From 8 Seconds to 370ms: Kernel-Fused SAR Imaging on Apple Silicon via Single-Dispatch FFT Pipelines

2026/04/04 by Mohamed Amine Bergach · 1 voice · 2 citations
Computer Science · #cs.PF

paper · pdf · doi:10.48550/arxiv.2604.03585

Abstract

We present the first kernel-fused SAR Range Doppler pipeline on any GPU platform. By fusing FFT, matched-filter multiply, and IFFT into a single Metal compute dispatch -- keeping all intermediate data in 32 KiB on-chip memory -- we process a 4096 × 4096 complex SAR scene in 370 ms on an Apple M1 GPU, a 22× speedup over the multi-dispatch baseline (8.16 s). We further report the first FFT to exploit Apple's simdgroup_matrix 8×8 hardware MMA, enabled by an in-place Cooley--Tukey decimation-in-frequency formulation that halves the memory footprint versus Stockham. Radar image quality is preserved: all five point targets show 0.0 dB SNR deviation from the unfused FP32 reference.

Citations

• TurboFNO: High-Performance Fourier Neural Operator with Fused FFT-GEMM-iFFT on GPU
• TurboFFT: Co-Designed High-Performance and Fault-Tolerant Fast Fourier Transform on GPUs
• Efficient FFT mapping on GPU for radar processing application: modeling and implementation

Cited by

• Range, Not Precision: Block-Floating-Point Half-Precision FFT and SAR Imaging on Apple Silicon

Discussions

• Mohamed Amine Bergach: From 8 Seconds to 370ms: Kernel-Fused SAR Imaging on Apple Silicon via Single-Dispatch FFT Pipelines https://arxiv.org/abs/2604.03585 https://arxiv.org/pdf/2604.03585 https://ar [bsky, 0 points, 0 comments]

Related

• Range, Not Precision: Block-Floating-Point Half-Precision FFT and SAR Imaging on Apple Silicon
• Efficient FFT mapping on GPU for radar processing application: modeling and implementation
• Design and Implementation of a Real-Time Imaging Processor for Spaceborne Synthetic Aperture Radar With Configurability
• Adaptation of Range-Doppler Algorithm for Efficient Beamforming of Monostatic and Multistatic Ultrasound Signals
• BLINK: an End-To-End GPU High Time Resolution Imaging Pipeline for Fast Radio Burst Searches with the Murchison Widefield Array