Skip to content

hwloc_rsmi_get_device_osdev() likely doesn't work for partitioned GPUs #774

Description

@bgoglin

ROCm and AMD SMI both expose fake PCI BDF (with the partition ID in the BDF func, not exactly as documented). Hence getting the PCI BDF in hwloc_rsmi_get_device_osdev() will cause us to fail to find the corresponding PCI device. We could set func=0 but we would only find the first GPU partition.

ROCm SMI support is going anyway, and the bug doesn't exist in AMD SMI hence it's doesn't matter much.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions