[PATCH] D146523: [AMDGPU]: Add new intrinsic llvm.amdgcn.convergent.copy

[Brian Sumner via Phabricator via llvm-commits]

b-sumner added a comment.

In D146523#4210894 <https://reviews.llvm.org/D146523#4210894>, @foad wrote:

> In D146523#4210843 <https://reviews.llvm.org/D146523#4210843>, @b-sumner wrote:
>
>> FWIW, there is no desire to read from inactive lanes.  The loop is supposed to only be reading from, and writing to, lanes that were active before the for loop is executed by a select single lane.
>
> Then I'm back to not understanding what this convergent copy is for. I'd need to see a more complete example.

As I understand it, if an expression is involved, the computation of the expression is sinking into the loop, and that must be prevented.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D146523/new/

https://reviews.llvm.org/D146523