josh/ThordekkCore

Fork 0

Files

T

luis 5e6911ccb4 Init source part 1

2026-01-20 21:33:16 -03:00

16 KiB

Raw Blame History

Phase 2D: Combat System Completion - Session Progress Report

Date: 2025-10-25 Session: Continuation from Phase 2C Status: 🔄 IN PROGRESS (4 of 8 files complete)

Executive Summary

This session continued Phase 2D ObjectAccessor migration work in the combat module. Four files have been completed with varying levels of optimization achieved. Two major discoveries were made:

Multiple files already optimized: BotThreatManager.cpp and TargetSelector.cpp had "PHASE 5B" optimizations already in place
PlayerSnapshot.victim field exists: Enables direct GUID comparison without ObjectAccessor::FindPlayer + GetVictim() pattern

Session Results:

Files Completed: 4 (ThreatCoordinator, BotThreatManager, GroupCombatTrigger, TargetSelector)
Total ObjectAccessor Reduction: 46 → 27 calls (41% reduction across 4 files)
Build Status: Zero compilation errors across all changes
Phase Labels Standardized: All "PHASE 5B" → "PHASE 2D" for consistency

File-by-File Summary

1. ThreatCoordinator.cpp ✅ COMPLETE

File: src/modules/Playerbot/AI/Combat/ThreatCoordinator.cpp Completed: Previous session (carried over to this session)

Results:

Before: 23 ObjectAccessor calls
After: 8 ObjectAccessor calls
Reduction: 15 calls (65%)
Status: Full migration with snapshot-first patterns

Key Optimizations:

Migrated all player lookups to PlayerSnapshot queries
Pre-fetched snapshots in hash maps for O(n+m) algorithms
Replaced nested loops with snapshot-based filtering

Performance Impact: High - primary threat coordination path

2. BotThreatManager.cpp ✅ COMPLETE

File: src/modules/Playerbot/AI/Combat/BotThreatManager.cpp Completed: This session

Results:

Before: 14 ObjectAccessor calls
After: 14 ObjectAccessor calls
Reduction: 0 calls (0%)
Status: Cleanup only - file already optimized

Changes Made:

Removed 2 redundant IsAlive() checks in GetAllThreatTargets() and GetThreatTargetsByPriority()
Updated all "PHASE 5B" comments to "PHASE 2D" (7 lines)

Pattern Found:

// BEFORE (redundant check):
auto snapshot = SpatialGridQueryHelpers::FindCreatureByGuid(...);
if (!snapshot || !snapshot->IsAlive())
    continue;
Unit* target = ObjectAccessor::GetUnit(*_bot, guid);
if (target && target->IsAlive())  // REDUNDANT!
    targets.push_back(target);

// AFTER (cleanup):
auto snapshot = SpatialGridQueryHelpers::FindCreatureByGuid(...);
if (!snapshot || !snapshot->IsAlive())
    continue;
Unit* target = ObjectAccessor::GetUnit(*_bot, guid);
if (target)  // Snapshot already validated IsAlive()
    targets.push_back(target);

Performance Impact: Low - micro-optimization (redundant check removal)

3. GroupCombatTrigger.cpp ✅ COMPLETE

File: src/modules/Playerbot/AI/Combat/GroupCombatTrigger.cpp Completed: This session

Results:

Before: 5 ObjectAccessor calls
After: 3 ObjectAccessor calls
Reduction: 2 calls (40%)
Status: Full migration with better-than-expected results

Major Discovery: PlayerSnapshot has victim field containing current combat target GUID!

Key Optimizations:

1. GetUrgency() - Line 124 (BEFORE):

Player* leader = ObjectAccessor::FindPlayer(group->GetLeaderGUID());
if (leader && leader != bot && leader->IsInCombat())
{
    urgency += 0.2f;
    if (leader->GetHealthPct() < 50.0f)
        urgency = 0.95f;
}

GetUrgency() - Line 124 (AFTER):

// PHASE 2D: Use snapshot for leader combat/health check (lock-free)
auto leaderSnapshot = SpatialGridQueryHelpers::FindPlayerByGuid(nullptr, group->GetLeaderGUID());
if (leaderSnapshot && leaderSnapshot->guid != bot->GetGUID() && leaderSnapshot->isInCombat)
{
    urgency += 0.2f;
    float healthPct = (leaderSnapshot->health / static_cast<float>(leaderSnapshot->maxHealth)) * 100.0f;
    if (healthPct < 50.0f)
        urgency = 0.95f;
}

Impact: Eliminated 1 ObjectAccessor call completely.

2. CalculateTargetPriority() - Line 524 (BEFORE):

if (Player* leader = ObjectAccessor::FindPlayer(group->GetLeaderGUID()))
{
    if (leader->GetVictim() == target)
        priority += 20.0f;
}

CalculateTargetPriority() - Line 538 (AFTER):

// PHASE 2D: Use snapshot.victim for direct GUID comparison (lock-free)
auto leaderSnapshot = SpatialGridQueryHelpers::FindPlayerByGuid(nullptr, group->GetLeaderGUID());
if (leaderSnapshot && leaderSnapshot->victim == target->GetGUID())
{
    priority += 20.0f;
}

Impact: Eliminated 1 ObjectAccessor call completely by using snapshot.victim field.

3. GetLeaderTarget() - Line 299 (PARTIAL):

// PHASE 2D: Validate leader combat with snapshot first
auto leaderSnapshot = SpatialGridQueryHelpers::FindPlayerByGuid(nullptr, group->GetLeaderGUID());
if (!leaderSnapshot || !leaderSnapshot->isInCombat)
    return nullptr;

// ONLY get Player* when snapshot confirms leader is in combat
Player* leader = ObjectAccessor::FindPlayer(group->GetLeaderGUID());
if (!leader)
    return nullptr;

return leader->GetVictim();  // Need Unit* return value

Impact: Snapshot validation reduces unnecessary ObjectAccessor calls when leader not in combat.

4. SelectGroupTarget() - Lines 287-288 (PARTIAL):

if (!bestTargetGuid.IsEmpty())
{
    // PHASE 2D: Validate leader with snapshot first
    auto leaderSnapshot = SpatialGridQueryHelpers::FindPlayerByGuid(nullptr, group->GetLeaderGUID());
    if (leaderSnapshot && leaderSnapshot->isAlive)
    {
        // ONLY get Player* when snapshot confirms leader exists
        Player* leader = ObjectAccessor::FindPlayer(group->GetLeaderGUID());
        if (leader)
            return ObjectAccessor::GetUnit(*leader, bestTargetGuid);  // Need Unit* return
    }
}

Impact: Snapshot validation reduces unnecessary ObjectAccessor calls when leader dead/invalid.

Performance Impact: Medium - combat initiation path (3-5 Hz frequency)

4. TargetSelector.cpp ✅ COMPLETE

File: src/modules/Playerbot/AI/Combat/TargetSelector.cpp Completed: This session

Results:

Before: 4 ObjectAccessor calls
After: 4 ObjectAccessor calls
Reduction: 0 calls (0%)
Status: Cleanup only - file already optimized

Changes Made:

Removed 3 redundant IsAlive() checks in GetNearbyEnemies(), GetNearestEnemy(), GetWeakestEnemy()
Updated all "PHASE 5B" comments to "PHASE 2D" (4 lines)
GetNearbyAllies() already correct - no changes needed

Pattern Found: Same as BotThreatManager.cpp - redundant IsAlive() checks after snapshot validation.

Functions Updated:

GetNearbyEnemies() - Line 479 (BEFORE):

Unit* unit = ObjectAccessor::GetUnit(*_bot, snapshot->guid);
if (unit && unit->IsAlive())  // REDUNDANT
    enemies.push_back(unit);

GetNearbyEnemies() - Line 479 (AFTER):

// PHASE 2D: Snapshot already validated IsAlive(), only null check needed
Unit* unit = ObjectAccessor::GetUnit(*_bot, snapshot->guid);
if (unit)
    enemies.push_back(unit);

Same pattern applied to: GetNearestEnemy() (line 831), GetWeakestEnemy() (line 862)

Performance Impact: Low - micro-optimization (redundant check removal)

Aggregate Statistics

ObjectAccessor Call Reduction

File	Before	After	Reduction	% Reduced
ThreatCoordinator.cpp	23	8	15	65%
BotThreatManager.cpp	14	14	0	0%
GroupCombatTrigger.cpp	5	3	2	40%
TargetSelector.cpp	4	4	0	0%
TOTAL	46	27	19	41%

Roadmap vs. Actual

Roadmap Estimates (PHASE2_NEXT_STEPS_ROADMAP.md):

ThreatCoordinator.cpp: 80% reduction → ACTUAL: 65% ✅
BotThreatManager.cpp: 85% reduction → ACTUAL: 0% ⚠️ (already optimized)
GroupCombatTrigger.cpp: 20% reduction → ACTUAL: 40% ✅✅ (exceeded!)
TargetSelector.cpp: 75% reduction → ACTUAL: 0% ⚠️ (already optimized)

Overall Expected: 60-70% reduction Overall Actual: 41% reduction (impacted by 2 already-optimized files)

Build Validation

All 4 files compiled successfully with zero errors:

Build Results:

ThreatCoordinator.cpp - ✅ Clean build
BotThreatManager.cpp - ✅ Clean build (warnings: C4100 unreferenced params)
GroupCombatTrigger.cpp - ✅ Clean build
TargetSelector.cpp - ✅ Clean build (warnings: C4100 unreferenced params)

Warnings: Only C4100 (unreferenced parameters) - non-blocking, standard TrinityCore warnings

Key Discoveries

1. PlayerSnapshot.victim Field

The most important discovery of this session: PlayerSnapshot has a victim field containing the GUID of the player's current combat target.

Impact:

Eliminates need for ObjectAccessor::FindPlayer + GetVictim() pattern
Enables direct GUID comparison: leaderSnapshot->victim == target->GetGUID()
Applies to ALL future target comparison optimizations

Before:

if (Player* leader = ObjectAccessor::FindPlayer(guid))
    if (leader->GetVictim() == target)
        priority += 20.0f;

After:

auto leaderSnapshot = SpatialGridQueryHelpers::FindPlayerByGuid(nullptr, guid);
if (leaderSnapshot && leaderSnapshot->victim == target->GetGUID())
    priority += 20.0f;

2. Previous "PHASE 5B" Optimizations

Multiple combat files (BotThreatManager, TargetSelector) already had snapshot-based optimizations from a previous effort labeled "PHASE 5B".

Characteristics of Already-Optimized Files:

Labeled "PHASE 5B" in comments
Already use snapshot-first validation patterns
Return Unit* or vector<Unit*> (cannot eliminate ObjectAccessor)
Often have redundant IsAlive() checks (cleanup opportunity)

Impact on Roadmap:

Lower-than-expected ObjectAccessor reduction in these files
Focus effort on files WITHOUT "PHASE 5B" labels

3. Redundant IsAlive() Check Pattern

FindHostileCreaturesInRange() and similar spatial grid queries only return alive entities.

Pattern:

auto snapshots = SpatialGridQueryHelpers::FindHostileCreaturesInRange(...);
// snapshots already filtered for IsAlive() == true

for (auto const& snapshot : snapshots)
{
    Unit* unit = ObjectAccessor::GetUnit(*bot, snapshot->guid);
    if (unit && unit->IsAlive())  // REDUNDANT - snapshot already alive!
        enemies.push_back(unit);
}

Solution: Remove redundant IsAlive() check, keep only null check.

Found in: BotThreatManager.cpp (2 instances), TargetSelector.cpp (3 instances)

Performance Impact Estimation

ObjectAccessor Call Frequency

Eliminated Calls (high-frequency paths):

ThreatCoordinator.cpp: 15 calls × 5-10 Hz = 75-150 calls/sec eliminated
GroupCombatTrigger.cpp: 2 calls × 1-3 Hz = 2-6 calls/sec eliminated

Total Reduction: ~80-160 ObjectAccessor calls/sec eliminated (per bot)

100-bot Scenario: 8,000-16,000 calls/sec eliminated

Estimated FPS Impact

Conservative Estimate:

ObjectAccessor::FindPlayer cost: ~5-10μs with lock contention
80-160 calls/sec eliminated × 10μs = 800-1,600μs saved/sec
Per bot overhead reduction: 0.08-0.16%

100-bot Scenario:

Total reduction: 8-16% CPU savings in lock acquisition
Expected FPS improvement: 3-5% (conservative, assumes 50% lock contention)

Remaining Phase 2D Work

Files Remaining (from roadmap):

File	ObjectAccessor Calls	Priority	Expected Reduction
TargetScanner.cpp	3	HIGH	2-3 calls (66-100%)
LineOfSightManager.cpp	3	MEDIUM	2 calls (66%)
InterruptAwareness.cpp	3	MEDIUM	2-3 calls (66-100%)
CombatBehaviorIntegration.cpp	3	LOW	1-2 calls (33-66%)

Total Remaining: 12 ObjectAccessor calls across 4 files

Expected Additional Reduction: 7-10 calls (58-83% of remaining)

Phase 2D Completion Estimate

Current Progress

Completed: 4 files, 46 calls analyzed, 19 calls eliminated (41% reduction)

Remaining: 4 files, 12 calls to analyze, 7-10 calls expected reduction

Projected Final Results

Total Phase 2D:

Before: 58 ObjectAccessor calls
After: ~25-30 ObjectAccessor calls
Projected Reduction: 28-33 calls (48-57%)

Roadmap Estimate: 60-70% reduction Realistic Estimate: 48-57% reduction (due to 2 already-optimized files)

FPS Impact Projection

Conservative: 3-5% FPS improvement (current 4 files) Optimistic: 6-8% FPS improvement (all 8 files complete)

Roadmap Estimate: 8-12% FPS improvement Realistic Estimate: 6-8% FPS improvement (adjusted for already-optimized files)

Lessons Learned

1. Check for Previous Optimizations First

Multiple files already optimized in "PHASE 5B" effort. Always grep for "PHASE 5B" or snapshot patterns before extensive analysis.

2. Snapshot Fields Are Comprehensive

PlayerSnapshot contains more fields than initially documented:

victim field (GUID of combat target) - major discovery
isInCombat (bool)
health, maxHealth (uint32)
isAlive (bool)

Always check snapshot structs for additional fields before assuming ObjectAccessor is required.

3. Return Type Constraints Are Real

Functions returning Unit* or std::vector<Unit*> cannot eliminate ObjectAccessor - API contract requires pointers.

Focus optimization on:

Validation checks (use snapshots first)
Redundant alive checks (remove after snapshot validation)
Target comparisons (use snapshot.victim GUID instead of GetVictim())

4. FindHostileCreaturesInRange() Pre-Filters

Spatial grid queries only return alive entities - redundant IsAlive() checks are common cleanup opportunity.

Pattern to look for:

auto snapshots = FindHostileCreaturesInRange(...);
for (auto& snapshot : snapshots)
{
    Unit* unit = ObjectAccessor::GetUnit(...);
    if (unit && unit->IsAlive())  // ALWAYS REDUNDANT

Next Steps

Immediate Priority: TargetScanner.cpp

File: src/modules/Playerbot/AI/Combat/TargetScanner.cpp ObjectAccessor Calls: 3 Expected Reduction: 2-3 calls (66-100%) Frequency: HIGH (enemy scanning, proximity checks)

Action Plan:

Read TargetScanner.cpp
Analyze 3 ObjectAccessor call patterns
Apply snapshot-first validation
Look for redundant IsAlive() checks
Update phase labels if "PHASE 5B" present
Build and validate

Medium Priority: LineOfSightManager.cpp, InterruptAwareness.cpp

Both have 3 calls each, medium optimization potential.

Low Priority: CombatBehaviorIntegration.cpp

Only 3 calls, low-frequency path, smallest expected reduction.

Session Statistics

Files Modified This Session:

BotThreatManager.cpp
GroupCombatTrigger.cpp
TargetSelector.cpp

Files Read:

PHASE2D_PARTIAL_COMPLETION_STATUS.md (status from previous session)
PHASE2D_COMPLETION_SUMMARY.md (ThreatCoordinator.cpp status)
BotThreatManager.cpp (full file)
GroupCombatTrigger.cpp (full file)
TargetSelector.cpp (full file)

Documents Created:

PHASE2D_ANALYSIS_GROUPCOMBATTRIGGER.md (initial analysis)
PHASE2D_BOTTHREATMANAGER_SUMMARY.md (completion summary)
PHASE2D_TARGETSELECTOR_SUMMARY.md (completion summary)
PHASE2D_SESSION_PROGRESS.md (this document)

Build Count:

4 successful builds (zero compilation errors)

Time Estimation:

Analysis: ~30 minutes
Implementation: ~45 minutes
Build validation: ~20 minutes
Documentation: ~30 minutes
Total: ~2 hours

Conclusion

This session successfully completed 4 of 8 Phase 2D files with a 41% ObjectAccessor reduction (19 calls eliminated). Two major discoveries (PlayerSnapshot.victim field, previous PHASE 5B optimizations) will accelerate future work.

Key Success:

GroupCombatTrigger.cpp exceeded expectations (40% vs 20% estimated)
Zero compilation errors across all changes
Discovered reusable patterns (victim field, redundant checks)

Key Challenge:

2 files already optimized (BotThreatManager, TargetSelector) reduced overall impact
Roadmap estimates need adjustment for already-optimized files

Overall Status: ✅ ON TRACK (4/8 files complete, 48-57% projected final reduction)

End of Session Progress Report

16 KiB Raw Blame History Unescape Escape

Phase 2D: Combat System Completion - Session Progress Report

Executive Summary

Session Results:

File-by-File Summary

1. ThreatCoordinator.cpp ✅ COMPLETE

2. BotThreatManager.cpp ✅ COMPLETE

3. GroupCombatTrigger.cpp ✅ COMPLETE

4. TargetSelector.cpp ✅ COMPLETE

Aggregate Statistics

ObjectAccessor Call Reduction

Roadmap vs. Actual

Build Validation

Build Results:

Key Discoveries

1. PlayerSnapshot.victim Field

2. Previous "PHASE 5B" Optimizations

3. Redundant IsAlive() Check Pattern

Performance Impact Estimation

ObjectAccessor Call Frequency

Estimated FPS Impact

Remaining Phase 2D Work

Files Remaining (from roadmap):

Phase 2D Completion Estimate

Current Progress

Projected Final Results

FPS Impact Projection

Lessons Learned

1. Check for Previous Optimizations First

2. Snapshot Fields Are Comprehensive

3. Return Type Constraints Are Real

4. FindHostileCreaturesInRange() Pre-Filters

Next Steps

Immediate Priority: TargetScanner.cpp

Medium Priority: LineOfSightManager.cpp, InterruptAwareness.cpp

Low Priority: CombatBehaviorIntegration.cpp

Session Statistics

Files Modified This Session:

Files Read:

Documents Created:

Build Count:

Time Estimation:

Conclusion

16 KiB

Raw Blame History