perf(diag): per-frame history export + checkpoint LOH/cache counters + soak capped mode (2026-07-24 audit review)

An adversarial performance review found our own instruments cannot
measure the project's own performance gates:

- FrameProfiler aggregated CPU/GPU/alloc/stage samples into ~5-second
  windows and reset the ring buffers after each report, so route-wide
  p50/p95/p99 distributions across a whole soak could not be
  reconstructed after the fact. ACDREAM_FRAME_HISTORY=<path> now opts
  into a separate per-frame history (one record per frame, ~72
  bytes/record, accumulated in memory with zero frame-thread I/O) that
  a shutdown-only Dispose() writes as CSV. The aggregated [frame-prof]
  report format and its existing metrics are unchanged.

- The canonical checkpoint JSON tracked cache residency (entry/byte
  counts) but never LOH size/fragmentation, process-wide allocated
  bytes, or cache hit/miss/eviction traffic — a committed audit JSON
  showed 65% LOH fragmentation that no tracked instrument recorded,
  and "does a revisit portal hit or miss the caches" was unanswerable
  from an artifact alone. WorldLifecycleResourceSnapshot now carries
  loh_size_bytes/loh_fragmentation_bytes (GCMemoryInfo.GenerationInfo
  index 3), process_total_allocated_bytes (GC.GetTotalAllocatedBytes),
  and Interlocked hit/miss/eviction counters for the CPU mesh cache,
  decoded-texture cache, and the four bounded DAT-object caches
  (portal/cell/highRes/language, aggregated).

- run-connected-r6-soak.ps1 unconditionally forced
  ACDREAM_UNCAPPED_RENDER=1 with no capped mode, while its sibling
  lifecycle-gate script correctly gated it behind a switch. Added
  -Uncapped (default capped, matching the sibling script's pattern),
  fixed the stationary dwell (12s -> 26s, past the 25s
  LiveEntityLivenessController deadline the adjacent comment already
  cited), and now write an env-disclosure.json into the automation
  artifact directory before every launch listing every ACDREAM_* var
  the script sets plus -Uncapped, since the prior audit could only see
  ACDREAM_DUMP_MOVE_TRUTH and nothing else was ever recorded anywhere.

Cache counters are wired via the existing composition path
(ObjectMeshManager already owns the CPU mesh cache and the mesh
extractor directly; content.Dats is threaded into
WorldLifecycleResourceSnapshotSource the same way every other
composition consumer receives it). The DAT-object cache lives behind
IDatReaderWriter, a third-party interface from the DatReaderWriter
package that cannot be extended; RuntimeDatCollection (the one
production implementation) exposes the aggregate stats directly and a
pattern match reads them, degrading to zero for any test double —
no new static registry was introduced (GpuMemoryTracker remains the
one precedented process-wide static).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
(cherry picked from commit 1da2c33c875b41fa383dd79694ee2765f0e21896)
This commit is contained in:
Erik 2026-07-24 11:58:47 +02:00
parent bfc5e47365
commit 7b456b49d6
20 changed files with 477 additions and 9 deletions

View file

@ -4,6 +4,7 @@ param(
[string]$Account = $env:ACDREAM_TEST_USER,
[string]$Password = $env:ACDREAM_TEST_PASS,
[switch]$SkipBuild,
[switch]$Uncapped,
[int]$LoginTimeoutSeconds = 90
)
@ -602,7 +603,10 @@ $env:ACDREAM_TEST_USER = $Account
$env:ACDREAM_TEST_PASS = $Password
$env:ACDREAM_RETAIL_UI = '1'
$env:ACDREAM_FRAME_PROF = '1'
$env:ACDREAM_UNCAPPED_RENDER = '1'
# Capped by default (matches the retail presentation cadence the connected
# visual matrix is gated against); pass -Uncapped to opt into uncapped
# render, matching run-connected-world-lifecycle-gate.ps1's pattern.
$env:ACDREAM_UNCAPPED_RENDER = if ($Uncapped) { '1' } else { $null }
$env:ACDREAM_DEVTOOLS = '0'
$env:ACDREAM_UI_PROBE_DUMP = '0'
$env:ACDREAM_UI_PROBE_SCRIPT = $routePath
@ -611,6 +615,24 @@ $env:ACDREAM_DUMP_MOVE_TRUTH = '1'
$env:ACDREAM_NO_AUDIO = $null
$env:ACDREAM_WB_DIAG = $null
# Machine-readable disclosure of the exact launch configuration (2026-07-24
# measurement-tooling review — the prior audit could only see
# ACDREAM_DUMP_MOVE_TRUTH because nothing else the script sets was
# recorded anywhere). Written into the automation artifact directory
# BEFORE launch so a failed/aborted run still leaves the disclosure behind.
$null = New-Item -ItemType Directory -Force -Path $artifactDir
$acdreamEnvVars = Get-ChildItem Env: | Where-Object { $_.Name -like 'ACDREAM_*' } |
Sort-Object Name |
ForEach-Object { [pscustomobject]@{ Name = $_.Name; Value = $_.Value } }
$envDisclosure = [pscustomobject][ordered]@{
Script = 'run-connected-r6-soak.ps1'
GeneratedUtc = [DateTime]::UtcNow.ToString('O')
Uncapped = [bool]$Uncapped
EnvironmentVariables = @($acdreamEnvVars)
}
$envDisclosure | ConvertTo-Json -Depth 4 |
Set-Content -LiteralPath (Join-Path $artifactDir 'env-disclosure.json') -Encoding utf8
try {
$process = Start-Process -FilePath $exe -WorkingDirectory $Repository `
-RedirectStandardOutput $stdoutLog -RedirectStandardError $stderrLog -PassThru
@ -644,7 +666,7 @@ try {
# LiveEntityLivenessController's authoritative absence deadline is 25
# seconds. Sample only after that deadline so a previous destination's
# records cannot masquerade as retained memory in the return gate.
Start-Sleep -Seconds 12
Start-Sleep -Seconds 26
$stationaryProfileStart = (Get-FrameProfiles $stdoutLog).Count
$stationaryResult = Wait-ForNextFrameProfile $process $stdoutLog $stationaryProfileStart 12
$stationarySample = Capture-Sample $process $destination.Name 'stationary' $stationaryResult.Line
@ -719,6 +741,7 @@ finally {
SamplesCsv = $sampleCsv
ReportJson = $reportJson
CheckpointTimeline = $checkpointTimeline
EnvDisclosure = Join-Path $artifactDir 'env-disclosure.json'
}
}
$report | ConvertTo-Json -Depth 8 | Set-Content -LiteralPath $reportJson -Encoding utf8