ALL RIGHTS RESERVED. PAGE 2 Understanding Key points Mapping in LEC
Purpose When performing formal verification using the Conformal Equivalence Checker (EC), mapping plays a key role. This document studies the various aspects of mapping, including the types of mapping, and the available options that cover the various scenarios of mapping. This application note also gives a brief picture regarding handling of not mapped key points.
Audience This document is intended for conformal users who are looking for better picture of mapping in terms of mapping of key points, mapping types, handling of not mapped points and categories of not mapped points along with usage of available mapping options.
ALL RIGHTS RESERVED. PAGE 4 Understanding Key points Mapping in LEC
What is key point mapping?
Key point mapping pairs Golden key points to their corresponding Revised key points.
You cannot pair different types of key points. For example, a DFF cannot be mapped to DLAT or BBOX. You can map a Golden DFF key point ff_reg1 to its counterpoint ff_reg1 in the Revised design.
Golden Revised PI PI PO PO DFF DFF DLAT DLAT BBOX BBOX CUT CUT Z Z E E Corresponding key points are mapped
ALL RIGHTS RESERVED. PAGE 5 Understanding Key points Mapping in LEC
Why is mapping necessary?
Mapping is necessary because it identifies the corresponding logic cones
between the Golden and Revised designs.
Corresponding logic cones are compared
Mapping is done so that the tool knows which Golden combinational logic cone to compare to which Revised combinational logic cone. When mapping is complete, the two designs are ready for comparison. The comparison is done between two corresponding key points (called the compare points) in the Golden and Revised designs for the combinational logic feeding these compare points. If the logic cones between the two compare points are not the same, the tool returns a non-equivalent point (NEQ).
ALL RIGHTS RESERVED. PAGE 6 Understanding Key points Mapping in LEC
Types of Mapping There are two types of mapping: name based and function based.
Name-Based Mapping In name-based mapping, the tool maps key points based on their name— looking for key points with the same name in both the Golden and Revised designs. Name-based mapping is the preferred mapping approach because, as compared to function-based mapping, it takes less time and the chances of an incorrect mapping are negligible.
In addition, the ANALYZE SETUP command tends to be more successful
with name-based mapping versus function-based mapping. Also, the name effort (set in the SET MAPPING METHOD command) is high by default and can take care of most changes in delimiters.
After name-based mapping is performed, the left over key points will have different names between the Golden and Revised designs. There are several ways you can make their names similar in both designs:
Naming rules
Naming rules will be applied in RTL and facilitates in name based mapping with netlist. Naming rules should be applied before reading designs and libraries.
For instance, naming rules are highly useful in case of generate blocks.
ALL RIGHTS RESERVED. PAGE 7 Understanding Key points Mapping in LEC
When for-generate statement is used in VHDL and/or Verilog, the naming
convention of the resulting instances depend on the synthesis tool as well as the name setting.
For example, consider the following piece of code:
{{{ generate begin: blkA for (i=0;i<=0;i=i+1) begin: forblkB or n1(a,b,c); for (j=23;j<=23;j=j+1) begin: forblkC and n2(d,e,f); end begin : blkD nor n3 (g,h,i); end8ifdf
end end endgenerate }}}
• Using {{{set naming rule "%s" "%s_%d" "%s" -instance}}} (the default Conformal setting) would rename the instances as: n1, n2 and n3 as n1_0, n2_0_23, and n3_0 • Using {{{set naming rule "%s" "%L[%d].%s" "%s_INS" -instance}}} would rename the instances as: n1, n2 and n3 as forblkB[0].n1_INS,forblkB[0].forblkC[23].n2_INS, and forblkB[0].n3_INS
• Using {{{set naming rule "%L_%s" "%L[%d].%s" "%s" -instance}}} would
rename the instances as:n1, n2 and n3 as blkA_forblkB[0].n1, blkA_forblkB[0].forblkC[23].n2, and blkA_forblkB[0].blkD_n3
• #(expr) – expr evaluates to a constant integer • %s- matches 1 or more alpha-numeric characters • Refer to the Conformal reference manual for more renaming structures.
Renaming rule for the above example will be as stated below:
ALL RIGHTS RESERVED. PAGE 9 Understanding Key points Mapping in LEC
many name changes, it is not practical to use regular expressions with the "add renaming rule" command or function-based mapping.
In such case one can use the change name report generated from the synthesis tool in order to retain the original names. In Synopsys DC, you can report the renamed signals using the following command:
There is a new switch associated with change_names, namely, -
log_changes. The format is as follows:
change_names -rules vendor_rules -hier -verbose \
-log_changes name.change.rpt
NOTE: Make sure that the Correct header format is intact.
Using the change name command
=============================
Execute the following LEC command right after you read in the design:
SETUP> change name name_changes.rpt -revised > /tmp/dump
This will eliminate screen dump and reduce significantly the amount of time needed for the name change operation.
This command forces the netlist with the modified names back to the original names implicitly or permanently. Retaining the original names aids in the debugging process.
The renaming will apply to cell names and net names for the purpose of mapping, and for port names for the purpose of optimal hierarchical verification.
ALL RIGHTS RESERVED. PAGE 10 Understanding Key points Mapping in LEC
The LEC "change name" command will also take in multiple modified names that have been applied in several steps through DC. So make sure to save the change name report at each step in DC. Since in Conformal LEC, we are trying to retrieve the original names, you must make sure that the "change name" commands are applied in the reverse sequence of the commands that were executed in DC.
Example ======= dc_shell> report_names -hierarchy > TI_core_rules.rpt dc_shell> change_names dc_shell> report_names -hierarchy > TI_module_rules.rpt dc_shell> change_names In Conformal LEC, you will use the following order for the "change name" commands: SETUP> change name TI_module_rules.rpt -revised SETUP> change name TI_core_rules.rpt -revised
Using the ADD MAPPED POINTS command
When the number of key points having different names do not have common pattern and are less in number then it is good to manually map those key points using add mapped points command.
ALL RIGHTS RESERVED. PAGE 11 Understanding Key points Mapping in LEC
So in the above case one can manually map the key points in the following manner:
add mapped points 465718 424316 –noinvert
For more information on above command refer to conformal reference
guide.
Using the UNIQUIFY command
Uniquify command can take care of some basic renaming rules when used in hierarchical compare. Hence it is recommended to use uniquify command as it aids in name based mapping.
Function-Based Mapping Function based mapping uses heuristic methods to map key points on the basis of logic cone functionality. The tool uses a built-in algorithm to map key points that have different names in the Golden and Revised designs.
Conformal EC cannot map PIs and POs functionally. If PIs and POs have different names in the Golden and Revised design, they will turn to extra key points. Extra key points are points that exist only on one side (Golden or Revised). In most cases, extra key points do not affect the circuit. However, extra PIs and POs are a cause for concern and should be mapped. To map PIs and POs with different names, make their names similar using renaming rules and then use name based mapping.
ALL RIGHTS RESERVED. PAGE 12 Understanding Key points Mapping in LEC
Abc0/X[1] D Q PO
Golden In this example, the names of the DFF
are different. The tool will first map the POs being driven by these flops, then it will back trace and will undergo the function based mapping of the DFFs Xyz0/X[1] having different names.
D Q PO
Revised
Since function-based mapping can produce incorrect mapping of key
points, it’s best to map most of the key points using name-based mapping. Mapping run time is also higher in the case of function-based mapping.
Available Mapping Method Options
There are various mapping options available which can be used in different scenarios as stated below.
• set mapping method –name first
This is the default method used by Conformal to map key points. Key points having same path names in both Golden and Revised are mapped first, then the remaining key points are mapped by function.
• set mapping method -name only
With this method, the tool maps only the key points with matching names in the Golden and Revised designs; the remaining key points are
ALL RIGHTS RESERVED. PAGE 13 Understanding Key points Mapping in LEC
not mapped. These notmapped points can be reported using command
report unmapped points > notmapped and accordingly apply renaming rules to map them.
This approach is generally used if function-based mapping generates
incorrect mappings results or takes long time to map the key points.
• set mapping method –noname
With this method, tool will not map the key points based on the path of the gates. This mapping method is used when the names do not yield good correlation (custom circuits, implementation tool completely renames instances), name-based mapping will not work, and functional based mapping is needed. Mapping run time can be long with this method due to the function based mapping.
• set mapping method –nobbox_name_match
By dhhhefault, Conformal checks both module names and instance
names to ensure only blackboxes with the same module and instance name are paired. So, even though a set of blackboxes have the same instance names, Conformal will not map them if their module names differ. This method maps blackboxes that have different module names, but the same instance names.
• set mapping method –sensitive
By default, Conformal is case insensitive with regards to key point
names. For example, by default, the following key points would not be mapped:
Use this mapping method so that Conformal is case sensitive.
• set mapping method –phase
This method maps the key point with an inverted phase. One way to evaluate if a pair of key points should be phase mapped is to consider the simulation result for the key point. If the simulation result at the end of the Golden or Revised cones is always inverted, they should be phase mapped. Phase mapping is recommended when the synthesized netlist has gone through sequential inversion or inverter push.
In this case, the DFF should be mapped using set mapping method -phase. Phase mapping can require a lot of CPU time. One strategy is to do non-phase mapping first, then—after the first comparison—delete the NEQ points from the mapped list and remap with phase-mapping turned on. So in above case DFF will be inverted equivalent after comparison.
By default, Conformal first maps key points by name followed by
function. To determine the current mapping method, use the following command:
report environment -mapping
which displays a report similar to the following:
Mapping method: Name usage : name first Inverted mapping : not used Case sensitivity : OFF Name effort : HIGH
ALL RIGHTS RESERVED. PAGE 15 Understanding Key points Mapping in LEC
Unmapped Key Point Categories
Extra: Extra key points are the points that exist only on one side of the design (Golden or Revised). Extra key points generally do not affect the functionality of the design; extra PIs and Pos, however, are matter of concern and should be mapped.
To view a list of extra key points, use the following command in LEC mode:
report unmapped points -extra
Unreachable: Unreachable points are key points that do not
propagate to any output or to any other key points (flops, latches).
For instance, a DLAT is marked as unreachable after gated clock modeling
because the signal controlling the latch was moved to the MUX control; therefore, the latch is no longer needed. In this case, Conformal reports it as an unreachable key point.
To view a list of unreachable key points, use the following command in LEC mode:
report unmapped points –unreachable
Not-mapped: Not-mapped key points are reachable, but do not
have corresponding point in the other design. For complete mapping, there should not be any not-mapped points. Not mapped points may be resolved with renaming rules if corresponding points exist in the other design under different names. Otherwise, not-mapped points, in particular DFF/DLAT, need to be remodeled away.
To view a list of not-mapped key points, use the following command in LEC mode:
ALL RIGHTS RESERVED. PAGE 16 Understanding Key points Mapping in LEC
Recommendations • Do not use set mapping method –unreach as a default command in production/generic scripts as it may lead to false NEQs.
• Add set mapping method -auto as a default command in generic
scripts as it can take care of commonly encountered setup issues, such as sequential constants, sequential merge, loop cutting, phase mapping, and clock gating. Better modeling helps in mapping.
• If mapping takes too long, use set mapping method –name only and then use renaming rules to map the remaining not-mapped points.
• If there are noneqs other than phase mapping then deleting the mapped points contributing to noneqs and remapping them does not work. In other words this technique only works f noneqs are due to phase inversion.
Voltus-Fi Custom Power Integrity Solution - Electromigration and IR Drop at The Transistor Level - Industry Insights - Cadence Blogs - Cadence Community PDF
Voltus-Fi Custom Power Integrity Solution - Electromigration and IR Drop at The Transistor Level - Industry Insights - Cadence Blogs - Cadence Community PDF
