Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

DC-topo and ICC environment comparison

Status
Not open for further replies.

r1caw ex ua6bqg

Member level 2
Member level 2
Joined
Jun 27, 2017
Messages
44
Helped
0
Reputation
0
Reaction score
0
Trophy points
6
Activity points
403
SPG flow: DC-topo and ICC STA results difference

Dear all!
I have two questions related to DC-topo to ICC BE flow.

I synthesized design in DC-topo with physical libraries (ref MW, tf, TLU+, libs etc) using "compile_ultra -spg" command. After synthesis has been done, I used ICC (with exactly same physical libraries as in DC) to source pre-placed design from DC-topo. I compared two pre-placed floorplans (from DC-topo and restored in ICC with "restore_spg_placement" command) - they are the same. But STA results are much worse in ICC, (slack is worse more than 2 times). Why I have this huge difference (Synopsys user guides told me about 10-20%, but not 200%)? Libraries are the same, floorplans with placed cells are the same.

I compared environments (variables, settings etc) from DC-topo and ICC using "write_environment -consistency" command and obtained two txt files (from DC-topo and from ICC) with environment variables and their values, etc. But comparing them (even using Meld or diff) is tedious. Maybe there is any DC/ICC command to automatically setup same environments to DC-topo/ICC? Or should I check only some keys variables from DC/ICC environment files to obtain STA consistency for restored pre-placed design within 10-20% in ICC?
 

The simplest way to crosscheck is to debug the most critical path in both environments. The difference should become evident very clearly: either from loads or from cell timing.

However, what do you mean by slack is worse by 2x? If you were missing timing by 10ps, and now you are missing by 20ps, that is probably correct.
 

Thank you for your answer!
Unfortunately, I am talking about ns, for example, -0.22ns for DC-topo and -0.47ns for ICC.
Why I get this huge difference?
 

Thank you for your answer!
Unfortunately, I am talking about ns, for example, -0.22ns for DC-topo and -0.47ns for ICC.
Why I get this huge difference?

I don't know, these numbers don't mean much without context. Post the critical paths and maybe it will become clearer why.
 

If you want to improve the colleration between ICC and DC-topo,
there are variable to switch the placer and router as ICC engine on DC-topo.
Run time will be longer, but colleration is better.
 

Thank you!
But I am always work in DC-topo.
What do you mean when talking about placer and router in ICC?
 

Thank you!
But I am always work in DC-topo.
What do you mean when talking about placer and router in ICC?
Something I ussually understand as "engine".
DC can use that engine of ICC to do placing and routing within DC-topo.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top