x86: optimize lock prefix switching to run less frequently
On VMs implemented using JITs that cache translated code changing the lock
prefixes is a quite costly operation that forces the JIT to throw away and
retranslate a lot of code.
Previously a SMP kernel would rewrite the locks once for each CPU which
is quite unnecessary. This patch changes the code to never switch at boot in
the normal case (SMP kernel booting with >1 CPU) or only once for SMP kernel
on UP.
This makes a significant difference in boot up performance on AMD SimNow!
Also I expect it to be a little faster on native systems too because a smp
switch does a lot of text_poke()s which each synchronize the pipeline.
v1->v2: Rename max_cpus
v1->v2: Fix off by one in UP check (Thomas Gleixner)
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
diff --git a/init/main.c b/init/main.c
index 3f8aba2..5843fe9 100644
--- a/init/main.c
+++ b/init/main.c
@@ -128,7 +128,7 @@
#ifdef CONFIG_SMP
/* Setup configured maximum number of CPUs to activate */
-static unsigned int __initdata max_cpus = NR_CPUS;
+unsigned int __initdata setup_max_cpus = NR_CPUS;
/*
* Setup routine for controlling SMP activation
@@ -146,7 +146,7 @@
static int __init nosmp(char *str)
{
- max_cpus = 0;
+ setup_max_cpus = 0;
disable_ioapic_setup();
return 0;
}
@@ -155,8 +155,8 @@
static int __init maxcpus(char *str)
{
- get_option(&str, &max_cpus);
- if (max_cpus == 0)
+ get_option(&str, &setup_max_cpus);
+ if (setup_max_cpus == 0)
disable_ioapic_setup();
return 0;
@@ -164,7 +164,7 @@
early_param("maxcpus", maxcpus);
#else
-#define max_cpus NR_CPUS
+#define setup_max_cpus NR_CPUS
#endif
/*
@@ -393,7 +393,7 @@
/* FIXME: This should be done in userspace --RR */
for_each_present_cpu(cpu) {
- if (num_online_cpus() >= max_cpus)
+ if (num_online_cpus() >= setup_max_cpus)
break;
if (!cpu_online(cpu))
cpu_up(cpu);
@@ -401,7 +401,7 @@
/* Any cleanup work */
printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
- smp_cpus_done(max_cpus);
+ smp_cpus_done(setup_max_cpus);
}
#endif
@@ -824,7 +824,7 @@
__set_special_pids(1, 1);
cad_pid = task_pid(current);
- smp_prepare_cpus(max_cpus);
+ smp_prepare_cpus(setup_max_cpus);
do_pre_smp_initcalls();