Allow SIMD types in generics. Closes #10604 #11717
Conversation
| @@ -424,19 +425,22 @@ fn generic_type_of(cx: &CrateContext, r: &Repr, name: Option<&str>, sizing: bool | |||
| let align = most_aligned.align; | |||
| let discr_ty = ll_inttype(cx, ity); | |||
| let discr_size = machine::llsize_of_alloc(cx, discr_ty) as u64; | |||
| let align_units = (size + align - 1) / align - 1; | |||
There was a problem hiding this comment.
What happens if size == n * align? Doesn't this give n - 1? (Is this correct?)
There was a problem hiding this comment.
It would be (n * align - 1) / align. Either way, all that was changed was I moved that line from 435 to 428 and added the minus one from 439. That was so I could use it in the match arms.
There was a problem hiding this comment.
It is n - 1, which is right — this patch also changes align_units from being the size of the entire object to the size of the padding.
|
cc @jdm, I think you're quite familiar with the adt module |
|
cc @sanxiyn too |
|
Looks good to me. |
| // option. This file may not be copied, modified, or distributed | ||
| // except according to those terms. | ||
|
|
||
| fn main() { |
There was a problem hiding this comment.
This will need to be pub fn main to pass the check-fast target on windows.
|
Updated the tests based on your comments, @alexcrichton; barring other issues, this PR should be good to go. |
|
Ah, I forgot, but these need an |
|
Ah kk, I was greatly confused. |
|
Both need |
|
Updated both to include |
Note that it still doesn't allow generic types to be marked with #[simd].
Note that it still doesn't allow generic types to be marked with #[simd].