dbgAutoCellFunction and RC scale factors

Hi Alex,

For 1: I think this was an old or beta way to enable the footprintless flow (meaning, you don't have to specify the buffer/inverter/delay cell footprints - FE can automatically tell what these cells are). In 6.2, the way to do this is: set dbgGPSAutoCellFunction 1. This is covered in the Timing Optimization chapter of the User Guide, under the section "Enabling the Footprintless Flow".

For 2: Look at the RC Extraction chapter in the User Guide. There is a section called "Correlating SPEF Files Using the Ostrich Utility". This is what you need to do to get the RC scale factors. Basically, you give Ostrich an FE SPEF and a signoff SPEF, and it calculates the factors for you.

- Kari

 Hi Kari,

I Extracted scaling factors using Ostrich tool,But how can use them for my design means at what stage I can use them?

What will happen if we use new scaling factors?

Can you give an explanation about shrink factor (If we make change this factor) for the design ?

Hi,

 You use the scaling factors when you set up RC corners in your MMMC setup. They are used before you get to signoff extraction, so that hopefully your parasitics match very closely to what the signoff extraction will be. For example,

create_rc_corner -name cworst_125c \

   -cap_table "cln65g+_1p08m+alrdl_5x2z_cworst.capTbl" \

   -T 125 \

   -preRoute_cap 1.070899 \

   -preRoute_res 0.885552 \

   -preRoute_clkcap 1.000928 \

   -preRoute_clkres 0.754607 \

   -postRoute_cap 0.934657 \

   -postRoute_xcap 1.032526 \

   -postRoute_res 0.992407 \

   -postRoute_clkcap 0.854106 \

   -postRoute_clkres 1.004779 \

   -qx_tech_file "cln65g+_1p08m+alrdl_5x2z_cworst.qrcTechFile"

I've never used a shrink factor, so unfortunately I don't know anything about that.

- Kari 

Hi Kari Thank you,

I got hold Violations (for Reg2Reg paths) for my block as follows:

@CTS:: -0.405ns,

@POST_CTS_OPTIMIZATION: -0.3ns,

@ROUTE: -0.45ns ,

@POST_ROUTE_OPT: -0.5ns ,

After I extracted spef and I correlate spef and detailed spef i derived new scaling factors(0.9 for both res and cap) by using Ostrich tool.With those new scaling factors how can i avoid hold violations for tha block.

And Where to use those Scaling factors exactly? What tool will do exactly if I run with new scaling factors?Will it take different Ressistance and capacitance values while calculating delays??

I use them as follows in config file::

postroute_res=0.9;

postroute_cap=0.9; 

My Primary doubt was as iam moving from post_cts_opt to routing stage the hold should be improved right.Then what will be the reason for hold degradation for my block? Is the tool doing RC extraction with very pessimism before routing the design??

2nd doubt::

Can the design density crosses more than 100%?? 

Hi,

The signoff extractor is not used throughout the flow (this would take too long). The extraction used until you switch to the signoff extractor is fast but not as accurate. That's why we use scaling factors. Sometimes the built-in extraction is too pessimistic, sometimes it's too optimistic.

Once you generate scaling factors, they you typically go back to the beginning of the design and start again. This time the scaling factors will guide the timing optimization to be closer to signoff timing. However, you may have some other issues going on here. Density CANNOT be more than 100%. Your floorplan might be too small, or you might have timing targets that are too aggressive, causing lots of cells to get added.

I would suggest using the full set of scaling factors, as in my last post. I would also suggest using MMMC to keep track of everything nicely, even if you don't have a lot of modes/corners.

 - Kari