Re: [PATCH 1/2] x86/CPU/AMD: Present package as die instead of socket

From: Borislav Petkov
Date: Tue Jun 27 2017 - 12:17:35 EST


On Tue, Jun 27, 2017 at 10:55:52PM +0700, Suravee Suthikulpanit wrote:
> The reason we are trying to present "package == NUMA node (die)" here is
> because the topology.txt defines package to contain a number of cores plus
> shared resources (e.g. DRAM controller, shared caches, etc). Since the
> cpuinfo_x86.phys_proc_id is also defined as the physical ID of the package,

Ok, it seems we will continue talking past each other here. So let's
look at the issues separately:

* irqbalance fails to allocating IRQs to individual CPU within the die.

Why does it fail? What is the root cause for this?

* The scheduler fails to load-balance across 8 threads within a die
(e.g. running 8-thread application w/ taskset -c 0-7 ) with
the DIE schedule domain omitted due to x86_has_numa_in_package.

Why do you need to load-balance within the die? Why not load-balance
within the 0-15 threads?

What are the disadvantages of the situation now?

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.