23 points | by luu 2 days ago ago
5 comments
I wonder how this will work for Rust where bytes are u8 - will they special-case them ? That would look weird.
Also there are some botched examples, e.g. :
> define void @src(ptr %a, ptr %b) > { > %mem = alloca i8 > call void @llvm.memcpy(ptr %mem, ptr %a, i32 1) > call void @llvm.memcpy(ptr %a, ptr %mem, i32 1) > ret void > }
The last %a should be a %b. But you can get the idea anyway.
wouldn't byte in rust be more like MaybeUninit<u8>, to account for uninit and pointer fragments?
That means MaybeUninit would not be library-only anymore, but would need some compiler help. But yes you're right.
It already has lots of "magic" semantics and never was just a library type
Yes.
I wonder how this will work for Rust where bytes are u8 - will they special-case them ? That would look weird.
Also there are some botched examples, e.g. :
> define void @src(ptr %a, ptr %b) > { > %mem = alloca i8 > call void @llvm.memcpy(ptr %mem, ptr %a, i32 1) > call void @llvm.memcpy(ptr %a, ptr %mem, i32 1) > ret void > }
The last %a should be a %b. But you can get the idea anyway.
wouldn't byte in rust be more like MaybeUninit<u8>, to account for uninit and pointer fragments?
That means MaybeUninit would not be library-only anymore, but would need some compiler help. But yes you're right.
It already has lots of "magic" semantics and never was just a library type
Yes.