Skip to content
KernelIndex
Search⌘K
Sign in

submission 512437

burtenshaw · python · License unknown
Kernel source · 16 lines ↓

Use it

Vendorable · source mirrored · license unknownView source →

No package. Vendor the mirrored source: 16 lines, June 9 Researcher Reciprocity License v1.0.

submission_team_l4_r1_compile_reduce_overhead.py
curl "https://kernelindex.com/api/v1/implementations/kernelbot-prefixsum-v2-512437?include=source"
interfacepython
Compatibility
measured onNVIDIA L4
declared hardwareNVIDIA L4
architecturessm_89
dtypesfp32

Benchmark evidence

1 measurement across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Inclusive prefix sumsuite of 11 cases
NVIDIA L4
9.15ms
#3 of 11
2026-03-02
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:93a180407034818ec18bba7fe042ddca010f845e9ddd0e2af382d583d39c715f
license declaredunknown
license concludedunknown
authorsburtenshaw
imported2026-08-15

Kernel source

submission_team_l4_r1_compile_reduce_overhead.py16 lines
#!POPCORN leaderboard prefixsum_v2
#!POPCORN gpu L4

import torch

from task import input_t, output_t


def _kernel(data: input_t) -> output_t:
    x, y = data
    torch.cumsum(x, dim=0, out=y)
    return y


custom_kernel = torch.compile(_kernel, mode="reduce-overhead")

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 512326.

  #!POPCORN leaderboard prefixsum_v2
- #!POPCORN gpu A100
+ #!POPCORN gpu L4
   
  import torch
+  
  from task import input_t, output_t
   
   
- def custom_kernel(data: input_t) -> output_t:
-     x, _y = data
-     return torch.cumsum(x, dim=0)
+ def _kernel(data: input_t) -> output_t:
+     x, y = data
+     torch.cumsum(x, dim=0, out=y)
+     return y
+  
+  
+ custom_kernel = torch.compile(_kernel, mode="reduce-overhead")

Best evidence level for this revision: reported

JSON