[llvm] [InstCombine] Don't fold integer-float bitcasts in store type combining (PR #212423)

[Nikita Popov via llvm-commits]

https://github.com/nikic requested changes to this pull request.

This has been extensively discussed in the past, and the consensus is that this is a backend issue. Just not one that anyone cares to fix, esp. given the costs involved and the irrelevance of non-sse2 targets.

https://github.com/llvm/llvm-project/pull/212423