Llvm

This one's for when you're working directly with LLVM's intermediate representation, not just compiling through it. You'll reach for it when you need to inspect IR, run specific opt passes to see what's happening between source and assembly, or debug why a loop didn't vectorize. It covers the whole pipeline: generating IR with clang, running transformations with opt, lowering to assembly with llc, and reading the IR constructs themselves. The pass reference is solid, mem2reg through loop-vectorize, and the debugging section on missed optimizations is exactly what you need when performance isn't what you expected. If you're doing compiler work, writing custom passes, or just trying to understand what your optimizer is actually doing, this gets you there without wandering through documentation.