[2024-11 LWG Motion 18] P1928R15 std::simd — merge data-parallel types from the Parallelism TS 2 #7440
Conversation
mattkretz
force-pushed
the
motion-2024-11-lwg-18
branch
from
472e4ad to
943a96d
Compare
|
mattkretz
force-pushed
the
motion-2024-11-lwg-18
branch
from
943a96d to
a408f67
Compare
jensmaurer
changed the title
|
I think the second commit is fine as-is. |
| \begin{itemize} | ||
| \item | ||
| \tcode{\exposid{deduced-simd-t}<T>}, if that type is not \tcode{void}, | ||
| otherwise |
There was a problem hiding this comment.
"otherwise" should be at the start of a (the next) bullet, not at the end of the previous one.
There was a problem hiding this comment.
This was a deliberate choice of LWG because the possible type of the alias follows right after each bullet. I don't care either way, but if the "otherwise" always goes at the start of a bullet (without exceptions) then I think I need to make a few more changes.
There was a problem hiding this comment.
@tkoeppe , your opinion here?
https://github.com/cplusplus/draft/wiki/Specification-Style-Guidelines isn't very clear on this topic.
| \tcode{value} identifies the alignment restrictions on pointers used for | ||
| (converting) loads and stores for the given type \tcode{T} on arrays of type | ||
| \tcode{U}. |
There was a problem hiding this comment.
That needs a normative statement somewhere. Cross-reference?
There was a problem hiding this comment.
The normative wording is in [simd.ctor] (the range ctor) and [simd.loadstore]. The cross-reference is directly preceding the note. Should it be in the note instead?
There was a problem hiding this comment.
I think I would prefer that, but it's ok either way.
… TS 2
* Add closing parenthesis in [simd.reductions] p3.
* Add namespace std to simd_flags, basic_simd, and basic_simd_mask
* Add comma after 'e.g.'
* Add \rSec3[simd.expos.defn]{Exposition-only helpers} (missing sibling)
* Remove \iref from simd.expos.abi] (5.3), already in (5.2)
Fixes cplusplus#7430
Fixes cplusplus/papers#670
The return type in the itemdecl/-descr was obviously wrong (ABI tag as first template argument to the simd_mask alias template and missing closing template bracket). The synopsis had the correct return type.
mattkretz
force-pushed
the
motion-2024-11-lwg-18
branch
from
a408f67 to
668fcd6
Compare
simd_flags,basic_simd, andbasic_simd_mask\rSec3[simd.expos.defn]{Exposition-only helpers}(missing sibling)\ireffrom simd.expos.abi] (5.3), already in (5.2)Fixes #7430
Fixes cplusplus/papers#670