Right DMA-Mask value for HD RDB?
Lars-Erik Holmquist
installed in my GVP HD8 controller. I think the problems
have something to do with the settings in the Rigid Disk
Block. I've lowered the max transfer rate, but that didn't
change anything.
Things got better when I changed the value for the DMA
Mask to the default given by RD Prep. But all games
still refuse to run! (They run OK from disk)
I would greatly appreciate if someone could tell me how
to calculate the right DMA Mask value, or anything else
that I might have missed. I haven't found anything
appropriate in the AmigaDos manual or Hardware Reference.
--
- Lars Erik Holmquist
----------------------------------------------------------------------
"We are so far ahead, we are in danger of lapping ourselves"
- Alan McGee, Creation Records, 1986
----------------------------------------------------------------------
The Hong Kong movies WWW-page: http://www.mdstud.chalmers.se/hkmovie/
Jim Hardy
>I'm having trouble with a second hand SCSI HD that I've
>installed in my GVP HD8 controller. I think the problems
>have something to do with the settings in the Rigid Disk
>Block. I've lowered the max transfer rate, but that didn't
>change anything.
>Things got better when I changed the value for the DMA
>Mask to the default given by RD Prep. But all games
>still refuse to run! (They run OK from disk)
>I would greatly appreciate if someone could tell me how
>to calculate the right DMA Mask value, or anything else
>that I might have missed. I haven't found anything
>appropriate in the AmigaDos manual or Hardware Reference.
Well in the my 3.0 manual it says this:
Mask= Displays hex number that defines areas of memory that can be
used with dma.
For example: A2091 and A590 0xFFFFFE, A3000 0x7FFFFFFC, and A4091T and
A4000T 0x7FFFFFFF.
It also says if you have a third party controller check their manual.
You could try the A2091 settings...
jdow on BIX
For that puppy the RDPrep default value is the best you can do. Maxtransfer
may want to be 65535 or 131071 or something like that. These values are
determined by the bus in use and the card's design. If the mentioned values
of maxtransfer do not work reduce it to a grinding slow 512 and see what
happens. Work up in powers of two until it gets strange. The higher the
number the faster the potential for the drive - until you hit numbers the
controller does not like.
(Unless the controller is REAL strange it wants a mask of $fffffe or less
likely $ffffff for most situations. If you have something dippy like a
Hurricaine accelerator which will not DMA into its 32 bit ram you may have
to set the mask as low as $1fffff. If you have only 3megs of DMA able
memory plus other memory then use $3fffff and pray the DMAable stuff is
configured there rather than AFTER the non-dma memory. If not you're stuck
doing EVERYTHING in chip ram.)
Now - for a GOOD hint.... REad off the values from your working drive and
use them as guides for setting the other drive.
{^_^} Joanne Dow, Editor Amiga Exchange, BIX
jd...@bix.com
Lars-Erik Holmquist
>Now - for a GOOD hint.... REad off the values from your working drive and
>use them as guides for setting the other drive.
Yeah, I would that - if I _had_ a working HD! :-) The whole
thing started when the old one blew up...
Anyway, thanks for the help - things seem to get a little
bit better for each thing I try!
Mr Philip Orr
>I'm having trouble with a second hand SCSI HD that I've
>installed in my GVP HD8 controller.
>I would greatly appreciate if someone could tell me how
>to calculate the right DMA Mask value, or anything else
>that I might have missed. I haven't found anything
>appropriate in the AmigaDos manual or Hardware Reference.
The DMA mask is a feature to allow the program FastPrep to control
32-bit RAM access by GVP's accellerators. This is what ultimately allows the
Combo boards and other 680(2,3,4)0 accellerators to DMA to RAM outside the
available address space of the 68000. Unless you are using 32-bit RAM
connected directly to the processor (ie a GVP Combo board), the default
setting should be used. (0FFFFFF I think).
Lars-Erik Holmquist
>The DMA mask is a feature to allow the program FastPrep to control
>32-bit RAM access by GVP's accellerators. This is what ultimately allows the
>Combo boards and other 680(2,3,4)0 accellerators to DMA to RAM outside the
>available address space of the 68000. Unless you are using 32-bit RAM
>connected directly to the processor (ie a GVP Combo board), the default
>setting should be used. (0FFFFFF I think).
I have tried all the settings suggested by various people.
It turn out that if I let the mask go outside chip ram (I
have two megs of DMA fast ram) the mask has to be long-
word aligned (that is, end with C) or I get lots of read
errors. There is no logical reason for this, and the manual
states quite clearly that the mask only has to be word
aligned (end with E).
If I only allow chip ram DMA, I don't get as many read errors
when I set the mask word aligned, but I still get errors.
Changing the max transfer doesn't change anything.
Could there be some special requirements for the "Faaast" ROM
that the GVP uses? I have a 1993 ROM, and I used RD Prep
to write the Rigid Disk Block. Perhaps GVP uses some
special parameters that RD Prep can't get at? Does anyone
know?
Jeff Walkup
: For example: A2091 and A590 0xFFFFFE, A3000 0x7FFFFFFC, and A4091T and
: A4000T 0x7FFFFFFF.
That's strange... I have an A3000, and when I put in a new drive
(couple days ago), HDToolBox set the default mask at 0x7FFFFFFE.
Well, I only got about 500k/sec, which is too slow since I've
seen xfer rates in excess of 850k/sec with my other drives. So
I changed the mask to 0xFFFFFFFC, which is what my other drives
are set to, and now get ~880k/sec. I got that 0xFFFFFFFC value
from someone on the net last time I asked.
Now, I've also got an old 2000HD w/A2091 SCSI board, and was getting
about 500k/sec, which is pretty good for the 68000 cpu, I guess.
I just put in a used GVP 030/50mhz and a new Maxtor 7590S, and
now I can only get ~120k/sec, which is MISERABLE. I've tried
a bunch of different mask values but none of them improve the
xfer rates.
The GVP sets up 4mb of it's RAM in autoconfig space, the other
20mb is somewhere else... Can I set the mask to that 4mb of
superfast GVP memory?? How? Even if I can't use the GVP's
memory, should it really be that slow using chipmem or the
2mb fastmem on the 2091??
The 2091 has v6.6 ROM.
--
Jeff Walkup . jwa...@sfsu.edu . San Francisco . 415.668.7312
Lars-Erik Holmquist
>Jim Hardy writes:
>: For example: A2091 and A590 0xFFFFFE, A3000 0x7FFFFFFC, and A4091T and
>: A4000T 0x7FFFFFFF.
>That's strange... I have an A3000, and when I put in a new drive
>(couple days ago), HDToolBox set the default mask at 0x7FFFFFFE.
>Well, I only got about 500k/sec, which is too slow since I've
>seen xfer rates in excess of 850k/sec with my other drives. So
>I changed the mask to 0xFFFFFFFC, which is what my other drives
>are set to, and now get ~880k/sec. I got that 0xFFFFFFFC value
>from someone on the net last time I asked.
"Curioser and curioser", as they say!
I've solved _my_ problem in an odd way now:
Since things got better when I changed from 0xFFFFFFF to 0xFFFFFFE,
and even better when I changed to 0xFFFFFFC, I changed to 0xFFFFF00.
And presto! Almost everything worked! So I went one better, and
changed to 0xFFFFE00 (which should mean that DMA uses only 512 byte
blocks) - and now all my read errors are gone.
The oddest thing is that all through my changing of max transfers
and DMA masks, the HD speed has stayed solid at around 750 K/sec.
I have no idea why my solution works, so if anyone can explain
I'd be very happy. :-)
Jan Holler
: Since things got better when I changed from 0xFFFFFFF to 0xFFFFFFE,
: and even better when I changed to 0xFFFFFFC, I changed to 0xFFFFF00.
Really 0xFFFFF00 ?? Or is it: 0x00FFFFF ! (The latter is correct).
-jan
Lars-Erik Holmquist
>Really 0xFFFFF00 ?? Or is it: 0x00FFFFF ! (The latter is correct).
Yes indeed - 0x00FFFFF may be correct but it DOES NOT WORK!!!
I get corrupted icons and random read errors with that one.
As I said, I have no idea why the value I'm currently using works,
and I guess it's just random luck. My HD setup is very mixed
(1993 ROM, unknown year board, very old Quantum HD).
--
- Lars Erik Holmquist
----------------------------------------------------------------------
Please note:
The Hong Kong movies WWW-page takes a VACATION from June 27th
and about four weeks onward. It will not be reachable during
that time.
John P Sheehy
> Since things got better when I changed from 0xFFFFFFF to 0xFFFFFFE,
> and even better when I changed to 0xFFFFFFC, I changed to 0xFFFFF00.
I have WB 2.1 and the version of HDToolBox that came with it won't allow
a mask ending in 'F'. Did a newer version of HDToolbox allow this, or
was this set with another program?
ohn <j...@cup.portal.com>
Lars-Erik Holmquist
>I have WB 2.1 and the version of HDToolBox that came with it won't allow
>a mask ending in 'F'. Did a newer version of HDToolbox allow this, or
>was this set with another program?
I did it with RDPrep, that I got from Aminet.
But as far as I understand things, you wouldn't want
a mask ending with "F" anyway - it won't work very well
with most controllers. But there's no harm trying of
course; you'll know when you've got a bad value,
believe me! :-)
--
- Lars Erik Holmquist
----------------------------------------------------------------------
PLE...@orplid.sh.sub.de
There are Problems with NEC3d using the CBM-Genlock...
When I plued in the CBM-Genlock the NEC doesn`t `Multisync`...
it shows a black screen..
..happens with A3000`s std 23pin connector/15pin VGA/Genlock23pin
I reworked the genlock and now there`s a picture but the genlock
should be a gen and not an FBAS-Modulator....I want to view
TV on the NEC....
Whats different with AGNUS`s Hsync/Vsync???
Thanks in advance / greetings from Germany!
+--------------------------------------------------------------+
| Jörn Plewka | // - Amiga 3000 /030/6/52 |
| PLE...@ORPLID.sh.sub.de | _ // - Amiga 3000T/030/6/520 |
|---------------------------+ \\ // - Amiga 4000/040/10/130 |
| "Amiga, the better choice!" \X/ - Amiga CDTV 2.04-A560 |
+--------------------------------------------------------------+
| Arcnet, Zyx, Tiga, Gen, HP4L, Syquest270, Envoy, UnixS5R4V2u |
+--------------------------------------------------------------+
Neil Richmond
point me in the direction of where I could get this info. I had some problems
a while back with this and I want to know what works, how it works and why.
Thanks.
neil
--
" Give a skeptic an inch and he'll measure it. "
Neil F. Richmond ne...@rhythm.com
Rhythm & Hues Inc.
Bjoern Lindgren
>--
Please tell me too.
--
* Björn Lindgren | - Hey Amiga, the Sun is rising ! *
* ^o with dots above | She thoughtfully took a byte in her Apple. *
* Be...@Solace.mh.se | It had allmost gone a DECade since she heard that *
* A500 T | the first time.... *
John P Sheehy
>Couls someone email me info on the dma masks and their meanings or a least
>point me in the direction of where I could get this info. I had some problems
>a while back with this and I want to know what works, how it works and why.
I've never read any official documentation on this, but from my
experimentation it seems that the mask is a value that determines which memory
addresses can receive DMA from the HD. This is determined by logically
"AND"ing the address with the Mask. For example, if a program has a buffer
at address 0x07ABFFFE, and the mask is 0xFFFFFFFE it will get it, because
0x07ABFFFE AND 0xFFFFFFFE = 0x07ABFFFE. If the mask were 0x00FFFFFE, it
would not get the direct DMA since 0x07ABFFFE AND 0x00FFFFFE = 0x00ABFFFE
rather than 0x07ABFFFE. I think what happens in that case is that AmigaDOS
allocates it's own buffer somewhere allowable by the mask and then copies
the data to the program's buffer. This is probably what's happening when
you hear of people getting floppy speeds on their HardDrives.
If you have a zorroII controller which allows only 24 bit DMA, your mask
should be no higher than 0x00FFFFFF (0x00FFFFFF = 2 to the 24th power,
minus one).
Various HD controllers suggest differing mask values and they often result
in slow performance.
If you have an A3000 or A4000, a mask of 0x7FFFFFFC seems to be the best
performance/reliability compromise. C= recommends 0x7FFFFFFE, but many
people have had problems with the final 'E', myself included. The 'C'
ending guarantees that DMA will only occur on longword aligned boundaries,
which can prevent data corruption that can occur on such things as un-DMSing
or loading IFFs into ImageFX or Viewtek, if the mask ended in 'E'.
Changing my mask to end in 'C' has solved many problems for me.
>Thanks.
>neil
>
>--
>
>" Give a skeptic an inch and he'll measure it. "
>Neil F. Richmond ne...@rhythm.com
>Rhythm & Hues Inc.
John <j...@cup.portal.com>