Recent changes to mpn_get_str/mpn_set_str

[Torbjörn Granlund]

  > I'll see if I can brush up the code and make use of it for machines with
  > pipelined multiply.
  
  If it extracts digits limb by limb, then progressively shorten the buffer
  used by the mul_1 phase, I believe it will reach the performances claimed
  by the two authors.
  
  Unfortunately I do not have access neither to your old code nor to the
  code of the two authors...

My old code lives in ~tege/GMP/setget_str on the main GMP dev machine.

I apprarently started this project around 2002 but the file inaptly
named newest-mpn_get_str.c is from 2007.  It links to the limb_out.c
file.  Both files are shock full of nested #if for testing various
pieces of code.  Not pretty, I think one might be better of starting
from scratch, even if looking at the old code for ideas is probably
wise.

  and I won't write a third implementation from scratch... not
  immediately at least :-)
  
How about now?  :-)

-- 
Torbjörn
Please encrypt, key id 0xC8601622