Skip to content
KernelIndex
Search⌘K
Sign in

RMSNorm

4 eligible runs
rmsnorm

KernelBench level1 problem 36: RMSNorm. The computation is the reference PyTorch module's forward pass; the output shape follows the module (mirrored as each implementation's source).

Fastest reported
7.84ms±0.01 · mean of 100 · 1.81× faster than baseline
torch.compile (inductor)PyTorch · MIT · python

Reported evidence · last observed 2026-03-05. Reported by source; not independently reproduced.

No package · vendor the source →View source →Run detail →

Current records

HardwareBest knownImplementationRuns
NVIDIA H100 · env 17.84 mstorch.compile (inductor)2NVIDIA H100 · env 212.3 mstorch.compile (inductor)2

Not measured on B200 for this workload. Challenges →

Source-native comparison · GPU NVIDIA H100 · Workload dim1 = 512 · dim2 = 512 · features = 64 · batch_size = 112 · fp32 · Protocol KernelBench timing scripts · mean · 2 results · last observed 2026-03-05Record history →
Estimated floor 2.24 ms · record 3.49× above itestimate, not evidence ›
DRAM 2.24 ms · 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
1
torch.compile (inductor)PyTorch
7.84ms±0.01
1.00×
Reported · MIT · source
2026-03-05stale

1.81× faster than the baseline. Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredMITno install recipeView source →Run detail →
2
PyTorch eagerbaselinePyTorch
14.2ms±0.05
1.81×
Reported · MIT · source
2026-03-05stale

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
PyTorch eagerPyTorch
python · torch_eager
14.2ms
1.81×
Reported
MIT · source
torch.compile (inductor)PyTorch
python · torch_compile_inductor
7.84ms
1.00×
Reported
MIT · source

Semantics

Inputs and outputs
xfp32 [batch_size, features, dim1, dim2]
outfloat [out]
Axes and behavior
outvariable
dim1variable
dim2variable
featuresvariable
batch_sizevariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
sha25688f4004d0694…
Sources: KernelBench baseline timings (2026-03-05) · MITlast observed 2026-03-05How records are decidedJSON