8 May
2008
8 May
'08
6:27 p.m.
Adrian Filipi adrian.filipi@eurotech.com wrote:
Yeah, I agree it's bogus for the pointer to be volatile. There shouldn't be anything unusual about that.
The assembler does show several additional memory accesses, so I think your theory is right. I'm at a loss for what to to on the sync().
I'm at a loss too, I'm afraid...I don't really know PXA.
Linux seems to put an mb() after every write to the flash, so perhaps include/asm-arm/system.h on Linux can provide some hints?
In any case, sync() should probably expand to at least a compiler barrier on all architectures...though I'm not sure if it will make a difference in this particular case.
Haavard