Beyond const: Production-Grade Performance Profiling & Impeller Optimization in Flutter

Most Flutter performance advice stops at basic surface-level tips; like adding `const` constructors or minimizing `setState` calls. However, in complex production applications, severe performance bottlenecks usually stem from low-level issues: shader compilation jank, unmanaged image caching, heavy garbage collection spikes, and inefficient widget tree rebuilds under load. This session cuts through surface-level advice to explore the internals of Flutter’s rendering pipeline and engine architecture. We will dissect how Impeller changes rendering execution, examine real-world CPU and GPU flame graphs, identify subtle memory leaks, and showcase actionable profiling strategies using Flutter DevTools. Attendees will leave with practical architectural patterns to systematically benchmark, profile, and guarantee 60/120 FPS performance across budget Android hardware and high-end iOS devices. Key Takeaways & Learning Points: 1. Impeller vs. Skia Deep Dive: Understanding Impeller's rendering pipeline and pre-compiled shaders to eliminate dropped frames. 2. Mastery of Flutter DevTools: Practical techniques for interpreting CPU profiles, memory allocation traces, and raster thread bottlenecks. 3. Memory & Asset Optimization: Architectural patterns for managing large image caches, reducing GC churn, and preventing memory leaks. 4. CI/CD Performance Gateways: Setting up automated performance regression benchmarking in CI/CD pipelines to catch frame drops before production. Experience Level: Intermediate to Advanced (tested on production applications deployed across diverse device specs). Previous Presentation History: New session created specifically for FlutterCon Kenya 2026. Keywords: Flutter Performance, Impeller, DevTools, Memory Optimization, Profiling, Raster Thread, Benchmarking, Mobile Engineering