High-performance SIMT code generation in an active visual effects library

Authors:
Jay L.T. Cornwall;Lee Howes;Paul H.J. Kelly;Phil Parsonage;Bruno Nicoletti
Affiliations:
Imperial College London, London, United Kingdom;Imperial College London, London, United Kingdom;Imperial College London, London, United Kingdom;The Foundry, London, United Kingdom;The Foundry, London, United Kingdom
Venue:
Proceedings of the 6th ACM conference on Computing frontiers
Year:
2009

Citing 9
Cited 2

A software environment for parallel computer vision

Computer
Data locality enhancement by memory reduction

ICS '01 Proceedings of the 15th international conference on Supercomputing
Skeletons for parallel image processing: an overview of the SKIPPER project

Parallel Computing - Special issue: Advanced environments for parallel and distributed computing
Brook for GPUs: stream computing on graphics hardware

ACM SIGGRAPH 2004 Papers
User transparency: a fully sequential programming model for efficient data parallel image processing: Research Articles

Concurrency and Computation: Practice & Experience
Program optimization space pruning for a multithreaded gpu

Proceedings of the 6th annual IEEE/ACM international symposium on Code generation and optimization
Scalable Parallel Programming with CUDA

Queue - GPU Computing
NVIDIA Tesla: A Unified Graphics and Computing Architecture

IEEE Micro
A compiler framework for optimization of affine loop nests for gpgpus

Proceedings of the 22nd annual international conference on Supercomputing

Generating GPU code from a high-level representation for image processing kernels

Euro-Par'11 Proceedings of the 2011 international conference on Parallel Processing
Halide: a language and compiler for optimizing parallelism, locality, and recomputation in image processing pipelines

Proceedings of the 34th ACM SIGPLAN conference on Programming language design and implementation

Quantified Score

Hi-index	0.00

Visualization

Abstract

SIMT (Single-Instruction Multiple-Thread) is an emerging programming paradigm for high-performance computational accelerators, pioneered in current and next generation GPUs and hybrid CPUs. We present a domain-specific active-library supported approach to SIMT code generation and optimisation in the field of visual effects. Our approach uses high-level metadata and runtime context to guide and to ensure the correctness of optimisation-driven code transformations and to implement runtime-context-sensitive optimisations. Our advanced optimisations require no analysis of the original C++ kernel code and deliver 1.3x to 6.6x speedups over syntax-directed translation on GeForce 8800 GTX and GTX 260 GPUs with two commercial visual effects.