A large binary's DWARF info can contain line tables with thousands of entries. Since dynamically resized vectors don't have exact capacities, there could be a large amount of wasted space. This patch changes the DWARF reader to store the line table rows in vectors with exact capacities.
Details
Details
Diff Detail
Diff Detail
Event Timeline
Comment Actions
Doesn't vector have "shrink_to_fit()" for this purpose?
Isn't there a way to reserve() appropriately initially?
Also two general issues:
- You need to add llvm-commits as a subscriber for the patch *on creation*. So now you should close it and reopen a new one, with llvm-commits (this is because for the archive, we want the mailing list to have a proper thread with the patch from the start).
- It is nicer when reviewing patches to have the full context (git diff -U9999, or use arc).