[llvm] [AggressiveInstCombine] Add shufflevector support to TruncInstCombine (PR #220218)

[Nikita Popov via llvm-commits]

nikic wrote:

This looks reasonable. @RKSimon Any concerns with unconditionally truncating shufflevectors like this? It looks like this transform always allows folding truncates into vector ops, so at least it's consistent.

@laxman-gupta1006 Please delete your PR description and rewrite it from scratch, in your own words, omitting any irrelevant information. Keep this in mind for all future PRs as well.

https://github.com/llvm/llvm-project/pull/220218