Virtual Reality (VR) is an emerging technique that provides immersive experience for users. Due to the high computation cost of rendering real-time animation twice (for both eyes) and the resource limitation of wearable devices, VR applications often face performance bottlenecks and performance optimization plays an important role in VR software development. Performance optimizations of VR applications can be very different from those in traditional software as VR involves more elements such as graphics rendering and real-time animation. In this paper, we present the first empirical study on 182 real-world performance optimizations from 45 VR software projects. In particular, we manually categorized the optimizations into 11 categories, and applied static analysis to identify how they affect different life-cycle phases of VR applications. Furthermore, we studied the complexity and potential negative effects of performance optimizations, and how optimizations are different between large organizational software projects and smaller personal software projects. Our major findings include: (1) graphics simplification (24.2%), rendering optimization (16.5%), language / API optimization (15.4%), heap avoidance (14.8%), and value caching (12.1%) are the most common categories of performance optimization in VR applications; (2) game logic updates (29.9%) and before-scene initialization (20.1%) are the most common life-cycle phases affected by performance issues; (3) 45.6% of the optimizations have potential negative effects and 30.8% of the optimizations are systematic changes; (4) the distributions of optimization classes are very different between organizational VR projects and personal VR projects.