submission 66466
shellsmile15795 · python · License unknown
Use it
Vendorable · source mirrored · license unknownView source →
No package. Vendor the mirrored source: 24 lines, June 9 Researcher Reciprocity License v1.0.
submission.py
curl "https://kernelindex.com/api/v1/implementations/kernelbot-grayscale-v2-66466?include=source"interfacepython
Compatibility
measured onNVIDIA B200
declared hardwareNVIDIA B200
architecturessm_100
dtypesfp32
Benchmark evidence
1 measurement across 1 GPU, fastest first.
Operation / workload
Hardware
Latency
Rank
Observed
Reported · How evidence levels are derived →
Source and license
sourceavailable
revision digestsha256:86705bf5ff096e203170f3038850e9e011a22feff442fdae3776be08d15b6a6c
license declaredunknown
license concludedunknown
authorsshellsmile15795
imported2026-08-15
Kernel source
submission.py24 lines
import torch
from task import input_t, output_t
# Standard luminance coefficients expressed as Python floats.
_WEIGHT_R = 0.2989
_WEIGHT_G = 0.5870
_WEIGHT_B = 0.1140
def custom_kernel(data: input_t) -> output_t:
data, output = data
# Avoid allocating temporary tensors by writing directly into the provided output buffer.
# Using in-place arithmetic keeps the computation bandwidth-bound and reuses the output memory.
r = data[..., 0]
g = data[..., 1]
b = data[..., 2]
torch.mul(r, _WEIGHT_R, out=output)
output.add_(g, alpha=_WEIGHT_G)
output.add_(b, alpha=_WEIGHT_B)
return output
Source code from GPU Mode and the KernelBot dataset · June 9 Researcher Reciprocity License v1.0
Changes from previous submission
Against this author's previous submission submission 66460.
- from task import input_t, output_timport torch+ from task import input_t, output_t+ # Standard luminance coefficients expressed as Python floats.+ _WEIGHT_R = 0.2989+ _WEIGHT_G = 0.5870+ _WEIGHT_B = 0.1140++def custom_kernel(data: input_t) -> output_t:data, output = data- weights = torch.tensor([0.2989, 0.5870, 0.1140],- device=data.device,- dtype=data.dtype)- output[...] = torch.sum(data * weights, dim=-1)++ # Avoid allocating temporary tensors by writing directly into the provided output buffer.+ # Using in-place arithmetic keeps the computation bandwidth-bound and reuses the output memory.+ r = data[..., 0]+ g = data[..., 1]+ b = data[..., 2]++ torch.mul(r, _WEIGHT_R, out=output)+ output.add_(g, alpha=_WEIGHT_G)+ output.add_(b, alpha=_WEIGHT_B)+return output
scrolls · 28 diff lines total
Best evidence level for this revision: reported
JSON