HS>DJGPP compiles different to MS C, and considering the fact the my message
HS>was/is/and always will be concerning DJGPP (not MS C, thats right 'NOT') I 
c
HS>only conclude to the fact that your results are nothing but invalid.
Puzzling.... Because the results you've been showing were not from
DJGPP.  And I've been using DJGPP and getting very very different
results than what you get....  Such as arrays being faster than your
pointer example.  If we WERE using the same compiler, then... we'd get
the same results etc. towards your low level optimizations.
And that table where you gave cycle counts....  That wasn't DJGPP.  That
was Watcom.
And all those discussions of *256 vs <<8 most definetly were not with
DJGPP, because you kept claiming your compiler didn't make that judgment
call optimization and that was why you were wanting to do it yourself.
I'm bowing out of your discussions. (I'm sure you'll be happy about
that.)  You are more than a little inconsistant and I'm not making
progress with you anyway.  You are going to end up getting hung from
your own gallows anyway, so it's not going to make much difference.
--- QScan/PCB v1.19b / 01-0162
---------------