From: Anton Blanchard <anton@samba.org>
It's basically the x86 version with cpumask fixes when compiling with
NR_CPUS > 64.
---
25-akpm/arch/ppc64/Kconfig | 10 +
25-akpm/arch/ppc64/kernel/smp.c | 222 ++++++++++++++++++++++++++++++++++
25-akpm/include/asm-ppc64/processor.h | 5
3 files changed, 237 insertions(+)
diff -puN arch/ppc64/Kconfig~ppc64-sched-domain-support arch/ppc64/Kconfig
--- 25/arch/ppc64/Kconfig~ppc64-sched-domain-support 2004-03-11 23:51:30.000000000 -0800
+++ 25-akpm/arch/ppc64/Kconfig 2004-03-11 23:51:30.000000000 -0800
@@ -173,6 +173,16 @@ config NUMA
bool "NUMA support"
depends on DISCONTIGMEM
+config SCHED_SMT
+ bool "SMT (Hyperthreading) scheduler support"
+ depends on SMP
+ default off
+ help
+ SMT scheduler support improves the CPU scheduler's decision making
+ when dealing with Intel Pentium 4 chips with HyperThreading at a
+ cost of slightly increased overhead in some places. If unsure say
+ N here.
+
config PREEMPT
bool
help
diff -puN arch/ppc64/kernel/smp.c~ppc64-sched-domain-support arch/ppc64/kernel/smp.c
--- 25/arch/ppc64/kernel/smp.c~ppc64-sched-domain-support 2004-03-11 23:51:30.000000000 -0800
+++ 25-akpm/arch/ppc64/kernel/smp.c 2004-03-11 23:51:30.000000000 -0800
@@ -796,3 +796,225 @@ static int __init topology_init(void)
return 0;
}
__initcall(topology_init);
+
+#ifdef CONFIG_SCHED_SMT
+#ifdef CONFIG_NUMA
+static struct sched_group sched_group_cpus[NR_CPUS];
+static struct sched_group sched_group_phys[NR_CPUS];
+static struct sched_group sched_group_nodes[MAX_NUMNODES];
+static DEFINE_PER_CPU(struct sched_domain, phys_domains);
+static DEFINE_PER_CPU(struct sched_domain, node_domains);
+__init void arch_init_sched_domains(void)
+{
+ int i;
+ struct sched_group *first_cpu = NULL, *last_cpu = NULL;
+
+ /* Set up domains */
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ struct sched_domain *phys_domain = &per_cpu(phys_domains, i);
+ struct sched_domain *node_domain = &per_cpu(node_domains, i);
+ int node = cpu_to_node(i);
+ cpumask_t nodemask = node_to_cpumask(node);
+ cpumask_t tmp1 = cpumask_of_cpu(i ^ 0x1);
+ cpumask_t tmp2 = cpumask_of_cpu(i);
+
+ *cpu_domain = SD_SIBLING_INIT;
+ cpus_or(cpu_domain->span, tmp1, tmp2);
+
+ *phys_domain = SD_CPU_INIT;
+ phys_domain->span = nodemask;
+
+ *node_domain = SD_NODE_INIT;
+ node_domain->span = cpu_possible_map;
+ }
+
+ /* Set up CPU (sibling) groups */
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ int j;
+ first_cpu = last_cpu = NULL;
+
+ if (i != first_cpu(cpu_domain->span)) {
+ cpu_sched_domain(i)->flags |= SD_FLAG_SHARE_CPUPOWER;
+ cpu_sched_domain(first_cpu(cpu_domain->span))->flags |=
+ SD_FLAG_SHARE_CPUPOWER;
+ continue;
+ }
+
+ for_each_cpu_mask(j, cpu_domain->span) {
+ struct sched_group *cpu = &sched_group_cpus[j];
+
+ cpus_clear(cpu->cpumask);
+ cpu_set(j, cpu->cpumask);
+ cpu->cpu_power = SCHED_LOAD_SCALE;
+
+ if (!first_cpu)
+ first_cpu = cpu;
+ if (last_cpu)
+ last_cpu->next = cpu;
+ last_cpu = cpu;
+ }
+ last_cpu->next = first_cpu;
+ }
+
+ for (i = 0; i < MAX_NUMNODES; i++) {
+ int j;
+ cpumask_t nodemask;
+ struct sched_group *node = &sched_group_nodes[i];
+ cpumask_t node_cpumask = node_to_cpumask(i);
+ cpus_and(nodemask, node_cpumask, cpu_online_map);
+
+ if (cpus_empty(nodemask))
+ continue;
+
+ first_cpu = last_cpu = NULL;
+ /* Set up physical groups */
+ for_each_cpu_mask(j, nodemask) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(j);
+ struct sched_group *cpu = &sched_group_phys[j];
+
+ if (j != first_cpu(cpu_domain->span))
+ continue;
+
+ cpu->cpumask = cpu_domain->span;
+ /*
+ * Make each extra sibling increase power by 10% of
+ * the basic CPU. This is very arbitrary.
+ */
+ cpu->cpu_power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE*(cpus_weight(cpu->cpumask)-1) / 10;
+ node->cpu_power += cpu->cpu_power;
+
+ if (!first_cpu)
+ first_cpu = cpu;
+ if (last_cpu)
+ last_cpu->next = cpu;
+ last_cpu = cpu;
+ }
+ last_cpu->next = first_cpu;
+ }
+
+ /* Set up nodes */
+ first_cpu = last_cpu = NULL;
+ for (i = 0; i < MAX_NUMNODES; i++) {
+ struct sched_group *cpu = &sched_group_nodes[i];
+ cpumask_t nodemask;
+ cpumask_t node_cpumask = node_to_cpumask(i);
+ cpus_and(nodemask, node_cpumask, cpu_possible_map);
+
+ if (cpus_empty(nodemask))
+ continue;
+
+ cpu->cpumask = nodemask;
+ /* ->cpu_power already setup */
+
+ if (!first_cpu)
+ first_cpu = cpu;
+ if (last_cpu)
+ last_cpu->next = cpu;
+ last_cpu = cpu;
+ }
+ last_cpu->next = first_cpu;
+
+ mb();
+ for_each_cpu(i) {
+ int node = cpu_to_node(i);
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ struct sched_domain *phys_domain = &per_cpu(phys_domains, i);
+ struct sched_domain *node_domain = &per_cpu(node_domains, i);
+ struct sched_group *cpu_group = &sched_group_cpus[i];
+ struct sched_group *phys_group = &sched_group_phys[first_cpu(cpu_domain->span)];
+ struct sched_group *node_group = &sched_group_nodes[node];
+
+ cpu_domain->parent = phys_domain;
+ phys_domain->parent = node_domain;
+
+ node_domain->groups = node_group;
+ phys_domain->groups = phys_group;
+ cpu_domain->groups = cpu_group;
+ }
+}
+#else /* CONFIG_NUMA */
+static struct sched_group sched_group_cpus[NR_CPUS];
+static struct sched_group sched_group_phys[NR_CPUS];
+static DEFINE_PER_CPU(struct sched_domain, phys_domains);
+__init void arch_init_sched_domains(void)
+{
+ int i;
+ struct sched_group *first_cpu = NULL, *last_cpu = NULL;
+
+ /* Set up domains */
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ struct sched_domain *phys_domain = &per_cpu(phys_domains, i);
+
+ *cpu_domain = SD_SIBLING_INIT;
+ cpu_domain->span = cpu_sibling_map[i];
+
+ *phys_domain = SD_CPU_INIT;
+ phys_domain->span = cpu_possible_map;
+ }
+
+ /* Set up CPU (sibling) groups */
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ int j;
+ first_cpu = last_cpu = NULL;
+
+ if (i != first_cpu(cpu_domain->span)) {
+ cpu_sched_domain(i)->flags |= SD_FLAG_SHARE_CPUPOWER;
+ cpu_sched_domain(first_cpu(cpu_domain->span))->flags |=
+ SD_FLAG_SHARE_CPUPOWER;
+ continue;
+ }
+
+ for_each_cpu_mask(j, cpu_domain->span) {
+ struct sched_group *cpu = &sched_group_cpus[j];
+
+ cpus_clear(cpu->cpumask);
+ cpu_set(j, cpu->cpumask);
+ cpu->cpu_power = SCHED_LOAD_SCALE;
+
+ if (!first_cpu)
+ first_cpu = cpu;
+ if (last_cpu)
+ last_cpu->next = cpu;
+ last_cpu = cpu;
+ }
+ last_cpu->next = first_cpu;
+ }
+
+ first_cpu = last_cpu = NULL;
+ /* Set up physical groups */
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ struct sched_group *cpu = &sched_group_phys[i];
+
+ if (i != first_cpu(cpu_domain->span))
+ continue;
+
+ cpu->cpumask = cpu_domain->span;
+ /* See SMT+NUMA setup for comment */
+ cpu->cpu_power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE*(cpus_weight(cpu->cpumask)-1) / 10;
+
+ if (!first_cpu)
+ first_cpu = cpu;
+ if (last_cpu)
+ last_cpu->next = cpu;
+ last_cpu = cpu;
+ }
+ last_cpu->next = first_cpu;
+
+ mb();
+ for_each_cpu(i) {
+ struct sched_domain *cpu_domain = cpu_sched_domain(i);
+ struct sched_domain *phys_domain = &per_cpu(phys_domains, i);
+ struct sched_group *cpu_group = &sched_group_cpus[i];
+ struct sched_group *phys_group = &sched_group_phys[first_cpu(cpu_domain->span)];
+ cpu_domain->parent = phys_domain;
+ phys_domain->groups = phys_group;
+ cpu_domain->groups = cpu_group;
+ }
+}
+#endif /* CONFIG_NUMA */
+#endif /* CONFIG_SCHED_SMT */
diff -puN include/asm-ppc64/processor.h~ppc64-sched-domain-support include/asm-ppc64/processor.h
--- 25/include/asm-ppc64/processor.h~ppc64-sched-domain-support 2004-03-11 23:51:30.000000000 -0800
+++ 25-akpm/include/asm-ppc64/processor.h 2004-03-11 23:51:30.000000000 -0800
@@ -618,6 +618,11 @@ static inline void prefetchw(const void
#define spin_lock_prefetch(x) prefetchw(x)
+#ifdef CONFIG_SCHED_SMT
+#define ARCH_HAS_SCHED_DOMAIN
+#define ARCH_HAS_SCHED_WAKE_BALANCE
+#endif
+
#endif /* ASSEMBLY */
#endif /* __ASM_PPC64_PROCESSOR_H */
_