arXiv subjects

Keren Zhou

Publications and source records attributed to Keren Zhou.

13 recordsLinked to original sources

arXiv2026

Syncopate: Efficient Multi-GPU AI Kernels via Automatic Chunk-Centric Compute-Communication Overlap

Communication has become a first-order bottleneck in large-scale GPU workloads, and existing distributed compilers address it mainly by overlapping whole compute and communication kernels at the stream level. This coarse granularity incurs extra kernel launches, forces device-wide synchronizations at kernel boundaries, and leaves substantial slack when the slowest tile or kernel stretches the communication tail. We present Syncopate, a compiler and runtime that enables automatic fine-grained overlap inside a single fused kernel. Syncopate introduces a communication chunk abstraction that decouples communication granularity from kernel structure and backend mechanisms, allowing chunk-level plans to be ported from existing distributed compilers, written directly by users, or instantiated from reusable templates. Given a local Triton kernel and a chunk schedule, Syncopate performs transformations to align computation with chunk availability. Implemented as a source-to-source compiler on Triton, Syncopate delivers an average end-to-end speedup of 1.3$\times$ and up to 4.7$\times$ on multi-GPU workloads. Our code is open-sourced at https://github.com/tie-pilot-qxw/syncopate.

cs.DC↗

arXiv2026

Exploring Cross-Scenario Generality of Agentic Memory Systems: Diagnostics and a Strong Baseline

LLM agents accumulate histories that outgrow their context windows, motivating a growing literature on memory systems. Yet most existing designs are tuned to a single scenario (multi-session chat or a single trajectory format), and there is little evidence that they generalize across the heterogeneous trajectories agents encounter in deployment. We revisit eight memory systems plus an agentic harness for search problems, on five scenarios: single-turn QA, multi-session chat, agentic-trajectory QA, memory stress tests, and long-horizon agentic tasks. The harness, which self-manages flat text-file storage via tool calls, achieves the best cross-task ranking, suggesting that memory performance hinges on giving the agent active control over storage and retrieval rather than on a passive store behind a fixed pipeline. We instantiate this insight in AutoMEM, an agentic memory harness with a self-managed tool interface that achieves the best cross-scenario generality among the systems we evaluate.

cs.AI↗

arXiv2026

OpenRFM: Dissecting Relational In-Context Learning

Relational Foundation Models (RFMs) promise a single pre-trained predictor that, given any relational database, returns predictions in one forward pass via relational in-context learning (ICL). Yet a substantial gap separates open RFMs from their commercial counterparts, and the origin of this gap has not been systematically understood. We dissect a representative framework, the Relational Transformer (RT), from two perspectives. Model side: we show that RT performs relation-level ICL, and a kernel regression view shows it fails when sparse label-cell coverage yields an underdetermined regression. Data side: we ablate RT's pre-training source and find that existing synthetic-only pre-training and in-distribution pre-training drive the same architecture into different regimes, lazy vs. feature-learning. Probing this gap reveals that the missing ingredient is a support-identifiable relational latent in the label-generation process. These two diagnoses translate into (1) a dual-stage ICL architecture that combines the relational backbone with a batch-level ICL layer lifted from a pre-trained tabular foundation model to overcome relation-level label scarcity, and (2) a homophily-aware synthetic plus continual real-data pre-training mixture, augmented with a prototype-based regularization. These choices define OpenRFM, a simple yet effective RFM that improves average task performance by approximately 30% over the RT backbone and surpasses the commercial model KumoRFMv1 on a large set of evaluation tasks.

cs.LG↗

arXiv2026

Linear Layouts: Robust Code Generation of Efficient Tensor Computation Using $\mathbb{F}_2$

Efficient tensor computation is a cornerstone of modern deep learning (DL) workloads, yet existing approaches struggle to achieve flexible and performant design and implementation of tensor layouts -- mappings between logical tensors and hardware resources. The increasing complexity of DL algorithms and hardware demands a generic and systematic approach to handling tensor layouts. In this work, we introduce Linear Layouts, a novel approach that models tensor layouts using linear algebra over $\mathbb{F}_2$. By representing tensor layouts as binary matrices acting on the bits of the hardware representation, our approach enables a generic layout definition -- as opposed to the classical case-by-case approach -- and allows for generic layout-to-layout conversions, eliminating the quadratic explosion that plagues existing solutions. We integrate linear layouts with Triton and demonstrate their effectiveness in optimizing individual Triton operators as well as kernels written in Triton. We also show that linear layouts reduce engineering effort in the compiler backend while fixing several bugs in Triton's legacy layout system.

cs.PL↗

arXiv2026

PASTA: A Modular Program Analysis Tool Framework for Accelerators

The increasing complexity and diversity of hardware accelerators in modern computing systems demand flexible, low-overhead program analysis tools. We present PASTA, a low-overhead and modular Program AnalysiS Tool Framework for Accelerators. PASTA abstracts over low-level profiling APIs and diverse deep learning frameworks, offering users a unified interface to capture and analyze runtime events at multiple levels. Its extensible design enables researchers and practitioners to rapidly prototype custom tools with minimal overhead. We demonstrate the utility of PASTA by developing several analysis tools, including a deep learning workload characterization tool and a UVM optimization tool. Through extensive evaluation on mainstream deep learning workloads tested on NVIDIA and AMD GPUs under both single- and multi-GPU scenarios, we demonstrate PASTA's broad applicability. On NVIDIA GPUs, we further show that PASTA provides detailed performance insights with significantly lower overhead, up to 1.3*10^4 faster than conventional analysis tools, thanks to its GPU-accelerated backend. PASTA strikes a practical balance between usability, extensibility, and efficiency, making it well-suited for modern accelerator-based computing environments.

cs.DC↗

arXiv2025

KPerfIR: Towards an Open and Compiler-centric Ecosystem for GPU Kernel Performance Tooling on Modern AI Workloads

In this work, we propose KPerfIR, a novel multilevel compiler-centric infrastructure to enable the development of customizable, extendable, and portable profiling tools tailored for modern artificial intelligence (AI) workloads on modern GPUs. Our approach integrates profiling capabilities directly into the compiler workflow, allowing profiling functionalities to be implemented as compiler passes, offering a programmable and reusable framework for performance analysis. This design bridges the gap between compilers and profilers, enabling fine-grained insights into complex optimization challenges such as overlapping the execution of fine-grained function units on GPUs. KPerfIR is integrated into the Triton infrastructure to highlight the power of a compiler-centric approach to advance performance analysis and optimization in the ever-evolving landscape of AI compilers. Our evaluation shows that our tool incurs low overhead (8.2%), provides accurate measurements (2% relative error), and delivers actionable insights into complicated GPU intra-kernel optimizations.

cs.DC↗

arXiv2025

Do Large Language Models Understand Performance Optimization?

Large Language Models (LLMs) have emerged as powerful tools for software development tasks such as code completion, translation, and optimization. However, their ability to generate efficient and correct code, particularly in complex High-Performance Computing (HPC) contexts, has remained underexplored. To address this gap, this paper presents a comprehensive benchmark suite encompassing multiple critical HPC computational motifs to evaluate the performance of code optimized by state-of-the-art LLMs, including OpenAI o1, Claude-3.5, and Llama-3.2. In addition to analyzing basic computational kernels, we developed an agent system that integrates LLMs to assess their effectiveness in real HPC applications. Our evaluation focused on key criteria such as execution time, correctness, and understanding of HPC-specific concepts. We also compared the results with those achieved using traditional HPC optimization tools. Based on the findings, we recognized the strengths of LLMs in understanding human instructions and performing automated code transformations. However, we also identified significant limitations, including their tendency to generate incorrect code and their challenges in comprehending complex control and data flows in sophisticated HPC code.

cs.DC↗

arXiv2024

DeepContext: A Context-aware, Cross-platform, and Cross-framework Tool for Performance Profiling and Analysis of Deep Learning Workloads

Effective performance profiling and analysis are essential for optimizing training and inference of deep learning models, especially given the growing complexity of heterogeneous computing environments. However, existing tools often lack the capability to provide comprehensive program context information and performance optimization insights for sophisticated interactions between CPUs and GPUs. This paper introduces DeepContext, a novel profiler that links program contexts across high-level Python code, deep learning frameworks, underlying libraries written in C/C++, as well as device code executed on GPUs. DeepContext incorporates measurements of both coarse- and fine-grained performance metrics for major deep learning frameworks, such as PyTorch and JAX, and is compatible with GPUs from both Nvidia and AMD, as well as various CPU architectures, including x86 and ARM. In addition, DeepContext integrates a novel GUI that allows users to quickly identify hotpots and an innovative automated performance analyzer that suggests users with potential optimizations based on performance metrics and program context. Through detailed use cases, we demonstrate how DeepContext can help users identify and analyze performance issues to enable quick and effective optimization of deep learning workloads. We believe Deep Context is a valuable tool for users seeking to optimize complex deep learning workflows across multiple compute environments.

cs.PF↗

arXiv2022

Efficient model compression with Random Operation Access Specific Tile (ROAST) hashing

Advancements in deep learning are often associated with increasing model sizes. The model size dramatically affects the deployment cost and latency of deep models. For instance, models like BERT cannot be deployed on edge devices and mobiles due to their sheer size. As a result, most advances in Deep Learning are yet to reach the edge. Model compression has sought much-deserved attention in literature across natural language processing, vision, and recommendation domains. This paper proposes a model-agnostic, cache-friendly model compression approach: Random Operation Access Specific Tile (ROAST) hashing. ROAST collapses the parameters by clubbing them through a lightweight mapping. Notably, while clubbing these parameters, ROAST utilizes cache hierarchies by aligning the memory access pattern with the parameter access pattern. ROAST is up to $\sim 25 \times$ faster to train and $\sim 50 \times$ faster to infer than the popular parameter sharing method HashedNet. Additionally, ROAST introduces global weight sharing, which is empirically and theoretically superior to local weight sharing in HashedNet, and can be of independent interest in itself. With ROAST, we present the first compressed BERT, which is $100\times - 1000\times$ smaller but does not result in quality degradation. These compression levels on universal architecture like transformers are promising for the future of SOTA model deployment on resource-constrained devices like mobile and edge devices

cs.LG↗

arXiv2021

Measurement and Analysis of GPU-accelerated Applications with HPCToolkit

To address the challenge of performance analysis on the US DOE's forthcoming exascale supercomputers, Rice University has been extending its HPCToolkit performance tools to support measurement and analysis of GPU-accelerated applications. To help developers understand the performance of accelerated applications as a whole, HPCToolkit's measurement and analysis tools attribute metrics to calling contexts that span both CPUs and GPUs. To measure GPU-accelerated applications efficiently, HPCToolkit employs a novel wait-free data structure to coordinate monitoring and attribution of GPU performance. To help developers understand the performance of complex GPU code generated from high-level programming models, HPCToolkit constructs sophisticated approximations of call path profiles for GPU computations. To support fine-grained analysis and tuning, HPCToolkit uses PC sampling and instrumentation to measure and attribute GPU performance metrics to source lines, loops, and inlined code. To supplement fine-grained measurements, HPCToolkit can measure GPU kernel executions using hardware performance counters. To provide a view of how an execution evolves over time, HPCToolkit can collect, analyze, and visualize call path traces within and across nodes. Finally, on NVIDIA GPUs, HPCToolkit can derive and attribute a collection of useful performance metrics based on measurements using GPU PC samples. We illustrate HPCToolkit's new capabilities for analyzing GPU-accelerated applications with several codes developed as part of the Exascale Computing Project.

cs.DC↗

arXiv2020

GPA: A GPU Performance Advisor Based on Instruction Sampling

Developing efficient GPU kernels can be difficult because of the complexity of GPU architectures and programming models. Existing performance tools only provide coarse-grained suggestions at the kernel level, if any. In this paper, we describe GPA, a performance advisor for NVIDIA GPUs that suggests potential code optimization opportunities at a hierarchy of levels, including individual lines, loops, and functions. To relieve users of the burden of interpreting performance counters and analyzing bottlenecks, GPA uses data flow analysis to approximately attribute measured instruction stalls to their root causes and uses information about a program's structure and the GPU to match inefficiency patterns with suggestions for optimization. To quantify each suggestion's potential benefits, we developed PC sampling-based performance models to estimate its speedup. Our experiments with benchmarks and applications show that GPA provides an insightful report to guide performance optimization. Using GPA, we obtained speedups on a Volta V100 GPU ranging from 1.01$\times$ to 3.53$\times$, with a geometric mean of 1.22$\times$.

cs.PF↗

arXiv2016

Quadboost: A Scalable Concurrent Quadtree

Building concurrent spatial trees is more complicated than binary search trees since a space hierarchy should be preserved during modifications. We present a non-blocking quadtree-quadboost-that supports concurrent insert, remove, move, and contain operations. To increase its concurrency, we propose a decoupling approach that separates physical adjustment from logical removal within the remove operation. In addition, we design a continuous find mechanism to reduce its search cost. The move operation combines the searches for different keys together and modifies different positions with atomicity. The experimental results show that quadboost scales well on a multi-core system with 32 hardware threads. More than that, it outperforms existing concurrent trees in retrieving two-dimensional keys with up to 109% improvement when the number of threads is large. The move operation proved to perform better than the best-known algorithm, with up to 47%.

cs.DS↗

arXiv2015

Parse Concurrent Data Structures: BST as an Example

Designing concurrent data structures should follow some basic rules. By separating the algorithms into two phases, we present guidelines for scalable data structures, with a analysis model based on the Amadal's law. To the best of our knowledge, we are the first to formalize a practical model for measuring concurrent structures' speedup. We also build some edge-cutting BSTs following our principles, testing them under different workloads. The result provides compelling evidence to back the our guidelines, and shows that our theory is useful for reasoning the varied speedup.

cs.DC↗