[llvm] `dwarf2yaml.cpp` optimizations (PR #179048)

[James Henderson via llvm-commits]

jh7370 wrote:

> @jh7370 Thanks for reviewing! I'll take address the code comment later.
> 
> > do you have any performance figures to back this up or is this fixing some other issue, like a clang-tidy check etc?
> 
> It's for performance. `wasm-emscripten-finalize` is really slow for me (~30s in total, but I'm addressing a subset) and I shaved of like 3 seconds of reallocating & copying with [WebAssembly/binaryen#8257](https://github.com/WebAssembly/binaryen/pull/8257). The PR you're looking at now applies a subset of these changes to LLVM upstream. Note that this upstream version is different from the Binaryen fork and has a couple more redundant copies than the fork does.

So to clarify, I see in [WebAssembly/binaryen#8257](https://github.com/WebAssembly/binaryen/pull/8257) that you've made changes outside the LLVM code. Do you still get the performance improvements in `wasm-emscripten-finalize` if you drop those changes? I'd just like to make sure we are attributing the gains to the right changes.

Also, please address the private email item in the automated comment [here](https://github.com/llvm/llvm-project/pull/179048#issuecomment-3828783306).

https://github.com/llvm/llvm-project/pull/179048