Conference hydra::amiga_v1

    Quick answer: NO
    
    Longer answer: Not in this release, supposed to be in there in the
    future

Hopefully 1.4 will do *something* with it. 

AMIX (Amiga UNIX) requires it.

John

    
    	The PD program SetCPU can copy the AMIGA ROMS into 32bit memory
    and use the MMU to map it in place of the ROMS. This is only a
    simple use of the MMU but it does make everything faster.
    
    	Steve
    

    Dave Haynie, the designer of this board, spoke early this year at
    a users group meeting in Norwood, Mass.  The MMU was put on the
    A2500 for UNIX, period.  Haynie, not wanting to it to go to waste
    when running AmigaDOS, wrote SetCPU (see .-1).   Mapping the Kickstart
    ROM to 32-bit RAM produces some modest improvement -- I think he
    said 20-30% doing heavy duty graphics stuff.
    
    Haynie did not at the time believe that there would be virtual memory
    support in AmigaDOS any time soon -- that's why the MMU is there
    for UNIX -- but lately on Usenet he's been hinting that some astute
    people will find interesting ways to use the chip RSN.
    
    Wes

RE:.4

>   Haynie did not at the time believe that there would be virtual memory
>   support in AmigaDOS any time soon -- that's why the MMU is there
>   for UNIX -- but lately on Usenet he's been hinting that some astute
>   people will find interesting ways to use the chip RSN.

Wes,

I would say there was more than mere hinting... There is quite a
lively discussion going on concerning Amigados VM...

	-SES

Has anybody seen the AMIX running yet ??? Better yet is there anybody that
has the AMIX on their system ???

I would be interested in some feedback on how they like it, Does Amiga-DOS 
run as a task of AMIX what wappens to WB,  how about performance wise is 
AMIX a DOG or does it run fast.....

I am thinking to get more into UNIX/ULTRIX....

Thanks...

...gss...

    Re: .6
    
    Yeah, I've seen & used it several times. 
    No, I don't have it on my system, because I still don't have that
    80MB HD capacity to spare. 
    
    No, Amix and AmigaOS are mutually exclusive at this time. With 1.5
    (maybe 1.4) this may change.
    Amix is slick, especially it's windowing system. But compared to
    AmigaOS any U*IX is a DOG...
    
    Regards,
    <CB>
    

CB, Any idea how much AMIX will cost us anxious users?

John

    
    Re: .8
    
    Nah, at least not anything I could reveal to the public. :-)
    Really, it's to early to make such guesses, right now only selected
    commercial developers can purchase an A2500UX, but this at a rather
    reasonable price. 
    My guess is that AMIX itself will be rather cheap, but that CBM
    will force you for at least some time to get an A2500UX with it,
    and that unit is rather expensive...
    
    As soon as something more substantial than these (educated) guesses
    shows up, I'll let you know.     
    
    <CB>
    

Does anyone know if the exec at least saves the 68851 registers
during context switches?

John

From what I've heard, not until WB 1.4 (along with virtual memory 
support for AMIX etc).

<CB> could better answer this one...

John

    Re: .10, .11
    
    Nah, as long as my memory isn't failing completly (but with that
    Beaujoulais overload this weekend ya never know;-), this isn't the
    case (YET!). But I'll check that again tonite. If nothing shows
    up in this note, take it as my official word, whatever that's worth. ;-)
                                                                        
    He John, who made me DEC's Amiga VM specialist? ;-)
    This is one thing in this company I never stop wondering about:
    Just tell 'em you read/saw/heard about a product or (gasp!) the
    manual and they'll happily declare you a first grade specialist.
    ;-) ;-) ;-)
    
    Regards,
    
    <CB>

>    He John, who made me DEC's Amiga VM specialist? ;-)

But you are, aren't you <CB> ? :-):-)

John

Kickstart/exec for V1.3 do not save mmu or FFP registers.

It is possible to "wedge" a routine into the function call to do this.

Perhaps <CB> could explain the process better than I can recall it at the
moment.

	-SES

    
    Re: .14
    
    Why, oh why always me??? ;-D
    
    There's function called SetFunction() in the AmigaOS that allows
    system functions to be replaced by something else. The current
    implementation of this function leaves a lot to be desired and it's
    going to be rewritten for 1.4. A nice example for SetFunction() is
    Carolyn Scheppners SetRequesterText program. 
    I wouldn't modify/extent ANY Exec function, if not absolutly necessary.
    Don't fix something that isn't broken, they always say. :-)
                                                               
    Regards,
    
    <CB>

>    implementation of this function leaves a lot to be desired and it's
>    going to be rewritten for 1.4. A nice example for SetFunction() is

Any idea when this [1.4] will be available?

John

    Re: .16
    
    I answered this question several times before, but here I go again...
    
    1.4 will be available when it's finished. No time sooner.
    
    OR
    
    1.4? RSN! ASAP!  
    This computer lingo translates roughly into:
    It's still so far away that I have no (your 4letter word here) idea
    when it's going to be released.
    
    'Nuff said??
    
    <CB>

    Re: .10, .11... (again) 
    
    
    Well, here's the official word from Commodore, or at least the
    inofficial one from their chief HW hacker... ;-)
    
From  Dave Haynie:    
                  
    ...
    
Now, about them questions:

1. No, Exec doesn't store 68851 registers on context swaps.  That really
   wouldn't make sense, anyway -- the '851 isn't simply register driven
   like the '881/'882.  In an OS that supports the MMU, you might never
   change the register setup in operation.  What would probably happen
   would be that portions of MMU tables would be swapped in and out on a
   task change.  There's going to always be shared memory, even in a 
   perfect protected world, so there's no swapping that out.  And probably
   lots of system stuff would be the same read-only memory for every task.
   So you might end up with just part of a table being changed for each
   task, that part which maps the memory blocks the current task has access
   to.  However they do support MMU stuff, I hope they do it soon.  Then I
   won't have to anymore....

       ...
    
    Makes perfect sense to me.
    
    Regards,
    
    <CB>