[PATCH] D27960: Use exact vector capacities to store DWARF line tables

[Simon Que via Phabricator via llvm-commits]

sque added a comment.

Her'es another 5 runs of each. None of these is the initial run after building llvm-symbolizer, which seems to take slightly longer because of loading the 69 MB binary for the first time.

With shrink_to_fit:
318800 5.64
318808 5.35
318796 5.74
318812 5.43
318816 5.76

With dry run:
314600 5.49
314592 4.85
314544 5.68
314600 5.14
314544 5.05

I think the difference in duration is just noise. But changing to a dry-run model seems like a lot of code churn for saving 4.2 MB of max RSS on a 200 MB input file.


https://reviews.llvm.org/D27960