Design space exploration of SW beamformer on GPU

Abstract

SummaryUltrasound imaging has become one of the most widely used modalities in medical diagnosis today. However, real‐time ultrasound imaging requires large amount of data transfer and massive computation and therefore mainly relies on a complex dedicated hardware system. A recent trend of a graphics processing unit (GPU) based software‐based approach offers the advantages of flexibility and quick implementation. The GPUs have been reported as excellent accelerators across a wide range of applications. For best exploiting outstanding computational power and high memory bandwidth of a GPU, the paper explores the design space of implementing an ultrasound beamformer on a GPU platform. The design spaces are expanded by applying different optimization strategies to the beamformer on a GPU platform, and we also discuss the performance evaluation results on the various GPUs whose architectural characteristics are different to each others. The performance analysis shows that by optimizing CUDA code, our real‐time‐GPU‐based beamformer can be successfully implemented with 181 frames per second (fps) and speedup of 230.6X compared with the single‐threaded implementation on a high‐performance CPU platform. Copyright © 2014 John Wiley & Sons, Ltd.

Software-based high-level synthesis design of FPGA beamformers for synthetic aperture imaging

“…system developers are required to have domain-specific skills that often lie beyond the expertise of ultrasound imaging researchers, and this has been acknowledged as a challenging factor for the development of new ultrasound imaging platforms [8], [12]. In this work, we have proposed the use of opencl-based highlevel synthesis tools to design FPGa-based sa beamformers: an approach that is premised on the translation of high-level c-like kernels to low-level rTl descriptions [18].…”

“…It would be more tractable to adopt an approach that can achieve rapid prototyping of novel ultrasound imaging techniques via the use of programmable computing hardware. one particular approach with demonstrated potential is to make use of graphics processing units (GPUs) [7], [8]. Its general merit is that, because GPUs are powerful parallel computing engines with thousands of processor cores, they can offer considerable efficiency gain in terms of energy-per-throughput and cost-per-throughput ratios [9].…”

A performance, power, and energy analysis of ultrasound B‐mode imaging on a GPU with VFS

“…Algorithm 1 shows the sequential version of the pseudo‐code for a B‐mode imaging algorithm. The complexity analysis of the beamforming algorithm has been performed in one study . A large amount of computation is required in a triple nested loop described from line 1 to line 12.…”

“…Figure shows the overall computation flow of an ultrasound system. A typical medical ultrasound imaging system consists of several processing blocks or stages: data acquisition, beamforming, B‐mode imaging, color Doppler, spectral Doppler, and scan conversion …”

“…B‐mode ultrasound imaging can be divided into 3 main functions, namely, beamforming , quadrature demodulation , and envelope detection . The workload distribution of these functions is illustrated in Table (the specific values for the parameters used for this profiling are presented in Table ).…”