[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CACRpkdbUye6RhbRNGn6sapARwVUyi5hKS-5VEVBr6ZR6W_KdQw@mail.gmail.com>
Date: Tue, 14 May 2024 09:49:22 +0200
From: Linus Walleij <linus.walleij@...aro.org>
To: Andy Shevchenko <andy.shevchenko@...il.com>
Cc: Patrick Rudolph <patrick.rudolph@...ements.com>, naresh.solanki@...ements.com,
linux-gpio@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] pinctrl: cy8c95x0: Cache muxed registers
On Tue, May 14, 2024 at 7:27 AM Andy Shevchenko
<andy.shevchenko@...il.com> wrote:
> Tue, Dec 19, 2023 at 01:53:49PM +0100, Patrick Rudolph kirjoitti:
> > Currently the port specific registers behind the PORTSEL mux aren't
> > cached in the regmap and thus the typical setup time for a single pin
> > on cy8c9560 is about 200msec on our system. The hotspot is the IRQ
> > (un)masking, which causes lots of R/W operations.
> >
> > Introduce a separate regmap for muxed registers and helper functions
> > to use the newly introduced regmap for muxed register access under
> > the i2c lock.
> >
> > With the new cache in place the typical pin setup time is reduced to
> > 20msec, making it about 10 times faster. As a side effect the system
> > boot time is also reduced by 50%.
>
> Interestingly that you have not Cc'ed me on your patches.
> This, btw, a good reinvetion of a wheel which regmap supports already.
Also sloppy reviewing from my side :(
Patrick can you look into making a patch on top of this
that switches the code to use the built-in caching in regmap
insted of rolling your own?
Yours,
Linus Walleij
Powered by blists - more mailing lists