Ok,
may be I'm here... :)
volatiles are ... difficult.
from the point of view of the compiler the 'volatile' qualifier means
nothing else but 'the compiler does know how the variable is evaluated'.
That's it.
So,
- one can find a perfect example which will break the compiler's behaviour.
- one may learn a bit from the past experience if the 'V' qualifier is
absolutely necessary (it there a chapter in docs about this? :) )
- one may find a way which is most suitable for the particular case.
Like interrupt source, semaphores, etc. other?
- if you fill the thing does not work properly - ask and show us why or
where :) then fix it :)
I did not use any other compiler than gcc for last 20 years (except one
from arm, but this is a different story)... Volatiles worked :)
Cheers,
~d
P.S. Chris, thanks for the good explanation. ...and the
stuffmaintenance. Keep doing this fine thing.
Chris Liechti ?????:
i don't have a smart comment except:
- on msp430 it's perfectly possible to read-modify-write to peripherals.
that's what the original 3.2.3 code was doing on all volatiles. Dmitry
implemented an optimization so that bis/bic instructions are emitted
where possible.
- except for the watchdog and flash controller... where the example they
are showing in the paper would never be used by a developer.
WATCHDOG=WATCHDOG is nonsense as the MSP430 watchdog by design reads a
different value that it is required to write, causing the WDTPW to be
wrong. so even if GCC would read the register and write it back, it
would still reset.
- volatile access only disables caching, it doesn't protect against
incomplete data for larger data types. that is, if one function reads
a 32 bit value while the interrupt function modifies that data, it can
read intermediate values. the correct thing to do is to disable
interrupts during the data structure access. that's easiest done with
a separate access function, marked with the "critical" attribute.
(see ringbuffer in libmspgcc for an example)
- if there is a bug, it's open source feel free to provide patches ;-)
chris
John Porubek schrieb:
Not sure if this has been discussed previously, but it's worth a
mention if not. This is not MSP430 specific (although the referenced
paper does mention the GCC port to the MSP430), but applies to anybody
using C or C++ for development. A colleague showed me an item in the
February 2009 issue of Embedded Systems Design magazine written by
Jack Ganssle that is both interesting and scary. In a nutshell,
compilers sometimes miscompile code using "volatile" qualifiers. See:
http://preview.tinyurl.com/cb3sk2
It references a paper that discusses the problem in detail. URL here
for convenience:
http://preview.tinyurl.com/6pmr42
It would be interesting to hear the viewpoints on this of the compiler
vendors who follow this group.
John
------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users
------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users
