Andheri Simulator Code Review Summary

Introduction

• Objective: Code review of Yandere Simulator.
• Concerns: Previously misunderstood points about no bullying.
• Background: Took 2 months, starting from May 15th build with subsequent builds on Jun 1, 15, and Jul 2.
• Method: Correspondence with Yandere Dev for fixes.

Key Issues Identified in Yandere Simulator

Performance and Optimization

• Initial Benchmarks: Used Wine via Lutra installer with missing text issues. Game didn’t hit 60fps despite high PC specs.
• Heavy Scene Load: Initial load into school scene is heavy, neither CPU nor GPU maxed out.
• Comparative Analysis: Ran Doom 2016 and Hitman 2 for comparison.
• Need for Profiling: Used Unity’s profiler, difficulties in getting source code from Yandere Dev.

Source Code Analysis

• Rendering Issues: Time spent rendering opaque geometry. High geometry on student models causing performance issue.
• Mesh Optimization: Need for better optimization, unnecessary vertices in models, use of combined meshes.
• LOD (Level of Detail): Partially implemented; a need for full implementation for performance improvements.
• Occlusion Culling: Proper occlusion culling implemented using Unity’s built-in feature.

Specific System Issues

• Heart Rate Monitor: Inefficient separate camera usage for heart rate line rendering.
• Physics: Many colliders and triggers without apparent use; improper assignment on layers.
• Animation: Time spent on animation calculations, “Stop Animation” script disables distant animations, but some remain active.
• Script Efficiency: Most time spent on rendering than script execution.

Script and UI Issues

• Common Theories Debunked: Else-if chains vs switch case show negligible performance impact. Same applies to vector3 distance myths.
• Prompt Scripts: Numerous UI prompt instantiations for each interactable causing inefficiency.

Game Object and Component Counts

• Counts: 49,300 game objects and 53,015 components in one scene; about 200 scripts run every frame leading to microsecond accumulation.
• Script Execution Time: Focus on disabling unnecessary scripts.
• Late Update: Dynamic bone updates and UI panel issues identified.

Suggested Architectural Improvements

• Current Structure: Key game components, such as YandereScript, StudentScript, and StudentManagerScript, discussed.
• Separation of Logic: Proposed separating generic and student-specific behaviors.
• New Architecture: Separate routine definitions, state storage, and simplified state updates.
• Optimization Recommendations: Focus on Unity job system, proper occlusion culling, efficient use of LOD, and caching methods.

Other Implementation Details

• Save System Critique: JSON format bloated, recommends a custom byte format. Player prefs misuse for save data.
• Interaction Prompts: Redesign suggested for interactable handling, including distance checks and dynamic prompts.
• School Map: Current camera setup identified as inefficient, suggests using a static image.
• Miscellaneous: Empty classes workaround, use of 2020 Unity version for serialization issues.

Cosmetic and Style Issues

• Code Style: String interpolation vs concatenation, elimination of redundant instance variables, better use of arrays and fields.
• Public Booleans: Too many, recommendation to use enums and bitwise operations for state representation.
• Animation System: Suggests using event triggers in animation clips.
• Extreme Technical Debt: Yandere Simulator’s technical debt has severely impacted development.

Conclusions

• Performance Tips: Focus on disabling unnecessary scripts, efficient use of Unity features, thorough profiling, and code organization.
• Architectural Overhaul: Game needs a significant refactor not done since 2016.
• Future Work: Good luck to Yandere Dev, need for managed technical debt, improvements in code maintenance and readability.

Additional Insights

• Unit Testing: Rare in game development but beneficial for catching bugs early.
• Frame Rate Calculation: Existing method inaccurate, suggests using current, average, min, and max frame rates display.