How chathuranus/gpu-rowhammer works
This is a research simulator project evaluating RowHammer defenses for shared-bank GDDR6 GPU memory. The main workflow generates synthetic request traces, runs them through a Ramulator-based experiment harness under selected mitigation policies, and analyzes the resulting safety, timing, fairness, and performance metrics. Separate suites cover baseline comparisons, hypotheses, and constrained-hardware defenses. The README documents DBIM-RR and other major defense capabilities, but the simulator fork’s controller implementation is not present in the supplied tree, so its internals are represented only at the documented level.
Main entry points: results Experiment Results · run_experiment.py Experiment Harness · run_hypothesis_suite.py Hypothesis Suite · run_minor_hardware_defenses.py Hardware Defense Suite
Components 18
Synthetic Workloads
- Trace Generators
scripts - Hypothesis Traces
scripts/generate_hypothesis_traces.py - Hardware Stress Traces
scripts/generate_hardware_stress_traces.py - Request Traces
traces
Simulation
- Experiment Harness
configs/run_experiment.py - Ramulator Simulator
- DBIM-RR Defense
Evaluation Suites
- Defense Suite
scripts/run_defense_suite.py - Hypothesis Suite
scripts/run_hypothesis_suite.py - Hardware Defense Suite
scripts/run_minor_hardware_defenses.py - Approach Comparison
scripts/run_approach_comparison.py
Results Analysis
- Experiment Results
results - Defense Analysis
scripts/analyze_defense_suite.py - Hypothesis Analysis
scripts/analyze_hypotheses.py - Hardware Analysis
scripts/analyze_minor_hardware_defenses.py - Comparison Summary
scripts/summarize_approach_comparison.py - Result Demonstration
scripts/show_supervisor_demo.py
Other components
- Researcher
Connections 24
- Researcher → Defense Suite: runs evaluation
scripts/run_defense_suite.py - Researcher → Hypothesis Suite: runs evaluation
scripts/run_hypothesis_suite.py - Researcher → Hardware Defense Suite: runs evaluation
scripts/run_minor_hardware_defenses.py - Researcher → Approach Comparison: runs evaluation
scripts/run_approach_comparison.py - Trace Generators → Hypothesis Traces: includes
scripts/generate_hypothesis_traces.py - Trace Generators → Hardware Stress Traces: includes
scripts/generate_hardware_stress_traces.py - Hypothesis Suite → Hypothesis Traces: generates
scripts/run_hypothesis_suite.py - Hardware Defense Suite → Hardware Stress Traces: generates
scripts/run_minor_hardware_defenses.py - Trace Generators → Request Traces: writes traces
scripts/generate_hypothesis_traces.py - Defense Suite → Experiment Harness: launches runs
scripts/run_defense_suite.py - Hypothesis Suite → Experiment Harness: launches runs
scripts/run_hypothesis_suite.py - Hardware Defense Suite → Experiment Harness: launches runs
scripts/run_minor_hardware_defenses.py - Experiment Harness → Request Traces: reads traces
configs/run_experiment.py - Experiment Harness → Ramulator Simulator: runs simulation
configs/run_experiment.py - Experiment Harness → DBIM-RR Defense: configures defense
README.md - Ramulator Simulator → Experiment Results: produces metrics
- Defense Suite → Experiment Results: records results
scripts/run_defense_suite.py - Hypothesis Suite → Experiment Results: records results
scripts/run_hypothesis_suite.py - Hardware Defense Suite → Experiment Results: records results
scripts/run_minor_hardware_defenses.py - Experiment Results → Defense Analysis: supplies metrics
scripts/analyze_defense_suite.py - Experiment Results → Hypothesis Analysis: supplies metrics
scripts/analyze_hypotheses.py - Experiment Results → Hardware Analysis: supplies metrics
scripts/analyze_minor_hardware_defenses.py - Experiment Results → Comparison Summary: supplies metrics
scripts/summarize_approach_comparison.py - Experiment Results → Result Demonstration: supplies summary
scripts/show_supervisor_demo.py