[llvm] [InstCombine] Match intrinsic recurrences when known to be hoisted (PR #149858)

[Antonio Frighetto via llvm-commits]

antoniofrighetto wrote:

> The question then would be whether handling multi-use is worth a dedicated fold. (Though possibly we could extend foldOpIntoPhi to handle this specific multi-use case, not sure...)

Comparing the results of the first run https://github.com/dtcxzyw/llvm-opt-benchmark/pull/2596 (w/o multi-use phi) and the one in https://github.com/llvm/llvm-project/pull/151115, and it seems like there are four test cases that were not folded via foldOpIntoPhi. Wouldn't this imply that #151115 does not entirely subsume this fold?

https://github.com/llvm/llvm-project/pull/149858