Query processing on tensor computation runtimes

The huge demand for computation in artificial intelligence (AI) is driving unparalleled investments hardware and software systems AI. This leads to an explosion the number of specialized devices, which are now offered by major cloud vendors. By hiding low-level complexity through a tensor-based interface, tensor runtimes (TCRs) such as PyTorch allow data scientists efficiently exploit exciting capabilities new hardware. In this paper, we explore how database management can ride wave innovation happening AI space. We design, build, evaluate Tensor Query Processor (TQP): TQP transforms SQL queries into programs executes them on TCRs. able run full TPC-H benchmark implementing novel algorithms relational operators routines. At same time, support various while only requiring fraction usual development effort. Experiments show that improve query execution time up 10$\times$ over CPU- GPU-only systems. Finally, accelerate mixing ML predictions end-to-end, deliver 9$\times$ speedup CPU baselines.