Mutation Cover with Yosys

MCY is a new tool to help digital designers and project managers understand and improve testbench coverage.

If you have a testbench, and it fails, you know you have a problem. But if it passes, you know nothing if you don’t know what your testbench is actually testing for.

Given a self checking testbench, MCY generates 1000s of mutations by modifying individual signals in a post synthesis netlist. These mutations are then filtered using Formal Verification techniques, keeping only those that can cause an important change in the design’s output.

All mutated designs are run against the testbench to check that the testbench will detect and fail for a relevant mutation. The testbench can then be improved to get 100% complete coverage.

mcy gui can be used to explore the results and find where coverage needs improvement.

mcy dash provides at a glance overview as the tool runs. It is browser-based to easily take advantage of cloud compute.

For more information please contact contact@yosyshq.com

Getting Started

Installing Tabby CAD Suite or OSS CAD Suite

MCY is part of the Tabby CAD Suite and the OSS CAD Suite! The easiest way to use MCY is to install the binary software suite, which contains all required dependencies.

Contact YosysHQ for a Tabby CAD Suite Evaluation License and download link
OR go to https://github.com/YosysHQ/oss-cad-suite-build/releases to download the free OSS CAD Suite
Follow the Install Instructions on GitHub

Make sure to get a Tabby CAD Suite Evaluation License if you need features such as industry-grade SystemVerilog and VHDL parsers!

For more information about the difference between Tabby CAD Suite and the OSS CAD Suite, please visit https://www.yosyshq.com/tabby-cad-datasheet.

Installing from Source

Install SBY and its dependencies first.

Install MCY:

sudo make install

Install requirements:

pip install click flask

Run an example project (may run for a while):

cd examples/bitcnt
mcy init
mcy run -j8

The examples additionally require Icarus Verilog (https://github.com/steveicarus/iverilog) to run the testbench under test.

Documentation

The full MCY documentation can be found at https://mcy.readthedocs.io/.

Short Tutorial

Create a new directory with a config.mcy file. For example:

[options]
size 100

[script]
read_verilog verilog/picorv32.v
prep -top picorv32

[files]
picorv32.v

[logic]
if result("sim_simple") == "FAIL":
    tag("COVERED")
    return

if result("eq_sim3 10") == "FAIL":
    tag("UNCOVERED")
    return

if result("eq_bmc") == "FAIL":
    tag("UNCOVERED")
    return

if result("eq_sim3 500") == "FAIL":
    tag("UNCOVERED")
    return

tag("NOC")

[report]
if tags("!NOC"):
    print("Coverage: %.2f%%" % (100.0*tags("COVERED")/tags("!NOC")))

[test sim_simple]
maxbatchsize 10
expect PASS FAIL
run bash $PRJDIR/scripts/sim_simple.sh

[test eq_bmc]
maxbatchsize 10
expect PASS FAIL
run bash $PRJDIR/scripts/eq_bmc.sh

[test eq_sim3]
maxbatchsize 10
expect TIMEOUT FAIL
run bash $PRJDIR/scripts/eq_sim3.sh

Create the referenced files and run mcy init.

This will create a database/ directory, run the Yosys script and create database/design.il, create size mutations and store them in database/db.sqlite3.

Running mcy reset re-runs the [logic] section for all mutations, imports additional mutations if the value of the size option has been increased, resets the "running" status of all tests, removes the tasks/ directory, and prints some statistics.

Running mcy status prints the statistics from mcy reset but will not make any changes to the database.

Running mcy list will list all mutations and their current tags.

Running mcy list --details will print more details for each mutation.

Running mcy run -j8 will run the mutation cover analysis, with a maximum of 8 concurently running jobs.

Running mcy task [-v] [-k] <test> <id_or_tag>.. will (re-)run the specified test with the given mutations. This is mostly used during development of test scripts. -v displays the command and its output on the console, and with -k the task directory tasks/<task_id> will not be removed when the task completes.

Running mcy source <filename> prints the original HDL source, annotated with coverage information. (This is hardcoded to use the tags COVERED and UNCOVERED.) This is using the copy of the specified source file that was placed in the database on mcy init. (This requires a [files] corresponding entry in the config file.) Use mcy source <attr_filename> <fs_filename> to use a source file from the file system instead.

Running mcy dash launches a web-based dashboard that can be used to launch jobs and monitor progress.

Running mcy gui launches a GUI application that can be used to investigate the database.

Finally, mcy purge removes the database/ and tasks/ directories.

Logic

The [logic] section contains python code that defines how MCY should run the individual tests. The following special functions are available:

tag(tagname): Mark the current mutation with the specified tag

result(test_with_args): Run the specified test and return the result, or use a cached result if the test had been run previously.

rng(N): Return a deterministic pseudo-random integer in the range 0..N-1.

Report

The [report] section contains python code for printing a report based on the data collect via the tagging mechanism described above.

tags() returns the total number of mutations that have any tag assigned to them.

tags("FOO") returns the number of mutations that have the tag FOO assigned.

tags("!FOO") returns the number of mutations that have at least one tag assigned, but not the FOO tag.

Tests

Each test has its own configuration file section, [test testname].

If maxbatchsize is set to a value > 1 then MCY will automatically create batches of up to the specified number of mutations, and then pass them all to a single invocation of the test. (The default maxbatchsize value is 1.)

Running the test means running the command specified with run, with any additional arguments to the test appended to the command.

The command is run in the tasks/<taskid> subdirectory, with the environment variable TASK set to the task id, TASKDIR set to the directory, and PRJDIR set to the project directory.

The environment variable KEEPDIR will be set to 1 if the task is run with mcy task -k, and 0 otherwise. This can be used for example to only create VCD traces if the task directory is kept.

The command is then given an input in input.txt, formatted as following:

1 <mutation_1>
2 <mutation_2>
3 <mutation_3>
...

And it will procude an output in output.txt, formatted as following:

1 <result_1>
2 <result_2>
3 <result_3>
...

The numbers are not mutation IDs but a task-local numbering, starting fresh at 1 for each task. The order of the output lines does not matter.

The environment variable MUTATIONS contains the list of mutation IDs for the task. This is mostly useful for debug purposes. Ideally a test should not need to know the mutation IDs of the mutations it tests.

Standard output of the command will be redirected to logfile.txt, unless the command is run from mcy task -v. Standard error always stays connected to the console.

A test with expect setting will cause a runtime error if the test returns anything else than one of the expected values.

YosysHQ/mcy