Overview
“Verifying Instruction Set Simulators using Coverage-guided Fuzzing” is cited as a DATE 2019 paper by V. Herdt, D. Große, H. M. Le, and R. Drechsler, appearing on pages 360–365. [publication]
The provided evidence excerpts mainly cover the paper’s future-work and conclusion material. They identify the work as applying coverage-guided fuzzing in the context of instruction set simulator verification, and discuss several directions for extending the approach. [publication]
Future-work themes
The paper discusses improving functional coverage around output operands. In particular, it notes that ideas for covering output-operand values might help maximize the functional V(RD) metric. [vrd]
It also points to machine-learning techniques for fuzzing as a promising direction to investigate in this application area. [ml-fuzzing]
Coverage metrics
The paper identifies stronger coverage metrics as an important direction for future work. It states that path coverage and cross-coverage of functional metrics can be effective, but are often challenging or impractical because of the large feature state space. [stronger-coverage]
As a scalability-oriented alternative, the authors propose considering selective path coverage and selective functional cross-coverage. These would be applied only to selected code regions, such as considering evaluation paths for each instruction separately in the ISS rather than across instructions, together with input operand values. The stated goal is to improve verification while maintaining scalability. [selective-coverage]
Broader scope
The paper also proposes broadening evaluation to additional architectures and instruction sets. Another stated direction is to apply the coverage-guided fuzzing approach to analysis of the whole platform rather than limiting analysis to the ISS component. [broader-scope]
Relationship to later work
The later paper Efficient Cross-Level Testing for Processor Verification: A RISC-V Case-Study cites “Verifying instruction set simulators using coverage-guided fuzzing” in its references. [cross-level-citation]