Hello guys,
I have a question regarding LLC occupancy recorded by Intel Cache Monitoring
Technology, on Linux.
I use perf_event_open(2) and read(2) to gather CMT values and I multiply by
a "scaling value" read from '/sys/devices/intel_cqm/events/llc_occupancy.scale'
in order to obtain a result in bytes (that's for instance what libvirt does).
But that file gives me a "scaling value" of 32768. When looking at the
corresponding code in the Linux kernel ('arch/x86/events/intel/cqm.c', around
line 1716), I understand that the returned value should be the processor's
cache line size:
cqm_l3_scale = boot_cpu_data.x86_cache_occ_scale;
[...]
get_online_cpus();
for_each_online_cpu(cpu) {
struct cpuinfo_x86 *c = &cpu_data(cpu);
if (c->x86_cache_max_rmid < cqm_max_rmid)
cqm_max_rmid = c->x86_cache_max_rmid;
if (c->x86_cache_occ_scale != cqm_l3_scale) {
pr_err("Multiple LLC scale values, disabling\n");
ret = -EINVAL;
goto out;
}
}
[...]
snprintf(scale, sizeof(scale), "%u", cqm_l3_scale);
str = kstrdup(scale, GFP_KERNEL);
if (!str) {
ret = -ENOMEM;
goto out;
}
event_attr_intel_cqm_llc_scale.event_str = str;
I understand that 'event_attr_intel_cqm_llc_scale.event_str' is the value I
should read from the file, i.e. the CPU's cache line size, which is 64 in
most cases and in mine.
Am I missing something? Thanks for any help!
Regards,
Thibaut S.
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@xxxxxxxxxxxxxxxxx
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
