OMAP4 network test

Would anyone with an OMAP4 board like to test networking and lanmanfs with the rom in OMAP4rom.zip. I’m interested if there is a speed problem (or any problem for that matter) and whether authentication works in LanmanFS. I’d like to know if there are any problems even if you don’t use LanMan.

For comparison, for me a 10MB file copies across to a remote shared ram disc at about 4MB/s in both directions. LanmanFS copies to an XP server at about 5.5MB’s and from an XP server at about 8MB/s. These speeds are dependent on the server Windows 8.1 for example uses a different system to XP and only transfers at about 3.5MB/s in both directions.

LanManFS… The one free with OmniClient. Right?
I’ll make tests Tuesday (my Panda is not in the same town than me this week-end).
Today, I’m pretty sure the speed is much lower than 8 MB.

LanManFS… The one free with OmniClient. Right?

Yes thats the one.

If anyones using Lanman98 I’d like to know it that has any problems too – the rom includes changes to EtherUSB as well as Omni.

I’d also be interested if anyone uses a plugin USB ethernet adapter on any riscos computer.

Regarding speed. If there is a wireless connection between the riscos machine and the server – maybe the server is on a wireless laptop – then the speed will plummet and you will be lucky to get 1MB/s

I’m using a plugin USB ethernet adapter on my touchbook. SOHOware NUB110 10/100 USB Network Adapter

Is that using a Pi?

Can you do a *ejinfo to find out which EtherUSB back end it uses?

I can’t test it now, but is has a PEGASUS_II chip. (I stay with my laptop in bed, recovering from flu.) Touchbook is OMAP3.

Colin – is this ROM significantly different to the one I tested last month. If it is, I’ll do some more tests.

Yes. I’ve been given the loan of a pandaboard and the rom now works properly for me on that where it didn’t before.

Have just tested the 14-Feb-2015 (PandaLand OS update 8) ROM against your test rom (03-Apr-2015):

RISCOSmark 1.01 (14 May 2003)
Comparison with RiscPC SA 202MHz running RISC OS 4.02 800x600,256
(HD benchmarks are in kilobytes/sec)
8Tbyte NAS via LanMan98
14-Feb-2015 rom:
HD Read - Block load 8MB file                               5389     180%
HD Write - Block save 8MB file                               297       9%
FS Read - Byte stream file in                                572     276%
FS Write - Byte stream file out                              405     210%
03-Apr-2015 rom:
HD Read - Block load 8MB file                               9637     323%
HD Write - Block save 8MB file                              6942     228%
FS Read - Byte stream file in                                966     466%
FS Write - Byte stream file out                              465     242%

ShareFS to VRPC Adjust
14-Feb-2015 rom:
HD Read - Block load 8MB file                               2202      73%
HD Write - Block save 8MB file                              2474      81%
FS Read - Byte stream file in                                100      48%
FS Write - Byte stream file out                               50      26%
03-Apr-2015 rom:
HD Read - Block load 8MB file                               2281      76%
HD Write - Block save 8MB file                              4762     156%
FS Read - Byte stream file in                                101      48%
FS Write - Byte stream file out                               50      26%

For comparison here is the ARMX6 equivalent benchmark (does not use EtherUSB):
17-Feb-2015 rom:
HD Read - Block load 8MB file                               4035     135%
HD Write - Block save 8MB file                             18478     607%
FS Read - Byte stream file in                               2048     989%
FS Write - Byte stream file out                              552     287%

Excellent so a big improvement. Does !Omni (LanManFS) log on to your NAS?

Yes, but it won’t log on to my 8Tbyte NAS using the 14-Feb-2015 rom so I tested it on my 1Tbyte NAS instead for that rom:

RISCOSmark 1.01 (14 May 2003)
Comparison with RiscPC SA 202MHz running RISC OS 4.02 800x600,256
(HD benchmarks are in kilobytes/sec)
1Tbyte NAS via LanManFS
14-Feb-2015 rom:
HD Read - Block load 8MB file                               3000     100%
HD Write - Block save 8MB file                               264       8%
FS Read - Byte stream file in                                 50      24%
FS Write - Byte stream file out                               50      26%

8Tbyte NAS via LanManFS
03-Apr-2015 rom:
HD Read - Block load 8MB file                               8714     292%
HD Write - Block save 8MB file                              5893     193%
FS Read - Byte stream file in                                970     468%
FS Write - Byte stream file out                              875     455%

Good. The encrypted authentication in my rom is new to LanManFS thats why it doesn’t work with the old rom.

Chris: Comparative figures using LanMan98 would be very interesting?

It’s the first block in his first post – big improvement

The encrypted authentication in my rom is new to LanManFS thats why it doesn’t work with the old rom.

I’m assuming there is nothing OMAP4 specific in there and that the alterations could be made to other ROMs

Faster LanManFS across the board would interest a fair number of people.

Yes. I was waiting to make sure that the OMAP4 test went ok (that had the most problems) before posting OMAP3, Pi and Iyonix tests. It’s complicated these days with all these machines.

Doh! Sorry I was being blind again.
Great work Colin. Which modules are updated?

I presume you can’t soft load them over a network. Changing the transport you are loading from could be interesting!

EtherUSB and LanManFS. EtherUSB is the main one and that is tricky to softload as it is used during startup.

Wouter.

Hope you are improving. If you are reading this I was just wondering if you thought networking with your adapter was slow? Do you know if you can still buy your adapter?

Hi Colin, have you changed things in the device-specific backends or in the main EtherUSB code? I am fairly sure I have at least two of the supported chipsets somewhere, the MosChip MCS7830 and either the Asix AX88172 or AX88772. Don’t know if networking was slow, I only used them during very early BeagleBoard days.

Yes I’ve changed all the backends though some have only minor changes – the ones which don’t support multiple packets per transfer. I’ve updated the link in the first post to include the changes for the backends I don’t have so if you can try them out anytime I’d appreciate it as I can’t test them. I have an AX88772 – thats as fast as the smsc95xx. Looking at the others I don’t expect them to be as fast but possibly faster than they were.

Ok would anyone like to try the Pi version PiRom_01.zip

I get Pi sharefs speeds of about 3.8MB/s and Lanman speeds of about 4MB/s write and 6MB/s read.

Again I’m interested if the authentication works – or not – with LanMan.

You’ve improved LanMan’s authentication support? Nice! A couple of months ago I finally made the jump from Win XP to Win 7 and had a look at what would be necessary to get LanMan working sensibly with latest(ish) Windows versions, but soon gave up when I realised that we were so far behind in terms of authentication/protocol support. So if you’ve got it working sensibly with Windows 8.1 then that’s a major achievement. (Of course now we’ll find that Windows 10 introduces another new protocol version ;-))

I’ll try and give this a go tonight.

Colin

I’ve tried your ROM my Pi, origonal version, attached to a TP link to give networking wirelessly and get the following:

LanManFS 0.62MB/s write , 0.62MB/s read
LanMan98 0.60MB/s write , 0.21MB/s read

Great to see authentication working to Windows 8.1 via LanManFS as well, great work.

My results for your !LanManTest:

• Linux machine is running Ubuntu 10.04
• Sunfish tests were using NFS3 TCP with 32K buffer + pipelining. Sunfish itself requires ARMv5 rotated load behaviour (and I didn’t feel like tracking down all the dependencies needed to recompile it), hence only testing with the Iyonix & Pi 1.
• Pandaboard ShareFS test using ROOL ROM required ShareFSWindow of 1 on the Iyonix to avoid lots of stalls during the read test (other ShareFS tests worked fine with a window of 2)
• All tests were over wired 100Mb network with SMSC backend (Pi/PandaBoard) or onboard LAN for the Iyonix
• LanMan auth worked perfectly without having to make any changes to the Linux or Windows machines (there’s no lanman tests for the current ROMs since I want to avoid registry hacks and the like – in order to “harden” my old Windows machine against the campus network at university I think I managed to break sharing completely, which wasn’t an issue at the time but caused problems later on when I actually wanted to use the feature!)
• Out of curiosity I tried toggling the “Use 128-bit encryption” option in Windows off (putting it into 40/56 bit mode) but that seemed to make things slightly slower. But I’m not actually sure what that setting does (e.g. whether it affects clients connecting to the machine or whether it’s for when the Windows machine itself is the client)

Excellent work!

However since you want to know if “there are any problems”, I will mention one minor niggle – when connecting using LanManFS the “Directory path” option in the OmniClient connection window doesn’t seem to actually accept a path. E.g. I wanted to connect to Users/Public (giving your test app a writable directory), but no combination of dots/slashes seemed to work (only “Users” seemed to work as that was the name of the share). Not sure if this falls into the category of “bug” or “unimplemented feature”. Ubuntu seemd to have two boxes for this – one for the share name, one for the folder.