[...] > atomic access means that you provide a way to warrant an exclusive access to > an address for Read or Write operation. I remind you that i 'm talking about > a uni, omni , single directionnal access : Again, i explain you : there is > one processor which always write while the other one is always reading. Atomic access means that you can get the current value of a memory location and modify it atomically and _that_ is what is needed for the lock-free ringbuffer that was discussed. Only writing to a memory location the worst thing that can happen is that the other processor reading it may not get the value that is in the writer's cache. > > in addition, a member of the VST plugins list (who i will not name > > here) worked extensively on the BeOS audio system and told me > > privately that he has seen other SMP systems (PPC if i recall > > correctly, but i'm not sure) with similar design issues to the > > SPARC. i trust his judgement and recounting of these experiences. > > i don't want the advise of your friends, i don't want the thinking of > students, i want an official statement saying that on SPARC system, the > cache management cannot warrant 32bits operation. And i want that an > electronician explain me how it is possible on 32 bits BUS (or a 64bits or > more) architecture. That was _not_ what we were talking about. It seems you have a different understanding of atomic, i.e. that either the memory location is updated or it is not, but never partially. And yes, that probably is the case when writing a 32bit value to a properly aligned memory location on 32bit (or higher) hardware. > > >> and lets just note that the SPARC is not the only system with this > > >> kind of issue. > > > > > >no longer the Intel based SMP system as far as i see ! :-) > > > > some linux kernel hackers have claimed that recent versions of the > > linux kernel have new checks for certain kinds of x86 XMP systems in > > which cache coherency no longer guarantees 32 bit atomicity. > > Linux Kernel Hacker !? what is that !? a kid who beleive that he is > programmer ! :-))) or it's the new name for IBM employee !? :-) You are quite annoying, no offence intended. --ms ---------------------------------------------------------------------- Generalized Music Plugin Interface (GMPI) public discussion list Participation in this list is contingent upon your abiding by the following rules: Please stay on topic. You are responsible for your own words. Please respect your fellow subscribers. Please do not redistribute anyone else's words without their permission. Archive: //www.freelists.org/archives/gmpi Email gmpi-request@xxxxxxxxxxxxx w/ subject "unsubscribe" to unsubscribe