[llvm-branch-commits] [llvm] AMDGPU: Add a subtarget feature for fine-grained remote memory support (PR #96442)

[Alex Voicu via llvm-branch-commits]

AlexVlx wrote:

Hmm, I might need to reparse this but I'm not entirely sure how this is 
expected to work in general since PCI-E attachment is not always / entirely 
knowable at compile time. Have I missed some preamble discussion here?

https://github.com/llvm/llvm-project/pull/96442
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

[llvm-branch-commits] [clang] release/20.x: [clang][CodeGen] `sret` args should always point to the `alloca` AS, so use that (#114062) (PR #127552)

[Alex Voicu via llvm-branch-commits]

AlexVlx wrote:

@tstellar apologies for the delay in circling back to this; here are the three 
options we have here, neither of which is ideal, sadly:

1. We backport this, #127528 and #128166 (if it gets approved & merged); these 
have to go together, and are, as you observed, a bit of a chonk;
2. We do the revert in #127771, as @arsenm suggests - my only concerns here are 
that it might be a rug pull for folks that might've implicitly taken a 
dependency on the behaviour that introduced some months ago and that the entire 
private-as-default hack is pretty egregious;
3. To unblock device libs on our end, I'm pretty certain it's enough to 
backport only the AS casting for IR args from here: 

 - I appreciate bespoke changes that only live in a release branch are icky, 
but I'll weakly retort that this is both individually sound (we probably should 
have been doing this ever since we moved away from typed pointers) and part of 
a larger change that'll land in 21 (the topic of this conversation), so perhaps 
less objectionable?

Personally, I don't have a strong preference, albeit I naively feel that (3) 
might be the safest.

https://github.com/llvm/llvm-project/pull/127552
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits