From: dxforth@gmail.com   
      
   On 10/07/2025 8:10 am, Paul Rubin wrote:   
   > dxf  writes:   
   >> As for SSE2 it wouldn't exist if industry didn't consider   
   >> double-precision adequate.   
   >   
   > SSE2 is/was first and foremost a vectorizing extension, and it has been   
   > superseded quite a few times, indicating it was never all that   
   > adequate.  I don't know whether any of its successors support extended   
   > precision though.   
   >   
   > W. Kahan was a big believer in extended precision (that's why the 8087   
   > had it from the start).  I believes IEEE specifies both 80 bit and 128   
   > bit formats in addition to 64 bit.  The RISC-V spec includes encodings   
   > for 128 bit IEEE but I don't know if any RISC-V hardware actually   
   > implements it.  I think there are some IBM mainframe CPUs that have it.   
      
   I suspect IEEE simply standardized what had become common practice among   
   implementers.  By using 80 bits /internally/ Intel went a long way to   
   achieving IEEE's spec for double precision.   
      
   What little I know about SSE2 it's not as well thought out or organized   
   as Intel's original effort.  E.g. doing something as simple as changing   
   sign of an fp number is a pain when NANs are factored in.  With the x87,   
   Intel 'got it right the first time'.  Except for the stack size and   
   efforts to fix it.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)