I wish to hear from you your choice of default extraction mode and
detailed extraction mode from placement to post-cts optimization.
The default Encounter flow is:
Start to pre-CTS : default extraction
CTS to pre-detailed routing : clock nets detailed extraction. The other use defualt extraction.
Do you see advantages in switching to detailed extraction during
placemet and CTS, and run optDesign with detailed extraction ? (Just in
case you are not aware, there is an undocumented swith to make
optdesign use detailed extraction (setOptMode -useExtractRCModeSetting).
I am assuming that both RC factor for default extraction and detailed
extraction are properly obtained from running Ostrich. I was told that
it is okay to follow the default flow for 0.13um or larger process, but
should use detailed extraction for nanometer process. I am not sure
my commentsIt depends how long do you are willing to wait for the run to finish. Regardless of default or detail mode, SOCE extraction only co-related less than 70-80% of nets (worst for default mode) as compared to signoff tool (StarRC or QRC). For setup check you could try to increase the multiplier to compensate for it, but for Hold time you are out of luck. SOCE (both default & detail) extraction is very inaccurate in small cap (parallel & xtalk cap) thus created a lot of problem in post route xalk hold check.for PreCts opt, the routing is not done yet, so default is as good as detail mode. You will save some run time using default mode. For postCts since the clk nets are routed, detail mode for clk nets make sense. For postRoute, you really want to use a signoff extraction flow so as to reduce the margin you put on SOCE to fix both setup & hold time.li siang