, ss->sum_y = 0; i < ss->dur; i++)
+				if (ss->state & __FIO_SS_IOPS)
 					ss->sum_y += ss->iops_data[i];
 				else
 					ss->sum_y += ss->bw_data[i];
+			ss->state |= __FIO_SS_BUFFER_FULL;
 		} else {		/* easy to update the sum */
 			ss->sum_y -= ss->oldest_y;
-			if (ss->mode & __FIO_SS_IOPS)
+			if (ss->state & __FIO_SS_IOPS)
 				ss->sum_y += ss->iops_data[ss->tail];
 			else
 				ss->sum_y += ss->bw_data[ss->tail];
 		}
 
-		if (ss->mode & __FIO_SS_IOPS)
+		if (ss->state & __FIO_SS_IOPS)
 			ss->oldest_y = ss->iops_data[ss->head];
 		else
 			ss->oldest_y = ss->bw_data[ss->head];
@@ -165,14 +167,14 @@ static bool steadystate_deviation(unsigned long iops, unsigned long bw,
 		ss->deviation = 0.0;
 
 		for (i = 0; i < ss->dur; i++) {
-			if (ss->mode & __FIO_SS_IOPS)
+			if (ss->state & __FIO_SS_IOPS)
 				diff = ss->iops_data[i] - mean;
 			else
 				diff = ss->bw_data[i] - mean;
 			ss->deviation = max(ss->deviation, diff * (diff < 0.0 ? -1.0 : 1.0));
 		}
 
-		if (ss->pct)
+		if (ss->state & __FIO_SS_PCT)
 			ss->criterion = 100.0 * ss->deviation / mean;
 		else
 			ss->criterion = ss->deviation;
@@ -209,7 +211,7 @@ void steadystate_check(void)
 		struct steadystate_data *ss = &td->ss;
 
 		if (!ss->dur || td->runstate <= TD_SETTING_UP ||
-		    td->runstate >= TD_EXITED || ss->attained)
+		    td->runstate >= TD_EXITED || (ss->state & __FIO_SS_ATTAINED))
 			continue;
 
 		td_iops = 0;
@@ -223,13 +225,14 @@ void steadystate_check(void)
 		prev_groupid = td->groupid;
 
 		fio_gettime(&now, NULL);
-		if (ss->ramp_time && !ss->ramp_time_over)
+		if (ss->ramp_time && !(ss->state & __FIO_SS_RAMP_OVER)) {
 			/*
 			 * Begin recording data one second after ss->ramp_time
 			 * has elapsed
 			 */
 			if (utime_since(&td->epoch, &now) >= (ss->ramp_time + 1000000L))
-				ss->ramp_time_over = 1;
+				ss->state |= __FIO_SS_RAMP_OVER;
+		}
 
 		td_io_u_lock(td);
 		for (ddir = DDIR_READ; ddir < DDIR_RWDIR_CNT; ddir++) {
@@ -248,7 +251,7 @@ void steadystate_check(void)
 		 * prev_iops/bw the first time through after ss->ramp_time
 		 * is done.
 		 */
-		if (ss->ramp_time_over) {
+		if (ss->state & __FIO_SS_RAMP_OVER) {
 			group_bw += 1000 * (td_bytes - ss->prev_bytes) / rate_time;
 			group_iops += 1000 * (td_iops - ss->prev_iops) / rate_time;
 			++group_ramp_time_over;
@@ -256,7 +259,7 @@ void steadystate_check(void)
 		ss->prev_iops = td_iops;
 		ss->prev_bytes = td_bytes;
 
-		if (td->o.group_reporting && !ss->last_in_group)
+		if (td->o.group_reporting && !(ss->state & __FIO_SS_DATA))
 			continue;
 
 		/*
@@ -281,41 +284,82 @@ void steadystate_check(void)
 			if (td->o.group_reporting) {
 				for_each_td(td2, j) {
 					if (td2->groupid == td->groupid) {
-						td2->ss.attained = 1;
+						td2->ss.state |= __FIO_SS_ATTAINED;
 						fio_mark_td_terminate(td2);
 					}
 				}
 			} else {
-				ss->attained = 1;
+				ss->state |= __FIO_SS_ATTAINED;
 				fio_mark_td_terminate(td);
 			}
 		}
 	}
 }
 
-void td_steadystate_init(struct thread_data *td)
+int td_steadystate_init(struct thread_data *td)
 {
 	struct steadystate_data *ss = &td->ss;
 	struct thread_options *o = &td->o;
+	struct thread_data *td2;
+	int j;
 
 	memset(ss, 0, sizeof(*ss));
 
-	if (!o->ss_dur)
-		return;
+	if (o->ss_dur) {
+		steadystate_enabled = true;
+		o->ss_dur /= 1000000L;
+
+		/* put all steady state info in one place */
+		ss->dur = o->ss_dur;
+		ss->limit = o->ss_limit.u.f;
+		ss->ramp_time = o->ss_ramp_time;
+
+		ss->state = o->ss;
+		if (!td->ss.ramp_time)
+			ss->state |= __FIO_SS_RAMP_OVER;
+
+		ss->sum_x = o->ss_dur * (o->ss_dur - 1) / 2;
+		ss->sum_x_sq = (o->ss_dur - 1) * (o->ss_dur) * (2*o->ss_dur - 1) / 6;
+	}
+
+	/* make sure that ss options are consistent within reporting group */
+	for_each_td(td2, j) {
+		if (td2->groupid == td->groupid) {
+			struct steadystate_data *ss2 = &td2->ss;
+
+			if (ss2->dur != ss->dur ||
+			    ss2->limit != ss->limit ||
+			    ss2->ramp_time != ss->ramp_time ||
+			    ss2->state != ss->state ||
+			    ss2->sum_x != ss->sum_x ||
+			    ss2->sum_x_sq != ss->sum_x_sq) {
+				td_verror(td, EINVAL, "job rejected: steadystate options must be consistent within reporting groups");
+				return 1;
+			}
+		}
+	}
+
+	return 0;
+}
+
+unsigned long long steadystate_bw_mean(struct thread_stat *ts)
+{
+	int i;
+	unsigned long long sum;
+
+	for (i = 0, sum = 0; i < ts->ss_dur; i++)
+		sum += ts->ss_bw_data[i];
 
-	steadystate_enabled = true;
-	o->ss_dur /= 1000000L;
+	return sum / ts->ss_dur;
+}
 
-	/* put all steady state info in one place */
-	ss->dur = o->ss_dur;
-	ss->limit = o->ss_limit.u.f;
-	ss->ramp_time = o->ss_ramp_time;
-	ss->pct = o->ss_pct;
-	ss->mode = o->ss;
+unsigned long long steadystate_iops_mean(struct thread_stat *ts)
+{
+	int i;
+	unsigned long long sum;
 
-	ss->ramp_time_over = (td->ss.ramp_time == 0);
-	ss->sum_x = o->ss_dur * (o->ss_dur - 1) / 2;
-	ss->sum_x_sq = (o->ss_dur - 1) * (o->ss_dur) * (2*o->ss_dur - 1) / 6;
+	for (i = 0, sum = 0; i < ts->ss_dur; i++)
+		sum += ts->ss_iops_data[i];
 
-	td->ts.ss = ss;
+	return sum / ts->ss_dur;
 }