From custom-fit to portable: Bridging the gap between synthesized and engineered GPU query execution

GPUs 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 Raja Appuswamy EURECOM Biot, France raja.appuswamy@eurecom.fr Arecent 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 and engineering effort. A recent line of work argues that query processing should be synthesized, not engineered. In this scenario, arXiv:2607.07632v1  [cs.DB]  8 Jul 2026 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? To 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 optimizations into SYCLDB,aperformanceportable 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.