set_global_routing_layer_adjustment

[oharboe]

So set_global_routing_layer_adjustment is essentially setting some global variables...

Is there a reason why global variables are used here instead of arguments to global_route?

I much prefer arguments to global route than a global variable, if those are the choices. It is easier to understand the ORFS .tcl scripts when I can see what information flows where...

Same applies to set_macro_extension.

For instance, I wonder why gui.tcl sets these global global route variables...

[maliberty]

The adjustments are needed by Rudy (global_place -routability & estimated congestion heatmap) as well.

[oharboe]

Except it has to be done correctly once by someone who has a chance of finding out(ORFS maintainers), instead of every time a user uses this information.

I did add some automation to try to force errors if a variable is used outside of its defined scope, so it is not entirely "subject to the same human error".

Other suggestions?

[maliberty]

No users of ORFS need this information - its purely a bazel requirement. If you have a check that seems sufficient.

[oharboe]

With the caveat that anyone using global placement needs to know that the global routing arguments are required for global placement to work correctly since it now relies on RUDY, right?

[oharboe]

Non-ORFS users that I know of are OpenLane, SiliconCompiler, bazel_rules_hdl and bazel-orfs.

[maliberty]

Also hammer and doubtless more than neither of us knows about. Developers of flows take this burden on themselves (and quite a lot more).

[oharboe]

Here bazel-orfs has been modified to map variables onto stages using a hardcoded mapping from arguments to stages: The-OpenROAD-Project/bazel-orfs#140