Skip to content
KernelIndex
Search⌘K

Multimodal RoPE position computation with grid based indexing

41 eligible runs
rope

Computes 3D rotary position embeddings for multimodal inputs by generating position indices from grid_thw (temporal, height, width) parameters, performing bilinear interpolation for variable resolution support, and creating interleaved MRoPE embeddings. Combines grid-based position index generation, learned position embedding lookup with bilinear interpolation weights, spatial permutation to merge_size blocks, and rotary embedding frequency computation with interleaving.

Fastest reported
4.00µsmean · 99% of SOL
doubleAILicense unknown · unknown

Reported evidence · last observed 2026-06-25. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhead_dim = 128 · num_heads = 12 · hidden_size = 1536 · quarter_head_dim = 32 · num_grid_per_side = 35 · spatial_merge_size = 2 · num_position_embeddings = 122516 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 · 21 results · last observed 2026-08-19Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
4.00µs99% SOL
1.00×
Reported · license unknown · no source
2026-06-25

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

no sourcelicense unknownno install recipeRun detail →
2
4.13µs99% SOL
1.03×
Reported · license unknown · no source
2026-07-06

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

no sourcelicense unknownno install recipeRun detail →
3
5.32µs98% SOL
1.33×
Reported · license unknown · no source
2026-07-01

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

no sourcelicense unknownno install recipeRun detail →
4
5.86µs98% SOL
1.46×
Reported · license unknown · no source
2026-06-18

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

no sourcelicense unknownno install recipeRun detail →
5
5.90µs98% SOL
1.47×
Reported · license unknown · no source
2026-06-20

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

no sourcelicense unknownno install recipeRun detail →
6
6.20µs98% SOL
1.55×
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 →
7
6.49µs98% SOL
1.62×
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 →
8
6.66µs98% SOL
1.66×
Reported · license unknown · no source
2026-06-06

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

no sourcelicense unknownno install recipeRun detail →
9
7.32µs97% SOL
1.83×
Reported · license unknown · no source
2026-06-14

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

no sourcelicense unknownno install recipeRun detail →
10
7.56µs97% SOL
1.89×
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 →
11
8.60µs96% SOL
2.15×
Reported · license unknown · no source
2026-07-08

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

no sourcelicense unknownno install recipeRun detail →
12
10.6µs96% SOL
2.66×
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 →
13
16.4µs93% SOL
4.10×
Reported · license unknown · no source
2026-07-09

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

no sourcelicense unknownno install recipeRun detail →
14
18.4µs93% SOL
4.60×
Reported · license unknown · no source
2026-07-08

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

no sourcelicense unknownno install recipeRun detail →
15
18.9µs92% SOL
4.73×
Reported · license unknown · no source
2026-07-08

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

no sourcelicense unknownno install recipeRun detail →
16
21.1µs91% SOL
5.28×
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 →
17
30.5µs88% SOL
7.62×
Reported · license unknown · no source
2026-05-18

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

no sourcelicense unknownno install recipeRun detail →
18
36.5µs86% SOL
9.13×
Reported · license unknown · no source
2026-04-27

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

no sourcelicense unknownno install recipeRun detail →
19
85.0µs72% SOL
21.23×
Reported · license unknown · no source
2026-04-25

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

no sourcelicense unknownno install recipeRun detail →
20
724.1µs25% SOL
180.92×
Reported · license unknown · no source
2026-04-22

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

no sourcelicense unknownno install recipeRun detail →
21
1.56ms14% SOL
389.95×
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
4.00µs
1.00×
Reported
license unknown · no source
unknown
4.13µs
1.03×
Reported
license unknown · no source
unknown
5.54µs
1.39×
Reported
license unknown · no source
unknown
5.89µs
1.47×
Reported
license unknown · no source
unknown
6.30µs
1.57×
Reported
license unknown · no source
unknown
6.22µs
1.55×
Reported
license unknown · no source
unknown
6.31µs
1.58×
Reported
license unknown · no source
unknown
6.49µs
1.62×
Reported
license unknown · no source
Show all 41 implementations ›
unknown
6.20µs
1.55×
Reported
license unknown · no source
unknown
7.32µs
1.83×
Reported
license unknown · no source
unknown
8.60µs
2.15×
Reported
license unknown · no source
unknown
10.6µs
2.66×
Reported
license unknown · no source
unknown
18.4µs
4.60×
Reported
license unknown · no source
unknown
32.3µs
8.06×
Reported
license unknown · no source
unknown
5.32µs
1.33×
Reported
license unknown · no source
unknown
5.80µs
1.45×
Reported
license unknown · no source
unknown
5.86µs
1.46×
Reported
license unknown · no source
unknown
5.90µs
1.47×
Reported
license unknown · no source
unknown
6.54µs
1.63×
Reported
license unknown · no source
unknown
6.66µs
1.66×
Reported
license unknown · no source
unknown
6.22µs
1.56×
Reported
license unknown · no source
unknown
6.79µs
1.70×
Reported
license unknown · no source
unknown
7.56µs
1.89×
Reported
license unknown · no source
unknown
8.58µs
2.14×
Reported
license unknown · no source
unknown
10.7µs
2.68×
Reported
license unknown · no source
unknown
16.4µs
4.10×
Reported
license unknown · no source
unknown
18.8µs
4.70×
Reported
license unknown · no source
unknown
18.9µs
4.73×
Reported
license unknown · no source
unknown
23.2µs
5.79×
Reported
license unknown · no source
unknown
21.1µs
5.28×
Reported
license unknown · no source
unknown
30.5µs
7.62×
Reported
license unknown · no source
unknown
36.5µs
9.13×
Reported
license unknown · no source
unknown
84.0µs
20.98×
Reported
license unknown · no source
unknown
84.2µs
21.04×
Reported
license unknown · no source
unknown
85.0µs
21.23×
Reported
license unknown · no source
unknown
724.1µs
180.92×
Reported
license unknown · no source
unknown
1.56ms
389.95×
Reported
license unknown · no source
unknown
2.75ms
686.29×
Reported
license unknown · no source
unknown
4.05µs
1.01×
Reported
license unknown · no source
unknown
4.14µs
1.03×
Reported
license unknown · no source
unknown
5.23µs
1.31×
Reported
license unknown · no source

Semantics

Inputs and outputs
grid_thwint64 [num_images, 3]
pos_embed_weightfp32 [num_position_embeddings, hidden_size]
inv_freqfp32 [quarter_head_dim]
patch_pos_embedsfp32 [total_tokens, hidden_size]
cos_embeddingsfp32 [total_tokens, head_dim]
sin_embeddingsfp32 [total_tokens, head_dim]
Axes and behavior
head_dimconstant = 128
num_headsconstant = 12
num_imagesvariable
hidden_sizeconstant = 1536
total_tokensvariable
quarter_head_dimquarter_head_dim = head_dim // 4
num_grid_per_sideconstant = 35
spatial_merge_sizeconstant = 2
num_position_embeddingsconstant = 1225
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias023_multimodal_rope_position_computation_with_grid_based_indexingmodelqwen3-vl-235b-a22b-instructsha256c126fe83d4cf…
Sources: NVIDIA SOL-ExecBench (2026-08-19)last observed 2026-08-19How records are decidedJSON