LLVM should have more end-to-end tests for IR opt+codegen
Open
llvm:instcombine
llvm:optimizations
llvm:SLPVectorizer
test-suite
- Dominant language
- LLVM
- Stars
- 40.5k
- Forks
- 18.7k
- PR merge metrics
- PR metrics pending
Description
Many InstCombine and SLP optimizations may seem superior in IR level, but in fact will lead to suboptimal codegen. LLVM should add end-to-end tests to prevent such cases from being missed.
https://github.com/llvm/llvm-project/issues/202245
https://github.com/llvm/llvm-project/issues/200538
https://github.com/llvm/llvm-project/issues/203756#issuecomment-4759845565
Contributor guide
Research direction
Start by reading the three linked issues and reviewing the InstCombine and SLP optimization areas named here. Identify cases where an IR-level optimization produces suboptimal codegen, then add end-to-end coverage for those cases and verify the tests prevent the regressions from being missed.
Written by the indexing model from the issue text.
Assessment
- Domain
- compilers, testing-qa
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100