Some builtin intrinsics that are not constexpr but could be

[RKSimon]

Bugzilla Link	47249
Version	unspecified
OS	Windows NT
CC	@erichkeane,@zygoloid
Fixed by commit(s)	e7d9182 , 2ceac91 , 00158ae

This isn't exhaustive, but all are low hanging fruit:

bit twiddling:

• __builtin_bitreverse D86339 e7d9182
• __builtin_rotateleft D86342 2ceac91
• __builtin_rotateright D86342 2ceac91

vectors:

• __builtin_convertvector
• __builtin_shufflevector

-fms-extensions variants:

• __lzcnt16, __lzcnt, __lzcnt64 D157420 00158ae
• __popcnt16, __popcnt, __popcnt64 D157420 00158ae
• _rotl8, _rotl16, _rotl, _rotl64, _lrotl
• _rotr8, _rotr16, _rotr, _rotr64, _lrotr

[RKSimon]

https://reviews.llvm.org/D86339 - __builtin_bitreverse support

[RKSimon]

https://reviews.llvm.org/D86342 - __builtin_rotateleft / __builtin_rotateright support

[RKSimon]

https://reviews.llvm.org/D86339 - __builtin_bitreverse support

e7d9182

[RKSimon]

https://reviews.llvm.org/D86342 - __builtin_rotateleft /
__builtin_rotateright support

2ceac91

[chfast]

Also __builtin_addc and __builtin_subc?

[Endilll]

Feels like a good first issue.

CC @shafik @AaronBallman

[RKSimon]

Feels like a good first issue.

Agreed although they need splitting into separate tasks - the ms-extension variants are pretty trivial, but convertvector might be harder and I expect shufflevector to be tricky.

[shafik]

@RKSimon Looking at the PRs above they look relatively straight forward.

[llvmbot]

Hi!

This issue may be a good introductory issue for people new to working on LLVM. If you would like to work on this issue, your first steps are:

1. Assign the issue to you.
2. Fix the issue locally.
3. Run the test suite locally.
   3.1) Remember that the subdirectories under test/ create fine-grained testing targets, so you can
   e.g. use make check-clang-ast to only run Clang's AST tests.
4. Create a git commit
5. Run git clang-format HEAD~1 to format your changes.
6. Submit the patch to Phabricator.
   6.1) Detailed instructions can be found here

For more instructions on how to submit a patch to LLVM, see our documentation.

If you have any further questions about this issue, don't hesitate to ask via a comment on this Github issue.

@llvm/issue-subscribers-good-first-issue

[alexguirre]

Hi, seems like the ROTL/ROTR ms-extension intrinsics were already made constexpr in a1dc3d2.

I would like to work on the remaining intrinsics.

[RKSimon]

Hi, seems like the ROTL/ROTR ms-extension intrinsics were already made constexpr in a1dc3d2.

I would like to work on the remaining intrinsics.

Great - I'd recommend starting with the the lzcnt/popcnt ms-extension intrinsics as they should be similar to the previous cases.

After that, __builtin_convertvector is probably easier than __builtin_shufflevector but both are very different to the other intrinsics, and will require some refactoring for missing vector type special case handling.

[alexguirre]

Submitted a patch for lzcnt/popcnt https://reviews.llvm.org/D157420

[BertalanD]

Also __builtin_addc and __builtin_subc?

I'm on it: D156156. I'll try to find time to finish it this week.

[RKSimon]

Also #51787 (elementwise and reduction builtins)

[Destroyerrrocket]

Could I take the last two missing expressions?

[shafik]

Could I take the last two missing expressions?

It does not look like anyone is currently working on them, so I can assign this to you if you want.

[Destroyerrrocket]

Thank you, I'll take it then!

[RKSimon]

@Destroyerrrocket Go for it! I'd recommend you start with __builtin_convertvector

[Destroyerrrocket]

Hi! I've made the changes for both intrinsics here: #76615
(I think that now you're meant to use github PRs? The bot seems to suggest to use Phabricator, but it seems down)
Can someone take a look? :D

[Endilll]

(I think that now you're meant to use github PRs? The bot seems to suggest to use Phabricator, but it seems down)

Yes, we moved over to GitHub PRs since that comment was posted.

[RKSimon]

Closing - __builtin_shufflevector and __builtin_convertvector were handled in #76615