Interesting, on my centrino laptop the BS method is actually 2x faster than p2bound4 but on my P4 its slower??

Excluding the raw lookup table approach, the p2Bound4 algorithm is the fastest so far, by about 20%.. nice.

Interesting that you say this is the fastest algorithm, which it is not. While it may be in the general case when one uses it in a certain context knowing the upper and lower bounds then a lookup table will be the fatest method (3-4x faster than BS) :-)

i.e.

   static final int[] lookupTable;
   static {
      lookupTable = new int[MAX_LOOKUP];
      for(int i = 0; i < MAX_LOOKUP; i++)
         lookupTable = get2FoldUnrolled(i);
   }
   
   public final static int get2FoldLookup(int val) {
      return lookupTable[val];
   }
   
Your application will probably hit a certain input range 90% of the time, in which case you use the lookup table method. For all other input values you use the BS method.

Its always a space/time tradeoff.

hmmm

convertToRGB() should probably have the higher order bits all on anyway (ie. alpha 255, or 1.0) as does AlphaComposite.getAlpha() when no alpha is set.