
processor issues
-----------------------------

  render.library  is  shipped  in  three versions: 68020, 68040, and
  68060.  if  your  system is supplied with a 68030 cpu, you can use
  both  the  68020 and the 68040 versions. the latter one requires a
  fpu, though.

  cpu type:             68020   68030   68030/68882    68040   68060
  ------------------------------------------------------------------
  render.library_060    -/-     -/-     -               +       ++
  render.library_040    -/-     -/-     ++             ++      +
  render.library_020    +      ++     ++              +       --

  -/-   won't work
  --    terrible combination
  -     bad combination
  +     good combination
  ++    very good combination
       slower but accuracy is better
       prior to v18.0, quantization
        did not work at all without a fpu


special annotations
-----------------------------

  > 68060
  there  was  a  lot  of work to be spent for the 060 version. hence
  this  version is most optimized in the render.library package. the
  code  benefits  from  large  caches, superscalarity, quick integer
  multiplication, even from branch prediction and other features.

  > 68040
  the  040  is  similar to the 060 version. also try the 060 version
  if your 68040 board has got poor memory transfer rates.

  > 68030
  there  was  no  need for a 030 version, since 68030 specificity is
  well  provided  with  both  the 020 and 040 versions. keep in mind
  that  the  040 version requires a fpu. the 68882 coprocessor might
  be  slower  than  appropriate  64  bit  integer   arithmetics, but
  it provides higher accuracy.

  > 68020
  the  68020  is  an  old-fashioned  cpu - no maths, poor caches, no
  special  features.  there  was  no particular need to optimize for
  this processor. the 68020 version is actually optimized for the MC
  68030.

  > 68000
  there never was and will never be a version for this cpu.

  > PPC
  not  yet  available. the re-portation of render.library to plain C
  code is in progress.
