Add switch controlling scheduler busy wait

5 files changed, 94 insertions, 6 deletions

diff --git a/erts/doc/src/erl.xml b/erts/doc/src/erl.xml
index cfbc38f176..0df16a551c 100644
--- a/erts/doc/src/erl.xml
+++ b/erts/doc/src/erl.xml
@@ -766,6 +766,16 @@
 	       <seealso marker="erlang#system_info_scheduler_bindings">erlang:system_info(scheduler_bindings)</seealso>.
 	    </p>
           </item>
+	  <tag><marker id="+sbwt"><c>+sbwt none|very_short|short|medium|long|very_long</c></marker></tag>
+	  <item>
+             <p>Set scheduler busy wait threshold. Default is <c>medium</c>.
+	        The threshold determines how long schedulers should busy
+		wait when running out of work before going to sleep.
+	     </p>
+             <p><em>NOTE:</em> This flag may be removed or changed at any time
+	        without prior notice.
+	     </p>
+	  </item>
           <tag><marker id="+scl"><c>+scl true|false</c></marker></tag>
           <item>
 	    <p>Enable or disable scheduler compaction of load. By default
diff --git a/erts/emulator/beam/erl_init.c b/erts/emulator/beam/erl_init.c
index ca4385dd3a..5236f41c8f 100644
--- a/erts/emulator/beam/erl_init.c
+++ b/erts/emulator/beam/erl_init.c
@@ -511,6 +511,8 @@ void erts_usage(void)
     erts_fprintf(stderr, "-rg amount  set reader groups limit\n");
     erts_fprintf(stderr, "-sbt type   set scheduler bind type, valid types are:\n");
     erts_fprintf(stderr, "            u|ns|ts|ps|s|nnts|nnps|tnnps|db\n");
+    erts_fprintf(stderr, "-sbwt val   set scheduler busy wait threshold, valid values are:\n");
+    erts_fprintf(stderr, "            none|very_short|short|medium|long|very_long.\n");
     erts_fprintf(stderr, "-scl bool   enable/disable compaction of scheduler load,\n");
     erts_fprintf(stderr, "            see the erl(1) documentation for more info.\n");
     erts_fprintf(stderr, "-sct cput   set cpu topology,\n");
@@ -1198,6 +1200,16 @@ erl_start(int argc, char **argv)
 		    erts_usage();
 		}
 	    }
+	    else if (has_prefix("bwt", sub_param)) {
+		arg = get_arg(sub_param+3, argv[i+1], &i);
+		if (erts_sched_set_busy_wait_threshold(arg) != 0) {
+		    erts_fprintf(stderr, "bad scheduler busy wait threshold: %s\n",
+				 arg);
+		    erts_usage();
+		}
+		VERBOSE(DEBUG_SYSTEM,
+			("scheduler wakup threshold: %s\n", arg));
+	    }
 	    else if (has_prefix("cl", sub_param)) {
 		arg = get_arg(sub_param+2, argv[i+1], &i);
 		if (sys_strcmp("true", arg) == 0)
diff --git a/erts/emulator/beam/erl_process.c b/erts/emulator/beam/erl_process.c
index 95d408f79d..651e48da46 100644
--- a/erts/emulator/beam/erl_process.c
+++ b/erts/emulator/beam/erl_process.c
@@ -52,10 +52,20 @@
 
 #define ERTS_SCHED_SPIN_UNTIL_YIELD 100
 
-#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT 10
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_VERY_LONG 40
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_VERY_LONG 1000
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_LONG 20
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_LONG 1000
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM 10
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_MEDIUM 1000
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_SHORT 10
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_SHORT 0
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_VERY_SHORT 5
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_VERY_SHORT 0
+#define ERTS_SCHED_SYS_SLEEP_SPINCOUNT_NONE 0
+#define ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_NONE 0
+
 #define ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT 1000
-#define ERTS_SCHED_TSE_SLEEP_SPINCOUNT \
-  (ERTS_SCHED_SYS_SLEEP_SPINCOUNT*ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT)
 #define ERTS_SCHED_SUSPEND_SLEEP_SPINCOUNT 0
 
 #define ERTS_WAKEUP_OTHER_LIMIT_VERY_HIGH (200*CONTEXT_REDS)
@@ -131,6 +141,12 @@ int erts_sched_thread_suggested_stack_size = -1;
 ErtsLcPSDLocks erts_psd_required_locks[ERTS_PSD_SIZE];
 #endif
 
+static struct {
+    int aux_work;
+    int tse;
+    int sys_schedule;
+} sched_busy_wait;
+
 #ifdef ERTS_SMP
 int erts_disable_proc_not_running_opt;
 
@@ -2046,7 +2062,7 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq)
 
 	erts_smp_runq_unlock(rq);
 
-	spincount = ERTS_SCHED_TSE_SLEEP_SPINCOUNT;
+	spincount = sched_busy_wait.tse;
 
     tse_wait:
 
@@ -2097,7 +2113,7 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq)
 	    }
 
 	    flgs = sched_prep_cont_spin_wait(ssi);
-	    spincount = ERTS_SCHED_TSE_SLEEP_SPINCOUNT;
+	    spincount = sched_busy_wait.aux_work;
 
 	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
 		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
@@ -2134,7 +2150,9 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq)
 	ASSERT(working);
 	sched_wall_time_change(esdp, working = 0);
 
-	spincount = ERTS_SCHED_SYS_SLEEP_SPINCOUNT;
+	spincount = sched_busy_wait.sys_schedule;
+	if (spincount == 0)
+	    goto sys_aux_work;
 
 	while (spincount-- > 0) {
 
@@ -3565,6 +3583,12 @@ erts_early_init_scheduling(int no_schedulers)
 {
     aux_work_timeout_early_init(no_schedulers);
     wakeup_other_limit = ERTS_WAKEUP_OTHER_LIMIT_MEDIUM;
+
+    sched_busy_wait.sys_schedule = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM;
+    sched_busy_wait.tse = (ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM
+			   * ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT);
+    sched_busy_wait.aux_work = (ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM
+				* ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_MEDIUM);
 }
 
 int
@@ -3585,6 +3609,46 @@ erts_sched_set_wakeup_limit(char *str)
     return 0;
 }
 
+int
+erts_sched_set_busy_wait_threshold(char *str)
+{
+    int sys_sched;
+    int aux_work_fact;
+
+    if (sys_strcmp(str, "very_long") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_VERY_LONG;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_VERY_LONG;
+    }
+    else if (sys_strcmp(str, "long") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_LONG;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_LONG;
+    }
+    else if (sys_strcmp(str, "medium") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_MEDIUM;
+    }
+    else if (sys_strcmp(str, "short") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_SHORT;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_SHORT;
+    }
+    else if (sys_strcmp(str, "very_short") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_VERY_SHORT;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_VERY_SHORT;
+    }
+    else if (sys_strcmp(str, "none") == 0) {
+	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_NONE;
+	aux_work_fact = ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_NONE;
+    }
+    else {
+	return EINVAL;
+    }
+
+    sched_busy_wait.sys_schedule = sys_sched;
+    sched_busy_wait.tse = sys_sched*ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT;
+    sched_busy_wait.aux_work = sys_sched*aux_work_fact;
+
+    return 0;
+}
 static void
 init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp)
 {
diff --git a/erts/emulator/beam/erl_process.h b/erts/emulator/beam/erl_process.h
index cff0783bc4..92ff99db6e 100644
--- a/erts/emulator/beam/erl_process.h
+++ b/erts/emulator/beam/erl_process.h
@@ -1097,6 +1097,7 @@ void erts_proclist_destroy(ErtsProcList *);
 int erts_proclist_same(ErtsProcList *, Process *);
 
 int erts_sched_set_wakeup_limit(char *str);
+int erts_sched_set_busy_wait_threshold(char *str);
 
 #if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
 int erts_dbg_check_halloc_lock(Process *p);
diff --git a/erts/etc/common/erlexec.c b/erts/etc/common/erlexec.c
index 19b3bb82ef..a49d10d6f3 100644
--- a/erts/etc/common/erlexec.c
+++ b/erts/etc/common/erlexec.c
@@ -121,6 +121,7 @@ static char *plusM_other_switches[] = {
 /* +s arguments with values */
 static char *pluss_val_switches[] = {
     "bt",
+    "bwt",
     "cl",
     "ct",
     "wt",