Files
vllm-omni/benchmarks/edge_harness/results/profile_ring2_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
11 KiB
JSON

{"execution_summary": {"estimated_inference_time": 1480, "estimated_inference_peak_memory": 140103680, "first_load_time": 111241, "first_load_peak_memory": 113639424, "warm_load_time": 115382, "warm_load_peak_memory": 125554688, "compile_time": 0, "compile_peak_memory": 0, "compile_memory_increase_range": [0, 0], "compile_memory_peak_range": [0, 0], "first_load_memory_increase_range": [0, 2062624], "first_load_memory_peak_range": [1265664, 10942624], "warm_load_memory_increase_range": [0, 4168816], "warm_load_memory_peak_range": [3784704, 13497104], "inference_memory_increase_range": [0, 2121776], "inference_memory_peak_range": [0, 8779808], "all_compile_times": [], "all_first_load_times": [111241], "all_warm_load_times": [115382], "all_inference_times": [2204, 1622, 1592, 1598, 1612, 1603, 1548, 1581, 1571, 1685, 2255, 1731, 1547, 1667, 1673, 1719, 1543, 1534, 1513, 1505, 1495, 1494, 1515, 1498, 1492, 1507, 1503, 1529, 1519, 1490, 1489, 1494, 1487, 1491, 1490, 1483, 1495, 1488, 1490, 1489, 1505, 1496, 1497, 1538, 1518, 1532, 1532, 1525, 1530, 1505, 1519, 1521, 1827, 1747, 1985, 1681, 2027, 1567, 1609, 1607, 1751, 1855, 1866, 1829, 1804, 1615, 1550, 1606, 1556, 1560, 1597, 1576, 2375, 1547, 1522, 1544, 1519, 1483, 1494, 1482, 1482, 1497, 1480, 1485, 1503, 1510, 1486, 1511, 1491, 1481, 1498, 1484, 1562, 1670, 1582, 1666, 1716, 1618, 1711, 1642]}, "execution_detail": [{"name": "Input", "type": "Input", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 6592}, {"name": "QNN_DATATYPE_FLOAT_32_converted_x_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 107275}, {"name": "rms_norm_/layers.0/Mul_1_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 23416}, {"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_347", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 251480}, {"name": "node_Slice_350", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 60490}, {"name": "node_Slice_353", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 39159}, {"name": "/layers.0/Reshape", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 482}, {"name": "/layers.0/Reshape_1", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 258}, {"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": 20778}, {"name": "/layers.0/Reshape_3", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 8519}, {"name": "node_Slice_521", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_524", "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": 22527}, {"name": "/layers.0/Mul_2", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1026}, {"name": "/layers.0/Neg", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1024}, {"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": 2223}, {"name": "/layers.0/Mul_3", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 979}, {"name": "/layers.0/Add_1", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 8600}, {"name": "node_Slice_476", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "node_Slice_479", "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": 4362}, {"name": "/layers.0/Neg_1", "type": "/layers.0/Neg", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 15065}, {"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": 341}, {"name": "/layers.0/Add_2", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 2615}, {"name": "/layers.0/Reshape_4", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 250}, {"name": "QNN_DATATYPE_FLOAT_32_converted_k_cache_0_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 146812}, {"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": 315975}, {"name": "/layers.0/Mul_6", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 56548}, {"name": "QNN_DATATYPE_FLOAT_32_converted_mask_sw_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 6455}, {"name": "/layers.0/Add_3", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 72308}, {"name": "node_Reshape_2", "type": "node", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 0}, {"name": "/layers.0/MatMul_1", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 24850}, {"name": "/layers.0/Mul_7", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 11607}, {"name": "/layers.0/ReduceMax", "type": "/layers.0/ReduceMax", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5899}, {"name": "/layers.0/Max", "type": "/layers.0/Max", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 3265}, {"name": "/layers.0/Sub", "type": "/layers.0/Sub", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 18352}, {"name": "/layers.0/Exp", "type": "/layers.0/Exp", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 6647}, {"name": "/layers.0/Sub_1", "type": "/layers.0/Sub", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 910}, {"name": "/layers.0/Exp_1", "type": "/layers.0/Exp", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 3280}, {"name": "/layers.0/ReduceSum", "type": "/layers.0/ReduceSum", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 9700}, {"name": "/layers.0/Add_4", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 509}, {"name": "QNN_DATATYPE_FLOAT_32_converted_v_cache_0_QNN_DATATYPE_FLOAT_16", "type": "QNN", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 95965}, {"name": "/layers.0/MatMul_2", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 40017}, {"name": "/layers.0/MatMul_3", "type": "/layers.0/MatMul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 8930}, {"name": "/layers.0/Add_5", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 890}, {"name": "/layers.0/Div_1", "type": "/layers.0/Div", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 8841}, {"name": "/layers.0/Reshape_5", "type": "/layers.0/Reshape", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 1007}, {"name": "/layers.0/Mul_8", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 893}, {"name": "/layers.0/out_proj/MatMul", "type": "/layers.0/out", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 158075}, {"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_6", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 85708}, {"name": "rms_norm_/layers.0/Mul_10_output_0", "type": "rms", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 5215}, {"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": 796342}, {"name": "/layers.0/up_proj/MatMul", "type": "/layers.0/up", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 787899}, {"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_13", "type": "/layers.0/Mul", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 12502}, {"name": "/layers.0/down_proj/MatMul", "type": "/layers.0/down", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 796088}, {"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_9", "type": "/layers.0/Add", "compute_unit": "NPU", "execution_time": 0, "execution_cycles": 6893}, {"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": 21746}]}