A measure of which features of the program have been tested .

1 Why do you need functional coverage ？

If it's a directional test , Coverage in testing is implicit , The design specification lists 100 Features , What needs to be done is to complete 100 One test , If it's done 50 individual , Then the coverage is 50%.

For complex designs , Need to use a constrained random test method （CRT）, At this time, you do not need to enter the incentive line by line manually , But you need to write code according to the verification plan to track the effectiveness of the test . The displayed functional coverage measures the progress of the test .

2 How to collect coverage ？

By changing the random number seed , You can run the same random test platform repeatedly to generate new incentives . Each simulation will produce a database with coverage information , Record the trajectory of random walk . By combining all this information together, we can get the function coverage , To measure the overall progress .

3 How to converge coverage ？

1. Change the random number seed to obtain the function coverage data .
2. Analyze the coverage data and decide how to modify the test set .
   1. The coverage rate has increased steadily ： Add a new random seed and continue running the existing test , Or longer running time .
   2. The growth rate of coverage slowed down ： Adding additional constraints creates more incentives .
   3. Coverage is stable and some parts have not been tested ： Create more new tests .
   4. Close coverage 100% And keep finding mistakes ： Not really covering some areas of the design .

---

4 What types of coverage ？

• Code coverage
• Functional coverage
• Vulnerability rate
• Assertion coverage

4.1 Code coverage

4.1.1 Code coverage measures ？

Code coverage is a measure of how well a test can be applied to a code that involves a specification “ Realization ” How thorough the test is , Not for the validation plan . The code coverage reaches 100% It doesn't mean that the work has been completed .

4.1.2 What code coverage measures ？

Code coverage is the easiest way to measure validation progress . Measure... In this way ：

1. How many lines of code have been executed （ Line coverage ）.
2. In the path through the code and expression, what has been executed （ Path coverage ）.
3. Which orders bit Variable value is 0 or 1（ Flip coverage ）.
4. Which states and state transitions in the state machine have been accessed （ Finite state machine coverage ）.

4.1.3 How to test code coverage ？

The tool automatically analyzes the source code and adds hidden code to complete the statistics of code coverage .
When all tests are run , The code coverage tool will create the corresponding database .

4.2 Functional coverage

Functional coverage is closely related to the design intent , Sometimes called “ Specification coverage ”, Code coverage is the implementation of bright design .
If a block of code is missing from the design , Code coverage can't find this error , But the functional coverage can .

4.2.2 How to measure functional coverage ？

1. First, write the verification plan and the corresponding executable version for simulation .
2. stay SystemVerilog The values of variables and expressions are sampled in the test platform . Where these samples are taken is the coverage point .
3. At the same time （ For example, when a transaction is completed ） Multiple coverage points of are put together in a coverage group .

4.2.3 What is an overlay group ？

Coverage groups are similar to classes ： It can be instantiated many times after one definition .
Contains coverage points 、 Options 、 Formal parameters and optional triggers （trigger）. An overlay group contains one or more overlay points , Collect at the same time .

4.3 Vulnerability rate

An indirect way to measure coverage is to look at the rate of new vulnerabilities .

4.4 Assertion coverage

Assertions are used for one-time or For a while check Declarative code for the relationship between two design signals .

As opposed to using SystemVerilog Procedural code check , Assertion （SVA） Easier to express .

---

5 Strategy of function coverage ？

1. Collect information instead of data .
   FIFO Examples of verification , There is no need to measure the data in the read-write address index , There are many possibilities . If you can make FIFO From empty to full, then from full to empty , It has covered all situations in data reading and writing , Plus other interested States, such as full address index 1 And all 0 But between .
2. Only measure what will be used
3. Completeness of measurement