Speed is one of Chrome’s four core principles, enabling web developers to provide users with faster, more engaging web experiences. While many components in the browser contribute to overall speed, the rendering pipeline is primarily responsible for ensuring websites display at 60 frames per second (fps), which feels fast and responsive to users. Over the last few months we’ve rolled out several performance improvements to Chrome’s rendering pipeline, making it even smarter about the work it completes. Chrome now more intelligently skips redundant tasks, chooses optimal rendering algorithms, and better utilizes system hardware, causing websites to load faster, run smoother, and use less power.

In order to display content at 60fps, Chrome has only 16ms to render each frame, including JavaScript execution, style, layout, painting, and pushing the resulting pixels to the user’s screen. Of that 16ms time budget, the less Chrome uses, the more time web developers have to run scripts, load content, and delight their users. Many of our recent improvements to the rendering pipeline focus on reducing the per-frame workload, rather than simply improving Chrome’s raw speed.

For example, when Chrome is preparing to paint pixels to the screen, it must first identify which elements on the page need to be redrawn and which can be copied from the previous frame’s cache. On highly dynamic pages with frequent DOM changes, this performance cost can add up quickly. To simplify this task, Chrome now tracks the draw commands generated for each element and can identify visually non-overlapping subsets. If one of these subsets hasn’t been modified, the entire block can be copied directly from the cache without any additional work. This optimization reduces the time it takes to paint a new frame to the screen by up to 35%.

Chrome also groups the pixels into tiles to enable smaller and faster updates to the screen. Previously, Chrome would redraw any of these tiles that had been modified by a DOM update, but this is only optimal if the majority of a tile’s area needs to be redrawn. If only a few pixels have changed, it’s faster to copy the entire tile from the previous frame and then update the new pixels. Chrome can now intelligently choose the redraw pipeline that will be faster, reducing tile redraw time by up to 40%.

   

Any update to the screen, such as the input cursor blinking, would previously require the whole tile to be re-rendered (left). When only a few pixels have changed, Chrome can now redraw only the modified region (right).

In addition to intelligently optimizing its workload, Chrome is now better at choosing how it completes that work given the underlying hardware. While video and canvas elements have been GPU accelerated for a long time, Chrome is constantly getting better at utilizing the GPU for more challenging rendering tasks. On Android, Mac, and Windows, Chrome now better utilizes GPU acceleration for complex site content rendering. This improves animation performance, input latency, and scroll smoothness for modern SVG and HTML5 pages.

There are many different dimensions to speed, and we’re committed to continually improving Chrome’s performance and enabling developers to optimize their user experience to hit 60fps. The rendering pipeline is only one piece of the puzzle, and we’ve got a lot more coming. Stay tuned as we continue taking deep dives into performance and steadily make the web faster and more responsive for all Chrome users.

Posted by Chris Harrelson, Painting Professional