Researchers from NVIDIA have introduced GPIR, a system that accelerates Private Information Retrieval (PIR) on GPUs by addressing the high server-side computation and memory traffic inherent in lattice-based protocols. The system employs a stage-aware hybrid execution model that dynamically switches between operation-level and stage-level kernels to maximize on-chip data reuse.

  • GPIR resolves performance mismatches in the RowSel phase using a transposed-layout design with fine-grained pipelining for large-scale matrix-matrix multiplications.
  • The architecture extends to multi-GPU systems, scaling throughput and database capacity with negligible communication overhead.
  • GPIR achieves up to 297.2x higher throughput compared to PIRonGPU, the previous state-of-the-art GPU implementation.

This approach enables practical private database queries by significantly reducing the computational bottlenecks that typically hinder PIR adoption.