Back to portfolio

SR-IOV & GIM Virtualization

Extended the AMD RAS tool to support SR-IOV and GIM (GPU-IOV Module) virtualization use cases — simulating hardware error injection and validating kernel driver compatibility under multi-tenant GPU deployments.

Key Contributions

Virtualization Error Injection

Enhanced RAS tool to simulate hardware error cases and validate AMDGPU driver behavior when multiple VMs share a physical GPU via SR-IOV.

Multi-VM Test Scenarios

Developed test features supporting server-class infrastructure with multiple VMs running ROCm workloads concurrently.

CVS Team Collaboration

Worked with Cloud Virtualization Support (CVS) team to extend testability for GIM driver testing and enterprise cluster deployments.

Kernel Driver Validation

Ensured RAS tooling remained compatible with virtualized ROCm environments and Linux kernel SR-IOV code paths.