Abstract Searching for periodic non-accelerated signals in the presence of ideal white noise using the fully phase-coherent fast-folding algorithm (FFA) is theoretically established as a more sensitive search method than the fast Fourier transform (FFT) search with incoherent harmonic summing. In this paper, we present a comparison of the performance of an FFA search implementation using RIPTIDE and an FFT search implementation using PRESTO, over a range of signal parameters with white noise and with real telescope noise from the Giant Meterwave Radio Telescope (GMRT) High Resolution Southern Sky (GHRSS) survey with the upgraded GMRT (uGMRT). We find that the FFA search with appropriate de-reddening of the time series performs better than the FFT search with spectral whitening for long-period pulsars under real GHRSS noise conditions. We describe an FFA-search pipeline implemented for the GHRSS survey looking for pulsars over a period of 0.1–100 s and up to a dispersion measure of 500 pc cm⁻³. We processed GHRSS survey data covering ∼1500 deg² of the sky with this pipeline. We re-detected 43 known pulsars with a better signal-to-noise ratio in the FFA search than in the FFT search. We also report the discovery of two new pulsars, including a long-period pulsar with a short duty cycle, using this FFA-search pipeline. A population of long-period pulsars with periods of several seconds or higher could help constrain the pulsar death line.