Commits on Feb 7, 2024

1. Outsmart the LLVM optimizer …

   The old definitions of bool_1, bool_2, bool_3 in simd_op_check_x86 (etc) all referred to the same entry in in_f32; as of llvm/llvm-project#76367, the LLVM optimizer is smart enough to realize that (eg) bool1 != bool2 by construction, and optimizes away the code that tests their conditions, such as the one for andps and orps. Initing them from different locations is enough to outsmart the compiler.
   
   (bug was only noticed in the x86 test, but I updated the other tests to guard against future improvements there too.)

   

   steven-johnson committed Feb 7, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for df0033e
   • Browse repository at this point

   Copy the full SHA

   df0033e View commit details

   Browse the repository at this point in the history