doca-gpunetio-ib-write-lat

DOCA GPUNetIO ib_write_lat

Examples

• "What GPU-init RDMA-WRITE latency can GPUNetIO deliver?" — this is in scope. Capture median and p99; do not treat GPI as a shipped latency binary.
• "GPUNetIO vs GPI — which latency number should I trust?" — same physical operation, different runtime; route GPI semantics to the GPI skill.

Instructions

Follow Loading order below. Open the tasks companion for configure / run / debug. Open the capabilities companion for pairing and observability. Do not open every companion file.

Where to start: This is a tool skill for the GPUNetIO- flavored ib_write_lat benchmark shipped under doca/tools/gpunetio_ib_write_lat/ (a client + server pair, built from source against the installed DOCA via meson). It measures the latency of an RDMA WRITE work request when the WR is posted from a CUDA kernel through the doca-gpunetio device-side surface, in a ping-pong cadence. Open the tasks companion and start at ## configure for the GPU-NIC pairing precondition and the build pattern; jump to ## run for the single-iteration smoke flow. Open the capabilities companion when the question is what this tool actually measures, how it differs from the GPI sister tool on the same physical operation, or how to interpret the half-iter / full-iter / CUDA-side usec output and the median / p99 / jitter characterization. If DOCA is not installed yet, route to doca-setup/SKILL.md first; if the user is still deciding between GPUNetIO and GPI as a programming surface, the picture in doca-gpunetio/SKILL.md and doca-gpi/SKILL.md is the first stop.

Example questions this skill answers well

The CLASSES of doca-gpunetio-ib-write-lat questions this skill is built to answer, each with one worked example. The class is the load-bearing piece; the worked example is one instance.

• "What GPU-init RDMA-WRITE latency / jitter can the GPUNetIO path deliver for a real-time / control-loop workload?" — worked example: "measure per-iteration WRITE latency between two hosts with an H100 + ConnectX-7 on each side, target the median and the p99 separately". Answered by the GPU-NIC pairing precondition in the capabilities companion
  • the bring-up flow in the tasks companion + the tasks companion.
• "This is the GPUNetIO tool — how does the latency number differ from the GPI programming surface?" — worked example: "the team is using GPI; should I expect GPUNetIO to beat / tie / lose vs GPI?". Answered by the "same physical operation, different runtime framework" rule in the capabilities companion
  • the cross-link to the GPI library skill doca-gpi/SKILL.md (note: doca/tools/ ships no GPI ib_write_lat benchmark binary — GPI is a programming surface, not a shipped benchmark tool).
• "Median vs p99 vs jitter — which one is the actual answer for a real-time control loop?" — worked example: "my control loop has a deadline; the median is well under the budget but p99 spikes; do I quote the median or the p99?". Answered by the median-vs-p99-vs-jitter rule in the capabilities companion
  • the eval-loop overlay in the tasks companion.
• "What is the latency-vs-batching trade-off specific to GPU-init RDMA?" — worked example: "my CUDA kernel could batch multiple WRs to amortize the GPU-side overhead; what does that buy me on latency vs what does it cost me?". Answered by the latency-vs-batching trade-off in the capabilities companion.
• "What version of DOCA + CUDA Toolkit do I need for this binary to build and run?" — worked example: "my install has DOCA at one semver and CUDA at another; will the ToT-shipped gpunetio_ib_write_lat even link?". Answered by the version overlay in the capabilities companion.
• "How do I read the half-iter / full-iter / CUDA-side usec columns?" — worked example: "the binary printed half-iter, full-iter, and a CUDA-side number — what is the right column to quote for one-way latency vs round-trip vs cross-check?". Answered by the column- semantics rule in the capabilities companion.