[llvm] [X86] Lower vector integer division and remainder through float division (PR #205263)

[Simon Pilgrim via llvm-commits]

RKSimon wrote:

> Looks like there's no CostTable for variable division in X86TargetTransformInfo? the cost of vector SDIV/SREM/UDIV/UREM also needs to be included, otherwise vectorizers will not attempt to vectorize them.

Good luck with that - IIRC cleaning up DIV/REM x86 scalar and vector costs has a lot of strange effects on existing middle-end tests - expect a lot of yak shaving and a long PR chain :/

https://github.com/llvm/llvm-project/pull/205263