[PATCH] D27146: Merge strings using a probabilistic algorithm to reduce latency.

[Sean Silva via Phabricator via llvm-commits]

silvas added a comment.

One other approach (which might even complement this) is that we might be able to come up with a fast heuristic for determining if a string is likely unique or likely not-unique. I assume there is some sort of pattern; e.g. probably file paths are duplicated a lot in debug info. We could calculate these heuristics nearly for free at the same time as we hash the strings.

With such a heuristic, we can bypass the hash table lookup altogether.


https://reviews.llvm.org/D27146