Skip to content
STIMSMITH

Interruption and Exception Simulation

Technique WIKI v1 · 5/27/2026

Interruption and Exception Simulation is a CPU RTL fuzzing technique used in Instiller to exercise realistic CPU interruption and exception handling. The technique adds exception simulation, supports multiple simultaneous interruptions and exceptions, and accounts for their priorities to better cover CPU verification corner cases.

Overview

Interruption and Exception Simulation is a technique for CPU RTL fuzzing that simulates realistic CPU interruption and exception handling during test execution. In the Instiller work, interruptions and exceptions are treated as common CPU execution events whose simulation can help cover corner cases in CPU verification. [C1]

Motivation

The evidence identifies interruption and exception handling as a realism and coverage challenge for CPU fuzzing. Prior fuzzing work had considered interruptions only in a relatively simple way: exceptions were not simulated, and multiple interruptions and exceptions with their priorities were not included. According to the Instiller paper, omitting these situations prevents fuzzing from simulating real-world CPU execution and from covering all CPU states. [C2]

Technique

The technique described in Instiller has three main elements:

  1. Exception inclusion: exceptions are added to CPU fuzzing, which the paper states had not been proposed in the referenced prior work. [C3]
  2. Multiple events: more than one interruption and exception can be integrated into a test scenario, with the goal of simulating real-world CPU execution more comprehensively. [C4]
  3. Priority handling: priorities among different interruptions and exceptions are considered, which the paper says can fuzz the CPU more thoroughly. [C5]

Together, these elements are presented as a way to better simulate real-world interruption and exception handling than prior approaches. [C6]

Use in Instiller

Instiller is the prototype fuzzer that uses this technique. The paper lists support for multiple interruptions and exceptions, including their priorities, as one of Instiller's contributions, and states that these mechanisms can simulate realistic interruption and exception handling well. [C7]

Reported evaluation context

The cited Instiller evaluation reports overall tool-level improvements, including a 29.4% coverage increase and 17.0% more mismatches found in the targets. These results are reported for Instiller as a whole rather than isolated solely to interruption and exception simulation. [C8]

LINKED ENTITIES

1 links

CITATIONS

8 sources
8 citations
[1] C1: Interruptions and exceptions are common in CPU execution, and simulating them in CPU testing can cover CPU verification corner cases. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[2] C2: Prior fuzzing work handled interruptions relatively simply and did not include exceptions, multiple interruptions and exceptions, or their priorities, limiting realistic CPU execution simulation and CPU-state coverage. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[3] C3: Instiller includes exceptions in CPU fuzzing, which the paper states was not proposed in the referenced prior work. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[4] C4: Instiller integrates more than one interruption and exception to simulate real-world CPU execution more comprehensively. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[5] C5: Instiller considers the priorities of different interruptions and exceptions to fuzz the CPU more thoroughly. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[6] C6: Including exceptions, multiple interruptions and exceptions, and their priorities is presented as better simulation of real-world interruption and exception handling than prior work. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[7] C7: Instiller enables its fuzzer to handle multiple interruptions and exceptions and considers their priorities as a stated contribution. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing
[8] C8: The Instiller evaluation reports a 29.4% coverage increase and 17.0% more mismatches found in the targets for the tool overall. [2401.15967] Instiller: Towards Efficient and Realistic RTL Fuzzing