<ul data-eligibleForWebStory="true"><li>Hyperdimensional Computing (HDC) is a computing paradigm using high-dimensional hypervectors.</li><li>Recent HDC methods focus on iterative training for improved accuracy, accelerated on GPUs.</li><li>Efficient HDC inference has mostly been on specialized hardware, not multi-core CPUs.</li><li>ScalableHD is proposed for high-throughput HDC inference on multi-core CPUs.</li><li>ScalableHD uses a two-stage pipelined execution model parallelized across cores.</li><li>Intermediate results are streamed between stages to enhance cache locality.</li><li>Features like memory tiling and NUMA-aware worker-to-core binding are integrated for performance.</li><li>ScalableHD has variants for small and large batch sizes to exploit compute parallelism.</li><li>It achieves up to 10x speedup over TorchHD, maintaining accuracy for tasks like image classification.</li><li>ScalableHD shows robust scalability with throughput improvements as cores increase.</li></ul>

ScalableHD: Scalable and High-Throughput Hyperdimensional Computing Inference on Multi-Core CPUs

Discover more