On Fri, Sep 24, 1999 at 07:49:51PM +0100, David Laundon wrote: > PUSH HL takes 16 tstates (but only 24 during screen contention)
True. > and CALL nn > takes 24 tstates (but only 40 during screen contention). I'm not sure if > RSTs and interrupt calls are affected the same, but I expect so. I would assume so. > I think the end condition of DJNZ takes 12 tstates, not 8 as expected - > could this be it? Also, DJNZ is another instruction which performs > particularly well during screen contention; IIRC it takes just 16 tstates > for both looping *and* for the end condition. Also correct. :o) > maybe having a common routine for 'memory accesses' which adjusts to the > next 4-tstate or 8-tstate boundary *at that point* rather than trying to > work it all out when the instruction starts. > Mmm. You're right; a bit tricky! ;-) Actually this could be done and might be moderately straightforward... but it would be dog slow. imc