[PATCH] D100717: [InstCombine] Transform memcpy to ptr load/stores if TBAA says so

[Jeroen Dobbelaere via Phabricator via llvm-commits]

jeroen.dobbelaere added a comment.

Looks good.

Still two remarks:

- Can you add a langref update indicating that `vtable pointer` and `any pointer` names might be used to drive optimization passes/code generation ?
- Can you extend the testcase with a `vtable pointer` and an `int` case ?

And also two questions:

- The testcase tries out a p64:64:64 and a p32:32:32 case, assuming that the p32:32:32 case should use a i64. Is this indeed the expected lowering ? Or would we expect two pointer load/stores ?
- When I try out a struct with single pointer fields and array pointer fields, I see that clang produces separate pointer fields for the single pointers and a large 'char' field for the array of pointers. (See https://www.godbolt.org/z/PT3qorvnq ) Maybe that is something that could be changed later in a pointer field with a larger size ?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D100717/new/

https://reviews.llvm.org/D100717