Advanced Search

Would the 2.7.x version support a 7i96 ?

RPi4 will be cool (as much as learned colleague Tommy isn't a fan ), but I would do this slight mod to the image, sorry I missed this when creating it, any other issues blame me. One user had an issue with GMOCCAPY locking up after a period of idleness.
If it was a choice between buying an RPi (5 would be the best choice out of the lot) I'd recommend as SFF pc, better bang for buck.
forum.linuxcnc.org/9-installing-linuxcnc...only?start=30#322251

If it was a choice between buying an RPi (5 would be the best choice out of the lot, blazingly fast with an NVME drive) I'd recommend a 64 bit PC platform, 3rd gen I5 is fine, as they are better bang for buck compared to a RPi.

Actually the whole thread is a jolly good read and tries to answer a lot. If you have an other questions be sure to add to the thread.

Does Mesa ethernet work with the RTAI kernel or just RT_PREEMPT ?

tommylight wrote:

F@king google, can not find the chip that was used as parallel port on mainboards, it was something like 665 and some other numbers...or 8255 i think, been to long, and google isn't helping at all.
 

Early versions used discrete components (think XT 74LS374 according to the tomb linked below), usually these were ISA cards, 8255 sounds about right maybe some more dedicated chips later then the Super IO chips, which actually connect via LPC bus.
en.wikipedia.org/wiki/Super_I/O.

For those interested in ancient history
ia800901.us.archive.org/26/items/f15_Par...images_processed.pdf

You can still use 32Bit PC's, there are older LinuxCNC ISO's still available for download so give it a try
I think this is the latest 32Bit version
www.linuxcnc.org/iso/linuxcnc-2.7.14-wheezy.iso
And if it is even older, here are the some more to try
www.linuxcnc.org/iso/ubuntu-10.04-linuxcnc3-i386.iso
www.linuxcnc.org/iso/ubuntu-8.04-desktop-emc2-aj13-i386.iso
They all work from USB without installing them, so you can actually run machines from the USB, granted the PC can boot from USB, and most can, from ~2000

langdons wrote:

What if you try 0xD010 ?

I think a parallel port card is basically 3 latched, buffered, 8-bit shift registers attached to a very primitive MCU.

2 bidirectional shift registers.
1 Parallel in, serial out shift register.

And a CMOS not gate on 4 pins.

Seriously dude your techno babble never ceases to amaze me.
MCS9900 Diagram 


Can't link directly to datasheets but you can find it on this page.
asix.com.tw/en/product/Interface/PCIe_bridge/MCS9900

Interesting idea:

Connect one of the pins on your breakout board to a falling-edge-triggered shift register to turn one output into 8 outputs!
[/code]

Already an example of that on the wiki. Bit more involved than what you suggest.
wiki.linuxcnc.org/cgi-bin/wiki.pl?Shift_Register_Port_Expander

 

F@king google, can not find the chip that was used as parallel port on mainboards, it was something like 665 and some other numbers...or 8255 i think, been to long, and google isn't helping at all.

PCW gave a very valid solution, so that has been covered.
The driver in use covers both functions of the board, and if you look at the dependencies of the driver you'll see that the parallel part uses parallel port modules.
But that is neither here nor there, the thing is a solution has given and if you really really want a change you'll have to mention on GitHub , as that's where the changes happen. You can go on and on posting messages on the forum but it'll do SFA.
And in reality it is only a minor thing that doesn't affect the functionality. If you get no response from the Devs there's nothing to stop you from cloning the repo, editting the source file and rebuilding from source, the build procedure is very well documented in the docs. I linked to the file where the message comes from, PCW gave a solution, all the hardwork has been done. Now the ball is in your court, if you choose not to do anything,nothing will change, if you choose to get involved and not just complain something may happen.
Or you could attempt to completely rewrite the detection routine.
This is how open source projects move forward. The forums are just really for user support, code stuff usually gets done via the Devs mailing list or via GitHub. Links to these resources can be found on the Linuxcnc home page.

The IRQ32 plate is the one that triggers the ALRMA

That's all excellent to hear. I pulled the trigger and ordered one this afternoon, after I realized that my mill PC was not only out of date by years, but also not upgradeable, since it's a 32-bit machine. I have a spare RPi 4 that I'll try with the 7i96. If I'm not happy with that (unlikely), I'll pick up a used small form-factor PC.

paul

Reply redacted.

What if you try 0xD010 ?

RobotMatic wrote:

the point is that Linuxcnc does not find a parallel port, it finds a port that it can use but the kernel driver defines it as serial,

Hmmm, if it works, i am sure LinuxCNC can find it as parallel port, the driver in use being serial is strange if there is actually nothing serial on the board, but from experience most of those boards do show as serial and parallel or both, despite missing serial ports.
I would aim at missing "driver in use" when the error appears...

No "serialized parallel" or "parallel serialized" no no noooooooo!
I really do not like labeling stuff at random like that, it is a parallel port interface running on a PCI-E bus, not the same thing and absolutely not something to be mixed at random.
The one with serial ports is an UART running on a PCI-E bus, nothing in common except the "serial" part.
-
Watching RoboMatic's pictures it just dawned on me, the normal PCI ports use one of the 16 IRQ's (FYI PC has 16 IRQ from 0-15) but this in the picture is 16 and still shows like a normal IRQ, so i wonder.... but the other one is IRQ32 and in a group of IOMMU so it has no memory address.
Also most probably lack of DMA access might be triggering the warning...
This needs more testing, whenever i get a chance, and whenever i get more PCI-E ones as i am out of those for now.

the point is that Linuxcnc does not find a parallel port, it finds a port that it can use but the kernel driver defines it as serial, it does not matter what is the capacity of the micro that uses the PCIE board, nor does it interest all the capabilities that you mention. Linuxcnc does not know how to solve the situation and gives us an alarm with a dual message, everything works perfectly, we are the ones who have to follow. I'm asking if you can somehow eliminate this situation of alarm or Dual message, I'm not trying to discuss technology, Nothing in the cpus works in Parallel, everything is serial, if you think differently perfect, I have no problem with that, but please don't deviate the reason why this thread was created in the forum. Thanks a lot.

By your definition of the PCIe being serial, no arguments there, it would be better described as a serialised parallel port. Generally any PCIe card would be described as a serialised device.
A more correct definition is a parallel/serial chip with a PCIe interface. Mesa sells PCIe motion control cards, would you call them serialised motion control cards ? What about GPUs, multilaned serialised devices ? Or look at the RPi 5 with the new RP interface chip, the GPIOs are on this chip which is connected via the PCIe bus, are they part of your serialised definition ?
Generally most people would look at the end point, not the bus of the computer. It's just easier to increase the speed of a serial bus compared to a parallel bus on a PCB. Trace length really comes into play, reflection, termination and many other things are more difficult routing wise with high speed parallel buses.

With regards to the chips used I think you would get a better understanding if you had a look at the data sheets. You'd be surprised the actual functions that these cards can perform. Some go further than just serial or parallel.

As PCW mentioned a change in the definition of the message would sort this out. If you feel strongly enough this should be changed, mention this on the Linuxcnc GitHub, as this is where changes to the code get done.