From: inderaj@gmail.com   
      
   It depends on the programming language you are using If you are using   
   a language like cuda, the parallelism is expressed well and you need   
   no rocket science in compilation but basic good codegen   
      
   Reducing operation counts is important on GPU's as well So most of the   
   same high level optimizations are required   
      
   One biggest difference is low or no cache, due to which you need to be   
   especially mindful of not creating too many extra lifetimes.In that   
   sense is tradeoff point of computation vs lifetime is different from   
   regular processors   
      
   Inderaj   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)