Mock array multiple metal layers pins #2412
Conversation
|
Broke out github issue for failure The-OpenROAD-Project/OpenROAD#5853 |
No other test-case in ORFS before this of pins on multiple metal layers. Signed-off-by: Øyvind Harboe <[email protected]>
oharboe
force-pushed
the
mock-array-multiple-metal-layers-pins
branch
from
9109abc
to
c1dc39c
Compare
…-metal-layers-pins
Signed-off-by: Øyvind Harboe <[email protected]>
Signed-off-by: Øyvind Harboe <[email protected]>
|
@maliberty PDN works now, but lots of DRC errors on the sides:
|
|
Looks like you need The-OpenROAD-Project/OpenROAD#5854 for the pins to be on track. That said I wonder if we couldn't have routed these off-grid as there are no other wires to conflict with - @osamahammad21 thoughts?
|
|
Nvm, that layer has LEF58_RIGHTWAYONGRIDONLY |
Thanks! I'll keep this PR open and try again when it is merged. |
Signed-off-by: Øyvind Harboe <[email protected]>
Signed-off-by: Øyvind Harboe <[email protected]>
Signed-off-by: Øyvind Harboe <[email protected]>
Signed-off-by: Øyvind Harboe <[email protected]>
|
@maliberty FYI, works now. |
|
@maliberty I've reenabled pin placement annealing. That it is slow is a separate concern to this PR, issue filed. |
| "value": 0.0, | ||
| "value": -15.92, |
There was a problem hiding this comment.
Any insight into this degradation?
There was a problem hiding this comment.
I think we need the --annealing io fix first before we have something comparable, but that said things are quite different w.r.t. routing and metal layers. 15ns isnt much, CTS has more variations, etc
|
@maliberty Annealing runs "forever" (usually seconds, now hours). Seperate issue filed. |
|
@maliberty times out after 6 hours in PPL. Will retry after The-OpenROAD-Project/OpenROAD#5847 is fixed. |
Signed-off-by: Øyvind Harboe <[email protected]>
Head branch was pushed to by a user without write access
|
@maliberty @eder-matheus Lots of new DRC errors. It might be worth looking at recent pin placement changes. If the mirror option was broken, it could present itself as these DRC errors:
|
|
@maliberty @eder-matheus These pins should be connected to the bottom of the element, so the problem is in the pin placement of the element, it would seem.
|
|
@maliberty @eder-matheus
to reproduce with latest OpenROAD master: |
|
It is easier to track issues when filed as such. @eder-matheus please take a look. |
Filed here: The-OpenROAD-Project/OpenROAD#5926 |
Signed-off-by: Øyvind Harboe <[email protected]>
Signed-off-by: Øyvind Harboe <[email protected]>
|
@maliberty @eder-matheus Works :-) Ready to merge now, saw pr-head passed for mock-array by drilling into the run. |
|
@maliberty There are unrelated failures in pr-head. Those are presumably there because of my OpenROAD update. I will merge latest master with this PR after OpenROAD has been updated in master. |
Is there another PR doing this? |
The bot that updates OpenROAD in ORFS? |
|
Please address the conflicts |
Signed-off-by: Øyvind Harboe <[email protected]>
Fixed. |
This test case is best articulated as a pull request. There are some issues to be worked out with pins on multiple metal layers.