Subj : Re: 'volatile' Rules
To   : comp.programming.threads
From : Uenal Mutlu
Date : Wed Jun 08 2005 12:21 am

"David Schwartz" wrote
>
> "Uenal Mutlu" wrote
>
> >>     Unless you're just screwing around for the heck of it, why would you
> >> even mess with something that "just happens to work"? Would you advocate
> >> 'malloc'ing smaller blocks of memory that you are going to use because
> >> your
> >> implementation of 'malloc' happens to round them up?!
>
> > It was an attempt to find reliable rules when and when not to use
> > 'volatile'.
> > There seems indeed be no good reason to use volatile:
> > it's unsafe/unreliable/unpredictable and a performance killer,
> > in short: totally useless.
> > Maybe someone can list when it is ever useful, if any.
>
>     But this attempt is doomed from the start. You cannot find reliable
> rules by experimentation. You can only find unreliable rules this way.
>
>     Just because the rules work in every case in which you tried them does
> not make them reliable. The next compiler, CPU, or OS version may cause your
> code to break. Code is reliable if it relies upon *guarantees*.

Well said, I see and agree.

> Code is reliable if it relies upon *guarantees*.

I like this statement. Someone should tell this also to some software giants
and standardization bodies to document such guarantees... After all it's
all logic and maths; at least it should be.

.