(Partially) Stabilize stdarch_neon_f16#1922
Conversation
|
|
Is there any reason not to stabilize The fcma intrinsics that don't depend on immediate |
|
Also one more thing, ideally don't update the feature name of the non-stabilized intrinsics, instead change the feature name of the stabilized ones. Otherwise it introduces compile errors in user code, for no good reason |
91ffa69 to
8cb1918
Compare
No. It's included in this pull request now.
Done.
Intrinsics are stablized under the new feature gate |
480dc48 to
24df9ae
Compare
24df9ae to
fd621c6
Compare
Nothing in this list, or the unstable list, seems to use |
All of the attributes are too complex and hard to read, so my script only outputs a subset of them. All intrinsics that use generics employ |
That makes sense. I'm just saying
I don't think what C does is relevant here. We're not aiming for C code to be literally copy-and-pasteable into Rust. What matters is that const generics are kind of unergonomic to use. That's being discussed at rust-lang/rust#146613. |
fd621c6 to
213d617
Compare
fae6ced to
f910718
Compare
f910718 to
0f9516b
Compare
|
This pull request is unblocked, because of rust-lang/rust#149654 (comment). |
sayantn
left a comment
There was a problem hiding this comment.
LGTM, should this be merged now? (just confirming)
Yes. |
Tracking issue: rust-lang/rust#136306
(blocked on rust-lang/rust#149654)stdarch_neon_fp16.stdarch_neon_f16types are still unstable on arm. They're gated bystdarch_arm_neon_intrinsicsnow.stdarch_neon_f16intrinsics are still unstable on arm. They're gated bystdarch_arm_neon_intrinsicsnow.stdarch_neon_f16intrinsics that requiresfhmare stable.stdarch_neon_f16intrinsics that requiresfcmaare still unstable. They're gated bystdarch_neon_fcmanow.stdarch_neon_f16intrinsics that requireslutare still unstable. They're gated bystdarch_neon_feat_lutbefore and after this pull request.stdarch_neon_f16intrinsics that takes or returnsf16,*const f16or*mut f16are still unstable.List of stablized intrinsics.
List of unstable intrinsics.