Hyena depthwise conv1d split gate
Depthwise 1D convolution with padding followed by channel splitting and initial element-wise gating in HyenaOperator. Takes projected input (inner_width channels), applies depthwise grouped convolution with kernel size 3 and padding 2, truncates to sequence length, splits into (order+1) components, and performs first gating operation (v * x[0]).
Reported evidence · last observed 2026-06-07. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencyd_model = 256 · hyena_order = 2 · inner_width = 768 · short_filter_order = 316 cases
Not measured on H100 for this workload. Challenges →
16.37µs78% SOL1.00×Reported · license unknown · no source2026-06-07
Measured exactly what you asked. Reported by source; not independently reproduced.
26.81µs77% SOL1.07×Reported · license unknown · no source2026-07-08
Measured exactly what you asked. Reported by source; not independently reproduced.
37.02µs76% SOL1.10×Reported · license unknown · no source2026-06-29
Measured exactly what you asked. Reported by source; not independently reproduced.
47.06µs76% SOL1.11×Reported · license unknown · no source2026-06-28
Measured exactly what you asked. Reported by source; not independently reproduced.
57.36µs74% SOL1.15×Reported · license unknown · no source2026-05-20
Measured exactly what you asked. Reported by source; not independently reproduced.
67.39µs75% SOL1.16×Reported · license unknown · no source2026-07-05
Measured exactly what you asked. Reported by source; not independently reproduced.
77.89µs73% SOL1.24×Reported · license unknown · no source2026-06-19
Measured exactly what you asked. Reported by source; not independently reproduced.
87.99µs72% SOL1.25×Reported · license unknown · no source2026-07-08
Measured exactly what you asked. Reported by source; not independently reproduced.
98.41µs71% SOL1.32×Reported · license unknown · no source2026-07-08
Measured exactly what you asked. Reported by source; not independently reproduced.
109.11µs70% SOL1.43×Reported · license unknown · no source2026-07-03
Measured exactly what you asked. Reported by source; not independently reproduced.
119.46µs70% SOL1.49×Reported · license unknown · no source2026-04-14
Measured exactly what you asked. Reported by source; not independently reproduced.
1210.0µs68% SOL1.58×Reported · license unknown · no source2026-06-09
Measured exactly what you asked. Reported by source; not independently reproduced.
1310.1µs67% SOL1.58×Reported · license unknown · no source2026-04-25
Measured exactly what you asked. Reported by source; not independently reproduced.
1410.1µs67% SOL1.58×Reported · license unknown · no source2026-06-26
Measured exactly what you asked. Reported by source; not independently reproduced.
1510.4µs66% SOL1.64×Reported · license unknown · no source2026-04-27
Measured exactly what you asked. Reported by source; not independently reproduced.
1610.8µs66% SOL1.69×Reported · license unknown · no source2026-04-22
Measured exactly what you asked. Reported by source; not independently reproduced.
1720.0µs51% SOL3.15×Reported · license unknown · no source2026-06-26
Measured exactly what you asked. Reported by source; not independently reproduced.
1822.0µs50% SOL3.45×Reported · license unknown · no source2026-06-26
Measured exactly what you asked. Reported by source; not independently reproduced.
1926.9µs45% SOL4.22×Reported · license unknown · no source2026-06-25
Measured exactly what you asked. Reported by source; not independently reproduced.
2029.4µs42% SOL4.62×Reported · license unknown · no source2026-06-25
Measured exactly what you asked. Reported by source; not independently reproduced.
2143.2µs36% SOL6.78×Reported · license unknown · no source2026-05-28
Measured exactly what you asked. Reported by source; not independently reproduced.