We present robust high-performance implementations of signal-processing tasks performed by a high-throughput wildlife tracking system called ATLAS. The system tracks radio transmitters attached to wild animals by estimating the time of arrival of packets encoding known pseudo-random codes to receivers (base stations). Time-of-arrival estimation of wideband radio signals is computatoinally expensive, especially when it is not known when a transmitter transmits. These computation are a key bottleneck that limits the throughput of the system. The paper reports on two implementations of ATLAS's signal-processing algorithms, one for CPUs and the other for GPUs, and carefully evaluates their performance. The evaluations, performed on two CPU platforms and on three GPU platforms, show dramatic improvements relative to our baseline, a high-end desktop CPU that is typical of the computers in current base stations. The improvements are both in terms of absolute performance (more than 50X with a high-end GPU and more than 4X with a GPU platform consumes almost 5 times less power than the CPU platform), in terms of performance-per-Watt ratios (more than 16X), and in terms of price-performance ratios.

中文翻译：

---

用于反向 GPS 野生动物跟踪系统的高性能 GPU 和 CPU 信号处理

我们提出了由称为 ATLAS 的高吞吐量野生动物跟踪系统执行的信号处理任务的强大高性能实现。该系统通过估计编码已知伪随机码的数据包到达接收器（基站）的时间来跟踪连接到野生动物的无线电发射器。宽带无线电信号的到达时间估计在计算上是昂贵的，特别是当不知道发射机何时发射时。这些计算是限制系统吞吐量的关键瓶颈。该论文报告了 ATLAS 信号处理算法的两种实现，一种用于 CPU，另一种用于 GPU，并仔细评估了它们的性能。在两个 CPU 平台和三个 GPU 平台上执行的评估显示相对于我们的基线有显着的改进，高端台式机 CPU，是当前基站中计算机的典型代表。改进既是在绝对性能方面（高端 GPU 超过 50 倍，GPU 平台超过 4 倍，功耗几乎是 CPU 平台的 5 倍），在每瓦性能比方面（更多比 16 倍），并且在性价比方面。