PolSpice V03-00-01 NAN value

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Song Chen
Posts: 3
Joined: August 21 2014
Affiliation: University Bielefeld

PolSpice V03-00-01 NAN value

Post by Song Chen » February 17 2015

Hello,

I get NAN results in cl and cor.
When I process a Nside 32 haslam map ( ud_grade haslam Nside 512 to Nside 32)

I use Healpix 3.20 with PolSpice v03-00-01
compiled with gfortran

Thanks

Dragan Huterer
Posts: 27
Joined: July 18 2005
Affiliation: University of Michigan
Contact:

PolSpice V03-00-01 NAN value

Post by Dragan Huterer » February 28 2015

I have the same problem. It appears that for NSIDE [tex]\leq[/tex] 32 the Cl come out as NaN or Infinity, while for NSIDE [tex]\geq[/tex] 64 they appear fine. This is in Polspice v03 run either from terminal or from IDL (ispice).

Any ideas? - much appreciated!

Eric Hivon
Posts: 7
Joined: September 25 2004
Affiliation: IAP

PolSpice V03-00-01 NAN value

Post by Eric Hivon » March 02 2015

Thank you for these bug reports.
The latest version of PolSpice (v03-00-03) fixes these problems
http://www2.iap.fr/users/hivon/software/PolSpice/

Dragan Huterer
Posts: 27
Joined: July 18 2005
Affiliation: University of Michigan
Contact:

PolSpice V03-00-01 NAN value

Post by Dragan Huterer » March 03 2015

Thanks Eric! Works now.

There is one more, probably unrelated, quirkiness that is specific to the IDL version only (ispice). When I run it on some simple file
HIDL> ispice, 'map.fits', 'OUTPUT/spice.cl'
the output file looks weird: it has one very long single line, and in it first contents of the fits file header(s) followed by the angular power spectrum Cl without the corresponding multipoles l - all of these words and numbers in one line. Any ideas? The angular power spectrum does appear to be correct. I can provide the (ugly) output if it's helpful.

One possible hint is that in the idl-screen output it reports
FITS output...... YES
while the same is given as a NO in the terminal-run output (i.e. output of just running spice from terminal).

Eric Hivon
Posts: 7
Joined: September 25 2004
Affiliation: IAP

PolSpice V03-00-01 NAN value

Post by Eric Hivon » March 04 2015

Hi Dragan,

what you described looks very much like a FITS file.
You can look at it more conveniently by typing (assuming it is named file.fits):
Unix/Linux:

Code: Select all

stty cols 80 ; more file.fits
MacOSX:

Code: Select all

resize -s 40 80 > /dev/null ; more file.fits
The multipole l is not explicitly included in those files as it is assumed that C(l)s for all l starting at 0 are provided

You can read these files with either
- fits2cl.pro (from Healpix package), or
- read_spice.pro (in PolSpice/src next to ispice.pro, which should read any C(l), C(\theta), FITS, plain ASCII,... file produced by Spice).

Last twist: ispice.pro forces the fits_out flag to YES, for easier interface with Healpix

Cheers,
Eric

Dragan Huterer
Posts: 27
Joined: July 18 2005
Affiliation: University of Michigan
Contact:

PolSpice V03-00-01 NAN value

Post by Dragan Huterer » March 06 2015

Oh it's a fits file! I feel silly for not having thought of that... I was assuming it was supposed to be the ascii output as in the fortran version.
Fits2cl had no problem converting to cl. Thanks!

Post Reply