[llvm] [RFC][BOLT] Add a new parallel DWARF processing(2/2) (PR #197859)

[via llvm-commits]

Thrrreeee wrote:

> Left some comments, but haven't had a chance to do a super deep dive. Overall question for LTO. For builds with tons of debug info. Say 2GB+ how many buckets we end up with that have cross CU references?

At present, almost all binaries larger than 2 GB use split DWARF. However, there is one binary of more than 1 GB. when  BOLT process, it roughly contains 1616 buckets  totally and 762 buckets have cross CU references (bucket size > 1).

| FILE SIZE % | FILE SIZE | VM SIZE % | VM SIZE | Section |
|---|---:|---:|---:|---|
| 26.8% | 480Mi | 0.0% | 0 | `.debug_info` |
| 22.8% | 407Mi | 0.0% | 0 | `.debug_str` |
| 10.9% | 195Mi | 0.0% | 0 | `.rela.debug_str_offsets` |
| 6.7% | 119Mi | 0.0% | 0 | `.debug_loclists` |

https://github.com/llvm/llvm-project/pull/197859