[N8VEM: 19170] SCSI2IDE with Atari Falcon

[Martin Lukasek]

Hello,

did anyone try SCSI2IDE with Atari Falcon? I have just build the new 2.0 version and have quite strange problem. The SCSI2IDE works with debug firmware, just slow. When I try normal (non debug) firmware, Atari freezes during SCSI initialization. I will try another computer and SCSI controller later, but it will take some time and I am a little bit curious about this.

Kind regards

Martin

[Martin Lukasek]

One more thing. I found that the Atari Falcon boot interrupt during SCSI 0 detection in driver is caused by Pseudo DMA mode ON in NON DEBUG firmware. If I build firmware with debug or Polled I/O, everything works ok, just very slow, because of debug messages. Any suggestions?

Another thing is performance. Does anyone did some measurements, how fast the interface should be, so I could compere my results?

Happy new year!

Martin



From: ***@googlegroups.com [mailto:***@googlegroups.com] On Behalf Of Martin Lukasek
Sent: Wednesday, December 31, 2014 8:45 PM
To: ***@googlegroups.com
Subject: [N8VEM: 19170] SCSI2IDE with Atari Falcon

Hello,

did anyone try SCSI2IDE with Atari Falcon? I have just build the new 2.0 version and have quite strange problem. The SCSI2IDE works with debug firmware, just slow. When I try normal (non debug) firmware, Atari freezes during SCSI initialization. I will try another computer and SCSI controller later, but it will take some time and I am a little bit curious about this.

Kind regards

Martin

--
You received this message because you are subscribed to the Google Groups "N8VEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to n8vem+***@googlegroups.com<mailto:n8vem+***@googlegroups.com>.
To post to this group, send email to ***@googlegroups.com<mailto:***@googlegroups.com>.
Visit this group at http://groups.google.com/group/n8vem.
For more options, visit https://groups.google.com/d/optout.

[Wayne Warthen]

Hi Martin,

Well, this is the second report I have gotten of initialization issues when
using the non-debug build of the SCSI2IDE firmware. I have reviewed the
code repeatedly and other than slowing things down, I see no difference in
the debug version. I am perplexed about what causes this and am unable to
replicate it on any of my reference platforms.

I assume that in non-debug mode, something is going too fast for some
controllers. I just don't know what it is.

If you are willing to play with the code, you could comment out the debug
output calls one-by-one to see if you can isolate where there may be a
speed issue.

Regarding speed, I don't know how to provide a metric for this. I can tell
you that when used on a 12MHz 8088 PC, it seems to run as fast as a real
SCSI hard disk meaning that speed is mostly gated by the computer, not the
SCSI2IDE interface.

Thanks,

Wayne

Post by Martin Lukasek
One more thing. I found that the Atari Falcon boot interrupt during SCSI
0 detection in driver is caused by Pseudo DMA mode ON in NON DEBUG
firmware. If I build firmware with debug or Polled I/O, everything works
ok, just very slow, because of debug messages. Any suggestions?
Another thing is performance. Does anyone did some measurements, how fast
the interface should be, so I could compere my results?
Happy new year!
Martin
*Sent:* Wednesday, December 31, 2014 8:45 PM
*Subject:* [N8VEM: 19170] SCSI2IDE with Atari Falcon
Hello,
did anyone try SCSI2IDE with Atari Falcon? I have just build the new 2.0
version and have quite strange problem. The SCSI2IDE works with debug
firmware, just slow. When I try normal (non debug) firmware, Atari freezes
during SCSI initialization. I will try another computer and SCSI controller
later, but it will take some time and I am a little bit curious about this.
Kind regards
Martin
--
You received this message because you are subscribed to the Google Groups "N8VEM" group.
To unsubscribe from this group and stop receiving emails from it, send an
Visit this group at http://groups.google.com/group/n8vem.
For more options, visit https://groups.google.com/d/optout.

[Martin Lukasek]

Hi Wayne,

thanks, it looks like some timing issue. I modified the code slightly, so I found 3 places that have to have some serial output to make Atari Falcon boot. Even I used CALL to character output in following construction:

#IF (S2I_TRACE >= 2)
...
#ELSE ;Martin - change for Atari Falcon compatibility
CALL PC_LT
#ENDIF

The full modified source is in attachment, as well as Atari boot screens and serial console boot output logs.

Even after the Falcon boots, the driver gets stuck when try to copy data to the SCSI. I tried to add simple character output to all S2I_TRACE >= 2 places, but it helped just partially, the behavior was not reliable. Strange is that when using debug with full message output, everything works ok, just a slow. I got speed about 20 kB/s.

I use oscillator 14 MHz as this was the only one higher than 10 MHz, I had at home and the SCSI2IDE board itself seems to work ok.

Kind regards

Martin


From: ***@googlegroups.com [mailto:***@googlegroups.com] On Behalf Of Wayne Warthen
Sent: Thursday, January 01, 2015 9:52 AM
To: ***@googlegroups.com
Subject: [N8VEM: 19176] Re: SCSI2IDE with Atari Falcon

Hi Martin,

Well, this is the second report I have gotten of initialization issues when using the non-debug build of the SCSI2IDE firmware. I have reviewed the code repeatedly and other than slowing things down, I see no difference in the debug version. I am perplexed about what causes this and am unable to replicate it on any of my reference platforms.

I assume that in non-debug mode, something is going too fast for some controllers. I just don't know what it is.

If you are willing to play with the code, you could comment out the debug output calls one-by-one to see if you can isolate where there may be a speed issue.

Regarding speed, I don't know how to provide a metric for this. I can tell you that when used on a 12MHz 8088 PC, it seems to run as fast as a real SCSI hard disk meaning that speed is mostly gated by the computer, not the SCSI2IDE interface.

Thanks,

Wayne

On Wednesday, December 31, 2014 3:54:20 PM UTC-8, MartinL wrote:
One more thing. I found that the Atari Falcon boot interrupt during SCSI 0 detection in driver is caused by Pseudo DMA mode ON in NON DEBUG firmware. If I build firmware with debug or Polled I/O, everything works ok, just very slow, because of debug messages. Any suggestions?

Another thing is performance. Does anyone did some measurements, how fast the interface should be, so I could compere my results?

Happy new year!

Martin



From: ***@googlegroups.com<javascript:> [mailto:***@googlegroups.com<javascript:>] On Behalf Of Martin Lukasek
Sent: Wednesday, December 31, 2014 8:45 PM
To: ***@googlegroups.com<javascript:>
Subject: [N8VEM: 19170] SCSI2IDE with Atari Falcon

Hello,

did anyone try SCSI2IDE with Atari Falcon? I have just build the new 2.0 version and have quite strange problem. The SCSI2IDE works with debug firmware, just slow. When I try normal (non debug) firmware, Atari freezes during SCSI initialization. I will try another computer and SCSI controller later, but it will take some time and I am a little bit curious about this.

Kind regards

Martin

--
You received this message because you are subscribed to the Google Groups "N8VEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to n8vem+***@googlegroups.com<javascript:>.
To post to this group, send email to ***@googlegroups.com<javascript:>.
Visit this group at http://groups.google.com/group/n8vem.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups "N8VEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to n8vem+***@googlegroups.com<mailto:n8vem+***@googlegroups.com>.
To post to this group, send email to ***@googlegroups.com<mailto:***@googlegroups.com>.
Visit this group at http://groups.google.com/group/n8vem.
For more options, visit https://groups.google.com/d/optout.

[Wayne Warthen]

Thanks Martin. This is MUCH more info to work from than I have had in the past. I will review all of this and see if I can come up with something.

Wayne

[Wayne Warthen]

Hi Martin,

Spent most of the day on this. Your testing and captures definitely help
pin down the nature of the issue. While I am still not seeing a code
issue, it seems pretty clear that the code is getting stuck in one of the
loops that wait for DMA REQ. So, I refactored the code to include a simple
timeout in those loops. If the timeout occurs, it will dump the controller
register values to the console, free the bus, and restart the protocol.

If you could try the latest firmware, I would appreciate it. I have
attached the zipfile to this post.

Regarding performance... I finally found a copy of Check IT which has a
hard disk performance benchmark. It indicates that the SCSI2IDE is
achieving 83K bytes per second which is almost identical to the performance
of a true legacy XT hard disk (although the seek times are obviously much
better on the SCSI2IDE). This performance was measured with a 20MHz CPU
clock on the SCSI2IDE and using CF media.

Thanks,

Wayne

Post by Wayne Warthen
Thanks Martin. This is MUCH more info to work from than I have had in the
past. I will review all of this and see if I can come up with something.
Wayne

[Borut]

Hi Wayne,

I tried your latest version of firmware (0.5a) and i am getting the
following timeout:
*** Timeout @13A1H ***
This happens with production version.
With debug version everything works ok.

Best regards,
Bo/

Post by Wayne Warthen
Hi Martin,
Spent most of the day on this. Your testing and captures definitely help
pin down the nature of the issue. While I am still not seeing a code
issue, it seems pretty clear that the code is getting stuck in one of the
loops that wait for DMA REQ. So, I refactored the code to include a simple
timeout in those loops. If the timeout occurs, it will dump the controller
register values to the console, free the bus, and restart the protocol.
If you could try the latest firmware, I would appreciate it. I have
attached the zipfile to this post.
Regarding performance... I finally found a copy of Check IT which has a
hard disk performance benchmark. It indicates that the SCSI2IDE is
achieving 83K bytes per second which is almost identical to the performance
of a true legacy XT hard disk (although the seek times are obviously much
better on the SCSI2IDE). This performance was measured with a 20MHz CPU
clock on the SCSI2IDE and using CF media.
Thanks,
Wayne

Post by Wayne Warthen
Thanks Martin. This is MUCH more info to work from than I have had in
the past. I will review all of this and see if I can come up with
something.
Wayne

[Wayne Warthen]

Thanks Not it. Currently on vacation. Will take a look in a few days. Thanks for this info.

Wayne

[Wayne Warthen]

I have posted the v0.6 firmware. It is in the normal Wiki folder at
http://n8vem-sbc.pbworks.com/w/browse/#view=ViewFolder&param=SCSI2IDE and
the file is called S2I-v0.6.zip. Please be sure to review the updated
document http://n8vem-sbc.pbworks.com/w/page/62549548/SCSI2IDE primarily
because the P1 jumper bank is now being used by the firmware.

Post by Wayne Warthen
Thanks Not it. Currently on vacation. Will take a look in a few days.
Thanks for this info.
Wayne