Lines Matching defs:lag
650 * Fair schedulers conserve lag:
699 * maximal (virtual) lag induced in the system due to quantisation.
766 * and end up with a larger lag than we started with.
773 * -r_max < lag < max(r_max, q)
779 s64 lag, limit;
782 lag = avg_vruntime(cfs_rq) - se->vruntime;
785 se->vlag = clamp(lag, -limit, limit);
790 * eg. lag >= 0.
3710 * adjusted if re-weight at !0-lag point.
3713 * re-weight without changing vruntime at !0-lag point.
3719 * Since lag needs to be preserved through re-weight:
3721 * lag = (V - v)*w = (V'- v')*w', where v = v'
3736 * Since we are doing at !0-lag point which means V != v, we
3747 * at 0-lag point, thus breach assumption. Proof completed.
3769 * always occurs at 0-lag point, so V won't change. Or else
4457 * The lag estimation comes with a cost we don't want to pay all the
5199 s64 lag = 0;
5206 * adding tasks with positive lag, or removing tasks with negative lag
5207 * will move 'time' backwards, this can screw around with the lag of
5216 lag = se->vlag;
5219 * If we want to place a task and preserve lag, we have to
5221 * average and compensate for this, otherwise lag can quickly
5229 * the virtual lag:
5239 * Then, the weighted average after adding an entity with lag
5248 * And the actual lag after adding an entity with vl_i is:
5254 * Which is strictly less than vl_i. So in order to preserve lag
5255 * we should inflate the lag before placement such that the
5256 * effective lag after placement comes out right.
5259 * we need to use such that the lag after placement is the lag
5274 lag *= load + scale_load_down(se->load.weight);
5277 lag = div_s64(lag, load);
5280 se->vruntime = vruntime - lag;
5334 * XXX now that the entity has been re-weighted, and it's lag adjusted,