Skip to content
KernelIndex
Search⌘K
Sign in

Conv2d residual block

30 eligible runs
other

Convolutional residual block that combines input convolution (3->32 channels), output convolution (32->3 channels), and residual skip connection. This is the atomic building block for image restoration networks like NAFNet.

Fastest reported
21.1µsmean · 89% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-19. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencypadding = 1 · in_channels = 3 · kernel_size = 3 · out_channels = 3 · hidden_channels = 3216 cases
batch_size
height
width
dtype
16464fp321128128fp321131131fp321512512fp3226464fp322256256fp322512512fp322541541fp324512512fp328373373fp328512512fp32166464fp3216128128fp3216256256fp3232256256fp3264128128fp32
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 121.1 µsRIAC as well17NVIDIA B200 · env 230.4 µsArgus13

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 v1.1 · mean · 17 results · last observed 2026-08-19Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
RIAC as well
21.1µs89% SOL
1.00×
Reported · license unknown · no source
2026-08-19

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

no sourcelicense unknownno install recipeRun detail →
2
Geometric
25.9µs87% SOL
1.23×
Reported · license unknown · no source
2026-08-12

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

no sourcelicense unknownno install recipeRun detail →
3
Geometric
26.1µs87% SOL
1.24×
Reported · license unknown · no source
2026-08-12

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

no sourcelicense unknownno install recipeRun detail →
4
Geometric
27.7µs87% SOL
1.32×
Reported · license unknown · no source
2026-08-12

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

no sourcelicense unknownno install recipeRun detail →
5
Amir M. Mir | SF Tensor
28.9µs86% SOL
1.37×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
6
Hyra
29.5µs86% SOL
1.40×
Reported · license unknown · no source
2026-07-15

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

no sourcelicense unknownno install recipeRun detail →
7
Argus
30.4µs85% SOL
1.44×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
8
doubleAI
32.2µs84% SOL
1.53×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
9
Amir M. Mir | SF Tensor
32.2µs85% SOL
1.53×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
10
ac4k
33.2µs84% SOL
1.57×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
11
Databricks
33.8µs84% SOL
1.60×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
12
Recursive
36.2µs83% SOL
1.72×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
13
Geometric
65.3µs73% SOL
3.10×
Reported · license unknown · no source
2026-08-06

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

no sourcelicense unknownno install recipeRun detail →
14
Cursor
75.3µs69% SOL
3.57×
Reported · license unknown · no source
2026-07-10

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

no sourcelicense unknownno install recipeRun detail →
15
Geometric
80.0µs68% SOL
3.80×
Reported · license unknown · no source
2026-07-23

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

no sourcelicense unknownno install recipeRun detail →
16
TH3CHARLie | TileGym
135.6µs57% SOL
6.43×
Reported · license unknown · no source
2026-08-05

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

no sourcelicense unknownno install recipeRun detail →
17
TH3CHARLie | TileGym
555.2ms0% SOL
26333.87×
Reported · license unknown · no source
2026-08-05

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

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
RIAC as well
unknown
21.1µs
1.00×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
28.9µs
1.37×
Reported
license unknown · no source
Hyra
unknown
29.5µs
1.40×
Reported
license unknown · no source
Argus
unknown
30.4µs
1.44×
Reported
license unknown · no source
Databricks
unknown
32.2µs
1.53×
Reported
license unknown · no source
ac4k
unknown
33.2µs
1.57×
Reported
license unknown · no source
doubleAI
unknown
32.2µs
1.53×
Reported
license unknown · no source
Recursive
unknown
35.6µs
1.69×
Reported
license unknown · no source
Show all 30 implementations ›
Cursor
unknown
72.9µs
3.46×
Reported
license unknown · no source
TH3CHARLie | TileGym
unknown
135.6µs
6.43×
Reported
license unknown · no source
gnehiz
unknown
157.6µs
7.47×
Reported
license unknown · no source
Igor Bragaia
unknown
248.5µs
11.79×
Reported
license unknown · no source
Argus
unknown
30.4µs
1.44×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
32.2µs
1.53×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
32.3µs
1.53×
Reported
license unknown · no source
doubleAI
unknown
32.4µs
1.54×
Reported
license unknown · no source
Databricks
unknown
33.8µs
1.60×
Reported
license unknown · no source
ac4k
unknown
34.9µs
1.65×
Reported
license unknown · no source
Recursive
unknown
36.2µs
1.72×
Reported
license unknown · no source
doubleAI
unknown
36.6µs
1.74×
Reported
license unknown · no source
Geometric
unknown
65.3µs
3.10×
Reported
license unknown · no source
Cursor
unknown
75.3µs
3.57×
Reported
license unknown · no source
Geometric
unknown
80.0µs
3.80×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
141.9µs
6.73×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
154.9µs
7.35×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
246.7µs
11.70×
Reported
license unknown · no source
TH3CHARLie | TileGym
unknown
555.2ms
26333.87×
Reported
license unknown · no source
Geometric
unknown
25.9µs
1.23×
Reported
license unknown · no source
Geometric
unknown
26.1µs
1.24×
Reported
license unknown · no source
Geometric
unknown
27.7µs
1.32×
Reported
license unknown · no source

Semantics

Inputs and outputs
xfp32 [batch_size, in_channels, height, width]
conv_in_weightfp32 [hidden_channels, in_channels, kernel_size, kernel_size]
conv_in_biasfp32 [hidden_channels]
conv_out_weightfp32 [out_channels, hidden_channels, kernel_size, kernel_size]
conv_out_biasfp32 [out_channels]
outputfp32 [batch_size, out_channels, height, width]
Axes and behavior
widthvariable
heightvariable
paddingconstant = 1
batch_sizevariable
in_channelsconstant = 3
kernel_sizeconstant = 3
out_channelsconstant = 3
hidden_channelsconstant = 32
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias040_conv2d_residual_blockmodelnafnetsha256fe5498b659b3…
Sources: NVIDIA SOL-ExecBench (2026-08-19)last observed 2026-08-19How records are decidedJSON