Inefficient codegen with ternary and additive inverse #135647

Activity

1. dotnet-policy-service

   added

   untriagedNew issue has not been triaged by the area ownerNew issue has not been triaged by the area owner

   on Oct 12, 2026
2. github-actions

   added

   area-CodeGen-coreclrCLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMICLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI

   on Oct 12, 2026
3. dotnet-policy-service commented on Oct 12, 2026

   

   dotnet-policy-servicebot

   on Oct 12, 2026 – with Dotnet Policy Service

   Contributor

   More actions

   Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
   See info in area-owners.md if you want to be subscribed.
4. hamarb123 commented on Oct 13, 2026

   

   hamarb123

   on Oct 13, 2026

   Last edited by hamarb123

   Contributor

   More actions

   I don't see how that is valid codegen for this IL btw - bool can be any value 0-255, and that codegen is only correct for 0-1 - 2-255 bool is UB in C# but not in ECMA-335 - you would have to write C# like num - 1 + (r ? 1 : 0) and change roslyn to get it to emit IL more like:

   ldarg.0
   ldc.i4.1
   sub
   ldarg.1
   add
   ret

   (that is, extend the bool ? 1 : 0 -> (int)bool optimization to any bool, not just for direct IL comparison sequences like x >= 0 ? 1 : 0).

   As you can see, it works as expected if I insert the cast using Unsafe.BitCast instead: https://godbolt.org/z/78PMe5cPd

   Program:M1(int,bool):int (FullOpts):
          movzx    rax, sil
          lea      eax, [rdi+rax-0x01]
          ret      
5. hamarb123 commented on Oct 13, 2026

   

   hamarb123

   on Oct 13, 2026

   Contributor

   More actions

   https://github.com/dotnet/roslyn/blob/main/docs/compilers/Boolean%20Representation.md