kernel: Decide whether to use simple or affine scheduler

Simple scheduler is used when we do not have to worry about cache affinity
(i.e. single core with or without SMT, multicore with all cache levels
shared).

When we replace gSchedulerLock with more fine grained locking affine
scheduler should also be chosen when logical CPU count is high (regardless
of cache).

headers/private/kernel/cpu.h

@@ -76,6 +76,7 @@ typedef struct cpu_ent {
 
 //extern cpu_ent gCPU[MAX_BOOT_CPUS];
 extern cpu_ent gCPU[];
+extern uint32 gCPUCacheLevelCount;
 
 
 #ifdef __cplusplus

src/system/kernel/arch/x86/arch_cpu.cpp

@@ -131,7 +131,6 @@ static uint32 sHierarchyShift[CPU_TOPOLOGY_LEVELS];
 
 /* Cache topology information */
 static uint32 sCacheSharingMask[CPU_MAX_CACHE_LEVEL];
-static uint32 sCacheLevelCount;
 
 
 static status_t
@@ -611,7 +610,7 @@ detect_amd_cache_topology(uint32 maxExtendedLeaf)
 
 	for (int i = 0; i < maxCacheLevel; i++)
 		sCacheSharingMask[i] = ~uint32(hierarchyLevels[i] - 1);
-	sCacheLevelCount = maxCacheLevel;
+	gCPUCacheLevelCount = maxCacheLevel;
 }
 
 
@@ -725,7 +724,7 @@ detect_intel_cache_topology(uint32 maxBasicLeaf)
 	for (int i = 0; i < maxCacheLevel; i++)
 		sCacheSharingMask[i] = ~uint32(hierarchyLevels[i] - 1);
 
-	sCacheLevelCount = maxCacheLevel;
+	gCPUCacheLevelCount = maxCacheLevel;
 }
 
 
@@ -789,7 +788,7 @@ detect_cpu_topology(int currentCPU, cpu_ent* cpu, uint32 maxBasicLeaf,
 		= get_topology_level_id(topologyID, CPU_TOPOLOGY_PACKAGE);
 
 	unsigned int i;
-	for (i = 0; i < sCacheLevelCount; i++)
+	for (i = 0; i < gCPUCacheLevelCount; i++)
 		cpu->cache_id[i] = topologyID & sCacheSharingMask[i];
 	for (; i < CPU_MAX_CACHE_LEVEL; i++)
 		cpu->cache_id[i] = -1;
@@ -800,14 +799,14 @@ detect_cpu_topology(int currentCPU, cpu_ent* cpu, uint32 maxBasicLeaf,
 		cpu->topology_id[CPU_TOPOLOGY_CORE],
 		cpu->topology_id[CPU_TOPOLOGY_SMT]);
 
-	if (sCacheLevelCount > 0) {
+	if (gCPUCacheLevelCount > 0) {
 		char cacheLevels[256];
 		unsigned int offset = 0;
-		for (i = 0; i < sCacheLevelCount; i++) {
+		for (i = 0; i < gCPUCacheLevelCount; i++) {
 			offset += snprintf(cacheLevels + offset,
 					sizeof(cacheLevels) - offset,
 					" L%d id %d%s", i + 1, cpu->cache_id[i],
-					i < sCacheLevelCount - 1 ? "," : "");
+					i < gCPUCacheLevelCount - 1 ? "," : "");
 
 			if (offset >= sizeof(cacheLevels))
 				break;

src/system/kernel/cpu.cpp

@@ -21,6 +21,7 @@
 
 /* global per-cpu structure */
 cpu_ent gCPU[MAX_BOOT_CPUS];
+uint32 gCPUCacheLevelCount;
 
 static spinlock sSetCpuLock;
 

src/system/kernel/scheduler/scheduler.cpp

@@ -59,20 +59,39 @@ scheduler_remove_listener(struct SchedulerListener* listener)
 }
 
 
+static bool
+should_use_affine_scheduler(int32 cpuCount)
+{
+	if (cpuCount < 2)
+		return false;
+
+	for (int32 i = 1; i < cpuCount; i++) {
+		for (int32 j = 0; j < gCPUCacheLevelCount; j++) {
+			if (gCPU[i].cache_id[j] != gCPU[i - 1].cache_id[j])
+				return true;
+		}
+	}
+
+	return false;
+}
+
+
 void
 scheduler_init(void)
 {
 	int32 cpuCount = smp_get_num_cpus();
-	dprintf("scheduler_init: found %" B_PRId32 " logical cpu%s\n", cpuCount,
-		cpuCount != 1 ? "s" : "");
+	dprintf("scheduler_init: found %" B_PRId32 " logical cpu%s and %" B_PRId32
+		" cache level%s\n", cpuCount, cpuCount != 1 ? "s" : "",
+		gCPUCacheLevelCount, gCPUCacheLevelCount != 1 ? "s" : "");
 
 	status_t result;
-#if 0
-	dprintf("scheduler_init: using affine scheduler\n");
-	result = scheduler_affine_init();
-#endif
-	dprintf("scheduler_init: using simple scheduler\n");
-	result = scheduler_simple_init();
+	if (should_use_affine_scheduler(cpuCount)) {
+		dprintf("scheduler_init: using affine scheduler\n");
+		result = scheduler_affine_init();
+	} else {
+		dprintf("scheduler_init: using simple scheduler\n");
+		result = scheduler_simple_init();
+	}
 
 	if (result != B_OK)
 		panic("scheduler_init: failed to initialize scheduler\n");