Songsong Fan
APAC Solutions Director, Incredibuild
With over 10 years of experience in CI/CD, data center infrastructure, and software development productivity, he currently serves as the APAC Solutions Director at Incredibuild, where he drives and implements large-scale CI/CD and build performance optimization projects across industries including gaming, automotive ADAS, and intelligent cockpit systems. Prior to joining Incredibuild, he worked at Huawei as a general software engineer and IT solutions architect, focusing on system architecture, performance optimization, and emerging technologies.
Topic
Beyond Parallelization - How Caching and Distribution Are Redefining Compute Acceleration
Demonstrated through solving large-scale Android (AOSP) build challenges. As modern software systems grow in size and complexity, traditional parallelization alone can no longer meet the demands of compute-intensive workloads, from simulations, financial models, and algorithmic computation to massive build pipelines involving compilation, testing, and code analysis. This session explores a new paradigm that combines shared caching and distributed processing to eliminate redundant work and scale computation efficiently. Using the Android Operating System Platform (AOSP) project as a real-world case study, we’ll examine how caching can accelerate over twenty toolchains (C/C++, Rust, Javac, R8, Metalava, and more) that are part of an AOSP build, and how distributed processing complements it by dispatching remaining tasks to hundreds of remote cores. You’ll see live build visualizations, benchmark data, and practical lessons on how these principles can cut build times by up to x5, from almost two hours down to 17 minutes, on standard hardware. While the example focuses on an AOSP build, the insights directly apply to large-scale C++ development and general computational problems, offering a glimpse into the future of scalable, deterministic compute acceleration.
