Fill in initial details #1
Conversation
CHARTER.md
Outdated
|
|
||
| The initial implementation will be built on LLVM intrinsics directly rather than `core::arch`. | ||
| This is so we don't have to block the portable API on filling in a lot of missing intrinsics in `core::arch` on non-x86 platforms. | ||
| Migrating from LLVM intrinsics to `core::arch` is kept as a future possibility. |
There was a problem hiding this comment.
I think this would also require some sort of detection of #[target_feature], since LLVM intrinsics already interact with it correctly. Not sure if it's worth noting
There was a problem hiding this comment.
As far as I'm aware, you tell LLVM what you want and you tell it the features available and it solves the problem as best it can. Which is what makes intrinsics so powerful.
There was a problem hiding this comment.
Sorry--I meant that switching to core::arch would probably require a language change as well, to allow that same interaction.
There was a problem hiding this comment.
Ah, yes. Moving to a core::arch based implementation would be a Tall Task, to say the least.
There was a problem hiding this comment.
I guess what I was trying to get at with this was consciously not painting ourselves into any corners that would make moving to core::arch totally infeasible instead of just lots of work. It probably goes without saying. Should I reword it or just cut it out?
There was a problem hiding this comment.
I agree that we should make it clear it's not exclusively possible with LLVM intrinsics, but I think this is potentially misleading when the language feature is probably a bigger hurdle than expanding core::arch. Not sure the best way of phrasing that, though! For reference, rust-lang/rust#42515 seems to be the associated issue.
There was a problem hiding this comment.
It sounds like we’ll want to capture this, but probably not in the charter 🙂 I’ll just remove this note 👍
This fills in our initial group details.
r? @calebzulawski @Lokathor @workingjubilee