Files
vllm-omni/benchmarks/edge_harness/results/profile_ring3_sliding_fp16.json
T
tzhouamandClaude Opus 5 9264ff81ab Include the baselines, comparisons and study documents, not just the fork
The packaging commit shipped the code and a summary of the numbers. It left out
everything the numbers are measured *against*, which is most of what makes them
checkable.

Added:

  baselines/                 llama.cpp runs (thread sweeps, NUMA sweeps, the
                             interleaved A/B passes behind the headline decode
                             figure) and the quality references: each build
                             against its own bf16 over 12 greedy prompts.
  benchmarks/.../results/    102 artifacts backing the published figures, each
                             with a provenance block naming the model,
                             quantization, thread set and build SHAs.
  docs/engines/              the engine analyses -- llama.cpp, ExecuTorch, MNN,
                             ncnn, vLLM, vLLM-Omni -- and the cross-engine and
                             device-simulation comparisons.
  docs/analysis/             the full study documents the summaries distil.
  docs/edge/comparisons.md   where this fork sits against llama.cpp and why:
                             decode is llama.cpp's, prefill is ours by ~2x, and
                             since neither engine exceeds 39% of the measured
                             222 GB/s ceiling the decode gap is the execution
                             model rather than the kernels.
  docs/edge/environment.md   pins and machine, and the correction below.

On "the modified vllm": there isn't one, and the doc says so plainly because it
is the obvious assumption. The vLLM checkout is clean upstream (0 commits
ahead, 0 uncommitted) and the installed wheel is stock 0.28.0+cpu. This fork
changes vLLM's CPU *behaviour* through vLLM's own extension points at import
time -- three CPU patches in vllm_omni/patch.py, the quantization-config
registry, the IR op registry, platform hooks -- specifically so the changes
survive a wheel upgrade rather than being lost on it. One of those patches is
why "KV sizing is not a memory lever" looked true for months: CpuPlatform
overwrote kv_cache_memory_bytes unconditionally, so the setting had no effect.
There is no patch series to apply; pin the version and let patch.py do the rest.

.gitignore gains explicit exceptions: upstream's blanket `*.json` and `results/`
rules were silently eating the measurement artifacts, which in this fork are the
substance rather than build output. Binary captures stay ignored.

No vLLM or llama.cpp source is redistributed -- only output produced with them.
NOTICE records both.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L1ax5wYLA6XbqEL1i8L9gZ
2026-09-14 14:01:12 +08:00

1 line
9.6 KiB
JSON

{"execution_summary": {"estimated_inference_time": 1486, "estimated_inference_peak_memory": 139599872, "first_load_time": 115214, "first_load_peak_memory": 109510656, "warm_load_time": 114073, "warm_load_peak_memory": 125624320, "compile_time": 0, "compile_peak_memory": 0, "compile_memory_increase_range": [0, 0], "compile_memory_peak_range": [0, 0], "first_load_memory_increase_range": [14961536, 24144032], "first_load_memory_peak_range": [19734528, 28917024], "warm_load_memory_increase_range": [0, 827344], "warm_load_memory_peak_range": [737280, 10669568], "inference_memory_increase_range": [0, 2129968], "inference_memory_peak_range": [0, 9903648], "all_compile_times": [], "all_first_load_times": [115214], "all_warm_load_times": [114073], "all_inference_times": [2310, 1720, 1648, 1682, 1698, 1625, 1628, 1619, 1561, 1514, 1490, 1986, 1815, 1857, 1898, 1837, 1814, 1797, 1788, 1727, 1612, 1642, 1615, 1601, 1504, 1627, 1506, 1494, 1531, 1499, 1491, 1513, 1499, 1565, 1510, 1493, 1497, 1500, 1494, 1496, 1499, 1488, 1494, 1505, 1488, 1494, 1493, 1488, 1493, 1493, 1491, 1520, 1523, 3012, 1500, 1512, 1507, 1502, 1495, 1501, 1496, 1493, 1508, 1494, 1533, 1518, 1497, 1508, 1520, 1501, 1503, 1500, 1499, 1501, 1497, 1577, 1749, 1645, 1707, 1657, 1586, 1646, 1564, 1676, 1557, 1587, 1524, 1519, 1594, 1578, 1517, 1517, 1586, 1545, 1501, 1543, 1492, 1516, 1486, 1496]}, "execution_detail": [{"name": "Input", "type": "Input", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 7431}, {"name": "QNN_DATATYPE_FLOAT_32_converted_x_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 121888}, {"name": "rms_norm_/layers.0/Mul_1_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 18417}, {"name": "/layers.0/q_k_v_proj/MatMul", "type": "/layers.0/q", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/q_k_v_proj/MatMul_post_reshape", "type": "/layers.0/q", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_497", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 262714}, {"name": "node_Slice_500", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 69757}, {"name": "node_Slice_503", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 54734}, {"name": "/layers.0/Reshape", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 488}, {"name": "/layers.0/Reshape_1", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 242}, {"name": "/layers.0/Reshape_2", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/g_proj/MatMul", "type": "/layers.0/g", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/g_proj/MatMul_post_reshape", "type": "/layers.0/g", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Sigmoid", "type": "/layers.0/Sigmoid", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 18138}, {"name": "/layers.0/Reshape_3", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 9962}, {"name": "node_Slice_393", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_396", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_32_converted_cos_sw_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 24465}, {"name": "/layers.0/Mul_2", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1032}, {"name": "/layers.0/Neg", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1068}, {"name": "/layers.0/Concat", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_32_converted_sin_sw_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2060}, {"name": "/layers.0/Mul_3", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 958}, {"name": "/layers.0/Add_1", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 13664}, {"name": "node_Slice_442", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_445", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Mul_4", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1716}, {"name": "/layers.0/Neg_1", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 7234}, {"name": "/layers.0/Concat_1", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Mul_5", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 326}, {"name": "/layers.0/Add_2", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2602}, {"name": "/layers.0/Reshape_4", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 210}, {"name": "QNN_DATATYPE_FLOAT_32_converted_k_cache_0_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 87204}, {"name": "/layers.0/Concat_2", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_32_converted_v_cache_0_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 69504}, {"name": "/layers.0/Concat_3", "type": "/layers.0/Concat", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 25594}, {"name": "/layers.0/Transpose", "type": "/layers.0/Transpose", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/MatMul", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 415559}, {"name": "/layers.0/Mul_6", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 70688}, {"name": "QNN_DATATYPE_FLOAT_32_converted_mask_sw_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1383}, {"name": "/layers.0/Add_3", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 69045}, {"name": "/layers.0/Softmax", "type": "/layers.0/Softmax", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 14385}, {"name": "/layers.0/MatMul_1", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 57686}, {"name": "/layers.0/Reshape_5", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 475}, {"name": "/layers.0/Mul_7", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2854}, {"name": "/layers.0/out_proj/MatMul", "type": "/layers.0/out", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 190749}, {"name": "/layers.0/out_proj/MatMul_post_reshape", "type": "/layers.0/out", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Add_4", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 86211}, {"name": "rms_norm_/layers.0/Mul_9_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 3788}, {"name": "/layers.0/gate_proj/MatMul", "type": "/layers.0/gate", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/gate_proj/MatMul_post_reshape", "type": "/layers.0/gate", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Gelu_0", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 836520}, {"name": "/layers.0/up_proj/MatMul", "type": "/layers.0/up", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 821543}, {"name": "/layers.0/up_proj/MatMul_post_reshape", "type": "/layers.0/up", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Mul_12", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 10772}, {"name": "/layers.0/down_proj/MatMul", "type": "/layers.0/down", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 821530}, {"name": "/layers.0/down_proj/MatMul_post_reshape", "type": "/layers.0/down", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Add_7", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 6788}, {"name": "/layers.0/Add_2_identity", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/Reshape_2_identity", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_16_converted_output_0_QNN_DATATYPE_FLOAT_32", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_16_converted_output_1_QNN_DATATYPE_FLOAT_32", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "QNN_DATATYPE_FLOAT_16_converted_output_2_QNN_DATATYPE_FLOAT_32", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "Output", "type": "Output", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 17655}]}