test: set thread count before grt to use parallel repair_antennas #5345
Conversation
Signed-off-by: Eder Monteiro <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
test/flow.tcl
Outdated
| @@ -191,6 +191,8 @@ write_verilog $verilog_file | |||
| ################################################################ | |||
| # Global routing | |||
|
|
|||
| set_thread_count [exec getconf _NPROCESSORS_ONLN] | |||
There was a problem hiding this comment.
Is there a reason not to put this at the start of the flow? It won't make much difference today but it could in the future.
There was a problem hiding this comment.
I thought it would affect macro placement because mpl2 produces different results when using multithreaded (@AcKoucher, please correct me if I'm wrong), but we're not using it in the regressions, so it should be fine.
There was a problem hiding this comment.
The problems with mpl2 generating different results for different number of threads were solved a while ago. It should be fine.
Signed-off-by: Eder Monteiro <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
Signed-off-by: Eder Monteiro <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
|
@maliberty I see this failure in the CI: But I can't reproduce it on my GCP machine. We have another source of possible non-determinism since this metric is generated before grt. I can't download the artifacts of this design from the CI. Is there a way to access it? It would be nice to see where the discrepancies begin. |
It has only for sky130hs/gcd. |
Signed-off-by: Eder Monteiro <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
Signed-off-by: Eder Monteiro <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
|
Closing it due to instability in the results. Will create a new PR from the latest master branch to check if the instability persists. |
No description provided.