[llvm] [x86] Optimize urem with a constant divisor to use multiply-by-reciprocal (PR #92669)

[Craig Topper via llvm-commits]

topperc wrote:

It only calls __umoddi3 when compiling for 32-bit x86. For x86-64 it already uses a multiply.

https://github.com/llvm/llvm-project/pull/92669