On 19 February 2013 14:50, Lukasz Majewski <l.majew...@samsung.com> wrote: > Hi Vincent, > >> On 5 February 2013 08:05, Lukasz Majewski <l.majew...@samsung.com> >> wrote: >> > Hi Vincent, >> > >> >> Hi Lukasz, >> >> >> >> I still need to get figures for the cover letter of the V3 of >> >> packing small tasks patches. Nevertheless, the patches series is >> >> available here: >> >> http://git.linaro.org/gitweb?p=people/vingu/kernel.git;a=shortlog;h=refs/heads/sched-pack-small-tasks-v3 >> >> >> > >> > Thanks for sharing code. I will look to it (and test) and provide >> > feedback. >> > >> >> Hi Lucasz, >> >> I have just found that a part of the 1st patch of packing small tasks >> (Revert "sched: Introduce temporary FAIR_GROUP_SCHED dependency for >> load-tracking") has disappeared during my rebase and push on my git >> tree. The consequence is that the branch sched-pack-small-tasks-v3 >> only working if CONFIG_FAIR_GROUP_SCHED is set in your kernel. I have >> pushed a new branch with the fixed >> http://git.linaro.org/gitweb?p=people/vingu/kernel.git;a=shortlog;h=refs/heads/sched-pack-small-tasks-v3-fixed >> > Hi Lucasz,
> I've applied the code to our development 3.8 kernel. It seems to work. > great! > On the other hand I'm wondering about the best set of tests. > > Currently, for basic tests I'm using preempt's-rt cyclic test suite > (./cyclictest -S) with various parameters. It is good enough to see > long therm system stability, but more fine grained code would be nice. > > On the linaro website and some big.LITTLE presentation I've found > information about the linsched tool used by Linaro for validation. > > The newest one which I've found is (LinSched for v3.3-rc7 -> > linsched-alpha): > > http://lwn.net/Articles/486635/ > > Apparently, it needs some porting to run with v3.8 linux kernel, but I'm > optimistic :-) yes, it's not a simple port regarding the lot of change in the scheduler code since v3.3 > > Do you use the same version? No, I don't use linsched but a TC2 versatile board. IIRC, Morten has already used linsched but he has switched to real HW too > Shall I use any special tests? I mainly use cyclictest and sysbench to make functional tests on the scheduler behavior and then i'm using real use case like mp3 and web browsing > > Thanks in advance, > Lukasz > > p.s. I've poked around the linaro website and I didn't find the > linsched code for download. Is there a plan to make it available for > download? AFAICT, we don't have special linsched version but we use the same than you point out above Regards, Vincent > >> >> >> I will send the patchset to LKML as soon as I will have got all >> >> figures of my tests on various platform >> >> >> >> Regards, >> >> Vincent >> >> >> >> On 30 January 2013 10:23, Lukasz Majewski <l.majew...@samsung.com> >> >> wrote: >> >> > Hi Vincent, >> >> > >> >> > Thanks for a very fast response. >> >> > >> >> >> On 30 January 2013 09:35, Lukasz Majewski >> >> >> <l.majew...@samsung.com> wrote: >> >> >> > Dear All, >> >> >> > >> >> >> > >> >> >> > I'd like to ask about the power aware scheduler development: >> >> >> > >> >> >> > https://blueprints.launchpad.net/linaro-power-kernel/+spec/power-aware- >> >> >> > scheduler >> >> >> > >> >> >> > >> >> >> > >> >> >> > Why I'm interested? >> >> >> > >> >> >> > >> >> >> > >> >> >> > I'd like to test and further develop methods to put CPU to >> >> >> > IDLE or changing its operating frequency. I'm especially >> >> >> > interested in packing as much as possible tasks to a CPU and >> >> >> > put the other one to deep idle (RFTS - policy). I'm also >> >> >> > curious how aggressive SCHED_POLICY_POWERSAVING is going to >> >> >> > be? (are there any special requirements) >> >> >> >> >> >> I'm going to release a v3 of packing small tasks patches this >> >> >> week >> >> >> >> >> >> > >> >> >> > >> >> >> > In the above page at the "Work items 2013.01" section states >> >> >> > that there is a work TODO in the "max_power and >> >> >> > current_power" for DVFS. >> >> >> > >> >> >> > Could you share code (if available) and plans for this >> >> >> > development? >> >> >> >> >> >> This step of the blueprint is not already done but there are RFC >> >> >> patch from Chris Redpath about frequency scaling invariance that >> >> >> could interest you >> >> >> http://git.linaro.org/gitweb?p=arm/big.LITTLE/mp.git;a=commit;h=6af10694b3a52f47f6814480e48993b0a9cfa4ef >> >> >> >> >> > >> >> > >> >> > >> >> >> Vincent >> >> >> > >> >> >> > >> >> >> > >> >> >> > I've looked to the: >> >> >> > >> >> >> > git://git.linaro.org/arm/big.LITTLE/mp.git tree (branch: >> >> >> > power-aware- scheduling-v4), but didn't find the power related >> >> >> > code (especially DVFS). >> >> >> > >> >> >> > >> >> >> > >> >> >> > Probably I've looked at wrong place, so any guidance would be >> >> >> > appreciate. >> >> >> > >> >> >> > >> >> >> > >> >> >> > >> >> >> > As a side question: >> >> >> > >> >> >> > On the linaro-dev mailing list there are some patches for >> >> >> > tuning cpufreq governor. Is there any roadmap for this effort? >> >> >> > >> >> >> > >> >> >> > >> >> >> > >> >> >> > -- >> >> >> > Best regards, >> >> >> > >> >> >> > Lukasz Majewski >> >> >> > >> >> >> > Samsung R&D Poland (SRPOL) | Linux Platform Group >> >> > >> >> > >> >> > >> >> > -- >> >> > Best regards, >> >> > >> >> > Lukasz Majewski >> >> > >> >> > Samsung R&D Poland (SRPOL) | Linux Platform Group >> > >> > >> > >> > -- >> > Best regards, >> > >> > Lukasz Majewski >> > >> > Samsung R&D Poland (SRPOL) | Linux Platform Group > > > > -- > Best regards, > > Lukasz Majewski > > Samsung R&D Poland (SRPOL) | Linux Platform Group _______________________________________________ linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev