Efficient Backprojection-Based Synthetic Aperture Radar Computation with Many-Core Processors

Abstract

Tackling computationally challenging problems with high efficiency often requires the combination of algorithmic innovation, advanced architecture, and thorough exploitation of parallelism. We demonstrate this synergy through synthetic aperture radar (SAR) via backprojection, an image reconstruction method that can require hundreds of TFLOPS. Computation cost is significantly reduced by our new algorithm of approximate strength reduction; data movement cost is economized by software locality optimizations facilitated by advanced architecture support; parallelism is fully harnessed in various patterns and granularities. We deliver over 35 billion backprojections per second throughput per compute node on an Intel® Xeon® processor E5-2670-based cluster, equipped with Intel® Xeon Phi™ coprocessors. This corresponds to processing a 3K×3K image within a second using a single node. Our study can be extended to other settings: backprojection is applicable elsewhere including medical imaging, approximate strength reduction is a general code transformation technique, and many-core processors are emerging as a solution to energy-efficient computing.

A unified Coq framework for verifying C programs with floating-point computations

“…The code of our Clight implementation is shown in Figure 10. We choose to dynamically adapt the computation of Our adaptive implementation allows reducing the number of accurate square root computations (and thus save energy) while staying within the total implementation error bound determined by the Taylor approximation of square root within a disc of constant radius, by contrast to Tang et al [34], who use the Taylor approximation everywhere and thus dramatically reduce the overall precision of their norm computation.…”

Modeling synthetic aperture radar computation with Aspen

“…While this paper focuses on results specific to the SAR scenario, a full description of the syntax of these models is available online (Spafford and Vetter, 2012). Further details on the scenario being modeled are available in prior publications as well (Campbell et al, 2011; Park et al, 2012). We will utilize many of the same scenario parameters for the present analysis.…”

“…We employ the Aspen language to construct a model for the requirements of a SSCP featuring synthetic aperture radar (SAR) processing. The scenario summarized herein is also described in a recent publication focused on optimizing a portion of the SAR processing on many-core architectures (Park et al, 2012). In this scenario, continuous data from a radar onboard a platform orbiting about a fixed scene center is converted into a series of images of that scene over time.…”

Automatic registered back‐projection approach based on object orientation for airborne repeat‐track interferometric SAR

“…Also, the BP algorithm can be parallelised easily due to processing data pixel‐by‐pixel. It has been implemented on graphics processing units (GPU) [20, 21] and central processing units (CPU) with many‐cores [22]. Besides, some fast BP (FBP) algorithms have been suggested.…”