On Wed, Feb 28, 2018 at 01:01:21PM +0530, Viresh Kumar wrote:
On 28-02-18, 15:23, Leo Yan wrote:
But we don't want to place a boosted task on one low capacity CPU, so finally select max value: max(min_util, new_util).
I am not sure I read the code as you stated above and so a bit confused no why you said it that way. The only thing we are making sure with this code is that we set new_util to boosted-task-util on a almost fully-idle CPU (whose utilization is almost 0).
Yeah, if boosted-task-util is big enough, it's bias to select big CPU; if boosted-task-util isn't big value, then it has side effect to eliminate util difference so cannot find a cpu with real lowest util.
Do I miss any other issue?
Thanks, Leo Yan