Machine-learning-based design-for-test (DFT) recommendation system for improving automatic test pattern generation (ATPG) quality of results (QoR)
A first set of features may be extracted from a first integrated circuit (IC) design. A trained machine learning (ML) model may predict a set of ranked test-case configurations for the first IC design based on the first set of features. A test-case configuration may correspond to a count of scan chain input and output ports and a scan chain length value.
1. A method comprising:
extracting a first set of features from a first integrated circuit (IC) design;
predicting, by a trained machine learning (ML) model which executes on a processor, a set of ranked test-case configurations for the first IC design based on the first set of features, wherein each test-case configuration corresponds to a count of scan chain input and output ports and a scan chain length value;
selecting a first test-case configuration from the set of ranked test-case configurations; and
inserting scan chains in the first IC design based on the first test-case configuration to obtain an updated first IC design.
2. The method of claim 1 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts an increase in a count of test cycles required for achieving a target test coverage relative to an optimal test-case configuration.
3. The method of claim 2 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts a confidence level corresponding to the increase in the count of test cycles required for achieving the target test coverage relative to the optimal test-case configuration.
4. The method of claim 1 , further comprising:
collecting training data based on a set of test-case configurations for each IC design in a set of IC designs, the training data comprising:
a set of features extracted from each IC design, and
a count of test cycles required for achieving a target test coverage for each test-case configuration; and
training the ML model using the training data to obtain the trained ML model.
5. The method of claim 4 , further comprising generating the set of test-case configurations for each IC design, wherein each test-case configuration includes a different combination of a count of scan chain input and output ports and a scan chain length value.
6. The method of claim 1 , wherein the trained ML model is a tree-based boosting regression model.
7. The method of claim 1 , wherein the first set of features extracted from the first IC design comprises one or more of: (1) a count of flip-flops, (2) a count of clock domains, (3) a count of faults, (4) a count of primitive gates, (5) static test coverage, and (6) random test coverage.
8. A non-transitory computer-readable storage medium comprising stored instructions, which when executed by a processor, cause the processor to:
extract a first set of features from a first integrated circuit (IC) design;
predict a set of ranked test-case configurations for the first IC design based on the first set of features using a trained machine learning (ML) model, wherein each test-case configuration corresponds to a count of scan chain input and output ports and a scan chain length value; and
select a first test-case configuration from the set of ranked test-case configurations; and
insert scan chains in the first IC design based on the first test-case configuration to obtain an updated first IC design.
9. The non-transitory computer-readable storage medium of claim 8 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts an increase in a count of test cycles required for achieving a target test coverage relative to an optimal test-case configuration.
10. The non-transitory computer-readable storage medium of claim 9 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts a confidence level corresponding to the increase in the count of test cycles required for achieving the target test coverage relative to the optimal test-case configuration.
11. The non-transitory computer-readable storage medium of claim 8 , further comprising stored instructions, which when executed by the processor, cause the processor to:
collect training data based on a set of test-case configurations for each IC design in a set of IC designs, the training data comprising:
a set of features extracted from each IC design, and
a count of test cycles required for achieving a target test coverage for each test-case configuration; and
train an ML model using the training data to obtain the trained ML model.
12. The non-transitory computer-readable storage medium of claim 11 , further comprising stored instructions, which when executed by the processor, cause the processor to generate the set of test-case configurations for each IC design, wherein each test-case configuration includes a different combination of a count of scan chain input and output ports and a scan chain length value.
13. The non-transitory computer-readable storage medium of claim 8 , wherein the trained ML model is a tree-based boosting regression model.
14. The non-transitory computer-readable storage medium of claim 8 , wherein the first set of features extracted from the first IC design comprises one or more of: (1) a count of flip-flops, (2) a count of clock domains, (3) a count of faults, (4) a count of primitive gates, (5) static test coverage, and (6) random test coverage.
15. A system comprising:
a memory storing instructions; and
a processor, coupled with the memory and to execute the instructions, the instructions when executed cause the processor to:
extract a first set of features from a first integrated circuit (IC) design, wherein the first set of features extracted from the first IC design comprises one or more of: (1) a count of flip-flops, (2) a count of clock domains, (3) a count of faults, (4) a count of primitive gates, (5) static test coverage, and (6) random test coverage;
predict a set of ranked test-case configurations for the first IC design based on the first set of features using a trained machine learning (ML) model, wherein each test-case configuration corresponds to a count of scan chain input and output ports and a scan chain length value; and
select a first test-case configuration from the set of ranked test-case configurations; and
insert scan chains in the first IC design based on the first test-case configuration to obtain an updated first IC design.
16. The system of claim 15 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts an increase in a count of test cycles required for achieving a target test coverage relative to an optimal test-case configuration.
17. The system of claim 16 , wherein for each suboptimal test-case configuration in the set of ranked test-case configurations, the trained ML model predicts a confidence level corresponding to the increase in the count of test cycles required for achieving the target test coverage relative to the optimal test-case configuration.
18. The system of claim 15 , wherein the instructions when executed cause the processor to:
collect training data based on a set of test-case configurations for each IC design in a set of IC designs, the training data comprising:
a set of features extracted from each IC design, and
a count of test cycles required for achieving a target test coverage for each test-case configuration; and
train an ML model using the training data to obtain the trained ML model.
19. The system of claim 18 , wherein the instructions when executed cause the processor to generate the set of test-case configurations for each IC design, wherein each test-case configuration includes a different combination of a count of scan chain input and output ports and a scan chain length value.
20. The system of claim 15 , wherein the trained ML model is a tree-based boosting regression model.