XPost: comp.parallel, comp.sys.super   
   From: terje.mathisen@hda.hydro.com   
      
   Ted Dunning wrote:   
   > Sorry to rain on your parade.  It is really exciting to have ideas   
   > like this.  The trick is to savor the euphoric moments, even with ones   
   > that turn out not as good as they seem at first and avoid being too   
   > depressed by the downturn that you experience when they turn out not   
   > so good.  The only way to have good ideas is to have lots of bad ones   
   > as well (and then prune efficiently).   
      
   Hear, hear!   
      
   This is _very_ true: Nearly all the really good programmers I've met   
   love trying all sorts of stuff, then testing/measuring the actual   
   results they get.   
      
   John Carmack (of id Software/Doom/Quake fame) have done an awful lot of   
   this.   
      
   Personally I had a great idea for HD-DVD decoding a few months ago; I   
   realized that I could decode up to 22-bit long tokens using a single   
   reciprocal multiplication which emulated a binary division loop.   
      
   When I managed to implement it, it turned out that at least on my sample   
   videos, the average length of such tokens were just one or two bits, so   
   the bit-serial loop was faster than my beautiful branchless code. :-(   
      
   Terje   
      
   --   
   -    
   "almost all programming can be viewed as an exercise in caching"   
      
   --   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)