perf(web): reuse incremental build and status caches

my-tools/show_blender_web_scheduler.py

@@ -10,6 +10,7 @@ from __future__ import annotations
 import argparse
 import hashlib
 import json
+import os
 import statistics
 import re
 import time
@@ -179,12 +180,22 @@ def project_progress() -> dict:
         batch_units = 0.0
     remaining_hours = unverified / unit_rate if unit_rate > 0 else None
     stage_durations = {}
+    stage_labels = {}
     for item in productive:
         for index, command in enumerate(item.get("commands", [])):
             duration = command.get("durationMs") if isinstance(command, dict) else None
             if isinstance(duration, (int, float)) and duration >= 0:
-                stage_durations.setdefault(str(index + 1), []).append(duration)
+                stage = str(index + 1)
+                stage_durations.setdefault(stage, []).append(duration)
+                if stage not in stage_labels and isinstance(command.get("command"), str):
+                    stage_labels[stage] = command["command"].split("&&", 1)[0].strip()[:120]
     stage_medians = {stage: round(statistics.median(values)) for stage, values in stage_durations.items() if values}
+    bottleneck_stage = max(stage_medians, key=stage_medians.get) if stage_medians else None
+    bottleneck = {
+        "stage": bottleneck_stage,
+        "label": stage_labels.get(bottleneck_stage) if bottleneck_stage else None,
+        "medianDurationMs": stage_medians.get(bottleneck_stage) if bottleneck_stage else None,
+    }
     historical = max(historical_candidates, default=("", {}), key=lambda item: item[0])[1]
     workflow_modules = product_backlog.get("workflowModules", []) if isinstance(product_backlog.get("workflowModules"), list) else []
     product_tasks = product_backlog.get("tasks", []) if isinstance(product_backlog.get("tasks"), list) else []
@@ -213,6 +224,7 @@ def project_progress() -> dict:
         "reusedBatches": sum(1 for item in receipts if item["reused"]),
         "observedExecutionMs": round(sum(item["durationMs"] for item in productive)),
         "stageMedianDurationMs": stage_medians,
+        "bottleneck": bottleneck,
         "medianIncrementalUnitsPerBatch": round(batch_units, 2),
         "medianIncrementalUnitsPerHour": round(unit_rate, 2),
         "estimatedRemainingHours": round(remaining_hours, 2) if remaining_hours is not None else None,
@@ -286,6 +298,14 @@ def wasm_module_status() -> dict:
     """Report module preload policy and artifact integrity without fetching or rebuilding."""
     manifest_path = ROOT / "web/app/public/wasm-module-manifest.json"
     manifest = _read_json(manifest_path)
+    cache_path = CONTROL_ROOT / "wasm-module-status-cache.json"
+    cache = _read_json(cache_path)
+    cached_entries = cache.get("entries", {}) if isinstance(cache.get("entries"), dict) else {}
+    next_cache = {"schemaVersion": 1, "manifestMtimeNs": None, "entries": {}}
+    try:
+        next_cache["manifestMtimeNs"] = str(manifest_path.stat().st_mtime_ns)
+    except OSError:
+        pass
     entries = []
     for entry in manifest.get("modules", []) if isinstance(manifest.get("modules"), list) else []:
         resources = entry.get("resources", {}) if isinstance(entry, dict) else {}
@@ -296,11 +316,26 @@ def wasm_module_status() -> dict:
             relative = resource["url"].lstrip("/")
             path = ROOT / "web/app/public" / relative
             actual = None
-            if path.is_file():
+            bytes_count = 0
+            mtime_ns = None
+            try:
+                stat = path.stat()
+                bytes_count = stat.st_size
+                mtime_ns = str(stat.st_mtime_ns)
+            except OSError:
+                stat = None
+            cache_key = str(path.relative_to(ROOT))
+            cached = next_cache["entries"].get(cache_key)
+            if not isinstance(cached, dict):
+                cached = cached_entries.get(cache_key, {}) if isinstance(cached_entries.get(cache_key), dict) else {}
+            if stat and cached.get("bytes") == bytes_count and cached.get("mtimeNs") == mtime_ns and isinstance(cached.get("sha256"), str):
+                actual = cached["sha256"]
+            elif stat:
                 actual = hashlib.sha256(path.read_bytes()).hexdigest()
+            next_cache["entries"][cache_key] = {"bytes": bytes_count, "mtimeNs": mtime_ns, "sha256": actual}
             variants[variant] = {
                 "url": resource["url"],
-                "bytes": path.stat().st_size if path.is_file() else 0,
+                "bytes": bytes_count,
                 "integrity": actual == resource.get("sha256") if actual else False,
             }
         entries.append({
@@ -309,6 +344,14 @@ def wasm_module_status() -> dict:
             "preload": entry.get("preload") is True,
             "variants": variants,
         })
+    try:
+        temporary = cache_path.with_suffix(f".tmp-{os.getpid()}")
+        temporary.write_text(json.dumps(next_cache, ensure_ascii=False, separators=(",", ":")) + "\n", encoding="utf-8")
+        temporary.replace(cache_path)
+    except OSError:
+        # Status reporting remains read-only from the caller's perspective if
+        # the control directory is unavailable; the next poll simply rehashes.
+        pass
     return {
         "manifest": "web/app/public/wasm-module-manifest.json" if manifest else None,
         "moduleCount": len(entries),
@@ -340,6 +383,7 @@ def format_progress(progress: dict, fast_path: dict | None = None) -> list[str]:
         f"  source traceability reference: {progress.get('verifiedUnits', 0)}/{progress.get('totalUnits', 0)} bound; excluded from product completion denominator",
         f"  task plan: {plan.get('done', 0)}/{plan.get('tasks', 0)} done, active={','.join(plan.get('active', [])) or 'none'}, blocked={plan.get('blocked', 0)}",
         f"  old verification estimate: remaining={estimate_text}, total={total_text} (reference only; strategy retired)",
+        f"  observed bottleneck: stage={progress.get('bottleneck', {}).get('stage') or 'n/a'} median={progress.get('bottleneck', {}).get('medianDurationMs') or 0}ms command={progress.get('bottleneck', {}).get('label') or 'n/a'}",
         f"  fast path: batch={batch.get('batchId') or 'none'} selected={batch.get('selectedRecordCount', 0)} completed={batch.get('completed', False)}",
         f"  context cache: {context_cache.get('status', 'MISSING')} task={context_cache.get('task') or 'none'} inputs={context_cache.get('inputCount', 0)}; failure ledger entries={ledger.get('entryCount', 0)} hydratedLogs={ledger.get('hydratedLogs', 0)}",
         f"  command budget: max={fast_path.get('commandBudget', {}).get('maxCommands', 48)}, read/exploration max={fast_path.get('commandBudget', {}).get('maxReadCommands', 16)}; verification={fast_path.get('verificationPolicy', 'n/a')}",

my-tools/start_unattended.sh

@@ -20,8 +20,10 @@ if test -d /dev/shm && test -w /dev/shm; then
   export EM_CACHE="${EM_CACHE:-$WEB_RAM_CACHE_ROOT/emcache}"
   export TMPDIR="${TMPDIR:-$WEB_RAM_CACHE_ROOT/tmp}"
 fi
-export CMAKE_BUILD_PARALLEL_LEVEL="${CMAKE_BUILD_PARALLEL_LEVEL:-8}"
-export NINJAFLAGS="${NINJAFLAGS:--j8}"
+WEB_BUILD_JOBS="${WEB_BUILD_JOBS:-$(WEB_BUILD_MEMORY_GIB_PER_JOB=4 WEB_BUILD_JOBS_CAP=16 bash "$ROOT_DIR/tools/web/compute-build-jobs.sh")}"
+export WEB_BUILD_JOBS
+export CMAKE_BUILD_PARALLEL_LEVEL="${CMAKE_BUILD_PARALLEL_LEVEL:-$WEB_BUILD_JOBS}"
+export NINJAFLAGS="${NINJAFLAGS:--j$WEB_BUILD_JOBS}"
 
 mkdir -p "$CONTROL_DIR"