Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-10 Thread Stratos Karafotis
On 10/11/2016 02:13 πμ, Rafael J. Wysocki wrote: > On Wed, Nov 9, 2016 at 6:55 AM, Viresh Kumar wrote: >> On 08-11-16, 21:25, Stratos Karafotis wrote: >>> But this is the supposed behaviour of conservative governor. We want >>> the CPU to increase the frequency in steps. The patch just resets >>

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-09 Thread Rafael J. Wysocki
On Wed, Nov 9, 2016 at 6:55 AM, Viresh Kumar wrote: > On 08-11-16, 21:25, Stratos Karafotis wrote: >> But this is the supposed behaviour of conservative governor. We want >> the CPU to increase the frequency in steps. The patch just resets >> the frequency to a lower frequency in case of idle. >>

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-09 Thread Stratos Karafotis
On 09/11/2016 07:55 πμ, Viresh Kumar wrote: > On 08-11-16, 21:25, Stratos Karafotis wrote: >> But this is the supposed behaviour of conservative governor. We want >> the CPU to increase the frequency in steps. The patch just resets >> the frequency to a lower frequency in case of idle. >> >> For

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-08 Thread Viresh Kumar
On 08-11-16, 21:25, Stratos Karafotis wrote: > But this is the supposed behaviour of conservative governor. We want > the CPU to increase the frequency in steps. The patch just resets > the frequency to a lower frequency in case of idle. > > For argument's sake, let's assume that the governor time

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-08 Thread Stratos Karafotis
On 08/11/2016 10:32 πμ, Viresh Kumar wrote: > On 8 November 2016 at 12:49, Stratos Karafotis wrote: >> I think we shouldn't. That's why the patch first decreases the frequency >> by n freq steps (where n the number of deferred periods). >> Then the normal processing takes place. > > The problem t

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-08 Thread Viresh Kumar
On 8 November 2016 at 12:49, Stratos Karafotis wrote: > I think we shouldn't. That's why the patch first decreases the frequency > by n freq steps (where n the number of deferred periods). > Then the normal processing takes place. The problem that I see is that the new algorithm will reduce the f

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-07 Thread Viresh Kumar
On 07-11-16, 19:27, Stratos Karafotis wrote: > Yes, it could be done only when we decrease frequency. But I thought that > maybe > this is against conservative governor principle. > > I initially observed this issue on a Snapdragon 808 using conservative on the > big cluster (A57). The CPU seemed

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-07 Thread Stratos Karafotis
Hi, Thanks for reviewing. On 07/11/2016 08:12 πμ, Viresh Kumar wrote: > For the record, I have never got the original mail with this subject. I'm sorry for inconvenience. It seems that I had an issue on my mail server. > On 06-11-16, 11:19, Stratos Karafotis wrote: >> Conservative governor cha

Re: [Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-06 Thread Viresh Kumar
For the record, I have never got the original mail with this subject. On 06-11-16, 11:19, Stratos Karafotis wrote: > Conservative governor changes the CPU frequency in steps. > That means that if a CPU runs at max frequency, it will need several > sampling periods to return at min frequency when t

[Resend][PATCH] cpufreq: conservative: Decrease frequency faster when the timer deferred

2016-11-06 Thread Stratos Karafotis
Conservative governor changes the CPU frequency in steps. That means that if a CPU runs at max frequency, it will need several sampling periods to return at min frequency when the workload is finished. If the timer that calculates the load and target frequency is deferred, the governor might need