Hugging Face's logo

Datasets:
arm-bench
/
arm-bench-trace

Languages:
English
Size:
n<1K
Tags:
code
kernels
benchmark
arm
aarch64
sve
License:
Dataset card Files
xet
 Community
5
arm-bench-trace / solutions /llama.cpp

Commit History

19 days ago

  1. Llama-3.1-8B e2e set (defs, real-activation workloads, Fable gate-v2 solutions, runs, measurements); Qwen kernel-vs-repack results; fixed baseline-sve2 loop_114/121/219

    awu12
    Verified signature
    def2c3e

23 days ago

  1. fix claude-code-claude-fable-5-1-gatev2: the 5 re-optimized kernels were previous-round sources (assembled by hand); regenerated from the agent run dirs

    awu12
    Verified signature
    0a68d23

  2. gate-v2 set complete: all 5 defective kernels re-optimized by Fable under the corrected gate (geomean 3.205x vs the shortcut versions' 3.160x)

    awu12
    Verified signature
    e675d0a

  3. Fable 5.1 re-optimization of 3 kernels under gate v2: 2.967x / 3.026x / 3.601x vs the shortcut versions' 3.00x / 3.03x / 3.39x

    awu12
    Verified signature
    571b667

  4. fable-exact-bsums: the 2026-09-21 Fable kernels with the batch-shared-exponent block sums repaired (5 of 11 patched); prefill perplexity cost 12.68 -> 8.80 against stock 8.59

    awu12
    Verified signature
    41c5f12

  5. gate v2: real-activation tensors for the 11 Qwen3.5-4B packed gemm definitions, baseline-relative SQNR floors on all 30, exact-bsums control solution

    awu12
    Verified signature
    ae3e8ff

24 days ago

  1. Add claude-code-claude-fable-5-1-sve2 kernels for the Qwen3.5-4B e2e run (11 packed-ABI gemm solutions + raw run dirs), additive

    awu12
    Verified signature
    e11ed9d

  2. Add Qwen3.5-4B e2e gemm definitions (packed ggml Q4_K/Q5_K/Q6_K ABI): 11 definitions + workloads + reference-scalar + baseline-llamacpp-arm (additive)

    awu12
    Verified signature
    fd33c4a

27 days ago

  1. llamacpp gemm bf16 kernels remove -march=sve+sve2 flag

    ArtificialRay
    518f45b

28 days ago

  1. update kleidiai sme2 kernels

    ArtificialRay
    b854496

  1. Fix weak bf16 attention baselines (mha/gqa/mla_prefill/mla_decode): cast ggml inputs to f32 so mul_mat uses NEON f32 SIMD (~3x over scalar); update baselines + traces

    awu12
    Verified signature
    fd0e643

  2. Fix weak expert baselines: bf16 GEMM -> NEON BFMMLA (~20x over scalar), ncnn gemm_w8a8ch create_pipeline hoist (~1.6x); update baselines + traces (8 defs)

    awu12
    Verified signature
    2dd1bd8

  1. Update llamacpp kernel's baseline by split kernel initialization with calculation, now kernel init per workload (but not per-run)

    ArtificialRay
    e09ca6e

  1. Attention kernel update:⁠

    ArtificialRay
    ad52de0

  1. Definition+workload update for rmsnorm, and baseline, reference-scalar, and claude-sonnet-4-6 solution update⁠

    ArtificialRay
    f55190a

  1. Rewrite llama.cpp q8_0 reference-scalar solution signature, now compatible with bench/dataset/raw.py

    ArtificialRay
    6b3c7a1

  2. Add kernel in different precision:⁠

    ArtificialRay
    c09d00d

  1. Add reference-scalar llama.cpp solutions (naive scalar baseline, 16 defs)

    awu12
    Verified signature
    c2269ab

  1. Add llama.cpp baseline solutions (baseline-llamacpp-arm, 16 defs)

    awu12
    Verified signature
    550f1b3