Matt Liberty
Matt Liberty
Issues or PRs should be filed with https://github.com/parallaxsw/OpenSTA if still relevant. This is effectively a fork (though not strictly for historical reasons).
Looks like https://github.com/The-OpenROAD-Project/OpenROAD/blob/1749609708ddfa210d26a18c3746ac6f0cc4764f/src/gpl/src/nesterovBase.cpp#L2690 is similar.
I've started a full test run to make sure the approximation is fine.
> @maliberty I'm thinking of switching the bias to the value `kExpC =45799`, which minimizes the maximum relative error. Let me know soon as I'll have to restart the run.
I see no check to ensure " The user must ensure that the argument is in the valid range (roughly, -700 to 700)."
> Do you care to make sure that exp(NaN) == NaN? not especially
I've started the tests
I had a wrong submodule and so had to restart the run this morning.
In the public CI I see aes_lvt asap7 and mock-array asap7 both fail with this change. It might be a case of a small diff leading to a magnified diff...
https://jenkins.openroad.tools/job/OpenROAD-flow-scripts-All-Tests-Private/job/secure-schraudolph/ I'm not sure if you'll be able to see that pipeline or not