[clang] [ObjC] Defer to the LLVM backend for unaligned atomic load and stores (PR #79191)

James Y Knight via cfe-commits cfe-commits at lists.llvm.org
Thu Jan 25 05:25:19 PST 2024


jyknight wrote:

Does this cause an ABI incompatibility? E.g. if we have a case where an existing object calls copyStruct on a given object (which presumably has its own internal mutex), and a newly compiled object file calls `__atomic_load` on the same object, implemented with its own internal mutex, the operations won't be atomic w.r.t. each-other.

I don't know enough about ObjC codegen stuff to be able to say if there are mitigating factors that make this OK (e.g. if it's all within one TU?).

Sidenote: please avoid rebase, amend, and force-pushing, because it makes it hard to review what's been changed -- instead, just keep adding more commits to your pull-request branch.


https://github.com/llvm/llvm-project/pull/79191


More information about the cfe-commits mailing list