Volatile were buggy in old compiler: <a href="https://llvm.org/pubs/2008-10-EMSOFT-Volatiles.html" rel="nofollow">https://llvm.org/pubs/2008-10-EMSOFT-Volatiles.html</a> [2008 paper]<p>Even if compilers are better now, if possible it is better to wrap any IO memory access into ioread/iowrite inline (no cost) functions.
This can then be implemented in a way known to be safe for the compiler used. And that can be changed easily in case the architecture or the compiler changes ( yeah abstraction ! )<p>As linux does for example: <a href="https://www.kernel.org/doc/html/v4.15/driver-api/device-io.html" rel="nofollow">https://www.kernel.org/doc/html/v4.15/driver-api/device-io.h...</a>
I have, in various projects over the years, resorted to inline Asm when I couldn't coerce the compiler to do exactly what I wanted; including very rare compiler bugs.
> the compiler vendor took care of it promptly in a paid upgrade to proprietary compiler 5.next<p>Ouch (assuming that this required them to pay to get the bugfix.)
It’s never the compiler, it can’t be the compiler. The first rule of debugging is that it can’t be the compiler…<p>Wow does it ever suck and waste a ton of time when it IS the compiler. I feel sorrow for whomever had to find this out in their workflow