Run 01a02032 · KernelIndex

Primary measurement

3.43msmean

Rank 3 in its comparison group · source-native comparison · observed 2026-03-28

Compare with #1 →

0.90× reference

Identity

implementationgpt-5_cuda_00b2dd

projectgpt-5-2025-08-07

revisionda915083d4c7c5e61aa3005e3d17ae488e0fc71c

workloadnum_pages = 32348 · batch_size = 16 · len_indptr = 17 · num_kv_indices = 23079 · bf16

comparison keysha256:b1ff61d40ca7609c…

sourceFlashInfer-Bench

sha256:0b759ac45a02a7ff99769ed…

Correctness

comparatorflashinfer_bench_eval

max absolute error0.00048828125

max relative error0.04795771837234497

resultpassed

Workload

num_pages32348

batch_size16

len_indptr17

num_kv_indices23079

qbf16 [16, 32, 128]

k_cachebf16 [32348, 1, 4, 128]

v_cachebf16 [32348, 1, 4, 128]

kv_indptrint32 [17]

kv_indicesint32 [23079]

definition comparatorsol_execbench_eval

Measurements

latency · mean3.43 ms

Protocol

harnessflashinfer-bench

timerharness_reported

primaryStatisticmean

correctnessReferencedefinition reference implementation

comparabilityFamilyflashinfer_bench_trace

Environment

gpuNVIDIA B200 (unknown)

cudaToolkit12.8

frameworkpytorch 2.9.1+cu128

Artifacts

logssha256:ff3d6e426981602d… · text/plain · unavailable

Replications and notes

No attestations yet.

Community attestations. They never change the evidence level; only a KernelIndex-controlled rerun does.

Add a reproduction or note

Canonical manifest

Show manifest

{
  "run": {
    "kind": "BenchmarkRun",
    "spec": {
      "status": "passed",
      "timing": {
        "latencyNs": {
          "mean": 3432629
        },
        "primaryStatistic": "mean"
      },
      "evidence": {
        "logs": {
          "uri": "kernelindex:artifact/sha256:ff3d6e426981602d1fd41b7c43cc5b5c23ac10f723e1c585d1505b4eb3aacdbc",
          "digest": "sha256:ff3d6e426981602d1fd41b7c43cc5b5c23ac10f723e1c585d1505b4eb3aacdbc"
        }
      },
      "observedAt": "2026-03-28T15:03:08.056Z",
      "correctness": {
        "comparator": "sol_execbench_eval",
        "maximumAbsoluteError": 0.00048828125,
        "maximumRelativeError": 0.04795771837234497
      },
      "sourceNative": {
        "source": "flashinfer-bench",
        "metrics": {
          "latency_ms": 3.4326292673746743,
          "speedup_factor": 0.8981880800895119,
          "reference_latency_ms": 3.0831466913223267
        },
        "benchmark": "gqa_paged_decode_h32_kv4_d128_ps1"
      },
      "protocolDigest": "sha256:7d44d206ad086b4bece35ce31f6531ebb11f07e0d9dcfb74c75be86d3ec02095",
      "workloadDigest": "sha256:6364d816f6cc87ac5199585699572af54852f5c08159e7997393e22559d2c3ae",
      "environmentDigest": "sha256:af7cb68a0bb0eb9671092a620668c2600f0431301fc46a71807ea75fb7de8bce",
      "implementationDigest": "sha256:3b7023b82a8abc5ad4f3b833025c2f32a6e068b4c14d60cb171d681a82b194ed"
    },
    "metadata": {
      "name": "sol-trace-gqa-paged-decode-h32-kv4-d128-ps1-85c1e8ef",
      "title": "gpt-5_cuda_00b2dd · gqa_paged_decode_h32_kv4_d128_ps1"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  },
  "protocol": {
    "kind": "BenchmarkProtocol",
    "spec": {
      "harness": {
        "name": "flashinfer-bench",
        "repository": "https://github.com/flashinfer-ai/flashinfer-bench"
      },
      "correctness": {
        "reference": "definition reference implementation",
        "comparator": "flashinfer_bench_eval"
      },
      "measurement": {
        "timer": "harness_reported",
        "primaryStatistic": "mean"
      },
      "comparability": {
        "family": "flashinfer_bench_trace"
      }
    },
    "metadata": {
      "name": "flashinfer-bench-trace",
      "title": "FlashInfer-Bench trace harness"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  },
  "environment": {
    "kind": "ExecutionEnvironment",
    "spec": {
      "hardware": {
        "vendor": "nvidia",
        "product": "NVIDIA B200",
        "architecture": "unknown"
      },
      "software": {
        "framework": {
          "name": "pytorch",
          "version": "2.9.1+cu128"
        },
        "libraries": {
          "triton": "3.5.1"
        },
        "cudaToolkit": "12.8"
      }
    },
    "metadata": {
      "name": "sol-trace-nvidia-b200-2-9-1-cu128-3-5-1-12-8",
      "title": "NVIDIA B200 (trace-reported software)"
    },
    "apiVersion": "kernelindex.dev/v1alpha1"
  }
}

Cite this record (permalink, digest, access date)

Report an issue with this run

Published 2026-08-20 · FlashInfer-Bench · Apache-2.0All results for GQA paged decode h32 kv4 d128 ps1 →JSON

gpt-5 / cuda00b2dd