Research map / Architectures and efficiency / RNNs and normalization

CHoRD: Coordinating Scheduling and Data Placement for Efficient Deep Neural Network Inference on Chiplet-Based GPUs

NeurIPS 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes CHoRD, a method that coordinates kernel scheduling and data placement to improve DNN inference efficiency on chiplet-based GPUs with NUMA memory.

Problem

Chiplet-based GPUs introduce hierarchical NUMA memory access, so DNN inference performance depends on alignment between kernel execution and data placement across chiplets.

Approach

Jointly coordinates scheduling of DNN kernels and placement of their data across chiplets, rather than treating them independently, to reduce NUMA penalties during inference.

Closest papers by meaning

• Practical Performance Guarantees for Pipelined DNN Inference (ICML 2024 · similarity 0.85)
• Integrated Hardware Architecture and Device Placement Search (ICML 2024 · similarity 0.85)
• Poplar: Efficient Scaling of Distributed DNN Training on Heterogeneous GPU Clusters (AAAI 2025 · similarity 0.85)
• Mozart: Modularized and Efficient MoE Training on 3.5D Wafer-Scale Chiplet Architectures (NeurIPS 2025 · similarity 0.85)
• FlashRNN: I/O-Aware Optimization of Traditional RNNs on modern hardware (ICLR 2025 · similarity 0.85)
• SpeedLoader: An I/O efficient scheme for heterogeneous and distributed LLM operation (NeurIPS 2024 · similarity 0.84)