Vector graphics are a versatile form of representing visual information. They are resolution independent and can represent a broad range of content: text fonts, maps, user interfaces, and games.

Rendering vector graphics is not an easy task. High quality rendering requires huge amount of processing power. To achieve real-time performance, many applications sacrifice the quality of their resulting images.  The solution is to parallelize rendering using GPUs.

In this thesis we describe a rendering pipeline that is fully parallel, and implemented on the GPU. We show how to create a novel acceleration data structure, called shortcut tree, that allow us to render high quality images using antialiasing filters of large support. Our results surpass the state-of-the-art in quality and performance.