From: no.email@nospam.invalid   
      
   anton@mips.complang.tuwien.ac.at (Anton Ertl) writes:   
   > So just use the same implementations of transcentental functions, and   
   > your results will be bit-identical   
      
   Same implementations = same FP operations in the exact same order?  That   
   seems hard to ensure, if the functions are implemented in a language   
   that leaves anything up to a compiler.   
      
   Also, in the early implementations x87, 68881, NS320something(?),   
   transcententals were included in the coprocessor and the workings   
   weren't visible.  There is a proposal to add this to RISC-V   
   (https://libre-soc.org/ztrans_proposal/).  It looks like there was an   
   AVX-512 ER subset that also does transcententals, but it only appeared   
   on some Xeon Phi processors now discontinued (per Wikipedia article on   
   AVX).  No idea about other processors.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)