[DAG] SDPatternMatch - match integer min/max patterns as well as the existing ISD nodes

[RKSimon]

We already have matchers for ISD::SMAX/SMIN/UMAX/UMIN nodes, but it'd be useful if we could match icmp+select patterns as well, preferably with the same existing m_SMax/m_SMin/m_UMax/m_UMin helpers.

This would be similar to the the IR PatternMatch MaxMin_match helpers which match the min/max intrinsics and the cmp+select patterns.

[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. Check that no other contributor has already been assigned to this issue. If you believe that no one is actually working on it despite an assignment, ping the person. After one week without a response, the assignee may be changed.
2. In the comments of this issue, request for it to be assigned to you, or just create a pull request after following the steps below. Mention this issue in the description of the pull request.
3. Fix the issue locally.
4. Run the test suite locally. 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.
5. Create a Git commit.
6. Run git clang-format HEAD~1 to format your changes.
7. Open a pull request to the upstream repository on GitHub. Detailed instructions can be found in GitHub's documentation. Mention this issue in the description of the pull request.

If you have any further questions about this issue, don't hesitate to ask via a comment in the thread below.

[llvmbot]

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

Author: Simon Pilgrim (RKSimon)

We already have matchers for ISD::SMAX/SMIN/UMAX/UMIN nodes, but it'd be useful if we could match icmp+select patterns as well, preferably with the same existing m_SMax/m_SMin/m_UMax/m_UMin helpers.

This would be similar to the the IR PatternMatch MaxMin_match helpers which match the min/max intrinsics and the cmp+select patterns.

[AmrDeveloper]

Interested to work on it

[c8ef]

Hi @AmrDeveloper, are you still working on this?

[nikic]

What's the motivation for this? The IR PatternMatch behavior is mostly a leftover from the time before we introduced min/max intrinsics. I wouldn't want SDAG pattern match to repeat that without good cause.

[RKSimon]

We have several patterns (add/sub sat, abd etc.) where we want to match min/max patterns but we might not have created a min/max node due to legality etc. (either in SelectionDAGBuilder::visitSelect or DAGCombiner)

[nikic]

I see. Do you think that all places using those patterns would also want to match the select form?

[RKSimon]

The uses so far in DAGCombine - yes - but to avoid problems in the future we could just create m_SMaxLike (etc) sibling matchers that handle all cases (similar to how we have m_SExt and m_SExtLike - m_SExtLike matches both sext + zext_nneg cases).