Filter usage timing overview to benchmark records

.github/workflows/result-server-tests.yml

@@ -79,6 +79,7 @@ jobs:
           bash scripts/tests/test_scheduler_extra_args.sh
           bash scripts/tests/test_send_results_profile_data.sh
           bash scripts/tests/test_send_estimate_artifacts.sh
+          bash scripts/tests/test_estimation_run_timing.sh
           bash scripts/tests/test_estimation_gpu_kernel_ensemble_average.sh
           bash scripts/tests/test_estimation_gpu_kernel_lightgbm_v10.sh
           bash scripts/tests/test_estimation_gpu_kernel_mlp_v15.sh

docs/cx/ESTIMATE_JSON_SPEC.md

@@ -176,6 +176,7 @@ In addition, `target_nodes` represents the estimated node count on each system s
 将来拡張として、Estimate JSON は以下の項目を持ってよい。
 
 - `estimate_metadata`
+- `estimation_timing`
 - `measurement`
 - `assumptions`
 - `input_artifacts`
@@ -310,7 +311,37 @@ When the source benchmark Result JSON carries `input_info`, that auxiliary input
 `estimation_package` and `estimation_package_version` identify the package that was actually applied.
 `requested_estimation_package` and `requested_estimation_package_version` identify the package initially requested before any fallback.
 
-### 6.2 measurement
+### 6.2 estimation_timing
+
+推定処理自体の実行時間を保持する任意項目。
+
+想定項目:
+
+- `schema_version`
+- `elapsed_time`
+- `unit`
+- `recorded_by`
+
+例:
+
+```json
+{
+  "estimation_timing": {
+    "schema_version": 1,
+    "elapsed_time": 42,
+    "unit": "s",
+    "recorded_by": "scripts/estimation/run.sh"
+  }
+}
+```
+
+`elapsed_time` は推定ジョブ内で app の `estimate.sh` 実行に要した wall-clock 秒数を表す。
+これは推定された benchmark 実行時間ではなく、推定処理そのものの運用コストを観測するための値である。
+
+`elapsed_time` records wall-clock seconds spent running the app's `estimate.sh` inside the estimate job.
+It is the operational cost of producing the estimate, not the estimated benchmark runtime.
+
+### 6.3 measurement
 
 推定入力となった計測方法や採取方式を保持する。
 
@@ -338,7 +369,7 @@ When the source benchmark Result JSON carries `input_info`, that auxiliary input
 
 This field stores how the measurement inputs used for estimation were obtained.
 
-### 6.3 model
+### 6.4 model
 
 推定モデルの識別情報を保持する。
 
@@ -378,7 +409,7 @@ For example, `current_system.model` may retain either an `intra_system_scaling_m
 
 When needed, a side-specific `model` may contain `source_system`, `target_system`, and `system_compatibility_rule`.
 
-### 6.4 assumptions
+### 6.5 assumptions
 
 推定時の仮定を保持する。
 
@@ -409,7 +440,7 @@ This field may include assumptions such as:
 - whether a communication-cost adjustment is applied
 - how problem size is increased
 
-### 6.5 applicability
+### 6.6 applicability
 
 推定方式に必要な入力が十分だったか、不足があったか、フォールバックが行われたかを保持する。
 
@@ -462,7 +493,7 @@ In such a case, `estimate_metadata.requested_estimation_package` identifies the
 
 This field records the final applicability state of the estimate, whether fallback was used, and what was missing.
 
-### 6.6 confidence
+### 6.7 confidence
 
 推定結果の信頼度や品質指標を保持する。
 

result_server/templates/_usage_report_performance_telemetry_section.html

@@ -1,32 +1,48 @@
 {% set performance_telemetry = performance_telemetry|default({
   "summary": {
     "result_count": 0,
+    "ignored_result_count": 0,
     "timing_record_count": 0,
     "profiled_result_count": 0,
+    "regular_run_timing_count": 0,
+    "profiled_run_timing_count": 0,
+    "estimate_record_count": 0,
+    "estimate_timing_record_count": 0,
     "build_cache_record_count": 0,
     "build_cache_hit_count": 0,
     "build_cache_miss_count": 0,
     "build_cache_store_count": 0,
     "avg_build_time": "-",
     "avg_queue_time": "-",
-    "avg_run_time": "-"
+    "avg_run_time": "-",
+    "avg_regular_run_time": "-",
+    "avg_profiled_run_time": "-",
+    "avg_estimate_time": "-"
   },
   "rows": []
 }, true) %}
 
 <section class="page-card table-card">
   <h2 class="section-title">Execution Timing Overview</h2>
-  <p class="section-intro">Best-effort timing and build-cache telemetry from stored Result JSON. These values summarize observed CI stages; profiler overhead requires a paired non-profile run.</p>
+  <p class="section-intro">Operator view for choosing trigger scope/frequency and improving CI and build-cache flow. It summarizes observed timing and build-cache telemetry from benchmark Result JSON; profiler overhead requires a paired non-profile run.</p>
 
   <div class="diagnostic-grid">
     <div class="diagnostic-card">
       <h3>Timing Records</h3>
-      <p>{{ performance_telemetry.summary.timing_record_count }} with timing / {{ performance_telemetry.summary.result_count }} results</p>
+      <p>{{ performance_telemetry.summary.timing_record_count }} with timing / {{ performance_telemetry.summary.result_count }} benchmark results</p>
     </div>
     <div class="diagnostic-card">
       <h3>Average Timing</h3>
       <p>build {{ performance_telemetry.summary.avg_build_time }} / queue {{ performance_telemetry.summary.avg_queue_time }} / run {{ performance_telemetry.summary.avg_run_time }}</p>
     </div>
+    <div class="diagnostic-card">
+      <h3>Run Split</h3>
+      <p>regular {{ performance_telemetry.summary.avg_regular_run_time }} / profiled {{ performance_telemetry.summary.avg_profiled_run_time }}</p>
+    </div>
+    <div class="diagnostic-card">
+      <h3>Estimation Runtime</h3>
+      <p>{{ performance_telemetry.summary.estimate_timing_record_count }} with timing / {{ performance_telemetry.summary.estimate_record_count }} estimates; avg {{ performance_telemetry.summary.avg_estimate_time }}</p>
+    </div>
     <div class="diagnostic-card">
       <h3>Build Cache</h3>
       <p>{{ performance_telemetry.summary.build_cache_hit_count }} hit / {{ performance_telemetry.summary.build_cache_miss_count }} miss / {{ performance_telemetry.summary.build_cache_store_count }} stored</p>
@@ -43,6 +59,8 @@ <h3>Build Cache</h3>
           <th>Results</th>
           <th>Average Timing</th>
           <th>Latest Timing</th>
+          <th>Run Split</th>
+          <th>Estimation</th>
           <th>Build Cache</th>
           <th>Latest Result</th>
         </tr>
@@ -65,7 +83,19 @@ <h3>Build Cache</h3>
           <td>
             build {{ row.latest_build_time }}
             <span class="profile-usage-subline">queue {{ row.latest_queue_time }}</span>
-            <span class="profile-usage-subline">run {{ row.latest_run_time }}</span>
+            <span class="profile-usage-subline">{{ row.latest_run_kind }} run {{ row.latest_run_time }}</span>
+          </td>
+          <td>
+            regular {{ row.avg_regular_run_time }}
+            <span class="profile-usage-subline">{{ row.regular_run_timing_count }} timing records</span>
+            <span class="profile-usage-subline">profiled {{ row.avg_profiled_run_time }}</span>
+            <span class="profile-usage-subline">{{ row.profiled_run_timing_count }} timing records</span>
+          </td>
+          <td>
+            {{ row.estimate_count }} estimates
+            <span class="profile-usage-subline">{{ row.estimate_timing_count }} timing records</span>
+            <span class="profile-usage-subline">avg {{ row.avg_estimate_time }}</span>
+            <span class="profile-usage-subline">latest {{ row.latest_estimate_elapsed_time }}</span>
           </td>
           <td>
             {{ row.build_cache_hit_count }} hit / {{ row.build_cache_miss_count }} miss

result_server/templates/usage_report.html

@@ -160,16 +160,18 @@
       white-space: normal;
     }
     .performance-telemetry-table {
-      min-width: 1080px;
+      min-width: 1420px;
       table-layout: fixed;
     }
     .performance-telemetry-table th:nth-child(1) { width: 140px; }
     .performance-telemetry-table th:nth-child(2) { width: 140px; }
     .performance-telemetry-table th:nth-child(3) { width: 150px; }
     .performance-telemetry-table th:nth-child(4) { width: 180px; }
     .performance-telemetry-table th:nth-child(5) { width: 180px; }
-    .performance-telemetry-table th:nth-child(6) { width: 170px; }
-    .performance-telemetry-table th:nth-child(7) { width: 220px; }
+    .performance-telemetry-table th:nth-child(6) { width: 190px; }
+    .performance-telemetry-table th:nth-child(7) { width: 160px; }
+    .performance-telemetry-table th:nth-child(8) { width: 170px; }
+    .performance-telemetry-table th:nth-child(9) { width: 220px; }
     .performance-telemetry-table td {
       vertical-align: top;
       white-space: normal;

result_server/tests/test_performance_telemetry.py

@@ -13,6 +13,8 @@ def _write_json(path, data):
 
 
 def test_performance_telemetry_summarizes_timing_and_build_cache(tmp_path):
+    estimated_dir = tmp_path / "estimated"
+    estimated_dir.mkdir()
     _write_json(
         tmp_path / "result_20260902_010101_aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee.json",
         {
@@ -45,46 +47,117 @@ def test_performance_telemetry_summarizes_timing_and_build_cache(tmp_path):
             "FOM": 1.0,
         },
     )
+    _write_json(
+        tmp_path / "result_20260901_030303_dddddddd-bbbb-cccc-dddd-eeeeeeeeeeee.json",
+        {
+            "code": "mtls-docker-runner",
+            "system": None,
+        },
+    )
+    _write_json(
+        tmp_path / "result_20260901_040404_eeeeeeee-bbbb-cccc-dddd-eeeeeeeeeeee.json",
+        {
+            "code": "diagnostic-tool",
+            "system": "Fugaku",
+            "pipeline_timing": {"build_time": 1},
+        },
+    )
+    _write_json(
+        tmp_path / "result_20260901_050505_ffffffff-bbbb-cccc-dddd-eeeeeeeeeeee.json",
+        {
+            "code": "../qws",
+            "system": "Fugaku",
+            "pipeline_timing": {"build_time": 1},
+        },
+    )
+    _write_json(
+        estimated_dir / "estimate_20260903_010101_11111111-bbbb-cccc-dddd-eeeeeeeeeeee.json",
+        {
+            "code": "qws",
+            "exp": "CASE1",
+            "estimate_metadata": {
+                "source_result": {
+                    "system": "Fugaku",
+                },
+            },
+            "estimation_timing": {
+                "elapsed_time": 42,
+                "unit": "s",
+            },
+        },
+    )
+    _write_json(
+        estimated_dir / "estimate_20260903_020202_22222222-bbbb-cccc-dddd-eeeeeeeeeeee.json",
+        {
+            "code": "diagnostic-tool",
+            "exp": "CASE0",
+            "estimate_metadata": {
+                "source_result": {
+                    "system": "Fugaku",
+                },
+            },
+            "estimation_timing": {
+                "elapsed_time": 999,
+                "unit": "s",
+            },
+        },
+    )
 
-    telemetry = build_performance_telemetry(str(tmp_path))
+    telemetry = build_performance_telemetry(str(tmp_path), str(estimated_dir))
 
-    assert telemetry["summary"]["result_count"] == 3
+    assert telemetry["summary"]["result_count"] == 2
+    assert telemetry["summary"]["ignored_result_count"] == 4
     assert telemetry["summary"]["timing_record_count"] == 2
     assert telemetry["summary"]["profiled_result_count"] == 1
+    assert telemetry["summary"]["regular_run_timing_count"] == 1
+    assert telemetry["summary"]["profiled_run_timing_count"] == 1
+    assert telemetry["summary"]["estimate_record_count"] == 1
+    assert telemetry["summary"]["estimate_timing_record_count"] == 1
     assert telemetry["summary"]["build_cache_record_count"] == 2
     assert telemetry["summary"]["build_cache_hit_count"] == 1
     assert telemetry["summary"]["build_cache_miss_count"] == 1
     assert telemetry["summary"]["build_cache_store_count"] == 1
     assert telemetry["summary"]["avg_build_time"] == "1m"
     assert telemetry["summary"]["avg_queue_time"] == "1m"
     assert telemetry["summary"]["avg_run_time"] == "3.5m"
+    assert telemetry["summary"]["avg_regular_run_time"] == "5m"
+    assert telemetry["summary"]["avg_profiled_run_time"] == "2m"
+    assert telemetry["summary"]["avg_estimate_time"] == "42s"
 
     rows = {(row["code"], row["system"]): row for row in telemetry["rows"]}
+    assert set(rows) == {("qws", "Fugaku")}
     qws = rows[("qws", "Fugaku")]
     assert qws["result_count"] == 2
     assert qws["timing_count"] == 2
     assert qws["profiled_count"] == 1
     assert qws["avg_build_time"] == "1m"
     assert qws["avg_run_time"] == "3.5m"
+    assert qws["avg_regular_run_time"] == "5m"
+    assert qws["avg_profiled_run_time"] == "2m"
     assert qws["latest_exp"] == "CASE1"
     assert qws["latest_build_time"] == "30s"
     assert qws["latest_queue_time"] == "1m"
     assert qws["latest_run_time"] == "2m"
+    assert qws["latest_run_kind"] == "profiled"
+    assert qws["estimate_count"] == 1
+    assert qws["estimate_timing_count"] == 1
+    assert qws["avg_estimate_time"] == "42s"
+    assert qws["latest_estimate_elapsed_time"] == "42s"
+    assert qws["latest_estimate_exp"] == "CASE1"
     assert qws["latest_build_cache_status"] == "hit"
     assert qws["build_cache_hit_count"] == 1
     assert qws["build_cache_miss_count"] == 1
     assert qws["build_cache_store_count"] == 1
 
-    genesis = rows[("genesis", "RIKYU")]
-    assert genesis["timing_count"] == 0
-    assert genesis["avg_build_time"] == "-"
-    assert genesis["latest_build_cache_status"] == "-"
-
 
 def test_performance_telemetry_handles_missing_directory(tmp_path):
     telemetry = build_performance_telemetry(str(tmp_path / "missing"))
 
     assert telemetry["summary"]["result_count"] == 0
+    assert telemetry["summary"]["ignored_result_count"] == 0
     assert telemetry["summary"]["timing_record_count"] == 0
+    assert telemetry["summary"]["estimate_record_count"] == 0
+    assert telemetry["summary"]["estimate_timing_record_count"] == 0
     assert telemetry["summary"]["avg_build_time"] == "-"
+    assert telemetry["summary"]["avg_estimate_time"] == "-"
     assert telemetry["rows"] == []

result_server/tests/test_portal_list_templates.py

@@ -669,13 +669,20 @@ def test_usage_report_evidence_snapshot_consolidates_coverage_and_quality():
                     "result_count": 1,
                     "timing_record_count": 1,
                     "profiled_result_count": 0,
+                    "regular_run_timing_count": 1,
+                    "profiled_run_timing_count": 0,
+                    "estimate_record_count": 1,
+                    "estimate_timing_record_count": 1,
                     "build_cache_record_count": 1,
                     "build_cache_hit_count": 1,
                     "build_cache_miss_count": 0,
                     "build_cache_store_count": 0,
                     "avg_build_time": "30s",
                     "avg_queue_time": "1m",
                     "avg_run_time": "2m",
+                    "avg_regular_run_time": "2m",
+                    "avg_profiled_run_time": "-",
+                    "avg_estimate_time": "42s",
                 },
                 "rows": [
                     {
@@ -684,12 +691,22 @@ def test_usage_report_evidence_snapshot_consolidates_coverage_and_quality():
                         "result_count": 1,
                         "timing_count": 1,
                         "profiled_count": 0,
+                        "regular_run_timing_count": 1,
+                        "profiled_run_timing_count": 0,
+                        "estimate_count": 1,
+                        "estimate_timing_count": 1,
                         "avg_build_time": "30s",
                         "avg_queue_time": "1m",
                         "avg_run_time": "2m",
+                        "avg_regular_run_time": "2m",
+                        "avg_profiled_run_time": "-",
+                        "avg_estimate_time": "42s",
                         "latest_build_time": "30s",
                         "latest_queue_time": "1m",
                         "latest_run_time": "2m",
+                        "latest_run_kind": "regular",
+                        "latest_estimate_elapsed_time": "42s",
+                        "latest_estimate_exp": "CASE0",
                         "build_cache_hit_count": 1,
                         "build_cache_miss_count": 0,
                         "build_cache_store_count": 0,
@@ -735,7 +752,10 @@ def test_usage_report_evidence_snapshot_consolidates_coverage_and_quality():
 
     assert "Evidence Snapshot" in html
     assert "Execution Timing Overview" in html
+    assert "Operator view for choosing trigger scope/frequency and improving CI and build-cache flow" in html
     assert "build 30s / queue 1m / run 2m" in html
+    assert "regular 2m / profiled -" in html
+    assert "1 with timing / 1 estimates; avg 42s" in html
     assert "1 hit / 0 miss" in html
     assert "Result / Quality" in html
     assert "Application/System Coverage" not in html