# CAMP: Context-Aware Mapping and Placement for LLMs on 3D-IMC Arrays

DevFeed: [CAMP: Context-Aware Mapping and Placement for LLMs on 3D-IMC Arrays](<https://devfeed.tech/articles/camp-context-aware-mapping-and-placement-for-llms-on-3d-imc-arrays-120237.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07707>)

Author: Orhun G\\"orkem, Athanasios Vasilopoulos, Georgios Pagonis, Victoria Clerico, Hadjer Benmeziane, Irem Boybat, Hsinyu Tsai, Abu Sebastian, William Andrew Simon

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [accelerators](<https://devfeed.tech/tags/accelerators.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [analog](<https://devfeed.tech/tags/analog.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [arrays](<https://devfeed.tech/tags/arrays.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [chips](<https://devfeed.tech/tags/chips.md>), [compute](<https://devfeed.tech/tags/compute.md>), [compute-resources](<https://devfeed.tech/tags/compute-resources.md>), [concurrency](<https://devfeed.tech/tags/concurrency.md>), [concurrently](<https://devfeed.tech/tags/concurrently.md>), [context](<https://devfeed.tech/tags/context.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [energy](<https://devfeed.tech/tags/energy.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [llms](<https://devfeed.tech/tags/llms.md>), [mapping](<https://devfeed.tech/tags/mapping.md>)

## AI overview

CAMP improves LLM inference throughput on 3D in-memory computing arrays by choosing weight placements that account for which matrix-vector multiplications may contend for resources. It builds a conflict matrix from the chip architecture, target network, and runtime context, then uses profiling feedback to prune the mapping search and refine placements. The method runs offline in software and requires no hardware changes. The abstract reports compute utilization rising from 42.4% to 60.3%, up to 2.0x higher 3D-IMC throughput, and 1.74x higher end-to-end throughput than state-of-the-art mapping algorithms.

## Source excerpt

arXiv:2611.07707v1 Announce Type: cross Abstract: Large Language Models (LLMs) incur severe memory-access overheads due to the cost of loading large weight matrices and key-value (KV) caches. Analog In-Memory Computing (AIMC) is an emerging solution that enables computation directly on static weights stored in memory, substantially reducing energy and latency. 3D-IMC further stacks memory arrays into vertical tiers, increasing capacity but revealing a new challenge: how weights are placed in memory determines which matrix-vector multiplications (MVMs) can execute concurrently, affecting latency and throughput. The concurrency aspect of 3D-IMC weight-mapping remains underexplored in the literature, leading to sub-optimal utilization of compute resources on 3D-IMC chips. In this context we propose CAMP (Context-Aware Mapping and Placement), an inference-context-aware weight mapping algorithm for 3D-IMC-based LLM accelerators. CAMP captures the potential resource contention resulting from the architecture, targeted network, and runtime context in a conflict matrix, enabling throughput-optimized weight mapping. This mapping is refined through a profiling-driven feedback loop that efficiently prunes the search space and returns optimized mappings within practical runtimes. CAMP is an offline, software-based solution requiring no hardware changes and improving 3D-IMC compute utilization from 42.4% to 60.3%, delivering up to 2.0x higher 3D-IMC throughput and 1.74x higher end-to-end throughput over state-of-the-art mapping algorithms.