Treat single CPU as if -Xgc:noConcurrentMark were set

There is little expected benefit with concurrentmark if you are running on a single-CPU machine/environment. On the other hand, you could gain some benefits with noConcurrentMark: 1) no interference from background marking thread; 2) shorter write-barrier path-length: for certain workloads, this provides substantial performance benefits (e.g. Spark TPC-DS q72 in particular).

Fix missing portLib declaration ...

Signed-off-by: Julian <[email protected]>

runtime/gc_modron_startup/mminit.cpp

@@ -2842,6 +2842,8 @@ configurateGCWithPolicyAndOptionsStandard(MM_EnvironmentBase *env)
 MM_Configuration *
 configurateGCWithPolicyAndOptions(OMR_VM* omrVM)
 {
+	J9JavaVM *vm = (J9JavaVM*) omrVM->_language_vm;
+	PORT_ACCESS_FROM_JAVAVM(vm);
 	MM_Configuration *result = NULL;
 	MM_GCExtensions *extensions = MM_GCExtensions::getExtensions(omrVM);
 	MM_EnvironmentBase env(omrVM);
@@ -2867,7 +2869,9 @@ configurateGCWithPolicyAndOptions(OMR_VM* omrVM)
 		extensions->gcModeString = "-Xgcpolicy:gencon";
 		omrVM->gcPolicy = J9_GC_POLICY_GENCON;
 		/* scavenge, concurrentMark, noConcurrentSweep, loa */
-		setDefaultConfigOptions(extensions, true, true, false, true);
+		setDefaultConfigOptions(extensions, true,
+			(1 != j9sysinfo_get_number_CPUs_by_type(J9PORT_CPU_TARGET)),
+			false, true);
 		result = configurateGCWithPolicyAndOptionsStandard(&env);
 		break;