A system for rapid, automatic shader level-of-detail

Abstract

Level-of-detail (LOD) rendering is a key optimization used by modern video game engines to achieve high-quality rendering with fast performance. These LOD systems require simplified shaders, but generating simplified shaders remains largely a manual optimization task for game developers. Prior efforts to automate this process have taken hours to generate simplified shader candidates, making them impractical for use in modern shader authoring workflows for complex scenes. We present an end-to-end system for automatically generating a LOD policy for an input shader. The system operates on shaders used in both forward and deferred rendering pipelines, requires no additional semantic information beyond input shader source code, and in only seconds to minutes generates LOD policies (consisting of simplified shader, the desired LOD distance set, and transition generation) with performance and quality characteristics comparable to custom hand-authored solutions. Our design contributes new shader simplification transforms such as approximate common subexpression elimination and movement of GPU logic to parameter bind-time processing on the CPU, and it uses a greedy search algorithm that employs extensive caching and upfront collection of input shader statistics to rapidly identify simplified shaders with desirable performance-quality trade-offs.

Approximate Program Smoothing Using Mean‐Variance Statistics, with Application to Procedural Shader Bandlimiting

“…Sitthi‐Amorn et al [SAMWL11] described a GP approach to the problem of automatic procedural shader simplification. Other researchers have also investigated automatic shader simplification by heuristic search methods that simplify programs [OKS, Pel, HFTF15], and by jointly modifying shaders and geometry [WYY*14]. Brady and colleagues [BLPW14] showed how to use GP to discover new analytic reflectance functions.…”

Runtime Shader Simplification via Instant Search in Reduced Optimization Space

“…He et al [HFTF15] proposed a system that automatically generates level‐of‐detail (LOD) shaders. They also introduced a heuristic model to estimate performance and a fast search algorithm to reduce the simplification process from hours to minutes.…”

“…where x and y are the integer pixel coordinates, v is a set of input geometry primitives with attributes, u is a set of uniform parameters (such as camera position and light position), and || · || 2 is the norm, (i.e., L 2 norm used in this paper). Based on the measured rendering time and image error, shader simplification is formulated as a multi‐objective problem in the two‐dimensional cost space (∊, t ), and seeks a Pareto‐optimal solution [SAMWL11,WYY*14,HFTF15]. To consider the variations of geometry primitives and uniform parameters, offline shader simplification usually integrates the error over the domain of each uniform parameter, and averages over different geometry models.…”

“…However, obtaining an accurate estimation of the cost of different instructions is a challenge. In this paper, we use the performance model proposed by He et al [HFTF15]. This model weighs the cost of DAG nodes by the number of scalar instructions needed to perform them (e.g., a float4 addition incurs cost 4), and assigns a cost of 100 units to texture operations.…”

“…In this case, the simplification process must be repeated to port it to new hardware. Given a large number of hardware customized shaders in current AAA games (Over 18000 in Bungie's Destiny [HFTF15]), optimizing all of them for various hardware platforms in a preprocess is impractical.…”