[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200416033715.hscztwkxie2o5i3r@vireshk-i7>
Date: Thu, 16 Apr 2020 09:07:15 +0530
From: Viresh Kumar <viresh.kumar@...aro.org>
To: Sumit Gupta <sumitg@...dia.com>
Cc: rjw@...ysocki.net, catalin.marinas@....com, will@...nel.org,
thierry.reding@...il.com, jonathanh@...dia.com, talho@...dia.com,
linux-pm@...r.kernel.org, linux-tegra@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
bbasu@...dia.com, mperttunen@...dia.com
Subject: Re: [TEGRA194_CPUFREQ Patch 2/3] cpufreq: Add Tegra194 cpufreq driver
On 15-04-20, 16:55, Sumit Gupta wrote:
>
>
> On 14/04/20 11:15 AM, Viresh Kumar wrote:
> > External email: Use caution opening links or attachments
> >
> >
> > On 13-04-20, 17:50, Sumit Gupta wrote:
> > > This was done considering long delay value as explained previously.
> > > Do you think that smp_call_function_single() would be better than work queue
> > > here?
> >
> > Don't work with assumptions, you should test both and see which one
> > works better. Workqueue should never be faster than
> > smp_call_function_single() with my understanding.
> Checked the time taken and its almost same in both cases.
> Earlier we used smp_call_function_single(), but delay time period was small
> in that SOC. In T194, the time period was more. So, this is an optimization
> done because using work queue has advantage as interrupts will not be
> disabled for that period.
Hmm, okay, keep the workqueue and mention the required details in a
comment for everyone to understand why the implementation is done that
way.
--
viresh
Powered by blists - more mailing lists