CosmoMC Mar-2013 Testing

[Sheng Li]

This is a report for test & verificaion, as well as for those who are going to configure and run this new version of CosmoMC by their own.

Requirements (as stated in makefile or otherwise on CosmoMC ReadeMe webpage):
{Fresh Installation:}

Code: Select all

python (2.6.6 in my local)
numpy v1.7.0 (latest)
PyFITS v3.1.1 (latest)

{Local Configurations Updated:}

Code: Select all

Ifort v2013;
Intel-MKL 11.0 (associated with ifort 2013 XE);
mpi: openmpi 1.6.4 (latest);

together with the following old but working packages:
{previously deployed for CosmoMC Oct 2012}

Code: Select all

cfitsio: v3.30
healpix: v2.20a

Location of Likelihood files (.clik)
[size-18]I put the .clik foldes and lensing likefile under [/size]

Code: Select all

/path/to/planck/plc-1.0

The test file and configurations is below:

Code: Select all

Test File: {test.ini}
DEFAULT(batch1/CAMspec_defaults.ini)
DEFAULT(batch1/lowl.ini)
DEFAULT(batch1/lowLike.ini)

**use_clik = F ** ---- I changed to false for test since I did not compile with clik.

Then the outputs for a single thread after run are:
Execute { ./cosmomc test.ini }

Code: Select all

 Number of MPI processes:           1
 Random seeds: 20919, 28823 rand_inst:   1
 WMAP options (beam TE TT) T T T
 WARNING: zero padding ext cls in LoadFiducialHighLTemplate
 Computing tensors: F
 Doing CMB lensing: T
 Doing non-linear Pk:           0
 lmax              = 6500
 lmax_computed_cl  = 2500
 max_eta_k         =    6625.00000000000     
 transfer kmax     =   0.800000011920929     
 Number of C_ls =    4
 Fast divided into            1  blocks
 Varying  6 parameters ( 6 slow ( 2 semi-slow),  0 fast ( 0 semi-fast))
 skipped unused params: aps100 aps143 aps217 acib143 acib217 asz143 psr cibr ncib cal0 cal2 xi aksz bm_1_1
 starting Monte-Carlo
 Initialising BBN Helium data...
           0 rat:   1.000000      in           50  (M) best:   0.0000000E+00
           0 rat:   1.000000      in          100  (M) best:   0.0000000E+00
 Chain1, MPI done 'burn', Samples =128, like =   0.0000000E+00
 Time:    114.818511962891      output lines=         127
 slow changes          87 semi-slow changes          41
           1 all_burn done
           0 rat:   1.000000      in          150  (M) best:   0.0000000E+00
           0 rat:  0.9950249      in          201  (M) best:   0.0000000E+00
           1 DoUpdates
           0 rat:  0.9727626      in          257  (M) best:   0.0000000E+00
           0 rat:  0.9771987      in          307  (M) best:   0.0000000E+00
           0 rat:  0.9803922      in          357  (M) best:   0.0000000E+00
           0 rat:  0.9708738      in          412  (M) best:   0.0000000E+00
 Chain           1  MPI communicating
 updating proposal density
           0 rat:  0.9656652      in          466  (M) best:   0.0000000E+00
           0 rat:  0.9541985      in          524  (M) best:   0.0000000E+00
           0 rat:  0.9515571      in          578  (M) best:   0.0000000E+00
           0 rat:  0.9493671      in          632  (M) best:   0.0000000E+00
           0 rat:  0.9489051      in          685  (M) best:   0.0000000E+00
 Chain           1  MPI communicating
 updating proposal density
 Reionization_zreFromOptDepth: Did not converge to optical depth
 tau =  0.157483404751234      optical_depth =   0.254022764599602     
   50.0000000000000        49.9984741210938     
 MpiStop:            0
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD 
with errorcode 23513272.

My question is:
If I did all thing correct, what to do for correcting reionization_zre so that it will converge?

At the writing moment, CosmoMC is running with the following;mpirun -np 2 ./cosmomc test.ini
But this time I compiled the cosmomc with clik.

And it seems that it works without convergence issue, though it is running only for about 30mins.

To Antony:
Please comment and advise if I made wrong, or any question on my configurations.
Thank you.

[Antony Lewis]

Looks like you are running with clik (PLanck likelihoods) off and no other data likelihoods at all? In which case the parameters have probably just wandered off into some crazy region of paremeter space since they are totally unconstrained.

[Sheng Li]

Alright, I think I got it work now.
Compiled with cliklike by setting in Makefile in source folder;
and switch

Code: Select all

use_clik=T

The outputs are:

Code: Select all

 Number of MPI processes:           1
 Random seeds: 17048,  1341 rand_inst:   1
 WMAP options (beam TE TT) T T T
 Using clik with likelihood file data/clik/CAMspec_v6.2TN_2013_02_26_dist.clik
----
clik version 5887 MAKEFILE
  CAMspec e61cec87-3a37-43ca-8ed1-edcfcaf5c00a
Checking likelihood 'data/clik/CAMspec_v6.2TN_2013_02_26_dist.clik' on test data. got -3908.71 expected -3908.71 (diff -3.76099e-08)
----
   TT from l=0 to l=        2500
 Clik will run with the following nuisance parameters:
 A_ps_100
 A_ps_143
 A_ps_217
 A_cib_143
 A_cib_217
 A_sz
 r_ps
 r_cib
 n_Dl_cib
 cal_100
 cal_217
 xi_sz_cib
 A_ksz
 Bm_1_1
 Using clik with likelihood file data/clik/commander_v4.1_lm49.clik
----
clik version 5887 MAKEFILE
  gibbs d462e865-e178-449a-ac29-5c16ab9b38f5
Checking likelihood 'data/clik/commander_v4.1_lm49.clik' on test data. got 3.2784 expected 3.2784 (diff -2.55579e-10)
----
   TT from l=0 to l=          49
 Using clik with likelihood file data/clik/lowlike_v222.clik
 Initializing Planck low-likelihood, version v2.1
----
clik version 5887 MAKEFILE
  lowlike "lowlike v222"
Checking likelihood 'data/clik/lowlike_v222.clik' on test data. got -1007.04 expected -1007.04 (diff -1.97381e-05)
----
   TT from l=0 to l=          32
   EE from l=0 to l=          32
   BB from l=0 to l=          32
   TE from l=0 to l=          32
 adding parameters for: lowlike_v222.clik
 adding parameters for: commander_v4.1_lm49.clik
 adding parameters for: CAMspec_v6.2TN_2013_02_26_dist.clik
 WARNING: zero padding ext cls in LoadFiducialHighLTemplate
 Computing tensors: F
 Doing CMB lensing: T
 Doing non-linear Pk:           0
 lmax              = 6500
 lmax_computed_cl  = 2500
 max_eta_k         =    6625.00000000000     
 transfer kmax     =   0.800000011920929     
 Number of C_ls =    4
 Fast divided into            1  blocks
 Varying 20 parameters ( 6 slow ( 2 semi-slow), 14 fast ( 0 semi-fast))
 starting Monte-Carlo
 Initialising BBN Helium data...
 Chain:0 mult:1 accept drag, accpt:   1.000000     fast/slow   31.00000    
 Chain:0 mult:2 accept drag, accpt:  0.7500000     fast/slow   42.50000    
 Chain:0 mult:2 accept drag, accpt:  0.6666667     fast/slow   42.75000    
 Chain:0 mult:1 accept drag, accpt:  0.7142857     fast/slow   40.20000    
 Chain:0 mult:1 accept drag, accpt:  0.7500000     fast/slow   39.83333    
 Chain:0 mult:3 accept drag, accpt:  0.6363636     fast/slow   43.55556    
 Chain:0 mult:1 accept drag, accpt:  0.6666667     fast/slow   45.90000    
 Chain:0 mult:1 accept drag, accpt:  0.6923077     fast/slow   46.63636    
 Chain:0 mult:4 accept drag, accpt:  0.5882353     fast/slow   46.58333    
 Chain:0 mult:3 accept drag, accpt:  0.5500000     fast/slow   46.60000    
 Chain:0 mult:1 accept drag, accpt:  0.5714286     fast/slow   45.93750    
 Chain:0 mult:2 accept drag, accpt:  0.5652174     fast/slow   46.17647    
 Chain:0 mult:2 accept drag, accpt:  0.5600000     fast/slow   46.89474    
 Chain:0 mult:5 accept drag, accpt:  0.5000000     fast/slow   49.66667    
 Chain:0 mult:1 accept drag, accpt:  0.5161290     fast/slow   49.12000    
 Chain:0 mult:2 accept drag, accpt:  0.5151515     fast/slow   48.53846    
 Chain:0 mult:4 accept drag, accpt:  0.4864865     fast/slow   47.25000    
 Chain:0 mult:2 accept drag, accpt:  0.4871795     fast/slow   47.31034    
 Chain:0 mult:2 accept drag, accpt:  0.4878049     fast/slow   48.03226    
 Chain:0 mult:1 accept drag, accpt:  0.5000000     fast/slow   48.68750    
 Chain:0 mult:1 accept drag, accpt:  0.5116279     fast/slow   49.03030    
 Chain:0 mult:7 accept drag, accpt:  0.4600000     fast/slow   48.18421    
 Chain:0 mult:4 accept drag, accpt:  0.4444444     fast/slow   47.85366    
 Chain:0 mult:6 accept drag, accpt:  0.4166667     fast/slow   48.08889    
 Chain:0 mult:2 accept drag, accpt:  0.4193548     fast/slow   48.63830    
 Chain:0 mult:2 accept drag, accpt:  0.4218750     fast/slow   48.97959    
 Chain:0 mult:1 accept drag, accpt:  0.4307692     fast/slow   49.18000    
 Chain:0 mult:1 accept drag, accpt:  0.4393939     fast/slow   49.43137    
 Chain:0 mult:4 accept drag, accpt:  0.4285714     fast/slow   50.43636    
 Chain:0 mult:9 accept drag, accpt:  0.3924051     fast/slow   52.04839    
 Chain:0 mult:1 accept drag, accpt:  0.4000000     fast/slow   52.04762    
 Chain:0 mult:4 accept drag, accpt:  0.3928571     fast/slow   51.53731    
 Chain:0 mult:1 accept drag, accpt:  0.4000000     fast/slow   51.54412

Thanks Antony.

[Antony Lewis]

Great. (If you set the exports in ~/.bashrc as described in the readme you shouldn't need to modify the Makefile; for actual run I assume you will have more than one chain)

[Sheng Li]

Great. (If you set the exports in ~/.bashrc as described in the readme you shouldn't need to modify the Makefile;

Yes, in my bashrc, I found I had actually the exports configurations.

Oh, I understand what you are suggesting.

for actual run I assume you will have more than one chain)

Sure, I test it on my local pc, but will run it with 8cores in HPC.

So I think I got it work now.

Thanks, and Good day..

[Sheng Li]

I want to assure one fact that
cosmomc MUST be compiled with specifying the option:

Code: Select all

PLANCKLIKE = cliklike

It will NOT work if just exporting the variable PLANCKLIKE to your login or working environment.

I can ensure that I have this env-variable, but cosmomc will NOT find the cliklike data if the option is *EMPTY*.

Can anyone else point if I did sth incorrect so that cosmomc wouldn't find cliklike?
What I have known so far is I can ONLY run cosmomc with the configuration which stated in code section, ie. quoted in

Code: Select all

.

I can verify this both in my own pc and in a HPC.

To Antony,
Did you compile without PLANCKLIKE in Makefile, but cosmomc will work with finding the right place for cliklike?

Also I am sure I have made the symbolical link to plancklike path.

[Antony Lewis]

The supplied Makefile should have

Code: Select all

PLANCKLIKE ?=

This will pull the default setting from your environment variables. But it needs ?= not =. Did you change it?

[Sheng Li]

The supplied Makefile should have

Code: Select all

PLANCKLIKE ?=

This will pull the default setting from your environment variables. But it needs ?= not =. Did you change it?

Oh, i did not understand this '?' .
Yes, I removed it so it is no wonder why I had not make it.

Thank you for correction.