On Thu, May 25, 2017 at 11:14:39PM +0300, Evgeniy Polyakov wrote: > Frankly saying, your non-regmap version was so much simpler, smaller and > cleaner. > I wonder why did people force you to use regmap.
> Guys, please speak up, if you want driver authors to use THIS, it is really > flawed. > Sebastien, iirc it was your idea to use regmaps, what was the reason for this? So, this is the first I've seen of this series but in general the main reason for using regmap is that it allows best practice to be factored out into common code both in terms of the I/O itself and also in terms of framework code being able to provide regmap based helpers for common operations so drivers need less code. The cache code is also really useful for a lot of devices, it's a performance boost for reads and is well optimized for restoring device state if you don't care about sequencing (this is where some of the MMIO users come from). You also get a bunch of debug support for free, things like the tracepoints and debugfs files (those do work better if you describe the register map). Looking at the code in the regmap patch I have to say I'm very surprised if this is adding much complexity, usually moving to regmap is basically a sed.
signature.asc
Description: PGP signature
