Parallel FFT implementation on TMS320c66x multicore DSP

Abstract

Fast Fourier transform (FFT) implementation on the TMS320C66x multicore digital signal processor (DSP) from Texas Instruments is considered in this paper. A FFT parallelization strategy is chosen and an expected performance gain is estimated. The algorithm is implemented in software for a four-core DSP architecture with interactions between cores based on inter-processor communication (IPC) and open multi-processing (OpenMP) techniques. Experimental results for a performance gain dependent on transform size are given. 2.9 times processing time reduction with 4 cores implementation is achieved.

Project Handover in Undergraduate Projects - Efficient Handover for Increased Learning Opportunities

“…As was already explained in Section 2.1, the students spent a large portion of the project time solving mundane issues with the hardware, resulting in significantly less time spent on the core of the project and hence fewer projects worthy of publication. For example, for projects beginning in the fall of 2012 and ending in the spring of 2013, there were only two publications [7,8] from 4 students. For the years from the fall of 2012 to the summer of 2015, the percentage of project students publishing papers were 50%, 0% and 0% from 4, 4 and 5 students respectively.…”

On the Feasibility of Fast Fourier Transform Separability Property for Distributed Image Processing

“…Some other FFT implementations take advantage of multicore processing. For example, Kharin et al [13] provide a DSP software implementation that concurrently computes radix-2 and radix-4 FFTs, achieving valid results up to 3 times faster than sequential implementations. Another implementation uses serial I/O ports to build a DSP network, where every DSP computes part of the entire FFT job.…”

Customizing FFT with FMA for China Accelerator

“…Fortunately, accelerators such as General Purpose Graphical Processing Unit (GPGPU), 1,2,[9][10][11] Many Integrated Core (MIC), [12][13][14] Digital Signal Processing (DSP), 3,4,[15][16][17][18] and Field Programmable Gate Array (FPGA) 5,6,[19][20][21][22][23][24][25][26][27][28] have recently proved to be a more promising platform to solve FFT prob-lems, since accelerators have much more parallel computing resources and can often achieve an order of magnitude performance improvement over CPUs.…”