On Mon, Mar 15, 2021 at 07:46:35PM -0600, Edmundo Carmona Antoranz wrote: > On Sun, Mar 14, 2021 at 10:01 AM Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > > > > > - byData = inb(0x61); \ > > > + inb(0x61); \ > > > > Are you sure that the compiler does not make the inb() now go away? > > Hey, Greg! > > Take what you are about to read with a BIG grain of salt because I'm > just learning about the details of what's going on and the tools at my > disposal. > > On v5.11 with an allyesconfig, by changing the value of > CB_DELAY_LOOP_WAIT in mac.h to 100 to force the compiler to include > the `else` path that includes the for with the inb() call, comparing > the resulting srom.o files before/after the patch: > > Compiling with only the value in mac.h changed: > > $ size drivers/staging/vt6655/srom.o > text data bss dec hex filename > 959 0 0 959 3bf drivers/staging/vt6655/srom.o > > By removing the variable in the macro: > > $ size drivers/staging/vt6655/srom.o > text data bss dec hex filename > 959 0 0 959 3bf drivers/staging/vt6655/srom.o > > By also removing the inb() call, in other words, keeping an empty for: > > $ size drivers/staging/vt6655/srom.o > text data bss dec hex filename > 865 0 0 865 361 drivers/staging/vt6655/srom. > > > That being the case, I think that inb is not fading away by removing > the variable... at least in my environment. > > Just in case, it's not like I _want_ the patch to be accepted. I took > it as an exercise and if you think that you want to play safe and > reject it, it's fine. It was still valuable to me because of your > question about the function going away and me trying to find out. Ok, I'll leave this as-is because the inb() needs to happen, so saving the value is normal for stuff like this. thanks, greg k-h
