FS> BM> second of all, why aren't you doing these as REP MOVSW instructions?
FS>It's slower than moving the data "manually".
think so?
let me compare here:
; "manual" data movement, basically equivalent to REP MOVSW
; assume cx > 1, since repetition is pointless otherwise.
; clock cycles               486 386 286 86 (from TASM 3.2 qref)
@@1:    MOV     AX, [SI]    ;  1   4   5 13
        MOV     ES:[DI], AX ;  1   2   3 16
        ADD     SI, 2       ;  1   2   3  4
        ADD     DI, 2       ;  1   2   3  4
        LOOP    @@1         ;  2  13   8 17 each time through the loop
                            ;--------------
; totals per iteration:     ;  6  23  22 54
                            ;  6  12   4  5 at the end of the loop
                            ;     (386 clock count assumes NOP follows)
; effective clocks (w/ cx)   486      386       286      86
(manual rep movsw)           6+6*CX   12+23*CX  4+22*CX  5+54*CX
; rep movsw version
; clock cycles               486      386       286      86
        REP     MOVSW       ;12+3*CX  5+4*CX    5+4*CX   9+18*CX
by my calculations, rep movsw is twice as fast on a 486, almost 6 times as
fast on a 386 or 286, and fully 6 times as fast on an 8086.  If you can make 

more efficient move routine, I'd love to see it.
FS>But I'm also pretty sure that's not the problem :). It seems as if Nick 
as
FS>previously programmed on an Amiga or a similar machine where
memory-to-memory
FS>operations are supported. That's not the case on the PC. You _need_ to 
ove
FS>the data through a register. Yes, the 80x86 has got to be the most 
worthless
FS>piece of silicone the world has ever seen :-\.
Actually, it DOES support memory-to-memory, but only with rep movsw/movsb
(movsd on 386+).
((Cloud))
 * OLX 2.2 * Time flies when you don't know what you're doing.
--- PCBoard (R) v15.3/M 10
---------------