From: Nick Piggin <piggin@cyberone.com.au>

The following patch builds a scheduling description for the i386
architecture using cpu_sibling_map to set up SMT if CONFIG_SCHED_SMT is
set.

It could be made more fancy and collapse degenerate domains at runtime (ie.
 1 sibling per CPU, or 1 NUMA node in the computer).



---

 arch/i386/Kconfig            |   10 +
 arch/i386/kernel/smpboot.c   |  229 +++++++++++++++++++++++++++++++++++++++++++
 include/asm-i386/processor.h |    5 
 3 files changed, 244 insertions(+)

diff -puN arch/i386/Kconfig~sched-domains-i386-ht arch/i386/Kconfig
--- 25/arch/i386/Kconfig~sched-domains-i386-ht	2004-01-24 16:20:44.000000000 -0800
+++ 25-akpm/arch/i386/Kconfig	2004-01-24 16:20:44.000000000 -0800
@@ -459,6 +459,16 @@ config NR_CPUS
 	  This is purely to save memory - each supported CPU adds
 	  approximately eight kilobytes to the kernel image.
 
+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 "Preemptible Kernel"
 	help
diff -puN arch/i386/kernel/smpboot.c~sched-domains-i386-ht arch/i386/kernel/smpboot.c
--- 25/arch/i386/kernel/smpboot.c~sched-domains-i386-ht	2004-01-24 16:20:44.000000000 -0800
+++ 25-akpm/arch/i386/kernel/smpboot.c	2004-01-24 16:20:44.000000000 -0800
@@ -38,6 +38,7 @@
 #include <linux/kernel.h>
 
 #include <linux/mm.h>
+#include <linux/sched.h>
 #include <linux/kernel_stat.h>
 #include <linux/smp_lock.h>
 #include <linux/irq.h>
@@ -1119,6 +1120,234 @@ static void __init smp_boot_cpus(unsigne
 		synchronize_tsc_bp();
 }
 
+#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;
+	cpumask_t all_cpus = CPU_MASK_NONE;
+	struct sched_group *first_cpu = NULL, *last_cpu = NULL;
+
+	for (i = 0; i < NR_CPUS; i++) {
+		if (!cpu_possible(i))
+			continue;
+
+		cpu_set(i, all_cpus);
+	}
+
+	/* Set up domains */
+	for_each_cpu_mask(i, all_cpus) {
+		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);
+
+		/* TODO: change me to SD_SIBLING_INIT */
+		*cpu_domain = SD_CPU_INIT;
+		cpu_domain->span = cpu_sibling_map[i];
+		cpu_domain->flags |= SD_FLAG_WAKE;
+		cpu_domain->cache_hot_time = 0;
+		cpu_domain->cache_nice_tries = 0;
+		cpu_domain->max_interval = 2;
+		cpu_domain->busy_factor = 8;
+		cpu_domain->imbalance_pct = 110;
+
+		*phys_domain = SD_CPU_INIT;
+		phys_domain->span = nodemask;
+		phys_domain->flags |= SD_FLAG_IDLE;
+
+		*node_domain = SD_NODE_INIT;
+		node_domain->span = all_cpus;
+	}
+
+	/* Set up CPU (sibling) groups */
+	for_each_cpu_mask(i, all_cpus) {
+		struct sched_domain *cpu_domain = cpu_sched_domain(i);
+		int j;
+		first_cpu = last_cpu = NULL;
+
+		if (i != first_cpu(cpu_domain->span))
+			continue;
+
+		for_each_cpu_mask(j, cpu_domain->span) {
+			struct sched_group *cpu = &sched_group_cpus[j];
+
+			cpu->cpumask = CPU_MASK_NONE;
+			cpu_set(j, cpu->cpumask);
+
+			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;
+		cpus_and(nodemask, node_to_cpumask(i), all_cpus);
+
+		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;
+
+			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;
+		cpus_and(nodemask, node_to_cpumask(i), all_cpus);
+
+		if (cpus_empty(nodemask))
+			continue;
+
+		cpu->cpumask = nodemask;
+
+		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_mask(i, all_cpus) {
+		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;
+	cpumask_t all_cpus = CPU_MASK_NONE;
+	struct sched_group *first_cpu = NULL, *last_cpu = NULL;
+
+	for (i = 0; i < NR_CPUS; i++) {
+		if (!cpu_possible(i))
+			continue;
+
+		cpu_set(i, all_cpus);
+	}
+
+	/* Set up domains */
+	for_each_cpu_mask(i, all_cpus) {
+		struct sched_domain *cpu_domain = cpu_sched_domain(i);
+		struct sched_domain *phys_domain = &per_cpu(phys_domains, i);
+
+		/* TODO: change me to SD_SIBLING_INIT */
+		*cpu_domain = SD_CPU_INIT;
+		cpu_domain->span = cpu_sibling_map[i];
+		cpu_domain->flags |= SD_FLAG_WAKE;
+		cpu_domain->cache_hot_time = 100000;
+		cpu_domain->cache_nice_tries = 0;
+		cpu_domain->max_interval = (HZ/500 ?: 1),
+		cpu_domain->busy_factor = 8;
+		cpu_domain->imbalance_pct = 110;
+
+		*phys_domain = SD_CPU_INIT;
+		phys_domain->span = all_cpus;
+		phys_domain->flags |= SD_FLAG_IDLE;
+	}
+
+	/* Set up CPU (sibling) groups */
+	for_each_cpu_mask(i, all_cpus) {
+		struct sched_domain *cpu_domain = cpu_sched_domain(i);
+		int j;
+		first_cpu = last_cpu = NULL;
+
+		if (i != first_cpu(cpu_domain->span))
+			continue;
+
+		for_each_cpu_mask(j, cpu_domain->span) {
+			struct sched_group *cpu = &sched_group_cpus[j];
+
+			cpu->cpumask = CPU_MASK_NONE;
+			cpu_set(j, cpu->cpumask);
+
+			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_mask(i, all_cpus) {
+		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;
+
+		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_mask(i, all_cpus) {
+		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 */
+
 /* These are wrappers to interface to the new boot process.  Someone
    who understands all this stuff should rewrite it properly. --RR 15/Jul/02 */
 void __init smp_prepare_cpus(unsigned int max_cpus)
diff -puN include/asm-i386/processor.h~sched-domains-i386-ht include/asm-i386/processor.h
--- 25/include/asm-i386/processor.h~sched-domains-i386-ht	2004-01-24 16:20:44.000000000 -0800
+++ 25-akpm/include/asm-i386/processor.h	2004-01-24 16:20:44.000000000 -0800
@@ -630,4 +630,9 @@ extern inline void prefetchw(const void 
 
 extern void select_idle_routine(const struct cpuinfo_x86 *c);
 
+#ifdef CONFIG_SCHED_SMT
+#define ARCH_HAS_SCHED_DOMAIN
+#define ARCH_HAS_SCHED_WAKE_BALANCE
+#endif
+
 #endif /* __ASM_I386_PROCESSOR_H */

_