add erl option to set schedulers by percentages

For applications where measurements show enhanced performance from the use
of a non-default number of emulator scheduler threads, having to accurately
set the right number of scheduler threads across multiple hosts each with
different numbers of logical processors is difficult because the erl +S
option requires absolute numbers of scheduler threads and scheduler threads
online to be specified.

To address this issue, add a +SP option to erl, similar to the existing +S
option but allowing the number of scheduler threads and scheduler threads
online to be set as percentages of logical processors configured and
logical processors available, respectively. For example, "+SP 50:25" sets
the number of scheduler threads to 50% of the logical processors
configured, and the number of scheduler threads online to 25% of the
logical processors available. The +SP option also interacts with any
settings specified with the +S option, such that the combination of options
"+S 4:4 +SP 50:50" (in either order) results in 2 scheduler threads and 2
scheduler threads online.

Add documentation for the +SP option.

Add tests for the +SP option to scheduler_SUITE.

Add tests and documentation for two existing features of the +S option: +S
0:0 resets the scheduler thread count and scheduler threads online count to
their defaults, and specifying negative numbers for +S results in those
values being subtracted from the default values for the host.

2 files changed, 183 insertions, 50 deletions

diff --git a/erts/emulator/beam/erl_init.c b/erts/emulator/beam/erl_init.c
index 8d137df7ae..f7bfa91b54 100644
--- a/erts/emulator/beam/erl_init.c
+++ b/erts/emulator/beam/erl_init.c
@@ -549,9 +549,12 @@ void erts_usage(void)
 		 ERTS_SCHED_THREAD_MAX_STACK_SIZE);
     erts_fprintf(stderr, "-spp Bool   set port parallelism scheduling hint\n");
     erts_fprintf(stderr, "-S n1:n2    set number of schedulers (n1), and number of\n");
-    erts_fprintf(stderr, "            schedulers online (n2), valid range for both\n");
-    erts_fprintf(stderr, "            numbers are [1-%d]\n",
+    erts_fprintf(stderr, "            schedulers online (n2), maximum for both\n");
+    erts_fprintf(stderr, "            numbers is %d\n",
 		 ERTS_MAX_NO_OF_SCHEDULERS);
+    erts_fprintf(stderr, "-SP p1:p2   specify schedulers (p1) and schedulers online (p2)\n");
+    erts_fprintf(stderr, "	      as percentages of logical processors configured and logical\n");
+    erts_fprintf(stderr, "	      processors available, respectively\n");
     erts_fprintf(stderr, "-t size     set the maximum number of atoms the "
 			 "emulator can handle\n");
     erts_fprintf(stderr, "            valid range is [%d-%d]\n",
@@ -631,6 +634,8 @@ early_init(int *argc, char **argv) /*
     int ncpuavail;
     int schdlrs;
     int schdlrs_onln;
+    int schdlrs_percentage = 100;
+    int schdlrs_onln_percentage = 100;
     int max_main_threads;
     int max_reader_groups;
     int reader_groups;
@@ -758,63 +763,128 @@ early_init(int *argc, char **argv) /*
 		    }
 		    break;
 		}
-		case 'S' : {
-		    int tot, onln;
-		    char *arg = get_arg(argv[i]+2, argv[i+1], &i);
-		    switch (sscanf(arg, "%d:%d", &tot, &onln)) {
-		    case 0:
-			switch (sscanf(arg, ":%d", &onln)) {
+		case 'S' :
+		    if (argv[i][2] == 'P') {
+			int ptot, ponln;
+			char *arg = get_arg(argv[i]+3, argv[i+1], &i);
+			switch (sscanf(arg, "%d:%d", &ptot, &ponln)) {
+			case 0:
+			    switch (sscanf(arg, ":%d", &ponln)) {
+			    case 1:
+				if (ponln < 0)
+				    goto bad_SP;
+				ptot = 100;
+				goto chk_SP;
+			    default:
+				goto bad_SP;
+			    }
 			case 1:
-			    tot = no_schedulers;
-			    goto chk_S;
+			    if (ptot < 0)
+				goto bad_SP;
+			    ponln = ptot < 100 ? ptot : 100;
+			    goto chk_SP;
+			case 2:
+			    if (ptot < 0 || ponln < 0)
+				goto bad_SP;
+			chk_SP:
+			    schdlrs_percentage = ptot;
+			    schdlrs_onln_percentage = ponln;
+			    break;
 			default:
-			    goto bad_S;
-			}
-		    case 1:
-			onln = tot < schdlrs_onln ? tot : schdlrs_onln;
-		    case 2:
-		    chk_S:
-			if (tot > 0)
-			    schdlrs = tot;
-			else
-			    schdlrs = no_schedulers + tot;
-			if (onln > 0)
-			    schdlrs_onln = onln;
-			else
-			    schdlrs_onln = no_schedulers_online + onln;
-			if (schdlrs < 1 || ERTS_MAX_NO_OF_SCHEDULERS < schdlrs) {
-			    erts_fprintf(stderr,
-					 "bad amount of schedulers %d\n",
-					 tot);
-			    erts_usage();
-			}
-			if (schdlrs_onln < 1 || schdlrs < schdlrs_onln) {
+                        bad_SP:
+                            erts_fprintf(stderr,
+                                         "bad schedulers percentage specifier %s\n",
+                                         arg);
+                            erts_usage();
+                            break;
+                        }
+
+                        VERBOSE(DEBUG_SYSTEM,
+                                ("using %d:%d scheduler percentages\n",
+                                 schdlrs_percentage, schdlrs_onln_percentage));
+                    } else {
+			int tot, onln;
+			char *arg = get_arg(argv[i]+2, argv[i+1], &i);
+			switch (sscanf(arg, "%d:%d", &tot, &onln)) {
+			case 0:
+			    switch (sscanf(arg, ":%d", &onln)) {
+			    case 1:
+				tot = no_schedulers;
+				goto chk_S;
+			    default:
+				goto bad_S;
+			    }
+			case 1:
+			    onln = tot < schdlrs_onln ? tot : schdlrs_onln;
+			case 2:
+			chk_S:
+			    if (tot > 0)
+				schdlrs = tot;
+			    else
+				schdlrs = no_schedulers + tot;
+			    if (onln > 0)
+				schdlrs_onln = onln;
+			    else
+				schdlrs_onln = no_schedulers_online + onln;
+			    if (schdlrs < 1 || ERTS_MAX_NO_OF_SCHEDULERS < schdlrs) {
+				erts_fprintf(stderr,
+					     "bad amount of schedulers %d\n",
+					     tot);
+				erts_usage();
+			    }
+			    if (schdlrs_onln < 1 || schdlrs < schdlrs_onln) {
+				erts_fprintf(stderr,
+					     "bad amount of schedulers online %d "
+					     "(total amount of schedulers %d)\n",
+					     schdlrs_onln, schdlrs);
+				erts_usage();
+			    }
+			    break;
+			default:
+			bad_S:
 			    erts_fprintf(stderr,
-					 "bad amount of schedulers online %d "
-					 "(total amount of schedulers %d)\n",
-					 schdlrs_onln, schdlrs);
+					 "bad amount of schedulers %s\n",
+					 arg);
 			    erts_usage();
+			    break;
 			}
-			break;
-		    default:
-		    bad_S:
-			erts_fprintf(stderr,
-				     "bad amount of schedulers %s\n",
-				     arg);
-			erts_usage();
-			break;
-		    }
 
-		    VERBOSE(DEBUG_SYSTEM,
-			    ("using %d:%d scheduler(s)\n", tot, onln));
-		    break;
-		}
+			VERBOSE(DEBUG_SYSTEM,
+				("using %d:%d scheduler(s)\n", tot, onln));
+		    }
+                    break;
 		default:
 		    break;
 		}
 	    }
 	    i++;
 	}
+
+#ifdef ERTS_SMP
+	/* apply any scheduler percentages */
+	if (schdlrs_percentage != 100 || schdlrs_onln_percentage != 100) {
+	    schdlrs = schdlrs * schdlrs_percentage / 100;
+	    schdlrs_onln = schdlrs_onln * schdlrs_onln_percentage / 100;
+	    if (schdlrs < 1)
+                schdlrs = 1;
+            if (ERTS_MAX_NO_OF_SCHEDULERS < schdlrs) {
+		erts_fprintf(stderr,
+			     "bad schedulers percentage %d "
+			     "(total amount of schedulers %d)\n",
+			     schdlrs_percentage, schdlrs);
+		erts_usage();
+	    }
+	    if (schdlrs_onln < 1)
+                schdlrs_onln = 1;
+            if (schdlrs < schdlrs_onln) {
+		erts_fprintf(stderr,
+			     "bad schedulers online percentage %d "
+			     "(total amount of schedulers %d, online %d)\n",
+			     schdlrs_onln_percentage, schdlrs, schdlrs_onln);
+		erts_usage();
+	    }
+	}
+#endif
     }
 
 #ifndef USE_THREADS
@@ -1312,7 +1382,10 @@ erl_start(int argc, char **argv)
 	    break;
 
 	case 'S' : /* Was handled in early_init() just read past it */
-	    (void) get_arg(argv[i]+2, argv[i+1], &i);
+	    if (argv[i][2] == 'P')
+		(void) get_arg(argv[i]+3, argv[i+1], &i);
+	    else
+		(void) get_arg(argv[i]+2, argv[i+1], &i);
 	    break;
 
 	case 's' : {
diff --git a/erts/emulator/test/scheduler_SUITE.erl b/erts/emulator/test/scheduler_SUITE.erl
index 8931562828..81539faa09 100644
--- a/erts/emulator/test/scheduler_SUITE.erl
+++ b/erts/emulator/test/scheduler_SUITE.erl
@@ -52,6 +52,7 @@
 	 update_cpu_info/1,
 	 sct_cmd/1,
 	 sbt_cmd/1,
+	 scheduler_threads/1,
 	 scheduler_suspend/1,
 	 reader_groups/1]).
 
@@ -66,7 +67,7 @@ all() ->
      equal_with_part_time_max,
      equal_and_high_with_part_time_max, equal_with_high,
      equal_with_high_max, bound_process,
-     {group, scheduler_bind}, scheduler_suspend,
+     {group, scheduler_bind}, scheduler_threads, scheduler_suspend,
      reader_groups].
 
 groups() -> 
@@ -1039,7 +1040,66 @@ sbt_test(Config, CpuTCmd, ClBt, Bt, LP) ->
 		      tuple_to_list(SB)),
     ?line stop_node(Node),
     ?line ok.
-    
+
+scheduler_threads(Config) when is_list(Config) ->
+    SmpSupport = erlang:system_info(smp_support),
+    {Sched, SchedOnln, _} = get_sstate(Config, ""),
+    %% Configure half the number of both the scheduler threads and
+    %% the scheduler threads online.
+    {HalfSched, HalfSchedOnln} = case SmpSupport of
+                                     false -> {1,1};
+                                     true ->
+                                         {Sched div 2,
+                                          SchedOnln div 2}
+                                 end,
+    {HalfSched, HalfSchedOnln, _} = get_sstate(Config, "+SP 50:50"),
+    %% Use +S to configure 4x the number of scheduler threads and
+    %% 4x the number of scheduler threads online, but alter that
+    %% setting using +SP to 50% scheduler threads and 25% scheduler
+    %% threads online. The result should be 2x scheduler threads and
+    %% 1x scheduler threads online.
+    TwiceSched = case SmpSupport of
+                     false -> 1;
+                     true -> Sched*2
+                 end,
+    FourSched = integer_to_list(Sched*4),
+    FourSchedOnln = integer_to_list(SchedOnln*4),
+    CombinedCmd1 = "+S "++FourSched++":"++FourSchedOnln++" +SP50:25",
+    {TwiceSched, SchedOnln, _} = get_sstate(Config, CombinedCmd1),
+    %% Now do the same test but with the +S and +SP options in the
+    %% opposite order, since order shouldn't matter.
+    CombinedCmd2 = "+SP50:25 +S "++FourSched++":"++FourSchedOnln,
+    {TwiceSched, SchedOnln, _} = get_sstate(Config, CombinedCmd2),
+    %% Apply two +SP options to make sure the second overrides the first
+    TwoCmd = "+SP 25:25 +SP 100:100",
+    {Sched, SchedOnln, _} = get_sstate(Config, TwoCmd),
+    %% Configure 50% of scheduler threads online only
+    {Sched, HalfSchedOnln, _} = get_sstate(Config, "+SP:50"),
+    %% Configure 2x scheduler threads only
+    {TwiceSched, SchedOnln, _} = get_sstate(Config, "+SP 200"),
+    %% Test resetting the scheduler counts
+    ResetCmd = "+S "++FourSched++":"++FourSchedOnln++" +S 0:0",
+    {Sched, SchedOnln, _} = get_sstate(Config, ResetCmd),
+    %% Test negative +S settings, but only for SMP-enabled emulators
+    case SmpSupport of
+        false -> ok;
+        true ->
+            SchedMinus1 = Sched-1,
+            SchedOnlnMinus1 = SchedOnln-1,
+            {SchedMinus1, SchedOnlnMinus1, _} = get_sstate(Config, "+S -1"),
+            {Sched, SchedOnlnMinus1, _} = get_sstate(Config, "+S :-1"),
+            {SchedMinus1, SchedOnlnMinus1, _} = get_sstate(Config, "+S -1:-1")
+    end,
+    ok.
+
+get_sstate(Config, Cmd) ->
+    {ok, Node} = start_node(Config, Cmd),
+    [SState] = mcall(Node, [fun () ->
+                                    erlang:system_info(schedulers_state)
+                            end]),
+    stop_node(Node),
+    SState.
+
 scheduler_suspend(Config) when is_list(Config) ->
     ?line Dog = ?t:timetrap(?t:minutes(5)),
     ?line lists:foreach(fun (S) -> scheduler_suspend_test(Config, S) end,