[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85011-en":3,"doc-seo-85011-105":29,"detail-sidebar-cat-0-en-105":83},{"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85011,13056703019404,"Miles","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","From Custom-Fit to Portable: Bridging the Gap Between Synthesized and Engineered GPU Query Execution","GPUs are increasingly used for analytical query processing, yet building GPU database engines that fully exploit hardware performance demands heavy engineering. Recent work proposes synthesizing query execution via LLM-generated, query- and machine-specific code, promising large speedups. This study revisits the synthesis-versus-engineering debate for GPU analytics by evaluating synthesized CUDA/HIP kernels, explaining the performance sources, and transferring generalizable optimizations into a performance-portable SYCL engine to narrow the gap.","From Custom-Fit to Portable: Bridging the Gap Between Synthesized and Engineered GPU Query Execution  \nIvan Donchev Kabadzhov  \nEURECOMBiot, France [ivan.kabadzhov@eurecom.fr](ivan.kabadzhov@eurecom.fr)  \nEugenio Marinelli EURECOMBiot, France  \n[eugenio.marinelli@eurecom.fr](eugenio.marinelli@eurecom.fr)  \nRaja Appuswamy EURECOMBiot, France  \n[raja.appuswamy@eurecom.fr](raja.appuswamy@eurecom.fr)  \narXiv :2607 .07632v 1 [ cs .DB] 8 Jul 2026  \nAbstract  \nGPUs are increasingly used for analytical query processing, but developing GPU-based database engines that achieve the peak performance of the underlying hardware requires substantial research and engineering effort. A recent line of work argues that query processing should be synthesized, not engineered. In this scenario, instead of tuning a general-purpose engine to fit a workload, a large language model (LLM) generates code specialized to one query, one dataset, and one machine, thereby achieving an order-of-magnitude improvement in performance. This thesis, however, has so far been tested only on CPUs. In this work, we revisit the synthesize-versusengineer debate for GPU analytics by answering three questions:(i) how good is synthesized GPU code?, (ii) why is it faster than engineered engines?, and (iii) how much of its advantage can be transferred back into a single, performance-portable engine?  \nTo answer the first question, we present SHADB, an LLM-based synthesis framework that generates optimized CUDA or HIP kernels using an automated, profile-guided optimization loop. Using SHADB, we show that the synthesized code approaches the memory-bandwidth ceiling and outperforms a state-of-the-art JITcompiled GPU database engine (HeavyDB) by 7.4× on SSB SF100 . To answer the second question, we decompose this performance gap and systematically classify optimizations as generalizable or workload-specific. Finally, to answer the third question, we integrate these generalizable optimizationsintoSYCLDB, a performanceportable engine written entirely in the open SYCL programming model. Using optimized SYCLDB, we show that it is possible to substantially bridge the gap to synthesized code (within 1.27× total execution time) while retaining workload-level generality and hardware-level performance portability.  \nArtifact Availability Notice  \nThe implementation and evaluation details for our synthesis loop are maintained at our repository workspace ([https://gitlab.eurecom](https://gitlab.eurecom). fr/dislab/SHADB) .  \n1 Introduction  \nThe past few years have witnessed a growth in adoption of GPUs for analytical workloads [8, 26, 38, 41] . However, the design of GPU databases is challenging due to the unique characteristics of GPU architectures, where fully exploiting the massive SIMT parallelism of GPUs requires attention to memory coalescing, occupancy, kernel-launch overhead, and inter-operator dataflow. Complicating this is the fact that GPUs from different vendors have different architectures, programming models, and performance characteristics, making it difficult to design a general-purpose GPU database engine that performs well across all workloads and hardware.  \nA recent line of work argues that query processing should be synthesized, not engineered [24, 42]. Instead of executing a query on a general-purpose engine whose operators are written once and tuned to be acceptable across every workload, these systems use large language models (LLMs) to generate code specialized to one query, one dataset, and one machine. For instance, GenDB synthesizes instance-optimized execution code through a chain of dedicated agents [24]. Bespoke OLAP [42] generates workloadspecific “one-size-fits-one” engines that outperform DuckDB [33] by more than 10× [42] . The common thesis is that the abstraction overhead a general engine pays is large enough that synthesizing query-specific code can provide substantial throughput/latency improvement especially under workloads with repetitive queries [37]","cbCaioN7kH75JPLx","https://ap.wps.com/l/cbCaioN7kH75JPLx","pdf",606620,1,13,"English","en",105,"# Abstract\n# Artifact Availability Notice\n# Introduction","[{\"question\":\"How is the advantage of synthesized code transferred to a general-purpose GPU engine?\",\"answer\":\"The study classifies optimizations into generalizable versus workload-specific, integrates generalizable optimizations into SYCLDB, and shows the portable engine can substantially narrow the runtime gap while keeping hardware-level performance portability.\"}]",1784200241,33,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":78,"head_meta":80,"extra_data":82,"updated_unix":27},"from-custom-fit-to-portable-bridging-the-gap-between-synthesized-and-engineered-gpu-query-execution","",{"@graph":35,"@context":77},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/from-custom-fit-to-portable-bridging-the-gap-between-synthesized-and-engineered-gpu-query-execution/85011/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"How is the advantage of synthesized code transferred to a general-purpose GPU engine?","Question",{"text":75,"@type":76},"The study classifies optimizations into generalizable versus workload-specific, integrates generalizable optimizations into SYCLDB, and shows the portable engine can substantially narrow the runtime gap while keeping hardware-level performance portability.","Answer","https://schema.org",{"og:url":51,"og:type":79,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":81,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,107,112,115,120,123,127],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":45,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":45,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":45,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":98,"slug":130},19,"General","general"]