Hyena FFT size padding RFFT
Fused FFT size padding and real FFT computation for Hyena convolution. Pads input to 2*seqlen for circular convolution, computes real FFT (rfft), and normalizes by fft_size. Input is (batch, channels, seqlen) and output is complex tensor (batch, channels, seqlen+1) in frequency domain.
Reported evidence · last observed 2026-08-12. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencyd_model = 25616 cases
Not measured on H100 for this workload. Challenges →
130.9µs82% SOL1.00×Reported · license unknown · no source2026-08-12
Measured exactly what you asked. Reported by source; not independently reproduced.
233.0µs81% SOL1.07×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
333.5µs80% SOL1.08×Reported · license unknown · no source2026-08-12
Measured exactly what you asked. Reported by source; not independently reproduced.
434.8µs80% SOL1.13×Reported · license unknown · no source2026-07-15
Measured exactly what you asked. Reported by source; not independently reproduced.
538.6µs77% SOL1.25×Reported · license unknown · no source2026-07-25
Measured exactly what you asked. Reported by source; not independently reproduced.
639.0µs78% SOL1.26×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
747.3µs74% SOL1.53×Reported · license unknown · no source2026-08-21
Measured exactly what you asked. Reported by source; not independently reproduced.
847.4µs74% SOL1.53×Reported · license unknown · no source2026-08-23
Measured exactly what you asked. Reported by source; not independently reproduced.
947.5µs74% SOL1.54×Reported · license unknown · no source2026-08-18
Measured exactly what you asked. Reported by source; not independently reproduced.
1048.1µs73% SOL1.56×Reported · license unknown · no source2026-08-15
Measured exactly what you asked. Reported by source; not independently reproduced.
1150.3µs72% SOL1.63×Reported · license unknown · no source2026-08-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1258.7µs69% SOL1.90×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1361.3µs68% SOL1.98×Reported · license unknown · no source2026-08-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1467.0µs66% SOL2.17×Reported · license unknown · no source2026-08-08
Measured exactly what you asked. Reported by source; not independently reproduced.
1567.4µs66% SOL2.18×Reported · license unknown · no source2026-08-04
Measured exactly what you asked. Reported by source; not independently reproduced.
1667.6µs66% SOL2.19×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
1779.7µs62% SOL2.58×Reported · license unknown · no source2026-08-07
Measured exactly what you asked. Reported by source; not independently reproduced.
1886.0µs60% SOL2.78×Reported · license unknown · no source2026-07-23
Measured exactly what you asked. Reported by source; not independently reproduced.
1986.1µs60% SOL2.78×Reported · license unknown · no source2026-08-11
Measured exactly what you asked. Reported by source; not independently reproduced.
20106.5µs56% SOL3.45×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
21242.5µs38% SOL7.85×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.