Does setting 'random_seed' from CLI guarantee deterministic routing? #583

danboyne · 2026-01-07T04:34:12Z

danboyne
Jan 7, 2026

I'm a new user to Freerouting, and am trying to benchmark this tool against another autorouter. For this purpose, I would like to ensure I can reproduce Freerouting's results, if necessary. That is, I would like to run Freerouting deterministically.

Using Freerouting's CLI, I set '-random_seed' to '1' for two consecutive runs with the same inputs (same DSN file, same '-mp' limit). However, the two resulting .SES files were not identical, and the log messages (to STDOUT) differed in the number of auto-router passes, number of unrouted nets, and other details..

I am running Freerouting v2.1.0 (build-date: 2025-04-12) under WSL2 on a Windows laptop. Am I missing something obvious in using the command below to force deterministic routing?

-Dan Boyne (Austin, Texas, USA)

$ java -jar /mnt/c/Users/danbo/AppData/Local/freerouting/app/freerouting-executable.jar --gui.enabled=false -de freerouting_CNH_Functional_Tester_1.dsn -do freerouting_CNH_Functional_Tester_1_run1.ses -drc freerouting_CNH_Functional_Tester_1_run1.rules -mp 1000 -random_seed 1 > freerouting_CNH_Functional_Tester_1_run1.log

Answered by andrasfuchs

Jan 7, 2026

Hey Dan,

I think there was an issue with the random seed parameter in CLI mode in version 2.1. There was a pull request that I merged, so it will be fixed in the next release.

You can try to test our latest snapshot build until that point, but that might produce questionable results, it hasn't gone through intensive testing. It represents the current development state.

You can find it here: https://github.com/freerouting/freerouting/releases/tag/SNAPSHOT

View full answer

andrasfuchs · 2026-01-07T15:46:20Z

andrasfuchs
Jan 7, 2026
Maintainer

Hey Dan,

I think there was an issue with the random seed parameter in CLI mode in version 2.1. There was a pull request that I merged, so it will be fixed in the next release.

You can try to test our latest snapshot build until that point, but that might produce questionable results, it hasn't gone through intensive testing. It represents the current development state.

You can find it here: https://github.com/freerouting/freerouting/releases/tag/SNAPSHOT

1 reply

danboyne Jan 7, 2026
Author

Thanks, Andras. I greatly appreciate your response. I'll await the next release, as the current snapshot did not generate a .SES file (completely understandable). In the meanwhile, I can become more familiar with Freerouting's features by using an older version.

-Dan Boyne

PS: For reference, my attempt with the snapshot is pasted below.

$ java -version
openjdk version "25.0.1" 2025-10-21
OpenJDK Runtime Environment (build 25.0.1+8-Ubuntu-124.04)
OpenJDK 64-Bit Server VM (build 25.0.1+8-Ubuntu-124.04, mixed mode, sharing)

$ java -jar /mnt/c/Users/danbo/AppData/Local/freerouting/app/freerouting-SNAPSHOT-20251216_081800.jar
-de freerouting_CNH_Functional_Tester_1.dsn
-do freerouting_CNH_Functional_Tester_1_run1.ses
--gui.enabled=false
-drc freerouting_CNH_Functional_Tester_1_run1.rules
-mp 1000
-random_seed 1 > freerouting_CNH_Functional_Tester_1_run1.log
(Carriage returned were added above only for readability.)

$ ls
freerouting_CNH_Functional_Tester_1.dsn # Input DSN file
freerouting_CNH_Functional_Tester_1_run1.log # Log messages (displayed below)
freerouting_CNH_Functional_Tester_1_run1.rules # Rules file looked good!
Note: No .SES file was created

$ cat freerouting_CNH_Functional_Tester_1_run1.log

2026-01-07 10:40:55.237 INFO   Freerouting v2.1.2-SNAPSHOT (build-date: 2025-12-16)
2026-01-07 10:40:55.336 WARN   No default constructor found for field: currentLocale
2026-01-07 10:40:55.340 INFO   Settings were loaded from freerouting.json
2026-01-07 10:40:56.729 INFO   New version available: v2.1.0
2026-01-07 10:40:57.020 INFO   DRC report written to: freerouting_CNH_Functional_Tester_1_run1.rules

andrasfuchs · 2026-01-07T17:44:29Z

andrasfuchs
Jan 7, 2026
Maintainer

I can see a small typo that could end up overwriting your rules files in the process.

You used the new -drc parameter for your .rules file, instead of the old -dr. I suspect that your intension was to use -dr to set your rules file as an input. The -drc parameter sets the following file as the output for the DRC (design rules checker).

This is a valuable feedback. Seeing this I might change the new parameter's name or at least show a warning if you try to pass an existing file to the -drc parameter.

Let me know if anything else comes up.

1 reply

danboyne Jan 8, 2026
Author

Thanks for the clarification between the -drc and -dr command-line switches. I did intend to have Freerouting generate as an output a .rules file. In other words, there was not a pre-existing .rules files in the directory when I launched Freerouting.

Regarding command-line switches, however: Freerouting v2.1.0 did not complain when I added nonsense command-line switches. For example, the following command did not result in any warning/error messages despite the -xyzzy command-line switch:

java -jar /mnt/c/Users/danbo/AppData/Local/freerouting/app/freerouting-executable-2.1.0.jar
--gui.enabled=false
-xyzzy
-de freerouting_CNH_Functional_Tester_1.dsn
-do freerouting_CNH_Functional_Tester_1_xyzzy.ses
-drc freerouting_CNH_Functional_Tester_1_xyzzy.rules
-mp 100
-random_seed 1 > freerouting_CNH_Functional_Tester_1_xyzzy.log

As shown below, the head and tail of the resulting log file did not mention the -xyzzy switch. Interestingly, Freerouting exceeded the 100-pass limit set by the -mp switch:

2026-01-07 20:26:17.623 INFO Freerouting v2.1.0 (build-date: 2025-04-12)
2026-01-07 20:26:17.718 WARN No default constructor found for field: currentLocale
2026-01-07 20:26:17.719 INFO Settings were loaded from freerouting.json
2026-01-07 20:26:19.149 INFO New version available: v2.1.0
2026-01-07 20:26:19.274 INFO [F630DA\6F30D0] Job 'F630DA\6F30D0' started at 2026-01-08T02:26:19.274173553Z
2026-01-07 20:26:34.133 INFO [F630DA\6F30D0] Auto-router pass #1 on board '9aa83c860c207566f00957dc41acbc59'
was completed in 14.66 seconds with the score of 901.94 (32 unrouted), using 111.32 CPU seconds and
7265 MB memory.
2026-01-07 20:26:38.729 INFO [F630DA\6F30D0] Auto-router pass #2 on board 'c126f69993e323b11ce2481de91710be'
was completed in 4.51 seconds with the score of 925.86 (22 unrouted), using 195.73 CPU seconds and
9647 MB memory.

...more lines here...

2026-01-07 20:42:21.590 INFO [F630DA\6F30D0] Auto-router pass #877 on board 'fe6774c3a38b0b1afb734fdc129977e1'
was completed in 0.99 seconds with the score of 941.22 (14 unrouted), using 455999.91 CPU seconds and
618943 MB memory.
2026-01-07 20:42:22.560 INFO [F630DA\6F30D0] Auto-router pass #878 on board 'fe6774c3a38b0b1afb734fdc129977e1'
was completed in 0.95 seconds with the score of 941.22 (14 unrouted), using 456927.41 CPU seconds and
619597 MB memory.
2026-01-07 20:42:23.514 INFO [F630DA\6F30D0] Auto-router pass #879 on board 'fe6774c3a38b0b1afb734fdc129977e1'
was completed in 0.94 seconds with the score of 941.22 (14 unrouted), using 457855.84 CPU seconds and
620269 MB memory.

For what it's worth, the rules were appended to the end of the above log file, starting with:

{
"host": "KiCad's Pcbnew,7.0.4",
"unit": "mm",
"board": {
"bounding_box": {
"x": 2541.0,
"y": -456.2,
"width": 507.0,
"height": -2109.2

...and ending with...

"clearance_violations": {
"total_count": 0
}
} (This was a closing brace in the log file, and not a smiley face.)

andrasfuchs · 2026-01-08T09:24:03Z

andrasfuchs
Jan 8, 2026
Maintainer

Excellent feedback, thank you Dan!

I remember that we had an issue with the -mp switch in CLI mode, I need to make sure that it is fixed now (I think it is in the snapshot build), so I need to write a unit test for it.

Also, warning the user about the unknown parameter(s) is important, that needs to be done.

The -drc parameter doesn't save the design rules, it only saves the design rules violations of the board (eg. wire clearance violations).
.rules files can be saved in the GUI by File / Save... and after you save the .ses file, you get the following dialog.
This isn't ideal at all, and I think we should have a command line argument for this function as well, but it's a low priority now.

Note to myself:

Create a unit test for -mp (--gui.enabled=false doesn't respect allowed maximum number of passes #522)
Add warning for unknown parameters
Add a command line argument to save .rules

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Does setting 'random_seed' from CLI guarantee deterministic routing? #583

Uh oh!

{{title}}

Uh oh!

Replies: 3 comments 2 replies

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Select a reply

Uh oh!

Uh oh!

Does setting 'random_seed' from CLI guarantee deterministic routing? #583

Uh oh!

danboyne Jan 7, 2026

Replies: 3 comments · 2 replies

Uh oh!

andrasfuchs Jan 7, 2026 Maintainer

Uh oh!

danboyne Jan 7, 2026 Author

Uh oh!

andrasfuchs Jan 7, 2026 Maintainer

Uh oh!

Uh oh!

danboyne Jan 8, 2026 Author

Uh oh!

Uh oh!

andrasfuchs Jan 8, 2026 Maintainer

danboyne
Jan 7, 2026

Replies: 3 comments 2 replies

andrasfuchs
Jan 7, 2026
Maintainer

danboyne Jan 7, 2026
Author

andrasfuchs
Jan 7, 2026
Maintainer

danboyne Jan 8, 2026
Author

andrasfuchs
Jan 8, 2026
Maintainer