[Eda-dev] Problems with qrouter getting stuck

Tim Edwards tim at opencircuitdesign.com
Mon Oct 16 08:22:37 EDT 2017

Hello Øyvind,

> I'm having a problem with qrouter getting stuck. It has been stuck on the
> last line for 12 hours.
> Any tips on how I might go about debugging this?
> I've managed to get smaller designs through. Am I running into some sort of
> limit?

Possibly.  From your output dump, I see you are using version 1.3.33,
which is almost two years out of date (on what is a very actively
developed program).  In the past year, I have made significant changes
that have increased speed and efficiency and lowered memory overhead.
The most recent version (1.3.93) will run designs with more than 100,000
gates (maybe much more;  I have not stress-tested it beyond that).

I expect the problem will go away with an update.  However, some of your
questions have answers that are more nuanced than just "go update
qrouter", so please see below.

> Thoughts:
> *Is there a limit to the number of input/output bits I can have on the top
> level design?* I'm using osu018 and I have LOTS of top level input/output
> pins. I'm primarily after numbers on size and clock frequency at this
> point, so I'm passing a mock design to qflow to try to get it to give me
> numbers.

There will be some limit at which there are more pins than room around
the layout perimeter to place them, in which case qflow may start stacking
them outside of the layout where they end up outside of the routing grid.
This is more likely to happen with pin constraints given to graywolf, such
as attempting to place all pins on one side of the layout.

> *Is the error message below relevant?* I don't think the error message is
> important. It's probably just from this line "if [ ${TERM:=""} = "cygwin"
> ]; then" statement in qrouter. If that statement fails, then it won't run
> the cygwin hack, which is fine, I'm running under Ubuntu 16.04.

No, that won't have an effect, but it is an interesting error.  What OS
are you using?  Most Linux systems map /bin/sh to /bin/bash, and the
syntax is probably a bash-ism.

>  From "top", I can see that the qrouterexec process is still busy. It is
> using 45GByte, I only have 64GByte swap on this machine. Too little?

The memory overhead problem was the biggest problem with qrouter until
I found a good way to overcome it in revision 76.  Since then, qrouter
runs with a much lower and more predictable memory overhead.

> I'm can't provide the entire design, but I'm happy to provide any other
> information that would be helpful. I couldn't find a way to create an
> account to register a bug in Bugzilla.

Sorry, I couldn't find a way to prevent people from creating accounts
in bugzilla and plastering it with spam, so I had to disable it.


| R. Timothy Edwards (Tim)       | email: tim at opencircuitdesign.com    |
| Open Circuit Design            | web:   http://opencircuitdesign.com |
| 19412 Cissel Manor Drive       | phone: (301) 528-5030               |
| Poolesville, MD 20837          | cell:  (408) 828-8212               |

More information about the Eda-dev mailing list