GQA ragged prefill causal h24 kv8 d128
20 eligible runs
gqa-ragged-attention
Batched Grouped Query Attention prefill with ragged (variable-length) inputs. Causal mask is applied. Captured from Llama 3.2 3B during incremental prefill. GQA ratio 3:1 (24 q, 8 kv, head_dim=128).
Source baseline · unbeaten
41.4µsmean
flashinfer / wrapperfe59e7FlashInfer-Bench baselines · Apache-2.0 · python
Reported evidence · last observed 2026-04-12. The source's designated baseline implementation. Reported by source; not independently reproduced.
Current records
Workloadtotal_q = 1 · total_kv = 1 · len_indptr = 2 · bf1620 cases
Not measured on H100, B200 for this workload. Challenges →
Source-native comparison · GPU NVIDIA GB200 · Workload total_q = 1 · total_kv = 1 · len_indptr = 2 · bf16 · CUDA 13.0 · Framework pytorch 2.9.1+cu130 · Protocol flashinfer-bench · mean · 1 results · last observed 2026-04-12Record history →
Estimated floor 1 ns · record 32270.67× above itestimate, not evidence ›
DRAM 1 ns · bandwidth-bound on GB200
every declared tensor crosses HBM exactly once (8,000 GB/s, GB200 datasheet)
no arithmetic formula for this family: bandwidth floor only
headroom-v1: a lower bound from declared tensors and datasheet peaks. A kernel can sit well above it for good reasons.
#
Implementation
Latency
vs #1
Trust
Observed
141.4µs1.00×Reported · Apache-2.0 · source2026-04-12
Measured exactly what you asked. The source's designated baseline implementation. Reported by source; not independently reproduced.
source mirroredApache-2.0no install recipeView source →Run detail →
Implementations
Implementation
Runtime
Best latency
Evidence
Availability
flashinfer / wrapperfe59e7FlashInfer-Bench baselines
python
41.4µs
1.00×
Reproduction-ready
Apache-2.0 · source
Semantics
Inputs and outputs
qbf16 [total_q, num_qo_heads, head_dim]
kbf16 [total_kv, num_kv_heads, head_dim]
vbf16 [total_kv, num_kv_heads, head_dim]
qo_indptrint32 [len_indptr]
kv_indptrint32 [len_indptr]
sm_scalefp32 scalar
outputbf16 [total_q, num_qo_heads, head_dim]
lsefp32 [total_q, num_qo_heads]
Axes and behavior
total_qvariable
head_dimconstant = 128
total_kvvariable
len_indptrvariable
num_kv_headsconstant = 8
num_qo_headsconstant = 24
determinismunspecified
constraintsNo mutation or aliasing
Identity
Sources: FlashInfer-Bench (2026-04-12) · Apache-2.0last observed 2026-04-12How records are decidedJSON