!! NOTE - This is a stable release. If you want newer features that are
   listed down there or even things that aren't listed, try the development
   version, although that is unsupported !!

-----------------------------------------------------------------------------

Here's what the letters down there mean:
  F = Finished (or smaller optimisations left)
  O = Not well optimised, things that could be done, like read multiple
      pixels (every routine writes multiple pixels though!)
    = A blank means still missing (but covered by generic converters!!)
  - = Not planned at the moment (or not needed because it's completely covered
      by other converters)
  C = Handled by direct copy routines
  
Note that none of the really frequent conversions are missing in this
release, only rare conversion will have an M. Let me stress again that
conversion not covered will normally still be done by generalised
converters, just a bit slower.

Finally, this file only lists C routines, assembly routines are not listed
here.

NEW: There are TWO values for each function now. The left one is for normal
conversion, the right one for stretching.


CLEAR FUNCTIONS:

Generic32         F
Generic24         O
Generic16         F
Generic8          F


    
GENERIC ROUTINES (slow):

           From (normal|stretch)

To         Generic32  Generic24  Generic16

Generic32      F F        F          F  
Generic24      F          F          F  
Generic16      F O        F          F  
Generic8       F          F          F  
(03/01/1999: Those F's are all wrong, the generic routines could be much
 faster!!)


SPECIALISED ROUTINES (fast):
        
           From (normal|stretch)

To         muhmu32   32rgb888  24rgb888  16rgb565  Index8

32rgb888      F         C C    *  O       ** O F     F F                   
32bgr888      F         F F       O          O F     F F
32rgba888     F         F F       F          O F     F F
32bgra888     F         F F       O          O F     F F
24rgb888      O         F         C C        O       F
24bgr888      O         O         F          O       O
16rgb565      F O       F O       O          C C     F O
16bgr565      F O       F O       O          F O     F O
16rgb555      F O       F O       O          F O     F O
16bgr555      F O       F O       O          F O     F O
8rgb332       F O       F O       O          F O     - -
Index8                                               C C

* The 24 bit routines could all read 4 pixels at ones and thus dword reads
but they aren't yet.
** Converters from 16 to 32 bit should be optimised in that they should read
two pixels at a time, i.e. a dword