On Dec 11, 2007, at 9:16 PM, Chris Lattner wrote: > > On Dec 11, 2007, at 9:08 PM, Christopher Lamb wrote: > >> >> On Dec 11, 2007, at 8:50 PM, Chris Lattner wrote: >> >>>>> That is a very strong argument to me, but it seems to argue even >>>>> more >>>>> strongly for: >>>>> >>>>> @G = constant float addrspace(5) 1.0, section "foo", align 4 >>>>> >>>>> Your example above would then be: >>>>> >>>>>> @foo = constant float addrspace(1)* addrspace(2) 1.0 >>>>> >>>>> which has type: >>>>> >>>>>> float addrspace(1)* addrspace(2)* >>>>> >>>>> What do you think? the downside is that this may cause bison to >>>>> have >>>>> issues :) >>>> >>>> This is ideally what I wanted, but it would mean seriously mucking >>>> with bison and changing how all CosntVal's are handled. I can give >>>> it a try. >>> >>> Hrm, I think bison will get extremely grumpy about this. >> >> I feared. >> >>> Quick question, what is the syntax for an external GV with an ASI? >> >> >> Sticky wicket, that. It should be: >> >> @foo = external global i32 addrspace(1) >> >> But that's not working right now.... > > Right, because it would require the same parsing logic as the proposal > above :).
I don't think so. I think the rules were simply missing the OptAddrSpace. > > How horrible would: > >> @foo = addrspace(1) external global i32 > > be? I like that the order in the decl is the order in the type lots. These now work with the rules being adjusted. @foo = extern_weak global i32 addrspace(1) @foo = external global i32 addrspace(1) @foo = dllimport global i32 addrspace(1) @foo = internal global i32 addrspace(1) @foo = weak global i32 addrspace(1) @foo = linkonce global i32 addrspace(1) @foo = appending global i32 addrspace(1) @foo = dllexport global i32 addrspace(1) -- Christopher Lamb _______________________________________________ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
