reduce core utilizaztion of rapidus2hp/hercules_is_int

[openroad-ci]

designs/rapidus2hp/hercules_is_int/rules-base.json updates:

Metric	Old	New	Type
finish__timing__setup__ws	-417.07	-411.86	Tighten
finish__timing__wns_percent_delay	-66.06	-65.28	Tighten

Update after changing core util to 52%:
designs/rapidus2hp/hercules_is_int/rules-base.json updates:

Metric	Old	New	Type
finish__timing__setup__ws	-411.86	-476.31	Failing

[eder-matheus]

@maliberty This should fix the timeout on the CI.

@jeffng-or FYI.

[jeffng-or]

@maliberty This should fix the timeout on the CI.

@jeffng-or FYI.

Hi guys,

The fix could fix the timeout, but it will degrade our area correlation. We need to be above 54% to hold the correlation.

[eder-matheus]

The fix could fix the timeout, but it will degrade our area correlation. We need to be above 54% to hold the correlation.

Oh, I see. I'm trying different values now. It's possible that a larger value could work too.

[eder-matheus]

@jeffng-or @maliberty I've tried multiple values above 54%, and all of them fails with placement issues (during DPL or with high congestion in GRT).

An alternative would be revert the Yosys update, but I think it would affect other projects that are running in parallel. Do you guys have any suggestion?

[maliberty]

Can we back off just 1-2%?

[eder-matheus]

Can we back off just 1-2%?

I'm running with 52% now, it should be ready in the next minutes.

[jeffng-or]

I think the Verific version (which is the front end that we are quoting Rapidus) still completes with util == 54%. So, go ahead with setting the util to 52% and then I'll add an ifeq to set the util to 54 for Verific.

[maliberty]

We were in the 3-4 hr runtime in the CI and it would be good to get back to something similar.

[eder-matheus]

We were in the 3-4 hr runtime in the CI and it would be good to get back to something similar.

Using 52% keeps this same runtime, at least on GCP. I'll update the PR.

[eder-matheus]

This should be good to go now!