Skip to content
KernelIndex
Search⌘K

Joint attention context projection

27 eligible runs
other

Joint attention context projection that concatenates image and context attention outputs along the feature dimension (duplicating the concatenated sequence) and projects through a linear layer, then splits back into image and context streams. This is a fused operation combining concatenation, linear projection, and sequence splitting.

Fastest reported
85.3µsmean · 97% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-21. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyinner_dim = 2432 · proj_input_dim = 486416 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 v1.1 · mean · 17 results · last observed 2026-08-21Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
85.3µs97% SOL
1.00×
Reported · license unknown · no source
2026-08-21

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

no sourcelicense unknownno install recipeRun detail →
2
86.5µs97% SOL
1.01×
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 →
3
87.2µs97% SOL
1.02×
Reported · license unknown · no source
2026-07-16

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

no sourcelicense unknownno install recipeRun detail →
4
94.1µs96% SOL
1.10×
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 →
5
96.7µs96% SOL
1.13×
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 →
6
96.8µs96% SOL
1.14×
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 →
7
102.6µs96% SOL
1.20×
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
109.0µs95% SOL
1.28×
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 →
9
114.6µs95% SOL
1.34×
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 →
10
119.0µs95% SOL
1.40×
Reported · license unknown · no source
2026-07-28

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

no sourcelicense unknownno install recipeRun detail →
11
144.0µs93% SOL
1.69×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
12
146.9µs93% SOL
1.72×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
13
147.8µs93% SOL
1.73×
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 →
14
149.7µs93% SOL
1.76×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
15
190.2µs90% SOL
2.23×
Reported · license unknown · no source
2026-07-16

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

no sourcelicense unknownno install recipeRun detail →
16
284.6µs86% SOL
3.34×
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 →
17
627.0µs72% SOL
7.35×
Reported · license unknown · no source
2026-07-24

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
84.5µs
1.00×
Reported
license unknown · no source
unknown
85.3µs
1.01×
Reported
license unknown · no source
unknown
87.5µs
1.04×
Reported
license unknown · no source
unknown
96.8µs
1.15×
Reported
license unknown · no source
unknown
98.0µs
1.16×
Reported
license unknown · no source
unknown
106.7µs
1.26×
Reported
license unknown · no source
unknown
119.0µs
1.41×
Reported
license unknown · no source
unknown
144.0µs
1.70×
Reported
license unknown · no source
Show all 27 implementations ›
unknown
146.9µs
1.74×
Reported
license unknown · no source
unknown
157.2µs
1.86×
Reported
license unknown · no source
unknown
190.2µs
2.25×
Reported
license unknown · no source
unknown
270.6µs
3.20×
Reported
license unknown · no source
unknown
94.1µs
1.11×
Reported
license unknown · no source
unknown
96.7µs
1.14×
Reported
license unknown · no source
unknown
99.6µs
1.18×
Reported
license unknown · no source
unknown
102.6µs
1.21×
Reported
license unknown · no source
unknown
109.0µs
1.29×
Reported
license unknown · no source
unknown
114.6µs
1.36×
Reported
license unknown · no source
unknown
147.8µs
1.75×
Reported
license unknown · no source
unknown
149.7µs
1.77×
Reported
license unknown · no source
unknown
284.6µs
3.37×
Reported
license unknown · no source
unknown
627.0µs
7.42×
Reported
license unknown · no source
unknown
1.53ms
18.05×
Reported
license unknown · no source
unknown
1.58ms
18.68×
Reported
license unknown · no source
unknown
87.2µs
1.03×
Reported
license unknown · no source
unknown
85.8µs
1.02×
Reported
license unknown · no source
unknown
86.5µs
1.02×
Reported
license unknown · no source

Semantics

Inputs and outputs
image_attention_outputfp32 [batch_size, image_seq_len, inner_dim]
context_attention_outputfp32 [batch_size, context_seq_len, inner_dim]
to_out_weightfp32 [inner_dim, proj_input_dim]
to_out_biasfp32 [inner_dim]
projected_imagefp32 [batch_size, image_seq_len, inner_dim]
projected_contextfp32 [batch_size, context_seq_len, inner_dim]
Axes and behavior
inner_dimconstant = 2432
batch_sizevariable
image_seq_lenvariable
proj_input_dimproj_input_dim = inner_dim * 2
context_seq_lenvariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias081_joint_attention_context_projectionmodelstable-diffusion-3-5-largesha2562fb157242ef0…
Sources: NVIDIA SOL-ExecBench (2026-08-21)last observed 2026-08-21How records are decidedJSON