Make wrapping_neg and overflowing_neg functions into const #57536

Activity

1. Centril

   added

   T-libs-api[DEPRECATED; DO NOT USE][DEPRECATED; DO NOT USE]

   on Jan 12, 2019
2. Centril commented on Jan 12, 2019

   

   Centril

   on Jan 12, 2019

   Contributor

   More actions

   cc @SimonSapin

   (I have no opinions on whether this is feasible to do branchless and w/o control flow, seems like something that needs to be perf tested to not regress non-CTFE...)
3. Centril

   added

   A-const-evalArea: Constant evaluation, covers all const contexts (static, const fn, ...)Area: Constant evaluation, covers all const contexts (static, const fn, ...)

   on Jan 12, 2019
4. Lokathor commented on Jan 13, 2019

   

   Lokathor

   on Jan 13, 2019

   ContributorAuthor

   More actions

   There appears to be the same performance once you go through optimization (with 1.31, and on x86_64, I didn't check every single target). The overflowing versions are slightly different, but the wrapping versions are the exact same.

   https://godbolt.org/z/IWJ2ev

   example::CORE_wrapping_neg_u8:
           neg     dil
           mov     eax, edi
           ret
   
   example::NEW_wrapping_neg_u8:
           neg     dil
           mov     eax, edi
           ret
   
   example::CORE_wrapping_neg_i8:
           neg     dil
           mov     eax, edi
           ret
   
   example::NEW_wrapping_neg_i8:
           neg     dil
           mov     eax, edi
           ret
   

   🎉React with 🎉2Reacted by Mazdak Farrokhzad and Paul Dicker
5. Centril commented on Mar 28, 2019

   

   Centril

   on Mar 28, 2019

   Contributor

   More actions

   This was done in #58044.
6. Centril

   closed this as completedon Mar 28, 2019