Conference hydra::amiga_v1

>    
>    Now, SYSINFO says my harddisk driver is running in slow memory and I
>    should move it.  HOW?


What controller are you using?

john

    ESCORT E5HDC-S with St277N
    
    Antique City

>    ESCORT E5HDC-S with St277N


So you probably have a mountlist for it. There is a parameter named
BufMemType which you set to the type of ram you want the drivers to 
load into.


set it to 5 for FAST memory or 3 for CHIP.

john

    Thanks, John.  It was set to 1, whatever that means.

1 is the just low bit which means PUBLIC memory.


here are the values (in 1.3):

	PUBLIC:	1
	CHIP:	2
	FAST:	4

5 is just FAST+PUBLIC, 3 = CHIP+PUBLIC

If you set just PUBLIC, I think you get whatever is available. If you set
CHIP or FAST, you get that in particular. 

PUBLIC (I believe) is a bit that does nothing now but was meant as a possible
future compatibility feature for providing process private address space.

Workbench 2.0 has added some new attributes like 

	ANY
	LOCAL
	24BITDMA


Hmm ... PUBLIC? LOCAL?

Guess VIRTUAL hasn't made it yet :-(

john

Doesn't PUBLIC cause a precidence to be invoked?

i.e Satisfy request from fast if available otherwise use chip.

This way it should work on machines either with or without fast memory



????????????    Stan.

Re: .all

Well, the BufMemType field actually does what it suggests, it defines the
type of memory any BUFFERS for that device are to be taken from.
It won't influence where the driver is loaded, this is left to the 
discretion of the OS the driver itself.

You might want to experiment some memory allocater (Fish or TAPE::) that
eats all chip memory before invoking the driver (the first mount), but if
the driver requests to be loaded to chip memory, you're outta luck.
You don't have by any chance an A2000 with an 512KB Agnus and don't use
fastmemfirst?

Cheers,

<CB>

    re .6 - I think you are correct.  It will use the Fast memory first,
    left to its own devices.
    
    re .7 - You have confirmed what I have suspected: the value BUFMEMTYPE
    only affects the buffers, not the driver.  After testing the initial
    suggestion, the driver was still in slow memory.
    
    The computer in question is an A500 with 2 meg of memory, slenderest of
    Agnuses with an A501 internal memory.
    
    I'd be interested in seeing any software entity that would consume CHIP
    memory, forcing a driver into FAST.
    
    Tony

    re: type ANY
    
    that might be the hook for virtual memory, after all, you are telling
    it that you don't care what kind of memory it is loaded into.
    
    What does BufMemType=0 do?  I noticed my CLtd controller mountlist had
    that.
    
    Dave

    
>    What does BufMemType=0 do?  I noticed my CLtd controller mountlist had
>    that.

it just so happens that the value for ANY (MEMF_ANY) is 0. It looks like
the MEMF_ANY type was added for backwards compatibility since many mountlists
and AllocMem calls request a "type" of 0.

I think C= has always stated that all memory requests should set the PUBLIC
bit for "future compatibility". Since not everyone followed this programming
guideline, they had to add the MEMF_LOCAL attribute to explicitly ask for
non-PUBLIC memory - but this is just my speculation...

john

    I just had my A500 upgraded to WB1.3 and a fat AGNUS.  Interestingly,
    the HD driver moved from slow memory to fast memory without any other
    influence!  Now, isn't that special!?

    Addition of the fat AGNUS caused several things to go faster. 
    Handshake is one example.  It must be spending a lot of time initially
    cleaning up chip memory for its display.
    
    

>   <<< Note 4386.0 by GOBAMA::WILSONTL "Lead Trumpet (Read that...LEED!)" >>>
>                             -< JUMPDISK NOV '90 >-
>
>    I've just perused the JUMPDISK for Nov '90 (Yeah, I know it's late!) 
>    It contains a couple of very useful tools: CDT (Change default tools)
>    and SYSINFO.  
>    
>    CDT allows you to change the default tools on a number of icons at
>    once.  This is very useful to me because I keep all my online
>    documentation in one directory ("drawer" for those who refuse to give
>    up the Workbench analogy.)
    

Are the programs on Jumpdisk PD?   Can you upload the CDT tool so I don't
have to write one:-) ?

john

    Will attempt it tonight.  Is TAPE::user2:[upload] OK?
    

great! tape is fine...

john

Re: Tony

Since your problems went away with the 1 MB Agnus, did you or didn't you
have a "FASTMEMFIRST" statement in your startup-sequence and if so, where?

Cheers,

<CB>

Just thought that I would mention this, since it bit me in the *ss once --
setting BUFMEMTYP to 5 is wonderful as long as you have fast RAM available.
However, if you have to disable the fast RAM or remove the board for some
reason, this can severely impact your ability to use the disk.  In my case,
the Quantum drive on the hardframe was unusable until I put the A2058 back
inside the box.  If you are playing musical boards and things don't work
the way you expect them to, this could be the reason.

Dave

    re: .17 - Youbetcha!  I sold my Minimegs, leaving me only only chip
    memory (boy, I'm proud of Agnus), then suddenly my harddrive had no
    buffer space.  I reverted to the BUFMEMTYP of 1 and it worked.
    
    re .16 -  I'll have to check the location of FASTMEMFIRST when I get my
    new power supply.  Yes, after 3+ years with no problems, it decided to
    retire.   BTW, after my local store wanted $139.00 for a new one, I
    told them they should check to magazine prices and ordered a 100W
    version from Computability for $85.  So what, if I have to wait a week.
    
    Tony