From d0cf0d64367df265415562e4b66ed2373f67a52d Mon Sep 17 00:00:00 2001 From: "Moe Charm (CI)" Date: Thu, 18 Dec 2025 09:37:55 +0900 Subject: [PATCH] docs: tone down Phase 75-5 PGO recovery estimates --- docs/analysis/PERFORMANCE_TARGETS_SCORECARD.md | 4 ++-- docs/analysis/PHASE75_4_FAST_PGO_REBASE_RESULTS.md | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/docs/analysis/PERFORMANCE_TARGETS_SCORECARD.md b/docs/analysis/PERFORMANCE_TARGETS_SCORECARD.md index 0ad3f7a1..6ba85849 100644 --- a/docs/analysis/PERFORMANCE_TARGETS_SCORECARD.md +++ b/docs/analysis/PERFORMANCE_TARGETS_SCORECARD.md @@ -149,8 +149,8 @@ Phase 75-3 validated C5+C6 inline slots optimization on Standard binary (+5.41%) 2. **HIGH PRIORITY - PHASE 75-5 (PGO Profile Regeneration)** - Regenerate PGO profile with C5=1, C6=1 training configuration - - Expected gain: +5-8% (if profile aligns with actual code optimization) - - Estimated recovery: 55.51 M ops/s → ~58-59 M ops/s + - Expected gain: unknown (likely positive if the training profile matches the actual hot path, but not guaranteed) + - Estimated recovery: treat any number as a hypothesis until re-measured (do not assume a return to Phase 69 levels) - Root cause analysis: Investigate 14% gap vs Phase 69 (layout, code bloat, or profile mismatch) **Documentation:** diff --git a/docs/analysis/PHASE75_4_FAST_PGO_REBASE_RESULTS.md b/docs/analysis/PHASE75_4_FAST_PGO_REBASE_RESULTS.md index 30c91c11..6423bd06 100644 --- a/docs/analysis/PHASE75_4_FAST_PGO_REBASE_RESULTS.md +++ b/docs/analysis/PHASE75_4_FAST_PGO_REBASE_RESULTS.md @@ -6,7 +6,7 @@ **Key Finding**: C5+C6 inline slots optimization shows **+3.16% gain** on FAST PGO binary, meeting the ideal threshold but significantly lower than Standard's +5.41% gain. -**Critical Concern**: FAST PGO baseline is **7.16% slower** than Standard baseline, suggesting potential PGO profile staleness or suboptimal training conditions. +**Critical Concern**: FAST PGO baseline is **7.16% slower** than Standard baseline, suggesting potential PGO profile staleness, training mismatch, or build/layout drift. --- @@ -155,7 +155,7 @@ 3. **⚠ Regenerate PGO Profile** - Train with C5=1, C6=1 (optimized config) - Use Phase 75 codebase for profiling - - Expected result: close gap to Standard baseline + - Expected result: uncertain; likely to improve if PGO was mismatched, but not guaranteed 4. **⚠ Root Cause Analysis: 14% Regression** - Compare Phase 69 vs Phase 75-4 binary characteristics @@ -176,7 +176,7 @@ 7. **Standard vs FAST PGO Convergence** - Investigate why Standard outperforms FAST PGO by 7-10% - - Consider unified build configuration + - Treat this as a measurement/forensics problem first (PGO profile, flags, link order), not an assumed “PGO must win” rule - Document PGO ROI vs complexity cost ---