TY - GEN
T1 - Metamorphic Testing for Processor Verification: A RISC-V Case Study at the Instruction Level
AU - Riese, Frank
AU - Herdt, Vladimir
AU - Große, Daniel
AU - Drechsler, Rolf
PY - 2021
Y1 - 2021
N2 - Metamorphic Testing (MT) has been shown to be a very effective technique in the Software (SW) domain. MT does
not require a reference model to compare against for testing
but instead relies on Metamorphic Relations (MR) to derive the expected result from relationships between several calls to thefunction under test. An example of an MR is the expectation that the sum of an arbitrary list of integers remain unchanged regardless of it being sorted or reversed. Thus, a key requirement for applying MT effectively is availability of MRs specific to the domain at hand.
In this paper, we propose MT to the domain of processor
verification. As a case study, we consider the RISC-V Instruction Set Architecture (ISA) and provide MRs tailored for RISC-V.
For evaluation purposes, we propose an efficient on-the-fly MT framework that integrates the MRs with an Instruction Set
Simulator (ISS). We measure the quality of those MRs by the
number of mutations they kill, also referred to as mutation
analysis. Our experiments demonstrate the effectiveness of the MRs to kill all mutations, which confirms our research question that MT is also a suitable technique for the domain of processor verification.
AB - Metamorphic Testing (MT) has been shown to be a very effective technique in the Software (SW) domain. MT does
not require a reference model to compare against for testing
but instead relies on Metamorphic Relations (MR) to derive the expected result from relationships between several calls to thefunction under test. An example of an MR is the expectation that the sum of an arbitrary list of integers remain unchanged regardless of it being sorted or reversed. Thus, a key requirement for applying MT effectively is availability of MRs specific to the domain at hand.
In this paper, we propose MT to the domain of processor
verification. As a case study, we consider the RISC-V Instruction Set Architecture (ISA) and provide MRs tailored for RISC-V.
For evaluation purposes, we propose an efficient on-the-fly MT framework that integrates the MRs with an Instruction Set
Simulator (ISS). We measure the quality of those MRs by the
number of mutations they kill, also referred to as mutation
analysis. Our experiments demonstrate the effectiveness of the MRs to kill all mutations, which confirms our research question that MT is also a suitable technique for the domain of processor verification.
UR - https://ieeexplore.ieee.org/document/9606997
UR - https://www.scopus.com/pages/publications/85122910238
U2 - 10.1109/VLSI-SoC53125.2021.9606997
DO - 10.1109/VLSI-SoC53125.2021.9606997
M3 - Conference proceedings
T3 - IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC
BT - International Conference on Very Large Scale Integration (VLSI-SoC)
ER -