[PATCH] D118804: Revert "[Clang] Propagate guaranteed alignment for malloc and others"

[James Y Knight via Phabricator via cfe-commits]

jyknight added a comment.

In D118804#3292185 <https://reviews.llvm.org/D118804#3292185>, @ychen wrote:

> I don't see why the patch is wrong... It uses the target/platform-specific `NewAlign`. If the platform allows customized memory allocation that assumes weak alignment, it should set the `NewAlign` accordingly, no?

NewAlign is set to the largest object alignment for which the compiler can call `::operator new(size_t)`, instead of calling `::operator new(size_t, align_t)`, so, no, you definitely wouldn't want to change that value.

That is: it is correctly set to 16 if `::operator new(16)`` will return 16-byte-aligned memory, even if `::operator new(8)` will return an 8-byte-aligned memory. That's because 8-byte-aligned memory is "suitably-aligned" for any possible object that can fit in 8 bytes.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118804/new/

https://reviews.llvm.org/D118804