[comp.sys.amiga] Just how FAST is that FAST ram?

bryce@COGSCI.BERKELEY.EDU.UUCP (05/06/87)

In article <> grr@cbmvax.UUCP (George Robbins) said:
>> In article <> someone@something.somewhere.abstraction
>>
>> [need fast fast ram for animation. A500 $c0000 no will do]
>>
> This is a software problem (say's a hardware person).  There's a
> program to delete the C0 memory from the "FAST" chain, if that's
> enought to solve your problem.  In the long run, I'm afraid the
> "FAST" notion may need refinement, as there is no simple way the
> system software can be sure some external (or internal) memory
> is really fast.  All it can really say it isn't chip memory.

(Say's a software/hardware person):  Noooo Problem...


[my mind is at work on an elegant solution; this will do for now:]

Here's a semi-kludge:  Allocate some ram and pass the address to it.
The program will copy itself to that RAM and execute.  Determining chip
ram is trivial.  Since this runs at system configure time there is no
worry about multi-tasking; Disable() then profile execution speed.  Fire
up Mr. Blitter and profile again.  The truth comes out.  Memory lists
could be reordered, memory tagged as so:

...........|.....attributes........|
------------------------------------
type       |  PUBLIC | CHIP | FAST |
------------------------------------
CHIP RAM   |   yes   | yes  |  no  |
FAST FAST  |   yes   | no   |  yes |
HYBRID     |   yes   | no   |  no  |
------------------------------------

MOST memory allocation is of attibute 0, which will match all types of RAM
including HYBRID.

The rest is of type CHIP, which will match only CHIP ram.

Once in a blue animation program FAST ram will be requested; this is NEVER
          ^^^^^^^^^^^^^^^^^^^^^^                                     ^^^^^
done casually since it will always fail on a stock A1000.  This was done
with the system descibed earlier, and probably with one or two other
programs such as Caligari.  Hybrid RAM will not be mistakenly procured.

]
] As far as I can tell this is clean for everyone. From the bean-counters
] to the hackers to the system designers.
]
] The remaining problems:
]
] 1> execbase and gfxbase will end up in HYBRID unless the $c0000 ram is
]    disabled/moved and FAST ram put in it's place.
]             ^^^^^
] 2> HYBRID ram will be at that start of the memory list.  Memory allocation
]    should go FAST, HYBRID, CHIP.
]
] Now a startup-sequence program could fix the order then make a copy of
] execbase/gfxbase and point the system to them.
] Better handled in ROM, but *sigh* that is too much to hope for.
]
] --->!!!??? Anything wrong with this ???!!!<---
]

# IMPORTANT STUFF OFF

>>[How about a] Tiny jumper on PCB to place +5 volts on the parallel port.
>There's a resistor there that you can replace with a jumper if
>you feel compelled to do so.  Don't try to plug in a standard
>IBM printer cable/printer if you do, because there is *NO* +5
>at all on the IBM PC standard printer connector.

These things must be goof-proof; that's why it's a jumper.  Or in this
case a resistor... what is the value in ohms, and does it go from +5 to
pin 23 with no other stops?

> Floppies are ok for program interchange, but [...]

exactly.