Skip to content
KernelIndex
Search⌘K

submission 856715

YUE SHUI · python · License unknown

Use it

Vendorable · source mirrored · license unknownView source →

No package. Vendor the mirrored source: 22 lines, June 9 Researcher Reciprocity License v1.0.

cand_route_cusolver_eq1024.py
curl "https://kernelindex.com/api/v1/implementations/kernelbot-eigh-856715?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
NVIDIA B200
53.8ms
#212 of 286
2026-07-05

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:e67f771164f0ccf8bc6a33f2f1ba692b5854e0666a94ba4921b8b06743e64846
license declaredunknown
license concludedunknown
authorsYUE SHUI
imported2026-08-26

Kernel source

cand_route_cusolver_eq1024.py22 lines
#!POPCORN leaderboard eigh
#!POPCORN gpu B200

import torch
from task import input_t, output_t

@torch.no_grad()
def custom_kernel(data: input_t) -> output_t:
    n = data.shape[-1]
    if n == 1024:
        try:
            torch.backends.cuda.preferred_linalg_library("cusolver")
        except Exception:
            pass
    else:
        try:
            torch.backends.cuda.preferred_linalg_library("magma")
        except Exception:
            pass
    values, vectors = torch.linalg.eigh(data)
    return vectors, values

Source code from GPU Mode and the KernelBot dataset · June 9 Researcher Reciprocity License v1.0

Best evidence level for this revision: reported

JSON