llvm-optimization

To develop a custom LLVM pass, use the modern Pass Manager to transform IR and guide analyses for scalar, loop, or interprocedural improvements. This approach allows you to reliably insert targeted transformations into your compiler pipeline.

Verify LLVM IR transformation correctness by using Alive2 to validate pass rewrites and catch incorrect transformations. This correctness-aware debugging ensures your optimizations do not degrade program behavior during the IR-to-machine-code workflow.

Choose LLVM optimization levels and pipeline stages by evaluating the requirements for scalar, loop, and interprocedural improvements like inlining, LICM, and vectorization. Selecting the right pipeline stages ensures you effectively improve generated code quality.

Yes, you can apply this to loop vectorization and dead-code removal tasks within your compiler toolchain. It supports scalar, loop, and interprocedural optimization workflows to improve generated program performance across IR transformations.

The best way to debug LLVM pass execution is through pass debugging and profiling workflows that inspect execution and validate transformations. This process helps identify performance degradation or incorrect rewrites during pipeline tuning.

LLVM optimization passes produce incorrect rewrites when transformations are not validated for correctness. Inspect pass execution and use Alive2 to verify that your IR transformations maintain program semantics and catch invalid rewrites.