What is the standalone RedHawk static IR/EM flow?
From PDVerse PnR Interview Handbook · pdVerse Mentor Guide
Short Answer
In the RedHawk TCL shell you import the design through a GSR file, build the database with setup design (RH), calculate power with perform pwrcalc (RH), extract the power and ground networks with perform extraction -power -ground (RH), and solve with perform analysis -static (RH). EM is a separate step, because RedHawk does not check it during simulation unless the ENABLE_AUTO_EM keyword is set, and perform emcheck (RH) reports it. Standalone RedHawk needs an Ansys licence.
Technical Explanation
- The GSR file lists the tech file, LEF, DEF, libraries, pad files and power settings.
import gsr(RH) reads it; separate commands such asimport tech(RH) andimport lef(RH) exist, but the manual recommends naming the files through GSR keywords. setup design(RH) builds the database and runs integrity checks, for example frequency not over 50 GHz and transition times not over 5 ns.perform pwrcalc(RH) computes instance power. Review power_summary.rpt in adsRpt, because wrong power makes every later drop number wrong.perform extraction -power -ground(RH) builds connectivity and extracts resistance for static analysis. Look for the warning that a net is not driven by any pad, and for missing connections in the adsRpt unconnect report.perform analysis -static(RH) solves the network. The log lists the five worst drops per net, and the design's .ir.worst report lists the worst nodes in decreasing order with actual and ideal voltage, net, location and layer.- For EM, set ENABLE_AUTO_EM to 1 in the GSR to check during post-simulation processing, or run
perform emcheck(RH) afterwards. Its modes are AVG, RMS, PEAK or all, the default is all, andperform emcheck -mode AVG -net VDD(RH) narrows it to one mode and net. Static EM uses average current density for wires and current per cut for vias. - Static EM violations go to adsRpt/Static as the design's .em.worst report: segments over 100% of the limit, top 1000 by default, adjustable with the EM_REPORT_PERCENTAGE and EM_REPORT_LINE_NUMBER keywords.
- Before power data exists,
perform gridcheck(RH) needs only the imported design and extraction, and ranks instances by relative grid resistance, which catches weak or floating connections early.
What To Check
- power_summary.rpt totals against the signoff power estimate.
- No undriven-net warnings and an empty unconnect report after extraction.
- Worst drop per net in the .ir.worst report against the budget.
- Map review for hotspots, unexpected colour jumps or black areas, which point to missing straps, missing data or missing connections.
- The .em.worst report produced with the rule set the foundry requires for each mode.
Command Checks & Actions
import gsr design.gsrReads the design files and settings named in the GSR.
setup designBuilds the database and runs the data integrity checks.
perform pwrcalcCalculates instance power, summarized in adsRpt/power_summary.rpt.
perform extraction -power -groundExtracts the power and ground networks for static analysis.
perform analysis -staticRuns static IR drop analysis; worst nodes go to the .ir.worst report.
perform emcheck -mode AVG -net VDDReports average-current EM on VDD.
Healthy, Suspicious & Hard-stop Results
- Healthy (illustrative): Total power within 5% of the signoff power number, no unconnect entries, VDD drop plus VSS bounce, for example 24 + 7 = 31 mV, inside the 5% budget of 45 mV at 0.9 V, and no EM ratio above 100%.
- Suspicious (illustrative): Black areas on the map, a few floating instances in the gridcheck report, or EM ratios between 90% and 100% on straps next to the pads.
- Hard stop: A net reported as not driven by any pad, drop over budget, or any EM ratio above 100% at signoff.
Common Mistake
The Trap: Assuming perform analysis -static (RH) also checked EM. By default RedHawk does not run EM as part of static or dynamic simulation, so a clean IR report goes into the signoff package with no EM check at all, and a late review finds ratios far above 100% on the strap stubs next to the pads.
What The Interviewer Is Testing
- Can the candidate list the steps in order and say what each produces?
- Knowing EM is a separate step and what the emcheck modes are.
- Knowing which reports to read: power summary, unconnect, .ir.worst and .em.worst.
Follow-up Question & Model Response
"What is the difference between gridcheck and a static IR run?"
Candidate Model Response: perform gridcheck (RH) needs only the imported design and P/G extraction, not power calculation or analysis, and reports each instance's grid resistance normalized from 0 to 100. It shows where the grid is weak regardless of how much current flows there. Static IR combines grid resistance with real current, so a weak spot under quiet logic may show little drop while a strong grid under very busy logic can still show a lot.
Practical Example
Tapeout Scenario: A 0.9 V block (illustrative): power calculation reports 412 mW against 405 mW from the power team, within 2%. Extraction lists 3 decap instances in the unconnect report; they are fixed in ICC2 and a new DEF is imported. Static analysis gives worst drop 24 mV on VDD and 7 mV bounce on VSS, 31 mV total against a 5% budget of 45 mV. perform emcheck -mode AVG (RH) lists 4 VDD segments on M2 at 112% to 128% of the limit, all next to one pad; widening those stubs brings the worst ratio to 86%.
PnR Flow Mentor Guide
Master the Physical Design Implementation Flow
Read the complete 8-chapter PnR Flow Mentor Guide free on the web — library setup through placement, clock tree synthesis, routing, chip finishing, hierarchical implementation, and ECO, all the way to stream-out.
Continue practising