CUDA implementation of Wagener's 2D convex hull PRAM algorithm

This paper describes a CUDA implementation of Wagener’s PRAM convex hull algorithm in R 2 [3,2]. It is presented in Knuth’s literate programming style. 1 Using this file The source of this document is a .nw file (for ‘noweb,’ an implementation of Knuth’s literate programming technique: see ‘Literate programming with noweb,’ by Andrew L. Johnson and Brad C. Johnson, Linux Journal, October 1st 1997). Noweb allows one to mix LaTeX with C (or pretty well any programming language), allowing a well-annotated program. One can extract ‘chunks’ from it. You need the noweb system, of course (that is, notangle to extract the C part and noweave to typeset the full document). This document includes a Makefile. To start the ball rolling, you can extract it as follows: notangle -t8 -Rwagener.Makefile wagener.nw > wagener.Makefile With it, you can make a CUDA source file (wagener.cu) or a DVI copy of this document (make dvi produces wagener.dvi) There is one problem with wagener.cu. The construct <<<. . .>>> is a necessary part of the cuda source code, and it conflicts with noweb’s construct <<. . .>>. Therefore wagener.cu contains match and merge LLL range, block RRR ( hood, newhood, scratch ); e-mail: [email protected]. Mathematics department website: http://www.maths.tcd.ie.