On Fri, Apr 7, 2017 at 10:48 AM, Charles Keepax ckeepax@opensource.wolfsonmicro.com wrote:
On Fri, Apr 07, 2017 at 10:30:12AM +0200, Linus Walleij wrote:
On Fri, Apr 7, 2017 at 10:27 AM, Linus Walleij linus.walleij@linaro.org wrote:
On Wed, Apr 5, 2017 at 12:07 PM, Richard Fitzgerald rf@opensource.wolfsonmicro.com wrote:
This patch adds a header file of register definitions for Cirrus Logic "Madera" class codecs. These codecs are all based off a common set of hardware IP so have a common register map (with a few minor device-to-device variations). These are complex devices with a large mber of features and so have a correspondingly large register set. The registers.h file has been auto-generated from the hardware register definitions, stripped down to only registers we need to access from the driver.
Signed-off-by: Richard Fitzgerald rf@opensource.wolfsonmicro.com
This: include/linux/mfd/madera/registers.h | 8832 ++++++++++++++++++++++++++++++++++
Get included in all subdrivers I suppose?
So you are broadcasting 8800+ lines into every subdriver across the entire kernel.
Just the time spent in the preprocessor parsing this will affect compilation time.
Or maybe this is a necessary sacrifice to get the regmap cache centralized in MFD. I don't know. I feel stupid.
I guess I should focus on "my" subsystems...
This only gets included in files that are part of this driver, it shouldn't affect compilation time for anyone not building the madera driver and even then it should only affect compilation times for the 10 or so C files that make up the driver. Also I don't really see any other way to specify the registers for the device.
No when using regmap cache this seems necessary. I was just wrong.
Yours, Linus Walleij