📰 Dev.to · Ingero Team

A trace of a hybrid Mamba-Transformer MoE inference run, broken down by layer type. The MoE...

On a shared GPU host, the kernel knows which container generated which event. The trace stays...

An LLM-written kernel benchmarked 38% faster on a microbench. Here is what kernel-level validation...

TL;DR A GPU that reports 97% utilization can still be the slowest part of a training step,...

A slow AllReduce on rank 5 lines up against TCP retransmits on rank 5’s NIC, four ms before the...

Three eBPF patterns hyperscalers run in production today, mapped to the equivalent patterns on the...

For an air-gapped GPU host, the trace is only useful if collection, storage, and query all happen...

Per-host overhead multiplied across N hosts, vs. one kernel-level instrumentation per host. The...

libhip.so is to ROCm what libcudart.so is to CUDA: the user-mode runtime API the framework calls...

DCGM stops at host-level GPU counters. Kernel-side eBPF adds the per-rank, per-tenant signals...

MCP exposes the agent’s tool calls. eBPF exposes the kernel events that explain why those tool...

An MCP tool call is a tiny line of agent code that fans out to syscalls, library calls, and kernel...

Eight ranks on two hosts. Every per-host metric reads healthy. Rank 5 enters the barrier 290ms...

TL;DR Ingero Fleet v0.10 FOSS is live. We validated the full pipeline end-to-end on two...

TL;DR: PyTorch's DataLoader can be 50-124x slower than direct tensor indexing for in-memory GPU...

TL;DR: A .cpu().numpy() call buried inside a forward pass was forcing a full CPU-GPU synchronization...