🚀 go-pugleaf

Here is the current status on GIPPP number 2.
 
After my request for comments last week, I received the suggestion that I
use "double" instead of "float" to ensure accuracy.  After hearing this
suggestion I decided to investigate the error resulting from using "float"
and found that it was in fact quite significant.  All of these results are
from a SparcStation:
  width height max_iter  pix_in_set  area       mantissa  type
   8192   4096   524288     8987154  1.50658906       25  float
   8192   4096   524288     8987086  1.50657769       54  double
 
The "mantissa" column shows the number of mantissa bits, and "type" shows
what data type was used in the computation.  I am currently generating
results for intermediate precisions, using a truncation formula to simulate
lesser precision.
 
As you can see, the difference between 25-bit "float" and 54-bit "double" is
about 0.000011 in the M-set area.
 
I am testing my program now.  It seems to be working well, but over the
weekend I thought of a new feature that I wanted to add: taking counts for
several different dwell limits at the same time.  
-- 
mrob's .sig #2 | \/                     RF(1/2(1/3B1)B1) -> _     .$^``   ^`7"
collect all 5! |      I name pieces of 	                    'm---,/          T[