Re: [avr-gcc-list] Save RAMPZ register in interrupt handler routines

[David Brown]

Paulo Marques wrote:
> Quoting Anatoly Sokolov <address@hidden>:
>
> > Hello.
>
> Hi,
>
> >  For FLASH memory access for 128KB avr-libc have pgm_read_*_far macros.
> > These macros modify RAMPZ register and is non-reenterable. Using these
> > macros in main loop/interrupts code can cause collisions of access to 
> > FLASH
> > memory.
> >
> >  For solving of this problem I offer save RAMPZ register in interrupt
> > handler routines for 128KB devices ('avr31' and 'avr51' 
> > architectures). The
> > '-mno-save-rampz' switch restore old behaviour.
>
> Wouldn't it be better to just disable interrupts / restore interrutps in 
> the macros themselves? Or maybe just create _atomic variants that do 
> this, or even just document that if you use pgm_read_*_far macros inside 
> interrupts you have to make sure that in the main code you do that from 
> atomic sections or at least with interrupts disabled...
>
> Otherwise you will be paying a latency penalty on every interrupt call 
> even when you don't need to (and that is probably most of the time).

Disabling and re-enabling interrupts during the pgm_read_*_far macros is 
also bad - it would introduce a space and time penalty on every 
pgm_read_*_far call.  I'd imagine that the use of pgm_read_*_far macros 
within interrupt functions is pretty rare - perhaps the best idea is to 
use the patch, but make "-mno-save-rampz" the default?  Or is there some 
practical way to let the interrupt function save rampz if and only if it 
is used (or if a function is called), just like the volatile registers?