On Fri, 2012-11-23 at 20:55 -0500, Peter Gavin wrote: > > Make this my vote that mtspr with an unimplemented address is just a > nop. >
+1 In my opinion, this is about knowing what processor features you've got (i.e. a documentation/discovery issue). An exception on write is just a heavyweight version of a feature bit telling you whether or not the SPR is present. Reading back 0 for non-existant regs is "nice", but I don't immediately see the advantage of 0 over "random garbage"... I would go as far as to saying saying "unused SPR space reads back undefined value", but there's possibly a 'hardware' advantage to not doing so..??? /Jonas (NB: the subject of this thread also refers to using NPC which is still a real issue... the PC should be determined with l.jal. I believe this on track, though...?) _______________________________________________ OpenRISC mailing list [email protected] http://lists.openrisc.net/listinfo/openrisc
