Piko

Abstract

InputStage A Stage B Stage C Stage D Piko Frontend Pipeline Schedule Pipeline Description Spatial Bins Multicore CPU Manycore GPU Piko Backend Figure 1: Piko is a framework for designing and implementing programmable graphics pipelines that can be easily retargeted to different application configurations and architectural targets. Piko's input is a functional and structural description of the desired graphics pipeline, augmented with a per-stage grouping of computation into spatial bins (or tiles), and a scheduling preference for these bins. Our compiler generates efficient implementations of the input pipeline for multiple architectures and allows the programmer to tweak these implementations using simple changes in the bin configurations and scheduling preferences. AbstractWe present Piko, a framework for designing, optimizing, and retargeting implementations of graphics pipelines on multiple architectures. Piko programmers express a graphics pipeline by organizing the computation within each stage into spatial bins and specifying a scheduling preference for these bins. Our compiler, Pikoc, compiles this input into an optimized implementation targeted to a massively-parallel GPU or a multicore CPU. Piko manages work granularity in a programmable and flexible manner, allowing programmers to build load-balanced parallel pipeline implementations, to exploit spatial and producer-consumer locality in a pipeline implementation, and to explore tradeoffs between these considerations. We demonstrate that Piko can implement a wide range of pipelines, including rasterization, Reyes, ray tracing, rasterization/ray tracing hybrid, and deferred rendering. Piko allows us to implement efficient graphics pipelines with relative ease and to quickly explore design alternatives by modifying the spatial binning configurations and scheduling preferences for individual stages, all while delivering real-time performance that is within a factor six of state-of-the-art rendering systems.

On-the-fly Vertex Reuse for Massively-Parallel Software Geometry Processing

“…A more current area of research where we encounter the problem of massively parallel vertex processing is software rendering on the modern GPU. Noteworthy examples of GPU software rendering pipelines include Freepipe [Liu et al 2010], CUDARaster [Laine and Karras 2011], and Piko [Patney et al 2015]. They all use the compute mode of the GPU (typically on top of the CUDA [NVIDIA 2016] ecosystem) to implement rasterization and fragment shading, but lack mechanisms for vertex reuse.…”