MVME5500 1 Gbit Ethernet update. It works ! (was Re: MVME5500 1 Gbit Ethernet no longer works)

[Kate Feng]

Hi all,

I made a new update at the same site:
http://sourceforge.net/projects/rtemsdiscobsp4mvme5500mvme6100/?source=navbar

Cheers,
Kate Feng

Hi all,
The mvme5500 BSP is updated and tested on RTEMS4.10.2 to support the
new 1 Gbit Ethernet interface chip: Intel 82545GM. It can be
downloaded as mvme5500.tar.gz at
http://sourceforge.net/projects/rtemsdiscobsp4mvme5500mvme6100/files/?source=typ_redirect
The new support is added for the disco BSP as well for the RTEMS4.10.2.
It can be downloaded as disco-1.2.tar.gz at the same URL. Packed in
the disco-1.2.tar.gz tarball,
the code for mvme6100 was tested in RTEMS4.10.0 a while ago. I
currently do not have the board
to test it for RTEMS4.10.2.
Cheers,
Kate Feng

In USA, the original version of mvme5500 is still available for

purchasing.

I was in the process of adding the software support for the new NIC on
the mvme5500 board, then here is the discussion.....
According to the manufacture specification of the Discovery
system controllers (i.e. controllers for mvme5500 and mvme6100),
"GPP interrupt value register" should be used to monitor the interrupts.
However, Beatnik BSP uses the "GPP interrupt cause register" to
monitor the interrupts. I started to write a disco BSP to add supports
for discovery based boards such as mvme6100 to be run-time
compatible with the mvme5500 BSP that I wrote.
1) I prefer to follow the manufacture's specification.
2) I read the code of the beatnik network drivers years ago.
The code for the 100 MHz NIC on mvme5500 and 1 GHz NIC
on mvme6100 was patched with workarounds. If I recall
correctly, I tested the network performance of the beatnik BSP on the
100 MHz NIC on an image processing application, it delivered only
50% of performance as compared with that of the mvme5500 BSP.
I do not recall if I tested the performance of beatnik BSP on mvme6100.
With the assigned tasks I had back then,
a) I needed to test the performance, especially that of the DMA,
on the mvme6100 without compromising its performance and the
code (i.e. no workarounds !!! ).
b) It is painless to add supports for other discovery based boards

because

I had the mvme5500 BSP handy.
The mvme5500 BSP supports the operations at NSLS smoothly. I never
needed to write any workarounds in my applications, including the code
for DMA. If there were any bug in the applications, I could be so

confident

that it was not due to the BSP. The mvme6100 functions smoothly with the
disco BSP, too. However, the performance delivered by the mvme5500 BSP
is fast enough for most of my applications.
Cheers,
Kate Feng

Till, does SLAC use a mixture of 5500 and 6100 painlessly with

Beatnik? I've told my client that uses the MVME5500 with Beatnik to get a
6100 in-house and test it out with their RTEMS software. As long as things
are changing on the MVME5500 (slower cache, different ethernet controller)
they should evaluate using the 6100 instead (faster processor, two gig-E
ethernets, etc) to see if that can be used on new systems and as a
replacement part (with the replacement maybe tied to recent software
releases, they can hoard the older boards for really old systems).

FWIW: the driver I ported for the beatnik BSP is almost 10 years old
and it seems to support the 82545GM (IDs 0x1026, 0x1027, 0x1028).
You could use this BSP or you could also use the driver from the
libbsdport addon package (which is newer than the one in beatnik).

Peter
-----------------
Peter Dufault
HD Associates, Inc. Software and System Engineering
_______________________________________________
rtems-users mailing list
http://www.rtems.org/mailman/listinfo/rtems-users