Efficiency Near the Edge: Increasing the Energy Efficiency of FFTs on GPUs for Real-Time Edge Computing

Abstract

The Square Kilometre Array (SKA) is an international initiative for developing the world's largest radio telescope with a total collecting area of over a million square meters. The scale of the operation, combined with the remote location of the telescope, requires the use of energy-efficient computational algorithms. This, along with the extreme data rates that will be produced by the SKA and the requirement for a real-time observing capability, necessitates in-situ data processing in an edge style computing solution. More generally, energy efficiency in the modern computing landscape is becoming of paramount concern. Whether it be the power budget that can limit some of the world's largest supercomputers, or the limited power available to the smallest Internet-of-Things devices. In this paper, we study the impact of hardware frequency scaling on the energy consumption and execution time of the Fast Fourier Transform (FFT) on NVIDIA GPUs using the cuFFT library. The FFT is used in many areas of science and it is one of the key algorithms used in radio astronomy data processing pipelines. Through the use of frequency scaling, we show that we can lower the power consumption of the NVIDIA A100 GPU when computing the FFT by up to 47% compared to the boost clock frequency, with less than a 10% increase in the execution time. Furthermore, using one common core clock frequency for all tested FFT lengths, we show on average a 43% reduction in power consumption compared to the boost core clock frequency with an increase in the execution time still below 10%. We demonstrate how these results can be used to lower the power consumption of existing data processing pipelines. These savings, when considered over years of operation, can yield significant financial savings, but can also lead to a significant reduction of greenhouse gas emissions.

Bits Missing: Finding Exotic Pulsars Using bfloat16 on NVIDIA GPUs

“…Lowering the memory traffic requirement of FDAS has led to an approximately 1.6× speedup. The improved compute performance of bfloat16 is not a factor in this case, as the pipeline is memory bandwidth bound, with a low computational overhead (Adámek 2021). The speedup enables users of GPU-accelerated FDAS to increase their pulsar parameter search space on a given set of hardware when performing a real-time search, or to reduce their upfront hardware requirement for a given search.…”

“…When performing FDAS, AA spends a significant fraction of the runtime performing GPU-accelerated FFTs. The performance of a GPU-accelerated FFT is limited by memory bandwidth on current-generation hardware (Adámek 2021), and so by reducing the numerical precision of the calculations from 32 to 16 bits, this bottleneck can be alleviated simply by halving the amount of data that need to be transported to and from the GPU processing cores.…”

Analysis and Recognition of Human Gait Activity Based on Multimodal Sensors

“…In this research, the Fast Fourier Transform (FFT) has been used to convert the received resultant signals into the frequency domain. Specifically, the Discrete Fourier Transform (DFT) [31,33,34] is applied to improve the computational complexity. The frequency content of the input signal x is then extracted using the following transformation:…”

Privacy Preservation of User Identity in Contact Tracing for COVID-19-Like Pandemics Using Edge Computing

“…Edge computing is a key enabling technology that reduces latency, thus significantly enabling many real-time applications. Adámek et al [35] demonstrated the benefit of using edge computing to reduce the total power consumption of data processing. In addition, the authors demonstrated an average of 43% reduction in power consumption by optimizing the frequency scaling of core clocks on NVIDIA A100 GPU.…”