[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83813-en":3,"doc-seo-83813-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},83813,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","Eiger An Efficient Library for GPU based Data Analytics","GPUs are increasingly used to accelerate analytical workloads through massive parallelism and high memory bandwidth, and recent findings show that with fast CPU–GPU interconnects and networks, bottlenecks shift toward query processing on the GPU rather than data movement. The widely used cuDF library limits performance by committing to statically chosen operator implementations and using little runtime data insight. This paper introduces Eiger, a GPU analytics library that improves single-GPU query processing via runtime workload adaptivity.","Eiger: An Efficient Library for GPU-based Data Analytics  \nBowen Wu  \nSystems Group, ETH Zurich Switzerland [bowen.wu@inf.ethz.ch](bowen.wu@inf.ethz.ch)  \nMarko Kabić  \nSystems Group, ETH Zurich Switzerland [marko.kabic@inf.ethz.ch](marko.kabic@inf.ethz.ch)  \nSven Hepkema  \nSystems Group, ETH Zurich Switzerland [sven.hepkema@inf.ethz.ch](sven.hepkema@inf.ethz.ch)  \nVasilis Mageirakos  \nSystems Group, ETH Zurich Switzerland [vmageirakos@inf.ethz.ch](vmageirakos@inf.ethz.ch)  \nChristos Kozyrakis  \nNVIDIA & Stanford University United States [ckozyrakis@nvidia.com](ckozyrakis@nvidia.com)  \nGustavo Alonso  \nSystems Group, ETH Zurich Switzerland [alonso@inf.ethz.ch](alonso@inf.ethz.ch)  \narXiv :2607 .04489v 1 [ cs .DB] 5 Jul 2026  \nABSTRACT  \nGPUs have become an increasingly attractive platform for accelerating analytical workloads due to their massive parallelism and high memory bandwidth. Recent studies show that in systems with fast CPU-GPU interconnects and fast networks, query processing within the GPU, rather than data movement, is the dominant bottleneck. This highlights the need for a more efficient implementation of relational operators on GPUs than the widely used library, cuDF. While offering rich functionality, cuDF commits to a single, statically chosen implementation for most operators and makes little use of runtime information about the data, limiting performance across diverse workloads and different GPUs. In this paper, we present Eiger, a high-performance library for GPU-based data analytics that improves single-GPU query processing through runtime workload adaptivity. Adaptivity in Eiger rests on two design principles. First, Eiger provides multiple implementation variants and tunable knobs for most operators, covering not only joins and group-bys but also expensive yet often overlooked operations, such as expression evaluation, string processing, and multi-key sorting, for which it contributes an array of new optimization techniques. Second, Eiger profiles intermediate data during query execution using lightweight statistics, such as value ranges and HyperLogLog++ sketches, and uses them to select implementations, tune configuration knobs, and compress data on the fly, thereby overcoming the limitations of traditional static query optimization. The breadth of operators and variants also enables a more comprehensive performance analysis, encompassing a wider range of operations and workloads than previous work. We evaluate Eiger with operator microbenchmarkson two GPU architectures and with the complete TPC-H benchmark (up to scale factor 100) . Across the 22 queries, Eiger reduces the total runtime by up to 1.8× compared to the state-of-the-art cuDF library; for individual queries, Eiger achieves up to 6.1× better performance.  \n1 INTRODUCTION  \nRecent years have seen rapid growth in the use of graphics processing units (GPUs) to accelerate database applications [12, 54, 57] . GPUs, with their massive parallelism and high-bandwidth memory, are well-suited for many, if not all, operations in data analytics. Recently, driven by large language model training and inference, GPUs have been deployed at large scales and are equipped with increasingly higher memory bandwidth, larger memory capacity, faster CPU-GPU interconnects, and high-bandwidth RDMA networks.  \nAll of these enable GPU-based databases to process data at one to two orders of magnitude faster than CPU-based databases [54, 57] .  \nRecent performance studies by Kabić et al. [19] and Wu et al. [54] evaluate GPU-based database systems on two different deployments: (1) relations initially stored in CPU DRAM requiring data transfer to the GPU, and (2) relations pre-partitioned and loaded into high-bandwidth memory (HBM) of multiple GPUs in a cluster. Across both settings using the TPC-H benchmark, they find that query processing time on the GPU dominates overall execution time when fast interconnects (e.g., NVLink-C2C) and RDMA networks (e.g., InfiniBand) are","cbCaisiJS7LBLfC6","https://ap.wps.com/l/cbCaisiJS7LBLfC6","pdf",1270292,4,1,15,"English","en",105,"# Introduction\n## GPU acceleration trends\n## Bottleneck shift to GPU query processing\n## Limits of existing GPU libraries","[{\"question\":\"Why does query processing on GPUs become the dominant bottleneck in modern systems?\",\"answer\":\"When systems have fast CPU–GPU interconnects and high-performance networks, performance studies show the time spent executing queries on the GPU outweighs data movement costs.\"},{\"question\":\"What performance limitations does cuDF have for GPU analytics workloads?\",\"answer\":\"cuDF offers a single statically chosen implementation for most operators and uses limited runtime information about intermediate data, reducing efficiency across diverse workloads and GPU configurations.\"},{\"question\":\"How does Eiger achieve runtime workload adaptivity?\",\"answer\":\"Eiger combines multiple operator implementation variants and tunable configuration knobs with lightweight profiling of intermediate data (e.g., value ranges and HyperLogLog++ sketches) to select implementations, tune settings, and compress data during execution.\"}]",1784190581,38,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"eiger-an-efficient-library-for-gpu-based-data-analytics","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/eiger-an-efficient-library-for-gpu-based-data-analytics/83813/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does query processing on GPUs become the dominant bottleneck in modern systems?","Question",{"text":75,"@type":76},"When systems have fast CPU–GPU interconnects and high-performance networks, performance studies show the time spent executing queries on the GPU outweighs data movement costs.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What performance limitations does cuDF have for GPU analytics workloads?",{"text":80,"@type":76},"cuDF offers a single statically chosen implementation for most operators and uses limited runtime information about intermediate data, reducing efficiency across diverse workloads and GPU configurations.",{"name":82,"@type":73,"acceptedAnswer":83},"How does Eiger achieve runtime workload adaptivity?",{"text":84,"@type":76},"Eiger combines multiple operator implementation variants and tunable configuration knobs with lightweight profiling of intermediate data (e.g., value ranges and HyperLogLog++ sketches) to select implementations, tune settings, and compress data during execution.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]