Abstract: This paper investigates the implementation of a frequency-domain high-resolution range profile (HRRP) synthesis algorithm for terahertz frequency-modulated stepped-frequency monopulse radar on an embedded hardware platform. Synthesizing the spectra of individual sub-pulses of a frequency-modulated pulse train into an ultra-wideband (UWB) spectrum is fundamental to generating HRRPs in the frequency domain. Since inter-pulse motion effects must be compensated, this process involves numerous processing steps and is computationally intensive. Furthermore, all relevant computations must be completed within a single sub-pulse repetition interval, imposing extremely high real-time requirements and placing significant demands on the hardware implementation design of the algorithm. In this paper, a computing architecture based on the TMS320C6678 multi-core DSP platform is proposed, innovatively adopting a "parallel processing across echo channels plus pipelined processing within each channel" approach. Among the eight computing cores of the C6678, seven are allocated: Core0 serves as the master core responsible for task scheduling and communication, while Cores 1~6 are used as computing cores and divided into three groups of two cores each, with each group processing the echo signals from one of the three channels: sum, elevation difference, and azimuth difference. The processing across the three groups is performed in parallel. Within each group, a pipeline architecture is employed, consisting of two pipeline stages. The processing tasks are partitioned into two batches according to the stages, with each core handling the computation of one stage. Further efficiency improvements are achieved through a series of optimization strategies, including memory hierarchy optimization, intrinsic instruction invocation, and loop unrolling. Experimental results demonstrate that the proposed design enables high-speed parallel processing of three-channel echo data under stringent real-time constraints (100μs), achieving high-quality spectrum stitching and high-resolution range profile generation.