Re: [RFC PATCH 3/4] sched/topology: remove smt_gain

From: Qais Yousef
Date: Mon Sep 10 2018 - 07:05:58 EST

On 04/09/18 11:37, Srikar Dronamraju wrote:
* Vincent Guittot <vincent.guittot@xxxxxxxxxx> [2018-09-04 11:36:26]:

Hi Srikar,

Le Tuesday 04 Sep 2018 Ã 01:24:24 (-0700), Srikar Dronamraju a Ãcrit :
However after this change, capacity_orig of each SMT thread would be
1024. For example SMT 8 core capacity_orig would now be 8192.

smt_gain was suppose to make a multi threaded core was slightly more
powerful than a single threaded core. I suspect if that sometimes hurt
Is there system with both single threaded and multi threaded core ?
That was the main open point for me (and for Qais too)

I dont know of any systems that have come with single threaded and
multithreaded. However some user can still offline few threads in a core
while leaving other cores untouched. I dont really know why somebody
would want to do it. For example, some customer was toying with SMT 3
mode in a SMT 8 power8 box.

What was the customer trying to achieve by this? Did this end up being useful?

If we get mixed SMT 3 and SMT 8 in the system we might have issues, but again I don't see how this makes sense from system point of view. I don't think power savings will be significant by turning off few hardware threads since the core which I'd expect to be the main energy consumer is still on. From performance perspective, SMT 3 might have less contention (hence better 'performance'), but we don't do anything special to decide to schedule work in this case to take advantage of that. So I don't think the setup is useful to cater for or encourage.

Qais Yousef