Papers for

machine learning system builders

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Gpu utilization for large language model inference breaks down efficiency factors

Dissecting GPU Utilization for LLM Inference on Nvidia Hopper

Abstract: A single SM utilization percentage can make an LLM inference workload look compute-saturated while hiding how much useful work is being done. The problem is not that the counter is wrong, but that it collapses several different mechanisms into one number. This is most severe during decode, where each request contributes only one new token and dense projection GEMMs become small-row matrix multiplications. On Hopper, the bfloat16 GMMA path executes these operations in fixed 64-row matrix fragments, so small-batch decode can fill only a small fraction of each fragment with real token rows. In this paper, we profile vLLM with FlashAttention-3 and cuBLASLt on an H100 NVL across cold prefill, warm prefill, and decode, sweeping sequence length and batch size. We replace the usual single utilization number with eight counter-validated views derived from raw Nsight Compute reports, each pinned to an NCU counter or explicit formula. Together, these views map utilization gaps to concrete mechanisms - fragment fill, occupancy limits, stall signatures, wave quantization, and kernel selection - across four production models and six per-layer kernel roles.

Fri 11 SeptPerformanceHardware ArchitectureDistributed, Parallel, and Cluster Computing
The gist
Measuring how busy a GPU is during large language model (LLM) use can be misleading because a single percentage doesn't show all the details. The authors explain that during tasks like generating text (decoding), GPU operations handle small pieces inefficiently, leading to underused GPU power. They studied Nvidia's Hopper GPU running models with different settings and used detailed counters to explain why the GPU isn't fully efficient. This helps reveal what exactly limits GPU performance during these tasks.
Open 2609.12923v1

Digital communication codes functions directly into wireless signals

The Computing Channel: How Modulation Programs the Airwaves

Abstract: Distributed computing and distributed artificial intelligence require frequent exchanges of intermediate results, although many applications need only an aggregate rather than messages from individual devices. Conventional systems recover each message before computing the aggregate, whereas over-the-air computation (OAC) exploits simultaneous transmission to obtain it directly. However, dominant OAC implementations rely on analog signaling, creating a mismatch with finite-precision data and digital communication procedures. This article presents digital function-oriented communication, in which finite-alphabet symbol representations and receiver decisions are jointly designed so that multiple-access superposition encodes the desired function without recovering individual inputs. We introduce its computational-constellation principle, main design approaches, extensions, and implementation challenges. Federated edge learning illustrates how the framework can reduce user-dependent data-bearing resources while operating directly on quantized model updates.

Thu 10 SeptDistributed, Parallel, and Cluster Computing
The gist
Sending messages between many computers or devices usually means sending each message separately, which is slow. The authors point out a new way to send data by letting devices transmit at the same time so the receiver can directly get answers to questions like sums or averages, without reading everything first. Their method works with digital signals, which are easier to handle in computers, by designing symbols and decoding rules together. This could help in applications like teaching AI models across many devices more efficiently.
Open 2609.11145v1