Table 4 Performance summary of CuFFT and proposed methods to improve it.
From: GPU-Based FFT Computation for Multi-Gigabit WirelessHD Baseband Processing
Method | FFT time ( |
|---|---|
CuFFT with no enhancements | 7 |
CuFFT enhancement #1: | 4.15 |
Large Batch Size | |
CuFFT enhancement #2: | 2.83 |
Page-locked memory + radix-2 | |
CuFFT enhancement #3: | 2.54 |
Asynchronous concurrency (streaming) | |
CuFFT enhancement #4: | 1.34 |
Reduced accuracy (16-bit) | |
CuFFT enhancement #5: | 0.534 |
Reduced accuracy (8-bit) |
s)