On Wed, Apr 17, 2019 at 08:26:22PM +0800, Zhaoyang Huang wrote: [quoting Johannes here] > As Matthew says, you are fairly randomly making refault activations > more aggressive (especially with that timestamp unpacking bug), and > while that expectedly boosts workload transition / startup, it comes > at the cost of disrupting stable states because you can flood a very > active in-ram workingset with completely cold cache pages simply > because they refault uniformly wrt each other. > [HZY]: I analysis the log got from trace_printk, what we activate have > proven record of long refault distance but very short refault time. You haven't addressed my point, which is that you were only testing workloads for which your changed algorithm would improve the results. What you haven't done is shown how other workloads would be negatively affected. Once you do that, we can make a decision about whether to improve your workload by X% and penalise that other workload by Y%.
