[LLVMdev] [RFC][PATCH][OPENCL] synchronization scopes redux

[Philip Reames]

On 01/13/2015 10:33 PM, Chandler Carruth wrote:
>
> On Tue, Jan 13, 2015 at 10:27 PM, Sameer Sahasrabuddhe 
> <sameer.sahasrabuddhe at amd.com <mailto:sameer.sahasrabuddhe at amd.com>> 
> wrote:
>
>     Ping! We need to close on whether everyone is convinced that
>     symbolic memory scopes have a significant advantage over opaque
>     numbers. Either of them will be examined by optimizations using a
>     target-implemented API. I personally don't think that readability
>     in the LLVM text format is worth the effort, especially given that
>     address spaces work well enough with opaque numbers.
>
>
> I am much more comfortable with symbolic memory scopes. The reason I 
> feel this way is actually because there *is* a particular ordering of 
> them that the target will mandate. Having an ordering but having it 
> *not* be the order of the numbers used seems too actively confusing to 
> me.
Chandler, while I agree with you, I'm not sure this should block forward 
progress.  Having the scope markers as opaque integers doesn't block us 
from later adding a side table of string names and changing the IR 
representation.  Forward serialization would be fairly straight forward 
as long as you didn't care about the resulting name.  Why don't we let 
this move forward with numeric IDs and then come back with string names 
at some point in the future?

Philip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20150114/de4fc2a6/attachment.html>