Cadence Liberate Characterization of Complex Logic Cells

Hi anurans

I wanted to generate automatically all cells definition (define_cell) and all arcs definition (define_arc) for all cells of the library in the template.tcl file. I put in this file after defining  slew, measure_slew, delay, set cell_list, indexes the following script:

 set  inputs {I A1 A2 A3 A4 I0 I1 S D SI SE CDN SD}    --- total inputs of cells in the library

 set  outputs {Z ZN Q}                                                          --- total ouputs of cells in the library  

 set  clockss { CP}

 set  asyncs { }    

define_cell \

          -input $inputs  -outputs  -clock $clocks  -async $asyncs \

          -constraint  constraint_template_11x11 \

          -delay         delay_template_11x11 \

          -power        power_template_11x11 \

        $cell_list

 define_arc \  

          - vector {RXXR} \                 -- assume the maximal inputs of a cell is 4  

          - related_pin $inputs \    

           - pin $outputs \    

           $cell_list  

 define_arc \  

         - vector {FXXF} \  

         - related_pin $inputs \    

         - pin $outputs \

        $cell_list  

It seems it does not function.

What is the best way to define all cells and its arcs?

Kind regards

BarPouy

Hello,

To generate all the cell and arc definitions, you may use (as Guangjun stated earlier): 

char_library -trial; //to generate a dummy .lib file, Then,

read_library existing.lib
write_template -verbose liberate_templates // to dump all the arcs found

This will generate all the arcs related to hidden, timing, constraints etc. If you had already done this, you should look at the log file for any possible issues. If its less descriptive, add following before char_library command : 

#####--------------------------------------------------------------------------------------------------------------------------------------------------
##### Debug variables
#####--------------------------------------------------------------------------------------------------------------------------------------------------
#select_index -style 1x1 ;# run only 1st point in table

set_var bisection_info 4 ;# print additional bisection search info to output log
set_var power_info 2 ;# print additional power calculation info of table's 1st point to file decks/po

set_var ski_enable 0
set_var extsim_save_passed all ;# save all run decks and output files
set_var extsim_save_failed all ;# save all run decks and output files
set_var extsim_save_verify 2 ;# save verify deck
set_var extsim_deck_dir [file normalize "Liberate_Out/decks"] ;# specify directory for SPICE decks and output files
set_var extsim_tar_cmd "" ;# disable tgz of simulation decks and run logs
#####-------------------------------------------------------------------------------------------------------------------------------------------------

liberate char.tcl |& tee char.log // your log file

This would give you a thorough look to the issues you have. May be you should do the debugging ONLY to the erroneous cells, otherwise you will end up having a very large deck file.

Anuradha

1.Check the manual for correct sytanx and use these commands,

set_var extsim_deck_dir full-path-here

set_var extsim_save_failed deck

set_var exism_save_passed none

2. If your run has failed characterization at simulation stage, go to the deck dir. You will see saved deck dir or a tarball. Inside each sub dir, you will see a map.list. open this file you will see where the failure is. Find the deck (sim.sp)for the failed char, remove save=nooutput, then run spectra on it. 

3. Start viva with ade license. Open the simulation results and check the waveform.

4. Review the settings and variables in the deck-- are they the same as your own test bench? If not, either your test bench or the characterization settings need to be corrected.

Tip. select_index 2x2 may use min/max indecis, which makes the debug easier.

Guangjun

Hi Guangjun, Hi anurans,

Thank you for your tips. I'm following it.

Kind regards

BarPouy

Hello Guangjun

Automatically generation of the template (with defined cells) for the cells having the same inputs /outputs, e.g. AN2d1 /OR2d1 (inputs: A1 A2) can be doen. However, If I mix such cells with the cells having more inputs /outpus or other inputs /outputs, e.g. AOI21.d1 (inputs: A1 A2 A3) or DFd1 (inputs: D, clock: CP, output: Q), then the template will not be generated and gives ERROR (see please below). It's clear, if I write for each class of cells, having the same inputs / outputs, a separate define_cell,  then this issue does not occur. There is an example of generating automatically template in the page 48 to 50 of 'Virtuoso Liberate Reference Maual', Version 18.1, section 'TCL Command File'. It shows that it might be possible to generate template of different class of cells like {NAND2X4 NOR2X2 DFFX1} just with one general define_cell.

What is my mistake?

Thanks!

Kind regards

BarPouy

ERROR (LIB-520): (char_library): The pin 'D' is not defined for cell 'AN2d1'. This cell will be ignored. Check the netlist and make sure it is consistent with 'define_cell' command.

ERROR (LIB-203): (char_library): Cell 'AN2d1' is scheduled for characterization but has no netlist, has an empty subckt or has no port on the subckt. This cell will be skipped. Check the netlist and rerun.

ERROR (LIB-520): (char_library): The pin 'CP' is not defined for cell 'DFd1'. This cell will be ignored. Check the netlist and make sure it is consistent with 'define_cell' command.

ERROR (LIB-203): (char_library): Cell 'DFd1' is scheduled for characterization but has no netlist, has an empty subckt or has no port on the subckt. This cell will be skipped. Check the netlist and rerun.

ERROR (LIB-520): (char_library): The pin 'D' is not defined for cell 'OR2d1'. This cell will be ignored. Check the netlist and make sure it is consistent with 'define_cell' command.

ERROR (LIB-203): (char_library): Cell 'OR2d1' is scheduled for characterization but has no netlist, has an empty subckt or has no port on the subckt. This cell will be skipped. Check the netlist and rerun.

WARNING (LIB-961): (char_library): Leakage deck initialization was requested using '.ic', but the 'leakage_sim_duration' was set to '0'. This may lead to unexpected leakage characterization results. Change 'leakage_sim_duration' to a positive non-zero value in seconds or change the setting of 'set_sim_init_condition' and rerun.

Building library database

Processing cell: AN2d1

Processing cell: DFd1

Processing cell: OR2d1

With the information you have provided, I could not make any useful suggestions to solve your issue.

If you do not want to waste time, I suggest you to use define_cell for each cell. I do not see any benefit of using one define_cell for all or a group cells.

Guangjun

First, clarify this ! Are you trying to generate a template/arcs automatically from a good (perfectly usable) existing library (That's what I had grasped) ?

Then you don't even need to worry about defining multiple "define_cells" ! 

read_library existing_good.lib
write_template -verbose Template_Set 

This "Template_Set" is the starting template for all the cells in the library. This only specifies the define_cells and load/slew index values for all the cells without delay/power/constraints arcs. You may need to adjust the min/max trans/caps for your need.

Then when running a "re-characterization" on the same set, let's say, for different corners, use char_library -trial instead of char_library -ccs (or whatever method). Then this generate a trial.lib file that should have additional information related to the missing arcs in earlier step. So that you can re-generate a template file from this trial .lib including all the arcs/define_cells for all the cells without bothering about I/O s of each cell.

Anuradha

Dear anurans,

Thank you for your suggestion.

Yes I'm generating a template/arcs automatically from a  library, generated by your suggested commands in last friday.

#char_library -trial

char_library

write_library cornerName.lib

read_library  cornerName.lib

write_template -verbose cornerName.template

I generated just for six cells in two groups (group1: Inputs: {A1 A2}, output: Z, group2: Inputs: {A1 A2 A3}, output: ZN). However I wrote two different define_cell. It's generated a correct template comprising all arcs of these cells. Als .lib file covers only these cells. In this file, all delay_template_4x4 and power_template_4x4 with index_1, index_2 and values are included .

My question from Guangjun was why it does not function when I write just one define_cell, which Liberate Manual advises. It's clear it does function when I write for each cell group one define_cell.

However for the rest of cells, existing in the directory where the mentioned six characterized cells are placed, delivers, e.g.

*Error* (char_library) : Cell xxxxd1 is scheduled for characterization but has no define_cell command. 

Even if I gave the following min /max values

set_var max_transition 1.601e-08
set_var min_transition 5.579e-12
set_var min_output_cap 9.25e-16
set_var scale_tran_by_template 1
set_var scale_load_by_template 1

but it's generated in the template the following values. maybe dut to the command set_var scale_* 1?

set_var max_transition 1e-09
set_var min_transition 1e-15
set_var min_output_cap 1e-15

It's also generated cornerName.trial.ldb.gz. However it's not generated further data (data sheet). Maybe due to the above mentioned ERROR?

Thanks a lot!

BarPouy

1. the set_var commands you listed has nothing to do with this error.

BarPouy said:

*Error* (char_library) : Cell xxxxd1 is scheduled for characterization but has no define_cell command

2. there is no point wasting your time on getting your scripts with one define_cell working. 

3. I can see you are still building up your knowledge on Liberate/characterization flow. In such a case, it is extreme important to stick to basics and not be carried away.

Guangjun

The problem is, as Guangjun stated, your post does not provide any insightful information to the issue you have. Atleast you could share the define_cell template information here or in a https://pastebin.com/ link. I bet, the pin order in your single define_cell command does not match with the pins you have in cells.

Anuradha

I do not believe this issue is worth further discussion, until one define_cell for each cell is used. There is no point to look into all the scripts. The approach, clearly as it has been, has caused more problem than offering benefit! 

IT IS NOT A GOOD APPROACH TO USE $VARIABES FOR CELL/PIN ETC, SPECIALLY BEFORE ONE BECOMES AN EXPERIENCE tcl/LIBERATE USER.

Guangjun

Guangjun Cao said:

EXPERIENCE

experienced.

Guangjun Cao said:

EXPERIENCE

experienced.