Skip to content
KernelIndex
Search⌘K

Conv gated projection with causal conv

34 eligible runs
mlp

Complete fused convolution layer operation that combines triple linear projection, element-wise gating, causal 1D grouped convolution, gating with output channel, and final output projection. This is the dominant operation in non-attention layers in LFM2 architecture.

Fastest reported
99.1µsmean · 76% of SOL
ac4kLicense unknown · unknown

Reported evidence · last observed 2026-07-02. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhidden_size = 2048 · triple_hidden = 6144 · conv_kernel_size = 416 cases

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 16 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.0 · mean · 17 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
99.1µs76% SOL
1.00×
Reported · license unknown · no source
2026-07-02

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
2
99.5µs76% SOL
1.00×
Reported · license unknown · no source
2026-06-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
3
100.4µs76% SOL
1.01×
Reported · license unknown · no source
2026-06-26

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
4
103.3µs75% SOL
1.04×
Reported · license unknown · no source
2026-04-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
5
103.4µs75% SOL
1.04×
Reported · license unknown · no source
2026-07-05

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
6
106.4µs74% SOL
1.07×
Reported · license unknown · no source
2026-05-20

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
7
106.6µs74% SOL
1.08×
Reported · license unknown · no source
2026-06-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
8
107.3µs73% SOL
1.08×
Reported · license unknown · no source
2026-07-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
9
111.8µs72% SOL
1.13×
Reported · license unknown · no source
2026-06-02

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
10
115.4µs71% SOL
1.16×
Reported · license unknown · no source
2026-07-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
11
122.2µs71% SOL
1.23×
Reported · license unknown · no source
2026-07-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
12
168.4µs58% SOL
1.70×
Reported · license unknown · no source
2026-05-26

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
13
195.9µs52% SOL
1.98×
Reported · license unknown · no source
2026-06-09

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
14
209.0µs50% SOL
2.11×
Reported · license unknown · no source
2026-04-20

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
15
238.4µs46% SOL
2.41×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
16
252.1µs44% SOL
2.54×
Reported · license unknown · no source
2026-04-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
17
253.4µs44% SOL
2.56×
Reported · license unknown · no source
2026-05-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
99.5µs
1.00×
Reported
license unknown · no source
unknown
100.4µs
1.01×
Reported
license unknown · no source
unknown
101.7µs
1.03×
Reported
license unknown · no source
unknown
103.3µs
1.04×
Reported
license unknown · no source
unknown
103.4µs
1.04×
Reported
license unknown · no source
unknown
106.6µs
1.08×
Reported
license unknown · no source
unknown
107.3µs
1.08×
Reported
license unknown · no source
unknown
107.3µs
1.08×
Reported
license unknown · no source
Show all 34 implementations ›
unknown
107.6µs
1.09×
Reported
license unknown · no source
unknown
122.2µs
1.23×
Reported
license unknown · no source
unknown
131.8µs
1.33×
Reported
license unknown · no source
unknown
195.9µs
1.98×
Reported
license unknown · no source
unknown
238.4µs
2.41×
Reported
license unknown · no source
unknown
252.1µs
2.54×
Reported
license unknown · no source
unknown
259.8µs
2.62×
Reported
license unknown · no source
unknown
99.1µs
1.00×
Reported
license unknown · no source
unknown
105.3µs
1.06×
Reported
license unknown · no source
unknown
106.4µs
1.07×
Reported
license unknown · no source
unknown
108.1µs
1.09×
Reported
license unknown · no source
unknown
109.3µs
1.10×
Reported
license unknown · no source
unknown
111.8µs
1.13×
Reported
license unknown · no source
unknown
112.1µs
1.13×
Reported
license unknown · no source
unknown
112.3µs
1.13×
Reported
license unknown · no source
unknown
115.6µs
1.17×
Reported
license unknown · no source
unknown
116.3µs
1.17×
Reported
license unknown · no source
unknown
115.4µs
1.16×
Reported
license unknown · no source
unknown
125.6µs
1.27×
Reported
license unknown · no source
unknown
168.4µs
1.70×
Reported
license unknown · no source
unknown
209.0µs
2.11×
Reported
license unknown · no source
unknown
253.4µs
2.56×
Reported
license unknown · no source
unknown
276.9µs
2.79×
Reported
license unknown · no source
unknown
99.1µs
1.00×
Reported
license unknown · no source
unknown
104.2µs
1.05×
Reported
license unknown · no source
unknown
105.3µs
1.06×
Reported
license unknown · no source

Semantics

Inputs and outputs
xbf16 [batch_size, seq_len, hidden_size]
in_proj_weightbf16 [triple_hidden, hidden_size]
in_proj_biasbf16 [triple_hidden]
conv_weightbf16 [hidden_size, 1, conv_kernel_size]
conv_biasbf16 [hidden_size]
out_proj_weightbf16 [hidden_size, hidden_size]
out_proj_biasbf16 [hidden_size]
outputbf16 [batch_size, seq_len, hidden_size]
Axes and behavior
seq_lenvariable
batch_sizevariable
hidden_sizeconstant = 2048
triple_hiddentriple_hidden = hidden_size * 3
conv_kernel_sizeconstant = 4
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias005_conv_gated_projection_with_causal_convmodellfm2-1-2bsha25667da2d9f4f11…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON