So one can measure SMP overhead.
Signed-off-by: Marcelo Tosatti <mtosa...@redhat.com>
Index: qemu-kvm-parallel-vmexit/kvm/user/test/x86/vmexit.c
===================================================================
--- qemu-kvm-parallel-vmexit.orig/kvm/user/test/x86/vmexit.c
+++ qemu-kvm-parallel-vmexit/kvm/user/test/x86/vmexit.c
@@ -80,22 +80,38 @@ static struct test {
void (*func)(void);
const char *name;
int (*valid)(void);
+ int parallel;
} tests[] = {
- { cpuid, "cpuid", },
- { vmcall, "vmcall", },
- { mov_from_cr8, "mov_from_cr8" },
- { mov_to_cr8, "mov_to_cr8" },
- { ipi, "ipi", is_smp },
- { ipi_halt, "ipi+halt", is_smp },
+ { cpuid, "cpuid", .parallel = 1, },
+ { vmcall, "vmcall", .parallel = 1, },
+ { mov_from_cr8, "mov_from_cr8", .parallel = 1, },
+ { mov_to_cr8, "mov_to_cr8" , .parallel = 1, },
+ { ipi, "ipi", is_smp, .parallel = 0, },
+ { ipi_halt, "ipi+halt", is_smp, .parallel = 0, },
};
+unsigned iterations;
+volatile int nr_cpus_done;
+
+static void run_test(void *_func)
+{
+ int i;
+ void (*func)(void) = _func;
+
+ for (i = 0; i < iterations; ++i)
+ func();
+
+ nr_cpus_done++;
+}
+
static void do_test(struct test *test)
{
int i;
unsigned long long t1, t2;
- unsigned iterations = 32;
void (*func)(void) = test->func;
+ iterations = 32;
+
if (test->valid && !test->valid()) {
printf("%s (skipped)\n", test->name);
return;
@@ -104,8 +120,17 @@ static void do_test(struct test *test)
do {
iterations *= 2;
t1 = rdtsc();
- for (i = 0; i < iterations; ++i)
- func();
+
+ if (!test->parallel) {
+ for (i = 0; i < iterations; ++i)
+ func();
+ } else {
+ nr_cpus_done = 0;
+ for (i = cpu_count(); i > 0; i--)
+ on_cpu_async(i-1, run_test, func);
+ while (nr_cpus_done < cpu_count())
+ ;
+ }
t2 = rdtsc();
} while ((t2 - t1) < GOAL);
printf("%s %d\n", test->name, (int)((t2 - t1) / iterations));
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
