Max Pooling 1D
4 eligible runs
reduction
KernelBench level1 problem 41: Max Pooling 1D. The computation is the reference PyTorch module's forward pass; the output shape follows the module (mirrored as each implementation's source).
Fastest reported
2.85ms±0.03 · mean of 100 · 3.75× faster than baseline
torch.compile (inductor)PyTorch · MIT · python
Reported evidence · last observed 2026-03-05. Reported by source; not independently reproduced.
Current records
HardwareBest knownImplementationRuns
NVIDIA H100 · env 12.85 mstorch.compile (inductor)2NVIDIA H100 · env 214.0 msPyTorch eager2Not measured on B200 for this workload. Challenges →
Source-native comparison · GPU NVIDIA H100 · Workload features = 192 · batch_size = 64 · sequence_length = 65536 · fp32 · Protocol KernelBench timing scripts · mean · 2 results · last observed 2026-03-05Record history →
Estimated floor 961.6 µs · record 2.96× above itestimate, not evidence ›
DRAM 961.6 µs · bandwidth-bound on H100 SXM
every declared tensor crosses HBM exactly once (3,350 GB/s, H100 SXM 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
1torch.compile (inductor)PyTorch2.85ms±0.031.00×Reported · MIT · source2026-03-05stale
3.75× faster than the baseline. Measured exactly what you asked. Reported by source; not independently reproduced.
source mirroredMITno install recipeView source →Run detail →
210.7ms±0.043.75×Reported · MIT · source2026-03-05stale
Measured exactly what you asked. The source's designated baseline implementation. Reported by source; not independently reproduced.
source mirroredMITno install recipeView source →Run detail →
Implementations
Implementation
Runtime
Best latency
Evidence
Availability
Semantics
Inputs and outputs
xfp32 [batch_size, features, sequence_length]
outfloat [out]
Axes and behavior
outvariable
featuresvariable
batch_sizevariable
sequence_lengthvariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
sha256583d4f318db5…
Sources: KernelBench baseline timings (2026-03-05) · MITlast observed 2026-03-05How records are decidedJSON