Skip to content
KernelIndex
Search⌘K

Rotary position embedding application

25 eligible runs
rope

Rotary Position Embedding (RoPE) application that rotates query and key vectors in attention mechanisms based on their positions. Splits the feature dimension into pairs, applies rotation matrices derived from position indices and frequency bands. Critical for diffusion transformers as it provides position information without learned parameters.

Fastest reported
23.3µsmean · 69% 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 latencyhead_dim = 128 · half_head_dim = 6416 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 · 13 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
23.3µs69% 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
23.3µs69% SOL
1.00×
Reported · license unknown · no source
2026-07-17

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

no sourcelicense unknownno install recipeRun detail →
3
23.6µs68% SOL
1.01×
Reported · license unknown · no source
2026-08-18

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

no sourcelicense unknownno install recipeRun detail →
4
24.3µs67% SOL
1.04×
Reported · license unknown · no source
2026-08-24

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

no sourcelicense unknownno install recipeRun detail →
5
24.5µs67% SOL
1.05×
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
24.5µs67% SOL
1.05×
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 →
7
24.8µs67% SOL
1.06×
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 →
8
24.8µs67% SOL
1.07×
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
25.1µs66% SOL
1.08×
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
25.2µs66% SOL
1.08×
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
25.3µs66% SOL
1.09×
Reported · license unknown · no source
2026-07-18

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

no sourcelicense unknownno install recipeRun detail →
12
28.1µs63% 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 →
13
39.7µs51% SOL
1.70×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
23.3µs
1.00×
Reported
license unknown · no source
unknown
24.3µs
1.04×
Reported
license unknown · no source
unknown
24.7µs
1.06×
Reported
license unknown · no source
unknown
24.8µs
1.06×
Reported
license unknown · no source
unknown
25.1µs
1.08×
Reported
license unknown · no source
unknown
25.0µs
1.07×
Reported
license unknown · no source
unknown
25.2µs
1.08×
Reported
license unknown · no source
unknown
25.5µs
1.09×
Reported
license unknown · no source
Show all 25 implementations ›
unknown
25.8µs
1.11×
Reported
license unknown · no source
unknown
25.9µs
1.11×
Reported
license unknown · no source
unknown
39.7µs
1.70×
Reported
license unknown · no source
unknown
23.6µs
1.01×
Reported
license unknown · no source
unknown
24.7µs
1.06×
Reported
license unknown · no source
unknown
24.8µs
1.07×
Reported
license unknown · no source
unknown
25.1µs
1.08×
Reported
license unknown · no source
unknown
25.2µs
1.08×
Reported
license unknown · no source
unknown
25.3µs
1.09×
Reported
license unknown · no source
unknown
26.2µs
1.12×
Reported
license unknown · no source
unknown
28.1µs
1.20×
Reported
license unknown · no source
unknown
28.2µs
1.21×
Reported
license unknown · no source
unknown
38.8µs
1.67×
Reported
license unknown · no source
unknown
39.9µs
1.72×
Reported
license unknown · no source
unknown
23.3µs
1.00×
Reported
license unknown · no source
unknown
24.5µs
1.05×
Reported
license unknown · no source
unknown
24.5µs
1.05×
Reported
license unknown · no source

Semantics

Inputs and outputs
queryfp32 [batch_size, num_heads, seq_len, head_dim]
keyfp32 [batch_size, num_kv_heads, seq_len, head_dim]
cosfp32 [seq_len, head_dim]
sinfp32 [seq_len, head_dim]
query_rotatedfp32 [batch_size, num_heads, seq_len, head_dim]
key_rotatedfp32 [batch_size, num_kv_heads, seq_len, head_dim]
Axes and behavior
seq_lenvariable
head_dimconstant = 128
num_headsvariable
batch_sizevariable
num_kv_headsvariable
half_head_dimhalf_head_dim = head_dim // 2
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias088_rotary_position_embedding_applicationmodelhunyuanimage-2-1sha2567f8c790b4016…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON