On Fri, 2014-03-14 at 11:21 +0000, Chris Redpath wrote:
On 14/03/14 09:58, Jon Medhurst (Tixy) wrote:
[...]
This isn't strictly a direct replacement of like-for-like because you've defectively replaced:
curr->avg.load_avg_ratio > hmp_up_threshold
with !(se->avg.load_avg_ratio < hmp_up_threshold)
which gives different results when the values compared are equal, I don't know if that makes much practical difference? In fact, new use possibly make more sense, as before there was a situation where hmp_force_up_migration could make different decisions to hmp_idle_pull. (I'm not particularly familiar with the code so I may not be talking sense.)
Well spotted :) This is intentional, and your comment makes perfect sense.
In which case it might be good to have the change explained in the commit message. Care to come up with some wording? I've not finished looking at the other patches, and we not had any feedback from Alex Shi about the patches yet, so I suggest waiting first to see if the patches needed updated and resending.
Cheers