One of the key scientific projects of the SKA radio telescope is a large survey for pulsars both in isolated and binary systems. The data rate of the pulsar search engine is expected to reach 0.6TeraSamples/sec. For the purposes of extracting hidden pulses from these streams, we need a complex search strategy which allows us to explore a three dimensional parameter space and it requires approximately lOPetaFlops. This problem is well suited for a parallel computing engine, but the dimensions of SKA bring this problem to a new level of complexity. An up-to-date study shows that this operation would require more than 2000 GPUs. In this report we will present possible mitigation strategies.
展开▼
机译：SKA Radio Telescope的关键科学项目之一是孤立和二元系统中的脉冲都气进行了大量调查。预计脉冲节搜索引擎的数据速率将达到0.6Terasamples /秒。为了从这些流中提取隐藏脉冲的目的，我们需要一个复杂的搜索策略，该策略允许我们探索三维参数空间，并且需要大约是Lopetaflops。这个问题非常适合并行计算引擎，但SKA的尺寸为这个问题带来了新的复杂程度。最新的研究表明，此操作需要2000多个GPU。在本报告中，我们将提出可能的缓解策略。
展开▼
