Skip to content

[Question]: Repeated stream creation during initialization phase #77

Description

@kwu130

Question

In the nvshmemi_common_init function, nvshmemi_get_cucontext is called first. Toward the end of nvshmemi_get_cucontext, a CUDA stream is already created and assigned to state->my_stream. Immediately afterward—still within nvshmemi_common_init—the function nvshmemi_setup_stream_priorities is invoked, which creates another stream and overwrites state->my_stream.

Does this sequence result in redundant stream creation, thereby introducing a potential resource leak?

Metadata

Metadata

Assignees

No one assigned

    Labels

    questionFurther information is requested

    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