[LLVMdev] Loads moving across barriers

[Chandler Carruth]

So, I don't have a horse in the OpenCL race, and I'd like to mostly keep it
that way, but I want to point out that some of these things have much
broader implications.

On Fri, Nov 8, 2013 at 11:45 PM, Matt Arsenault <arsenm2 at gmail.com> wrote:

> I don’t think think outright disallowing this transform is the right
> solution. This would be valid for OpenCL private or constant address
> spaces, it’s just global or local where this would be a problem.


Unless I've misunderstood (which is a possibility, I've not invested proper
time thinking about this, which I'll do if you indicate I'm missing
something), this would be tantamount to tying certain address to specific
rules regarding potential data races in the memory model. I'm not very
comfortable with that. Perhaps you're instead trying to say that with
certain address spaces "noalias" (and by inference, "restrict" at the
language level) has a different semantic model than other address spaces?
While it's less worrisome than the first interpretation, I still don't
really like it.


> This comes back to the questions about how to handle address space alias
> information which there was a long thread about a few months ago. There was
> debate over address spaces as a language vs. a target concept, and I don’t
> remember right now what the consensuses were there.


I think if OpenCL has special aliasing properties for specific address
spaces it would make much more sense (IMO) to either have a specialized
alias analysis that implements this, or to work out a proper metadata-based
system for expressing these constraints from the frontend.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20131109/c40bc1b3/attachment.html>