Bob, It all looks OK to me. The fact that both I/O and register space have unaligned read/write functions should provide enough flexibility to support the variety of VAX models now simulated. I still have a few details to work out for these other models. I may also need to do something about accesses to Qbus memory space (0x30000000 upwards) as I guess this should behave the same as Qbus I/O space? It's not really an issue for CVAX at the moment as accesses reflect to main memory.
Matt On 26/11/2013 01:33, Bob Supnik wrote: > I have a design with probe coding for fixing the unaligned IO space > problem recently uncovered in the VAX emulator. > > Because it tears up the core of unaligned processing, I'd like some > other eyes to look at it before I start editing code and debugging. > > My plan is: > > a) verify all cases of unaligned access to memory by simple tests > (read B, W, L, Q unaligned at every possible offset; write B, W, L, Q > unaligned at every possible offset) > b) run HCORE, the hard core diagnostic - this will test unaligned > references with memory management > c) verify that the VAX 780 still runs VMS (unaligned IO references are > not allowed, so only unaligned memory references need to be working > correctly) > d) verify all cases of unaligned access to Qbus IO space - a bit > tricky, as I need to find IO registers that are not "sparse" > e) verify that CVAX still runs VMS > f) turn over the changes to the community for further testing > > The design paper is attached, for review. > > /Bob > > > > _______________________________________________ > Simh mailing list > [email protected] > http://mailman.trailing-edge.com/mailman/listinfo/simh
_______________________________________________ Simh mailing list [email protected] http://mailman.trailing-edge.com/mailman/listinfo/simh
