This is a note how to port gnushogi to machines with scarce memory:
gnushogi minimal requirements are:
 - approximately 200 kByte memory for the executable program.
 - at least 300 kByte for data structures.
You dont want to port gnushogi to a machine with less memory than that.

gnushogi is optmized for speed and that means that memory has been used
when there has been a tradeoff between memory usage and speed. If you intend
to run gnushogi on a machine with less than 4 Mbyte memory the size of some
data structures have to be reduced. Here is a list of the largest data
structures in gnushogi, their sizes and a small comment on what can
be done to reduce their size:

ttable:		1.5	MByte	(#define vttblsz <something small>)
etab:		1.4	MByte	(can be removed)
Tree:		42 	kByte	(change f,t to unsigned char)
history:	131	kByte   (can be removed, not recommended)
nextpos:	98	kByte	(nothing save rewiting all move generation)
nextdir:	98	kByte	(nothing save rewiting all move generation)
distdata:	6	kByte	(can be changed to a macro)
ptype_distdata:	98	kByte	(can be changed to a procedure)
hashcode:	26	kByte   (no reduction possible)

First of all, start by reducing the transposition table size, this
is done by setting vttblsz in (gnushogi.h). If the transopsition table
does not fit entierly in memory it will have a detrimental effect on
performance. You can remove the transposition table by defining 
compiler option NOTTABLE.
You can remove the static evaluation cache by omitting the compiler
option CACHE.
You can remove the history table by omitting the compiler option HISTORY
(NOT recommended).
If this isn`nt enough, reconsider if you really want to do this port.
There is`nt really that much to gain by changing the other
data structures. 

There are some switches in order to enable space reduction:

#define SAVE_PTYPE_DISTDATA	(replace ptype_distdata by a procedure)
#define SAVE_DISTDATA		(replace distdata by a macro)