A Hybrid GPU and CPU Parallel Computing Method to Accelerate Millimeter-Wave Imaging

The range migration algorithm (RMA) based on Fourier transformation is widely applied in millimeter-wave (MMW) close-range imaging because of its few operations and small approximation. However, interpolation stage not effective due to the involved intensive logic controls, which limits speed performance a graphics processing unit (GPU) platform. Therefore, this paper, we present an acceleration optimization method hybrid GPU central (CPU) parallel computation for implementing RMA. proposed exploits strong logic-control capability CPU assist controls stage. common positions wavenumber-domain components be interpolated are calculated by stored constant memory broadcast at any time. This avoids repetitive consumed GPU-only scheme. Then responsible remaining matrix-related steps outputs needed values. experiments verify efficiency demonstrate that speedup ratio our more than 15 times CPU-only method, 2 method.