Table of Contents
Due to Covered's method of abstracting coverage from a dumpfile, a partial "resimulation" of the design is needed to obtain proper statistics for line, combinational logic and FSM metrics. Therefore, accurately resimulating what occurred in the actual simulation is a requirement for obtaining correct results. This can be easily achieved so long as the design doesn't contain any code that could result in potential race conditions.
To avoid this problem Covered performs automatic race condition checking on the specified design after parsing has been completed but before simulation/scoring is performed. All statement blocks in the design that don't adhere to certain coding guidelines are removed from coverage consideration by Covered. By following these coding guidelines, a design should be void of race condition scenarios that would lead to faulty simulation results. The following coding guidelines are applied by Covered when checking for race conditions:
All sequential logic and latches use non-blocking assignments.
All combinational logic in an always block uses blocking assignments.
All mixed sequential and combinational logic in the same always block uses non-blocking assignments.
Blocking and non-blocking assignments should not be used in the same always block.
Assignments made to a variable should only be done within one always block.
The $strobe system call should only be used to display variables that were assigned using non-blocking assignments. (Not currently implemented)
No #0 procedural assignments should exist. (Not currently implemented)
If Covered detects a block as violating any one of the above mentioned coding guidelines, it will do one of two things depending on options specified to the score command by the user.
If the -rS
, -rW
or none of the -r options are specified, Covered will simply remove the non-conforming always block from
coverage consideration and continue with scoring. -rW
will
display a message describing detected race conditions during the scoring command while the
-rS
option will display no messages.
If the -rE
option is specified, Covered will halt the score
command after all race condition checking has been performed with a message to the user specifying how many
always blocks were found to be in non-compliance to the race-condition-free coding guidelines.
The first mode of operation is the recommended mode of operation as this will allow Covered to continue calculating coverage information for the design while still providing accurate coverage information for the logic that is still under consideration. The second mode of operation is meant to allow the user to use Covered as a race-condition checking tool.
Covered has three means of avoiding race condition checking: (1) skip race condition checking for the entire design altogether, (2) skip race condition checking for one or more specified modules, or (3) place inline pragmas around specific logic blocks to avoid checking. The main advantage for using either of these methods is to allow Covered to score more of the design that would otherwise be excluded due to failing a static race condition checkpoint. The main disadvantage to avoiding race condition checking using either of these methods is that you could potentially be ignoring code that contains an actual race condition, causing Covered to potentially generate incorrect coverage information without warning. These types of "errors" are not assumed to be Covered bugs.
To exclude the entire design from race condition checking, simply specify the
-rI
option on the score command-line. This will cause Covered to skip
the race condition checking phase entirely, outputting a warning message (unless the global -Q option has been
specified). The use of this command-line option is highly discouraged due to its general assumption by the user that
the entire design is void of race conditions.
To exclude an entire module from race condition checking, simply specify the
-rI=
module name option on the score command-line.
This option may be specified one or more times on the score command-line (once for each module to be skipped).
To exclude portions of the design from race condition checking, the user may also place inline comment-based
pragmas in the design and specify the -rP
option to the score
command-line. The inline pragmas look like the following:
Example 4.1. Using a Race Condition Pragma
module foobar; ... // racecheck off always @(posedge clock) begin ... end // racecheck on ... endmodule
This example will cause the initial block surrounded by the racecheck
pragmas to be ignored from race condition
checking, thus making it available for coverage consideration. If there are any race condition rules that this block
contains or is a part of, they will not be reported.
The keyword racecheck
is the default pragma keyword for race condition checking avoidance; however,
if this name needs to be changed to something else by the user (i.e., if racecheck
is used by another
tool and has a different meaning or format), the user may make this keyword anything they please as long as the meaning is
preserved (i.e., keyword off
means that race condition checking should be turned
off for all subsequent code until its matching keyword on
pragma is seen). The
new name of this keyword is passed to the -rP option by specifying
-rP=keyword
. It should also be noted that race
condition pragmas can be embedded within each other (all code within the outside pair are removed from race
checking) and race condition avoidance pragmas (racecheck off
) are in effect on a file-basis -- that is,
they are in effect even beyond their current scope but are no longer in effect after the current file.
Even though race condition information is displayed (assuming the -rS option was not specified in the score command-line) during the scoring process, it may be convenient to view this information in a generated report as well. Because of this need, Covered saves all race condition information to the CDD file for use in using the report command. By specifying 'r' in the -m option to the report command (race condition report output is not turned on by default), the statement blocks which were eliminated from coverage consideration will be output to the coverage report file. Summary coverage will contain the total number of statement blocks eliminated for each module. Verbose coverage will contain each eliminated statement block, organized by module, specifying both the starting line of the eliminated statement block and the reason for why the statement block was removed.