[comp.sys.m68k] 68000 68010

ken@hpclkms.HP.COM (Kenneth Sumrall) (09/13/88)

/ hpclkms:comp.sys.m68k / Ian_Matthew_Smith@cup.portal.com / 11:52 pm  Sep 10, 1988 /

>   I have an Amiga computer with a 68000.  What I was wondering was what
>exactly are the diffrences between the 68000 and 68010?  I know that the
>010 dosen't have one instruction but I have a program to correct this. Would
>there be any real speedup in heavy number crunching?  Like ray-tracing or
>sound wave mixing?  Dosen't the 010 have better written math routines?
>   Thanks in advance for any information.
>
>Ian_Matthew_Smith@cup.portal

The 68010 isn't missing any instructions that the 68000 has, but their is
one instruction that used to be a user mode instruction that now is a super-
visor mode instruction.  The speedup that the 68010 would gives has been 
quoted as about 5%.  This is accomplished through a mode called
looping mode (or something similiar) in which small tight loops are held
entirely in a small on chip chache, and can be executed without loading 
opcodes from external memory.  This mode is most useful in operations
like memory block moves and such.

As for math routines, floating point math for the 68010 is done is software
or with a co-processor.  So I don't expect a 68010 would speed up math 
operations much.

Hope this information helps.

Kenneth Sumrall
ken%hpclkms@hplabs.hp.com
ken%hpclove@hplabs.hp.com

nerd@percival.UUCP (Michael Galassi) (09/16/88)

In article <300001@hpclkms.HP.COM> ken@hpclkms.HP.COM (Kenneth Sumrall) writes:

>The 68010 isn't missing any instructions that the 68000 has, but their is
>one instruction that used to be a user mode instruction that now is a super-
>visor mode instruction.  The speedup that the 68010 would gives has been 
>quoted as about 5%.  This is accomplished through a mode called
>looping mode (or something similiar) in which small tight loops are held
>entirely in a small on chip chache, and can be executed without loading 
>opcodes from external memory.  This mode is most useful in operations
>like memory block moves and such.

>Kenneth Sumrall

The loop mode tends not to be much of a win in compiler generated
code as it nearly never gets used other than in hand optimized
library routines, the win if ever comes from the faster mul[us],
div[us], and the presence of the rtd instruction in the '010.

This is contrasted by the fact that the most common exceptions are
slower in the '010.  Most common is what I have determined by profiling
the code I write at work, and inclues the av[1-7]  (+2 periods), the
trapv (+6 periods) illegal (+4 periods), your code may have different
ideas as to what is and is not "most common".

I have consistently gotten the same results out of the two parts
when I have benchmarked them on hand-written assembly though I realy
like the rtd instruction for general purpose code and have gotten
much mileage out of the bkpt instruction when I hacked together a
replacement for our AMS emulator.

These are of course the views of someone who works a lot with
assembly language and may not mean much to those who work in
high-level languages.
-michael
-- 
        Michael Galassi     | If my opinions happen to be the same as
...!tektronix!percival!nerd | my employer's it is ONLY a coincidence,
...!sun!nosun!percival!nerd | of course coincidences OFTEN DO happen.