[Liblas-devel] LAZ / LAS extraction performance ... why does scene size make a difference?

[Michael P. Gerlek]

One thing I'd like to see us do is create a set of reference files for the
community to use when benchmarking.  I picked my three from Howard's samples
pile more or less at random, but I'd like to make sure we have a set that
includes one with RGB data, one with nicely ordered points, one with
horribly ordered points, one with waveforms, etc.  Thoughts?

(Note this is not the same as a creating compliance or correctness reference
set, which would include one 1.2 file, one 1.3 file, one file with icky
VLRs, etc.  Although clearly that would be a Good Thing to do also.)

Also, it's already been suggested that a subsequent version of las2las
support a "-o null" convention to remove some of the I/O overhead.

-mpg




From: liblas-devel-bounces at lists.osgeo.org
[mailto:liblas-devel-bounces at lists.osgeo.org] On Behalf Of Michael Rosen
Sent: Wednesday, February 02, 2011 11:24 AM
To: liblas-devel at lists.osgeo.org
Subject: [Liblas-devel] LAZ / LAS extraction performance ... why does scene
size make a difference?



I also ran some benchmarks on laz and yes, MG4.  I’d like to be able to
collaborate with some of you on this effort.  I think MPG’s comments about
the different target workflows are spot on.  For now, I do have an
interesting observation about LAS and LAZ:  in both cases there seems to be
a linear relationship between the cropped area the time it takes to do the
extraction.  That would be expected if either
(a) there was some indexing support that told us where in the file the
points were.  But I don’t believe there is any.
(b) The process was dominated by output latency.  But I’m not creating
output (specifying “-o –“ 
 which is what that’s for I assume)

Absent either of those factors I would have expected the extraction time to
be roughly constant 
 a full scan through the file and pick out the points
we care about.

So, why is it doing this?  In both graphs, Y is time in seconds.  In the
first graph, X is the fraction of the total file cropped out.  In the
second, X is the number of points actually extracted.




To test, I installed OSGeo4W and ran las2las on increasingly large crop
areas from the center of MtStHelens.laz and .las (I encoded my own, but
sample are available from http://liblas.org/samples.

Here’s the script I used:

TimeThis las2las -e
1210343.88937174,315410.054252848,1215428.96062826,323308.825747152  -i
MtStHelens.las -o -
TimeThis las2las -e
1209290.7366317,313774.165113335,1216482.1133683,324944.714886665  -i
MtStHelens.las -o -
TimeThis las2las -e
1208482.6241118,312518.903227245,1217290.2258882,326199.976772755  -i
MtStHelens.las -o -
TimeThis las2las -e
1207801.35374348,311460.668505697,1217971.49625652,327258.211494303  -i
MtStHelens.las -o -
TimeThis las2las -e 1207201.1425,310528.345,1218571.7075,328190.535  -i
MtStHelens.las -o -
TimeThis las2las -e
1206658.51005792,309685.460122058,1219114.33994208,329033.419877942  -i
MtStHelens.las -o -
TimeThis las2las -e
1206159.50802811,308910.347481574,1219613.34197189,329808.532518426  -i
MtStHelens.las -o -
TimeThis las2las -e
1205695.0482634,308188.89022667,1220077.8017366,330529.98977333  -i
MtStHelens.las -o -
TimeThis las2las -e
1205258.81811523,307511.282758545,1220514.03188477,331207.597241455  -i
MtStHelens.las -o -
TimeThis las2las -e
1204846.22138258,306870.385680395,1220926.62861742,331848.494319605  -i
MtStHelens.las -o -
TimeThis las2las -e
1204453.78830496,306260.80932432,1221319.06169504,332458.07067568  -i
MtStHelens.las -o -
TimeThis las2las -e
1204078.82322361,305678.36645449,1221694.02677639,333040.51354551  -i
MtStHelens.las -o -
TimeThis las2las -e
1203719.18242262,305119.727182058,1222053.66757738,333599.152817942  -i
MtStHelens.las -o -
TimeThis las2las -e
1203373.12778539,304582.19164595,1222399.72221461,334136.68835405  -i
MtStHelens.las -o -
TimeThis las2las -e
1203039.22685462,304063.534773533,1222733.62314538,334655.345226467  -i
MtStHelens.las -o -
TimeThis las2las -e
1202716.28248697,303561.897011393,1223056.56751303,335156.982988607  -i
MtStHelens.las -o -
TimeThis las2las -e
1202403.2820478,303075.705408185,1223369.5679522,335643.174591815  -i
MtStHelens.las -o -
TimeThis las2las -e
1202099.3598951,302603.615340005,1223673.4901049,336115.264659995  -i
MtStHelens.las -o -
TimeThis las2las -e
1201803.76913607,302144.466639106,1223969.08086393,336574.413360894  -i
MtStHelens.las -o -
TimeThis las2las -e 1201515.86,301697.25,1224256.99,337021.63  -i
MtStHelens.las -o -