From: user5857@newsgrouper.org.invalid   
      
   BGB  posted:   
      
   > On 10/17/2025 1:48 AM, Lawrence D’Oliveiro wrote:   
   > > On Thu, 16 Oct 2025 22:19:21 GMT, MitchAlsup wrote:   
   > >   
   > >> But what gets me is the continual disconnect from actual vector   
   > >> calculations in source code--causing the compilers to have to solve many   
   > >> memory aliasing issues to use the vector ISA.   
   > >   
   > > Is this why C99 (and later) has the “restrict” qualifier   
   > > ?   
   >   
   > Ironically, this is also partly why I suspect if a C-like language could   
   > have a "T[]" type that was distinct from "T*" could be useful, even if   
   > they were the same representation internally (a bare memory pointer):   
   > "T[]" could be safely assumed to never alias   
      
   Restrict does nothing to make:: (from a few days ago)   
   {   
   I might note My 66000 vectorizes loops not instructions to avoid   
   this problem; For example::   
      
        for( i = 0; i < max; i++ )   
        {   
             temp     = a[i];   
             a[i]     = a[max-i];   
             a[max-i] = temp;   
        }   
      
   }   
      
   Runs fast and get the right answer. When i !~= max, the loop   
   runs at vector speeds, when i ~= (max-i) it runs slow to get   
   the right answer. Where ~= is within a cache line.   
      
   > At least in theory "restrict" works, when people use it   
      
   under the specification restrict has   
      
   >                                                         .   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)