Florian Hahn
Florian Hahn
ConstraintElimination is able to remove the checks in the loop If you replace the `size_t` with `unsigned`. The issue is that `getelementptr` indices are treated as signed, in LLVM IR...
Thanks @cynecx , your example highlighted several areas that needed generalizing a bit. All checks in the loop should be removed now using current `main`.
Posted https://discourse.llvm.org/t/retain-fact-that-index-offsets-for-gep-are-non-negative-for-size-t-uint64-t-offsets/65974 to discuss options for handling `size_t` index offsets.
@prakharbahuguna is working on this. I could not assign it to him, so I assigned myself to make sure it's not unassigned any more.
https://simd.godbolt.org/z/4z1vMfMGj is still crashing
Still crashing on current `main` with the same assertion: https://clang.godbolt.org/z/5rK8WM1q8
We could provide an API for clients to check if a given ISA version supports a given encoding. For example, half precision floating point arithmetic was only added in Armv8.2...
Yep, I just thought we should discuss the whole signup/social auth stuff once we've finished migrating to class based views.
> I would prefer not to disable this test, because it should start passing on Clang 18 (17.0.5 doesn't implement C23.) Would it be possible to check in cmake if...
d10245e added `fwrapv` to the flags for this test, so the overflow shouldn't be an issue, while preserving the original test. Is this change still needed?