WMAP DR5 Likelihood Software Change Log
---------------------------------------
CHANGE LOG For V5 (9-YEAR DATA) RELEASE
---------------------------------------
D. L. Larson, E. Komatsu, M. R. Greason, January 8, 2013
We made cosmetic changes to the likelihood which mostly consisted of adjusting
configuration options to the recommended values, as described below. We did not
increment the version number since no data files have changed.
Configuration flags have been changed to recommended values. This involved
commenting out flags in the Makefile:
#WMAPFLAGS += -DUSE_LOWELL_TBEB # turns on low-l maxlike TB/EB
#WMAPFLAGS += -DUSE_HIGHELL_TB # turns on high-l master TB
#WMAPFLAGS += -DFASTERTT # speed up matrix low-l TT
#WMAPFLAGS += -DTIMING # print out timing stats for profiling
and turning on the Gibbs sampling likelihood at low ell in WMAP_9yr_options.F90:
Logical :: use_gibbs = .True.
Integer :: lowl_max = 32
The obsolete use_gibbs_pol_cleaning flag has been removed.
The expected likelihood for the slower pixel TT likelihood was corrected in
test.F90 to the following:
expected_like_tot = 12349.777782d0
Several untested compilation options have been removed from the Makefile.
---------------------------------------
CHANGE LOG For V5 (9-YEAR DATA) RELEASE
---------------------------------------
D. L. Larson, September 20, 2012
The 7-year data files and routines have been replaced by 9-year ones.
---------------------------------------
CHANGE LOG FOR V4P1 (7-YEAR DATA) RELEASE
---------------------------------------
D.L.Larson, June 4, 2010
The diagnostic program test.F90 wsa updated to be consistent with the V4P1
release; and lines 660-665 of WMAP_7yr_tetbeebbeb_pixlike.F90, which write
debugging information to a file, were commented out to avoid problems with some
compilers. Neither change affects the actual likelihood function.
J.Dunkley, E.Komatsu, D.L.Larson, M.R.Nolta, May 27, 2010
** Two small bugs have been fixed, which involve the high-ell TT, TE, TB, and
point source power spectra:
1) The original computation of the TT spectrum used an incorrect monopole
subtraction which resulted in a small amount of excess power at high ell, and a
corresponding elevation of the best-fit residual point source amplitude.
Correcting the monopole subtraction reduced the high-ell power slightly which
produced a correspondingly lower residual point source amplitude, from 11.0 to
9.0 uK2 sr.
2) Due to a simulation configuration error, the TE fsky recalibration factor
used in version 4.0 was 1.049 instead of the factor 1.011 used in this version.
---------------------------------------
CHANGE LOG FOR V4 (7-YEAR DATA) RELEASE
---------------------------------------
J.Dunkley, E.Komatsu, D.L.Larson, M.R.Nolta, January 2010
** The 5-year data files and routines have been replaced by
the 7-year ones.
---------------------------
CHANGE LOG FOR V3P2 RELEASE
---------------------------
E.Komatsu, March 5, 2009
The orders of many do-loops have been changed to speed up
the calculations significantly. We thank Raphael Flauger for
making these improvements.
The V3P1 patch has not been integrated into this release.
The following codes have been modified:
br_mod_dist.f90
WMAP_5yr_likelihood.F90
WMAP_5yr_teeebb_pixlike.F90
WMAP_5yr_tetbeebbeb_pixlike.F90
WMAP_5yr_tt_pixlike.F90
The timing output for the V3 release:
....... wmap_likelihood_init:
7.6650 teeebb_lowl_like_setup
10.7410 wmap_likelihood_init
....... wmap_likelihood_compute:
0.1680 compute_tt_gibbslike
....... teeebb_lowl_likelihood:
0.4740 - finished Dp
0.4670 - finished spotrf
0.0000 - finished dpotri
0.9540 teeebb_lowl_likelihood
1.1900 wmap_likelihood_compute
The timing output for this release:
....... wmap_likelihood_init:
1.3230 teeebb_lowl_like_setup
2.9590 wmap_likelihood_init
....... wmap_likelihood_compute:
0.1640 compute_tt_gibbslike
....... teeebb_lowl_likelihood:
0.0970 - finished Dp
0.4640 - finished spotrf
0.0000 - finished dpotri
0.5740 teeebb_lowl_likelihood
0.7730 wmap_likelihood_compute
---------------------------
CHANGE LOG FOR V3P1 RELEASE
---------------------------
-- Patch to make the data directory selectable through an environment variable.
T. Kinser, Lawrence Berkeley National Laboratory, Oct 2008
This patch is available as a separate download from LAMBDA.
---------------------------------------
CHANGE LOG FOR V3 (5-YEAR DATA) RELEASE
---------------------------------------
J.Dunkley, E.Komatsu, D.L.Larson, M.R.Nolta, February 2008
** The 3-year data files and routines have been replaced by
the 5-year ones.
WMAP_5yr_tt_pixlike.F90
-- The resolution 3 option has been removed. This module supports
only the resolution 4. For a quicker estimation of low-l temperature
likelihood (with the same precision) the Gibbs sampling is recommended.
To use the Gibbs sampling (and disable pixlike option), set
"use_gibbs = .true." in wmap_options.F90.
** The following new routines have been added:
WMAP_5yr_tetbeebbeb_pixlike.F90
-- This module computes the low-l polarization likelihood of
TE/TB/EE/BB/EB spectra. To use this, compile the codes with
"WMAPFLAGS += -DUSE_LOWELL_TBEB" enabled in Makefile, and
call wmap_likelihood_compute(cl_tt,cl_te,cl_tb,cl_ee,cl_eb,cl_bb,like)
WMAP_5yr_gibbs.F90 [together with br_mod_dist.f90]
-- This module computes the low-l temperature likelihood by
using the Gibbs sampling. To use this, set "use_gibbs = .true."
in wmap_options.F90.
** The following new compiler options (see Makefile) have been added:
WMAPFLAGS += -DUSE_LOWELL_TBEB
-- enable the pixel-based polarization likelihood option for TB/EB at l=2-23
WMAPFLAGS += -DUSE_HIGHELL_TB
-- enable the master likelihood option for TB at l=2-450
-----------------------------
CHANGE LOG FOR V2P2P2 RELEASE
-----------------------------
M.R.Nolta, November 2006
WMAP_3yr_tt_pixlike.F90:
-- Sped up the TT pixel likelihood module by a factor of 10 using
signal-to-noise eigenvalue compression (suggested by Jon Sievers; see
http://www.cita.utoronto.ca/~sievers/WMAP3TTfaster/ for more details).
To enable, set -DFASTERTT during compilation.
WMAP_3yr_likelihood.F90:
-- Minor bug fix; TE sanity check is now abs(TE)/sqrt(TT*EE)<1 instead of
TE/sqrt(TT*EE)<1.
-- Moved fisher matrix from the stack to the heap to fix crash when using the
XLF compiler.
WMAP_3yr_teeebb_pixlike.F90:
-- Added deallocation statements to fix crash when using the XLF compiler.
-----------------------------
CHANGE LOG FOR V2P2P1 RELEASE
-----------------------------
M.R.Nolta, October 2006
WMAP_3yr_likelihood.f90:
-- A beam error is now treated as a warning instead of an error. Warnings are
not fatal to a run; instead the correction is set to zero and the issue is
reported.
WMAP_3yr_util.f90:
-- New status variables were added to track warnings and to indicate whether or
not there were errors and/or warnings.
-- A subroutine was added to report warnings, and the error reporting subroutine
was modified to also report warnings.
---------------------------
CHANGE LOG FOR V2P2 RELEASE
---------------------------
E.Komatsu & M.R.Nolta, October 2006
Makefile:
-- removed. The code no longer needs installation of Healpix.
-- Added compile switches -DOPTIMIZE and -DOPTIMIZE2, which enable
various speed and memory optimizations. -DOPTIMIZE streamlines the code
in WMAP_3yr_teeebb_pixlike.f90 without changing the logic. It is enabled
by default. When -DOPTIMIZE2 is defined some elements of the TT signal
covariance matrix (in WMAP_3yr_tt_pixlike.f90) are found by interpolating
the 2-pt correlation function C(theta). In tests this approximation changes
-2lnL by O(0.01), but to be conservative -DOPTIMIZE2 is not enabled by
default. If both -DOPTIMIZE and -DOPTIMIZE2 are enabled the code is twice
as fast:
pass2_compute_likelihood() run times (sec)
------------------------------------------
(no optimizations) 5.2722001
-DOPTIMIZE 4.0882001
-DOPTIMIZE -DOPTIMIZE2 2.6022999
These times are from a dual Intel Xeon 2.4GHz node.
-- Added compile switch -DTIMING. If enabled the code prints out
timing messages useful for gauging performance. Should not be
enabled when running chains.
WMAP_3yr_options.f90:
-- character(len=*), parameter :: WMAP_data_dir = './data/'
^^^^^^^^^
*** AN IMPORTANT CHANGE WITH REGARD TO THE RESOLUTION ***
The TT pixel likelihood code now supports either resolution 4 (nside=16) or
resolution 3 (nside=8) for the input temperature map. The resolution 4 map
should be used for the most accurate results. On the other hand, the
resolution 3 map runs much more quickly. Therefore, if you would like to use
the resolution 3 map for testing your code, please change
integer :: lowl_tt_res = 4
integer :: lowl_max = 30
to
integer :: lowl_tt_res = 3
integer :: lowl_max = 12
in WMAP_3yr_options.f90. MAKE SURE TO CHANGE BOTH lowl_tt_res AND lowl_max.
WMAP_3yr_likelihood.f90:
-- Now the code stops when the input clte/sqrt(cltt*clee) is greater than 1.
-- Invalidate the beam correction when ln(likelihood_beam) >=
ln(likelihood_total)/4
-- Removed ptsrc and SZ spectra from wmap_likelihood_inputs_tt.dat file, and
clps, clsz, and sz_amp from the code. The point-source marginalization is
done in "WMAP_3yr_tt_beam_and_ptsrc_corr.f90". If you wish to include the
SZ marginalization, download a "Predicted SZ Power Spectrum" file from
LAMBDA and use the V-band prediction with its amplitude being marginalized.
You may do it by adding an SZ contribution to the TT spectrum, cl_tt = cl_tt
+ sz_amp*clsz, and vary sz_amp.
-- A constant offset, te_lndet_offset defined in WMAP_3yr_options.f90,
is now subtracted from the raw TETE lndet term.
WMAP_3yr_teeebb_pixlike.f90:
-- Marginalization over polarized foreground errors has been added.
-- Pixel-window function is now provided as an ascii file.
-- Loss imbalance mode projected out of N^{-1}.
-- A constant offset, teeebb_pixlike_lndet_offset defined in
WMAP_3yr_options.f90, is now subtracted from the raw EE/BB lndet term.
-- Added speed and memory optimizations, enabled by defining -DOPTIMIZE during
compilation.
WMAP_3yr_tt_beam_and_ptsrc_corr.f90:
-- ptsrc correction reduced by 18% and uncertainty doubled.
-- fixed bug triggered when lmin,lmax changed from default values of 2,1000.
WMAP_3yr_tt_pixlike.f90:
-- Healpix functions have been replaced by the pre-computed data.
>> The ILC and V-band maps have been smoothed and degraded already.
>> V1 and V2 maps have been coadded.
-- The res4 (nside=16) smoothing scale is 9.1831 degrees.
(Note that the ILC map has already been smoothed by a 1-degree beam, so
the FWHM of an additional smoothing is 9.1285 degrees.)
-- The res3 (nside=8) smoothing scale is 18.3 degrees.
-- A white noise term (1uK per pixel at nside=8 or 16) is added to regularize
S+N matrix.
-- Foreground marginalization using the difference between ILC and V-band maps
is explicitly imposed.
-- Using res3 or res4 maps is an option (lowl_tt_res).
MAKE SURE TO CHANGE lowl_max ALSO! lowl_max=30 and 12 are recommended for
lowl_tt_res=4 and 3, respectively.
-- A constant offset, tt_pixlike_lndet_offset defined in WMAP_3yr_options.f90,
is now subtracted from the raw TTTT lndet term.
-- Added optional faster approximate calculation of TT signal covariance matrix,
enabled by defining -DOPTIMIZE2 during compilation.
NEW DATA FILES
./data/test_cls.dat
./data/wmap_likelihood_inputs_tt.dr2p2.dat
./data/clps.dr2p2.dat
./data/alm/alm_tt_fs_r9_ilc_nopixwin_3yr_v2.dat
./data/healpix_data/pixel_window_n0008.txt
./data/healpix_data/pixel_window_n0016.txt
./data/maps/low_resolution_map_fs_r3_ilc_smooth_18.3deg.fits
./data/maps/low_resolution_map_fs_r3_vband_smooth_18.3deg.fits
./data/maps/wmap_fs_r4_ilc_smooth_9.1285deg.fits
./data/maps/wmap_fs_r4_vband_smooth_9.1831deg.fits
./data/maps/wt_r3_p2_3yr_pxim.K1.QV.map_q
./data/maps/wt_r3_p2_3yr_pxim.K1.QV.map_u
./data/maps/wt_r3_p2_3yr_pxim.QV.map_q
./data/maps/wt_r3_p2_3yr_pxim.QV.map_u
./data/mask/mask_r3_kp2.fits
./data/mask/mask_r4_kp2.fits
./data/ninv/masked_bb_ninvplninv_qu_r3_p2_norm_corrected_3yr_pxim.QV.fits
./data/ninv/masked_ee_ninvplninv_qu_r3_p2_norm_corrected_3yr_pxim.QV.fits
./data/ninv/masked_ninv_qu_r3_p2_norm_corrected_3yr_pxim.QV.fits
./data/ninv/masked_ninvy_qu_r3_p2_norm_corrected_3yr_pxim.QV.fits
NEW DATA FILES (Optional; not being used)
./data/maps/low_resolution_map_fs_r3_vband_clean_smooth_18.3deg.fits
./data/maps/wmap_fs_r4_vband_clean_smooth_9.1831deg.fits
DATA FILES REMOVED
./data/wmap_likelihood_inputs_tt.p2.dat
./data/clps.p2.dat
./data/bias_corrected_pass2_3yr_ilc.fits
./data/pass2f_fltFiltBZFL_fs_spm_r9_ximb_xsd_yr1_itf.V1.map_i
./data/pass2f_fltFiltBZFL_fs_spm_r9_ximb_xsd_yr2_itf.V1.map_i
./data/pass2f_fltFiltBZFL_fs_spm_r9_ximb_xsd_yr3_itf.V1.map_i
./data/wmap_kp2_mask_2yr_v2p3.fits
./data/alm/alm_tt_fs_r9_ilc_nopixwin_3yr.dat
./data/maps/wt_r3_p2_3yr.QV.map_q
./data/maps/wt_r3_p2_3yr.QV.map_u
./data/ninv/masked_bb_ninvplninv_qu_r3_p2_norm_corrected_3yr.QV.fits
./data/ninv/masked_ee_ninvplninv_qu_r3_p2_norm_corrected_3yr.QV.fits
./data/ninv/masked_ninv_qu_r3_p2_norm_corrected_3yr.QV.fits
./data/ninv/masked_ninvy_qu_r3_p2_norm_corrected_3yr.QV.fits
---------------------------
CHANGE LOG FOR V2 RELEASE
---------------------------
E.Komatsu & M.R.Nolta, March 2006
Initial Release
|
