Homec4science
Diffusion RADIANCE Photon Map (pmapmerge) Browse

RadiancePmap/pmcontribsort.h
6fd5fde7e4adpmapmerge

Last Change
Raw File

pmcontribsort.h
View Options

/*
======================================================================
Header for N-way out-of-core merge sort of precomputed photon
contributions.
Roland Schregle (roland.schregle@{hslu.ch, gmail.com})
(c) Lucerne University of Applied Sciences and Arts,
supported by the Swiss National Science Foundation
(SNSF #179067, "Light Fields for Spatio-Temporal Glare Assessment")
======================================================================
$Id$
*/
#ifndef _PMAPCONTRIB_SORT_H
#define _PMAPCONTRIB_SORT_H
#include "pmapdata.h"
int contribPhotonSort (
const PhotonMap *pmap, FILE *leafFile, FVECT bbOrg, RREAL bbSize,
unsigned numBlk, unsigned long blkSize, unsigned numProc
);
/* Sort photons and their corresponding binned contributions by
* subdividing input files into small blocks, sorting these in-core, and
* merging them out-of-core via a priority queue. This code is adapted
* from oocsort.{h,c}
*
* The sort is implemented as a recursive (numBlk)-way sort; the input
* files are successively split into numBlk smaller blocks until these
* are of size <= blkSize, i.e. small enough for in-core sorting, then
* merging the sorted blocks as the stack unwinds. The in-core sort is
* parallelised over numProc processes.
*
* Parameters are as follows:
* pmap Photon Map with opened photon heap and precomputed
* contributions
* leafFile Opened output file for sorted photons (=OOC octree leaves)
* numBlk Number of blocks to divide into / merge from
* blkSize Max block size and size of in-core sort buffer, in bytes
* numProc Number of parallel processes for in-core sort
* bbOrg Origin of bounding box containing photons
* bbSize Extent of bounding box containing photons
*/
#endif
c4science · Help