ihp-sg13g2: Sync PDK #3562
ihp-sg13g2: Sync PDK #3562
Conversation
dnltz
commented
Oct 9, 2025
dnltz
commented
- Sync the PDK with the latest version from the PDK.
- Include the sg13g2.map file to the update script.
- Change the metal routing direction to match changes made in the PDK.
| export IO_PLACER_H ?= Metal2 | ||
| export IO_PLACER_V ?= Metal3 | ||
| export IO_PLACER_V ?= Metal2 | ||
| export IO_PLACER_H ?= Metal3 |
There was a problem hiding this comment.
These changes are important to match the changed routing direction.
|
There are lot of design failures. Please say more about the layer direction changes. I suspect the pdn setup will also need adjusting. |
@maliberty Yes, I'm working on this issue right now. I opened this PR to run all designs in the CI environment (making sure my OpenROAD set up is not broken) and in case I need to open an issue, I can link this PR. Do you have a "Do not merge" label? and sorry for misusing the PR flow :) |
|
Usually people just put that in the title temporarily, eg [DO NOT MERGE] |
dnltz
force-pushed
the
WIP/dnltz/bump-sg13g2-pdk
branch
2 times, most recently
from
310f0ea to
98adbe3
Compare
| }, | ||
| "detailedroute__route__drc_errors": { | ||
| "value": 0, | ||
| "value": 4, |
There was a problem hiding this comment.
@maliberty I had to move the PDN from M5+TM1 to TM1+TM2. Since IHP changed the routing direction, M5 had the same orientation as M1 and they can't be connected.
Some designs now have route DRC errors. I assume they need to be fixed, right?
There was a problem hiding this comment.
I think so as it routed cleanly before. What do they look like?
There was a problem hiding this comment.
The pin placer is completely ignoring those vias between Metal1 and the power ring. I will try to move the power ring a little bit away from the edge.
dnltz
force-pushed
the
WIP/dnltz/bump-sg13g2-pdk
branch
2 times, most recently
from
225af32 to
2eef655
Compare
Also sync the map file for SG13G2. Signed-off-by: Daniel Schultz <[email protected]>
The routing direction for all Metal layers recently changed the in the SG13G2 PDK. Update the config file to match with the latest tech file. Signed-off-by: Daniel Schultz <[email protected]>
Use sg13g2_udpate.py to sync with the latest changes from the main branch. Signed-off-by: Daniel Schultz <[email protected]>
Move Power Delivery Network to TopMetal1 and 2. Since IHP changed the orientation of all metal directions, Metal5 and TM1 can't be used anymore. Signed-off-by: Daniel Schultz <[email protected]>
Since the routing direction changed, pins at the macro edge and power ring vias might be too close to each other. Increase the core margin by 1um to prevent this issue. Signed-off-by: Daniel Schultz <[email protected]>
Move Power Delivery Network to TopMetal1 and 2. Since IHP changed the orientation of all metal directions, Metal5 and TM1 can't be used anymore. Use Metal4 and 5 for the PDN inside macro blocks. Signed-off-by: Daniel Schultz <[email protected]>
dnltz
force-pushed
the
WIP/dnltz/bump-sg13g2-pdk
branch
from
2eef655 to
be4398c
Compare
Update all ihp-sg13g2 designs to match the latest PDK files. Additionally, lower frequency for spi and gdc design to meet worst-case latency and core utilization for the ibex design. Signed-off-by: Daniel Schultz <[email protected]>
dnltz
force-pushed
the
WIP/dnltz/bump-sg13g2-pdk
branch
from
905ad1f to
af45b0d
Compare
3 participants
