On 05/02/2018 07:45 AM, Michael Matz wrote: > Hi, > > On Tue, 1 May 2018, Jeff Law wrote: > >> The very nature of a traditional asm implies that it can read or write >> anything visible to compiler. We can't realistically peek inside to see >> what's happening and the user hasn't provided appropriate dataflow >> information. One could make the argument that traditional asms should >> go the way of the dodo bird. I think we looked at that not terribly >> long ago, but didn't really get anywhere. > > Yeah, with traditional asms we have to assume dependencies to everything > (right now, i.e. if we don't want to enter the rathole of deprecating > them). But we were also talking about non-traditional asms, i.e. with > operands, where I think Alexander made the case that those should have > explicit dependencies if they access the same registers as global reg > vars. For a non-traditional ASM ISTM that they should provide the dependency, even for global register variables. So I'd agree with Alexander on that.
Jeff
