llvm/llvm-project

[DAG] computeKnownFPClass - add ISD::MAXNUM/MINNUM handling + test coverage

Open

#189,694 创建于 2026年3月31日

在 GitHub 查看
 (9 评论) (0 反应) (1 负责人)C++ (26,378 star) (10,782 fork)batch import
good first issuellvm:SelectionDAG

描述

ISD::FMAXIMUM/ISD::FMINIMUM + ISD::FMAXIMUMNUM/ISD::FMINIMUMNUM handling can be added as well

This will need to be based off the implementation in computeKnownFPClass in ValueTracking.cpp - DenormalMode access might take some additional patches / test coverage

This should be tested with IR tests and not unit tests

贡献者指南

技术栈
cpp
领域
backend
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
5
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
over 1 week
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
blocked
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
LLVM codegen experienceSelectionDAGC++
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
10
研究方向
Examine the existing computeKnownFPClass implementation in llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (or ValueTracking.cpp as mentioned). Understand how FMAXIMUM/FMINIMUM are handled and replicate the logic for MAXNUM/MINNUM. Check for any pending patches related to DenormalMode access that may be needed. Write IR tests for the new handling, ensuring coverage for different floating point scenarios.