Open mbalestrini opened 2 years ago
Here's a thread on the slack where I posted about this topic: https://skywater-pdk.slack.com/archives/C02K4RD241Y/p1640276417199900
Addressed with a preliminary input constraint. Not clear that the general problem is addressed with clear boundary constraints.
I'm running STA on the whole caravel+user_project_area to try to detect any issues we might get from the integreation of both. I have a question about the mgmt_core.user_irq inputs:
In the new core there seems to be a lot of hold delays cells to handle outputs from the user_project_wrapper (
la_data_out
,wbs_dat_o
) into the mgmt_core that help even out the difference in clocks to avoid hold violations. But it seems there are no delay cells on theuser_irq
paths, and that is causing some hold violations in our design. Taking a quick look at themgmt_core.sdc
I see that it sets aninput_delay
of 1.0 to all the inputs (mprj_dat_i
,la_input
) but not for theuser_irqs
. Is there a reason behind that? I saw that the user_irq inputs use double flip-flop to sync the signal but when the signal comes from a synced cell in the projet area it might probably always caused a hold violation on the first ff