Skip to content
KernelIndex
Search⌘K

submission 695367

MatrixGod-max · python · License unknown

Use it

Vendorable · source mirrored · license unknownView source →

No package. Vendor the mirrored source: 21 lines, June 9 Researcher Reciprocity License v1.0.

mxfp4_mm.py
curl "https://kernelindex.com/api/v1/implementations/kernelbot-amd-mxfp4-mm-695367?include=source"
interfacepython
Compatibility
measured onAMD Instinct MI355X
declared hardwareAMD Instinct MI355X
architecturesgfx950
dtypesbf16, mxfp4

Benchmark evidence

1 measurement across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
AMD MXFP4 GEMMsuite of 6 cases
AMD Instinct MI355X
23.8µs
#807 of 1143
2026-04-02

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:603aee14fb46d9a9dc5dd9c97d8f6bd1b27e4756f8b981b1d6934e6118e06ad5
license declaredunknown
license concludedunknown
authorsMatrixGod-max
imported2026-08-26

Kernel source

mxfp4_mm.py21 lines
from task import input_t, output_t
import aiter
from aiter import dtypes
from aiter.ops.triton.quant import dynamic_mxfp4_quant
from aiter.utility.fp4_utils import e8m0_shuffle


def custom_kernel(data: input_t) -> output_t:
    A, _, _, B_shuffle, B_scale_sh = data

    A_fp4, A_scale = dynamic_mxfp4_quant(A)

    return aiter.gemm_a4w4(
        A_fp4.view(dtypes.fp4x2),
        B_shuffle,
        e8m0_shuffle(A_scale).view(dtypes.fp8_e8m0),
        B_scale_sh,
        dtype=dtypes.bf16,
        bpreshuffle=True,
    )

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

Best evidence level for this revision: reported

JSON