Hello. I hope you are well. You may remember that we had a brief e-mail
discussion perhaps a year-and-a-half ago.
Recently, I discovered the new version and have been using it quite a bit on
my PC. I wanted to compute a variogram for a very large gridascii file
(729x729=5x10^5) and ran out of memory. I have 72 MB ram.
I thought a good solution would be to install the new program on our HP
Supercomputer. I tried first to compile it myself, but was not able to.
There was a problem with the definition of DINFINITY. Then I got some help
from our computer operator. He said _DINFINITY showed up in four
files: itersym.c, lufactor.c, qrfactor.c, and solve.c and that it is
supposed to replace HUGE_VAL. He found a default value for HUGE_VAL in the
code and defined it in those routines. He said that he didn't know why it
wasn't defined in the system includes. I'm not an expert on these things.
Never-the-less, he got the program compiled and I tried a test case that I
had done previously on the PC. The results were identical. However, when I
tried my large file, I got some incorrect results. The mean of the data is
correct, so I believe the file has been read correctly. The bin sizes are
probably correct too; we work in the 1x1 plane. The trouble starts in the
next column. I'm not sure if the trouble might be due to our problems at
compile-time or if we are just using a file that is too large. It may even
just be an output format issue. That is something I can probably modify
here. It took 20 hours to do the computation though, so I wanted to let you
know what was going on and possibly have your opinion before I try it again.
(Too bad if it turns out to be pure nugget, which it looks like it might!)
Sorry if this is addressed in the Computers and Geosciences article. Our
library cancelled its subscription in 1996. If you have a reprint, I would
love to have a copy.
Here is the output:
(~/gstat) cat b3i6d1.var
#gstat HP-UX 2.1.0 (August 1999) [./gstat 2.cmd]
#sample semivariogram
#Fri Apr 7 21:14:45 2000
#data(fractal): 'junk.dat';
#[1] mean: 0.00137174 variance: 0.00136986
#cutoff: 0.470752 interval width: 0.0313835
#direction: total
# from to n_pairs av_dist semivariogram
0 0.0313835 426273566 0.0208927 0.00132445
0.0313835 0.062767 1228117162 0.0487169 0.00133802
0.062767 0.0941505 1971363668 0.0794017 0.00132925
0.0941505 0.125534 -1655655558 0.110498 0.00131992
0.125534 0.156917 -1047944328 0.141701 0.00133207
0.156917 0.188301 -505141450 0.172957 0.00135515
0.188301 0.219684 -10324900 0.204276 0.00138111
0.219684 0.251068 402413102 2.75023 0.0164274
0.251068 0.282451 773850026 1.74834 0.00939973
0.282451 0.313835 1090109332 1.47341 0.00729726
0.313835 0.345218 1346266982 1.38121 0.00636835
0.345218 0.376602 1551918706 1.36001 0.0058593
0.376602 0.407985 1712255686 1.37655 0.00550089
0.407985 0.439369 1802717384 1.43323 0.00528542
0.439369 0.470752 1874457258 1.49781 0.00504564
(~/gstat)
Thanks for taking a look.
Mike Sukop
