How to Manually Place Different Sub-Modules at Specific Chip Locations Using OpenROAD Flow Script?

[wl764]

I’m currently working with the OpenROAD flow and have a design that contains several sub-modules. I’d like to manually place these sub-modules (for example, imul1 and imul2) at specific locations on the chip during the floorplan stage.

My goal is to assign each sub-module to a predefined physical region. I’m aware that OpenROAD allows both automated and manual placement approaches, but I’m not sure how to modify or extend the flow script to handle fixed placements for individual sub-modules.

Could someone please advise on the best way to achieve this?
• Do I need to encapsulate these sub-modules as macro cells first?
• How can I use or modify the floorplan or macro placement scripts to enforce placement into specific regions (for example, using fence regions or direct coordinate assignments)?
• Any sample scripts, guidelines, or documentation pointers would be extremely helpful.

Thanks in advance for your assistance!

[maliberty]

The easiest way is to encapsulate them as hard macros and then place those.

For manual placement you can use https://openroad.readthedocs.io/en/latest/main/src/mpl/README.html#place-macro

A small design using macro hardening is https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/tree/master/flow/designs/asap7/aes-block

[wl764]

When I try to make this example, I met with this problem:"make: *** No rule to make target 'designs/asap7/aes-block/aes_rcon/config.mk', needed by 'results/asap7/aes-block_aes_rcon/base/aes_rcon.lef'. Stop.
" and the aes_rcon and its config.mk didn't be generated. By the way, is there any other methods exclude encapsulate them as hard macros and then place.

[maliberty]

You want to run the top level config and it will drive running the subblocks

[wl764]

The problem still persists.

Is there an alternative solution that would allow me to place different sub-modules at specific chip locations without the need to wrap them as macros? Any insights, suggestions, or workarounds would be greatly appreciated! Additionally, is it possible to directly apply constraints on different modules in the constraint.sdc file to control their floorplan placement?

[maliberty]

This design runs in the CI successfully so it would be helpful to know exactly how you are running it. Let's at least get it working before jumping to other paths that are likely to have other challenges.

[wl764]

Hi @maliberty, thanks for the help! I now managed to wrap each sub-module as a macro, and I’m able to generate the LEF for each.

Now I’m wondering:
1. How can I manually control the placement of each macro at a specific coordinate (e.g., place macro A at (100,100), macro B at (200,100)) in OpenROAD?
2. Is there an alternative way besides hard macros? For example:
• Can I use create_place_area or create a fence region and constrain submodules inside those regions?
• Can we apply placement constraints to modules directly in the floorplan or placement phase without wrapping?

Would really appreciate an example or script snippet if possible!

[maliberty]

For manual macro placement see https://openroad.readthedocs.io/en/latest/main/src/mpl/README.html#place-macro

For the automatic macro placer you can create an overall fence but not a per-macro fence. See -fence* in https://openroad.readthedocs.io/en/latest/main/src/mpl/README.html#rtl-macro-placer

[wl764]

Thanks for the reply!

Just to clarify — is there any way to do this kind of floorplan without encapsulating submodules as macros? For example, if I have a hierarchical design where PE units are individual RTL submodules (not macros), is there any way to control their relative floorplan positions in OpenROAD?

Or is macro-level encapsulation the only way to achieve physical separation or placement control of logic blocks?

[maliberty]

You can define groups and regions for this purpose as well. Either through UPF (power domains), a floorplan DEF, or ODB API calls.

[wl764]

Thanks again! I’m using the manual place_macro method as described in the docs.
If I harden an RTL submodule (like a PE unit) into a macro and instantiate it multiple times at the top level (e.g., PE_0_0, PE_0_1, …, PE_3_3), can I use multiple place_macro commands to manually place each instance at a specific (x, y) coordinate?

[maliberty]

Yes. You might find https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/tree/master/flow/designs/asap7/mock-array useful as it places an array of harden blocks.