Deleted user

Hi @vanfanel , fyi regarding the audio buffer/latency setting in retroarch, when dynamic rate control is enabled then retroarch will target 50% saturation of the audio buffer, which means 64ms...

Mentioning #13262 as well.

> > My proposal: -barrier-placement=[fence-before-store | branch-after-load]. > > I agree with being explicit. However, the option applies only to Arm64 backend at the moment. Are there such architecture-specific configure...

Updated with KC's changes, moving the option to be arm64-only, and renaming it as suggested by Gabriel. I have thought a bit on how to prevent mixing code objects produced...

It would make sense to add information to the `cmx` files indeed. But wouldn't adding new fields require a change of magic numbers and regenerating the `boot` compilers as well?

Checkpoint of work in progress, `cmx` files now embed information and I'm working on adding code to prevent mixing incompatible cmx files at link time. ~~However my #.1 struggle at...

At this point, the strict rejection logic will prevent linking code compiled with `-mbarrier-placement=branch-after-load` against the stdlib, linked with the default `-mbarrier-placement=fence-before-store`. I suppose this calls for an opam option...

If you're positive that the opam option is good enough to prevent accidental mixing of both code flavours unless people are working really hard to shoot themselves in the feet,...

With that feedback, I've removed the `ocamlopt` option and kept everything else. The error message when attempting to mix conflicting objects might have become less understandable. Feel free to suggest...

The changes to the `.cmx` files have intentionally been as extensible as possible to let future changes not require changes to that format again. Having the state of the option...