fix - Change flops calculation logic

[baominghelly]

Description

Change flops calculation logic - add average latency calculation.

Test evidence

Op test succeeds.

python main.py format_input_matmul.json 
2026-03-10 17:01:09,503 - infinimetrics.utils.input_loader - INFO - Loaded 1 input(s) from format_input_matmul.json
2026-03-10 17:01:09,503 - infinimetrics.dispatcher - INFO - Processing 1 valid inputs (skipped 0 invalid)
2026-03-10 17:01:12,219 - infinimetrics.dispatcher - INFO - Validation complete: 1 valid, 0 skipped
2026-03-10 17:01:12,219 - infinimetrics.dispatcher - INFO - [1/1] Executing operator.InfiniCore.Matmul.Performance
2026-03-10 17:01:12,219 - infinimetrics.executor - INFO - Executor: Running operator.InfiniCore.Matmul.Performance
2026-03-10 17:01:12,219 - infinimetrics.operators.infinicore_adapter - INFO - InfiniCoreAdapter: Processing operator.InfiniCore.Matmul.Performance
🚀 Mode: Dynamic Execution
InfiniCore Operator Test Runner
Directory: Dynamic (1 cases)
Tests found: 1

✅  test_Matmul_0: PASSED (code: 0)

============================================================
Testing Matmul on NVIDIA
============================================================
TestCase(Auto-Gen: Matmul - inputs=[in_0: tensor(16, 4096, 4096), float16; in_1: tensor(16, 4096, 4096), float16], kwargs={out=output: tensor(16, 4096, 4096), float16})
    PyTorch    time - Host: 32.191522 ms, Device: 35.200121 ms
    InfiniCore time - Host: 32.398386 ms, Device: 34.822636 ms
✓ Passed

============================================================
TEST SUMMARY
Total tests: 1
Passed: 1
Success rate: 100.0%

All tests passed!
------------------------------------------------------------
BENCHMARK SUMMARY
PyTorch Host Total Time: 3219.152 ms
PyTorch Device Total Time: 3520.012 ms
InfiniCore Host Total Time: 3239.839 ms
InfiniCore Device Total Time: 3482.264 ms
Host Speedup (PyTorch/InfiniCore): 0.99x
Device Speedup (PyTorch/InfiniCore): 1.01x
============================================================
💾 Saving to: test_report_20260310_170129_863.json
   ✅ Saved.
----------------------------------------

================================================================================
CUMULATIVE TEST SUMMARY
================================================================================
Total tests run: 1
Passed: 1
Failed: 0
----------------------------------------
BENCHMARK SUMMARY (1 cases):
----------------------------------------

✅ PASSED OPERATORS (1):
  test_Matmul_0

Success rate: 100.0%

🎉 All tests passed!
2026-03-10 17:01:30,510 - infinimetrics.executor - INFO - Executor: operator.InfiniCore.Matmul.Performance completed with code=0
2026-03-10 17:01:30,512 - infinimetrics.dispatcher - INFO - Summary saved to summary_output/dispatcher_summary_20260310_170130.json

============================================================
Test Summary
============================================================
Total tests:   1
Successful:    1
Failed:        0
Success rate:  100.0%
============================================================

Test output

{
  "run_id": "test.matmul.perf.2048",
  "time": "2026-03-10 17:01:29",
  "testcase": "operator.InfiniCore.Matmul.Performance",
  "environment": {
    "cluster_scale": 1,
    "topology": "0x1 ring mesh",
    "cluster": [
      {
        "machine": {
          "cpu_model": "13th Gen Intel(R) Core(TM) i7-13700F",
          "memory_gb": 7,
          "accelerators": [
            {
              "model": "NVIDIA GeForce RTX 4070",
              "count": 0,
              "memory_gb_per_card": 11,
              "driver": "591.55",
              "cuda": "12.1",
              "type": "nvidia"
            }
          ]
        },
        "framework": [
          {
            "name": "unknown",
            "version": "unknown"
          }
        ]
      }
    ]
  },
  "result_code": 0,
  "config": {
    "operator": "matmul",
    "device": "nvidia",
    "torch_op": "torch.matmul",
    "infinicore_op": "infinicore.matmul",
    "inputs": [
      {
        "name": "in_0",
        "shape": [
          16,
          4096,
          4096
        ],
        "dtype": "float16"
      },
      {
        "name": "in_1",
        "shape": [
          16,
          4096,
          4096
        ],
        "dtype": "float16"
      }
    ],
    "attributes": [],
    "outputs": [
      {
        "name": "output",
        "shape": [
          16,
          4096,
          4096
        ],
        "dtype": "float16"
      }
    ],
    "warmup_iterations": 10,
    "measured_iterations": 100,
    "tolerance": {
      "atol": 0.001,
      "rtol": 0.001
    },
    "output_dir": "./output",
    "_testcase": "operator.InfiniCore.Matmul.Performance",
    "_run_id": "test.matmul.perf.2048",
    "_time": null
  },
  "metrics": [
    {
      "name": "operator.latency",
      "value": 34.822636,
      "type": "scalar",
      "raw_data_url": "",
      "unit": "ms"
    },
    {
      "name": "operator.tensor_accuracy",
      "value": "PASS",
      "unit": ""
    },
    {
      "name": "operator.flops",
      "value": 63.1492,
      "type": "scalar",
      "raw_data_url": "",
      "unit": "TFLOPS"
    },
    {
      "name": "operator.bandwidth",
      "value": 46.2519,
      "type": "scalar",
      "raw_data_url": "",
      "unit": "GB/s"
    }
  ]
}