Skip to content

Nvshmem configure nic pe mapping vs Nvshmem tune performance

Nvshmem configure nic pe mapping is an open-source agent skill for developer tools. Nvshmem tune performance is an open-source agent skill for developer tools.

  • Nvshmem configure nic pe mapping is best for pick the closest infiniband port for each nvshmem pe on a multi-gpu node.
  • Nvshmem tune performance is best for collect a baseline for a two-node nvshmem all-to-all before tuning.
What it is
Nvshmem configure nic pe mapping is an open-source agent skill for developer tools.
Nvshmem tune performance is an open-source agent skill for developer tools.
Best for
Pick the closest InfiniBand port for each NVSHMEM PE on a multi-GPU node
Collect a baseline for a two-node NVSHMEM all-to-all before tuning
Kind
Skill
Skill
Pricing
Open source
Open source
Platforms
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
GitHub stars
4K
4K
Last commit
Nov 2026
Nov 2026
License
Apache-2.0
Apache-2.0
Open source
Good for
Pick the closest InfiniBand port for each NVSHMEM PE on a multi-GPU node; Interpret nvidia-smi topo output alongside RDMA port state for NVSHMEM; Generate NVSHMEM_HCA_PE_MAPPING exports for a known GPU binding
Collect a baseline for a two-node NVSHMEM all-to-all before tuning; Choose an NVSHMEM remote transport for an InfiniBand cluster; Map NVSHMEM PEs to the correct NICs on a multi-GPU node

What Nvshmem configure nic pe mapping does

  • Target-node topology collection

    A read-only bash collector gathers NVSHMEM version, GPU PCI IDs, GPU-to-NIC topology and RDMA port state.

  • Closest-port ranking

    Candidate HCA ports are ranked by nvidia-smi topology distance from each bound GPU, with load balancing on ties.

  • Link layer filtering

    Inactive ports are excluded and InfiniBand and Ethernet/RoCE ports are kept in separate partitions by default.

  • Version-aware mapping rules

    Mapping variables and feature gates differ between NVSHMEM 3.7 and earlier and 3.8 or later, as set out in a reference matrix.

  • Explicit export blocks

    Recommendations use one selection mechanism and produce NIC-assignment environment exports only once inputs are resolved.

  • Conflict and ambiguity checks

    The agent asks for clarification instead of silently choosing when version, transport or PE binding evidence is missing or conflicting.

What Nvshmem tune performance does

  • Workflow routing

    Maps a performance request to one of four NVSHMEM specialist skills based on its concrete objective.

  • Underspecified request handling

    Presents baseline and targeted optimization options and asks for an explicit choice when intent is general.

  • Multi-stage sequencing

    Orders collection, transport selection and NIC-to-PE mapping so earlier results feed later steps.

  • Constraint preservation

    Carries transport settings, topology facts and tuning limits through each handoff and reports conflicts.

  • Missing specialist fallback

    Names any unavailable specialist skill and asks the user to enable it or pick another route.

Nvshmem configure nic pe mapping vs Nvshmem tune performance: questions

Is Nvshmem configure nic pe mapping or Nvshmem tune performance cheaper?

Both are open source. Compare current plans on each official site.

What is the difference between Nvshmem configure nic pe mapping and Nvshmem tune performance?

Nvshmem configure nic pe mapping is an open-source agent skill for developer tools. Nvshmem tune performance is an open-source agent skill for developer tools.

Keep comparing