Work @ CCT

Movies with shrimps

2007-09-06T18:40:00.000-07:00

Today I've finally succeeded compiling Cactus with HDF5. I have the simulation running on numrel03 and by tomorrow morning I should have the h5 files.

You can see the last movie I made - the one with the reduced color range for the water here:
http://www.cct.lsu.edu/~elena/ShrimpsBiomass.gif
I'll try to play a little bit more with the range, because I want to see more colors.

Past movies:
This has the range varying (initially is from -140 to 20). In terms of coloring is more beautiful than ShrimpsBiomass.gif: http://www.cct.lsu.edu/~elena/masura.gif
This is just with the biomass - no water/land: http://www.cct.lsu.edu/~elena/biomass_new.gif

*** command
screen - to open another terminal in the terminal
screen -x -to see the opened terminals or to log into the only one opened

How to create movies from 2D data

2007-05-29T10:05:00.000-07:00

1. How to create individual frames from 2D data
Using a perl script and gnuplot, generate the individual frames with the command:

./count.pl | gnuplot

(the gnuplot version recommended is 4.2; it seems that the previous versions do not support png format).

This is the perl script "count.pl" we used:

#! /usr/bin/perl -w

use strict;

my $itlast = 364;
javascript:void(0)
print "set pm3d map\n";
print "set size ratio -1\n";
print "set cbrange [0:1]\n";
print "set terminal png\n";
print "set title 'Shrimp Biomass'\n";
for (my $iter = 1; $iter <= $itlast; ++ $iter) {
printf "set output 'bmnew.%04d.png'\n", $iter;
printf "splot 'bmnew_2D.asc' index $iter\n";
}

The above script is for generating a frame for each day for the total cohort of 4.000.000 shrimps.

2. How to combine individual frames into a movie
- first, convert the png into tiff; there are two ways to do it: either use mogrify ( mogrify -format tiff *.png ), or convert.
- second, combine the frames into a movie called shrimp.mpg:
ffmpeg -f mpeg2video -pass 1/2 -b 20M -i bmnew.0%d.tiff shrimp.mpg

The small movie shrimp.mpg is ready to be visualized!
Enjoy it at www.cct.lsu.edu/~elena/index/shrimp.mpg or as a small gif file at www.cct.lsu.edu/~elena/biomass_new.gif

Useful commands for work with references

2007-05-01T13:39:00.001-07:00

* look for the keyword and performance in the Cactus.bib file.
grep keyword Cactus.bib | grep performance

* count how many times the word performance given as a keyword appears in the bibtex file
grep keyword Cactus.bib | grep performance | wc -l

* sort a bibtex file for Performance papers (the keyword for these is performance) and pipe the result into a new file Performance bib:
bibgrep keyword:performance Cactus.bib > Performance.bib

* list the entries by year in descending order:
./bibtex2html -r -d Performance.bib

Variables made parameters in Cactus

2007-04-04T07:59:00.000-07:00

1. Remove the declaration of the variable

2. In the param.ccl from ~/Cactus/arrangements/CactusShrimp/Shrimp/ declare the variables you want to make as Cactus parameters

i.e.

PRIVATE:
( !!Access level)

INT endday "The last day of the simulation or the number of days"
(!!The parameter type and name)

{
1:365::"range for the endday"
(!! The range or the allowed values)

} 80
(!!default value)

3. In the parameter file specify the value of the variable (~/Cactus/exe/shrimp.par):

i.e. - add this line:
Shrimp::endday=12

4. In the main file ( inside the code) :

- include the following library:
#include "cctk_Parameters.h"

- in the subroutines where the variable is used, add:
DECLARE_CCTK_PARAMETERS
(that would be after the implicit none)

Integrating the Shrimp code into Cactus

2007-04-03T07:50:00.000-07:00

1. Download Cactus

-download the script that is used to checkout Cactus:
wget http://www.cactuscode.org/download/GetCactus

-change the permision:
chmod 755 GetCactus

-execute:
./GetCactus

--------------------------------------------------------------------

Directory for Cactus installation [Cactus] :

Anonymous Flesh checkout? (y)es, n)o, h)elp) [yes] : yes

Checkout Flesh from
[1] Last stable release
[2] Standard development version (bug fixes/unstable)
Choose version: (1-2, h)elp, q)uit) [1] : 1

Verbose checkout (y)es, n)o, h)elp) [no] : no

--------------------------------------------------------------------

2. Create a new thorn

cd Cactus

gmake newthorn

________________________________________________________________________
Creating a new thorn
Thorn name
--> Shrimp

The following arrangements are available:
Pick one, or create a new one.
arrangement
--> CactusShrimp

Thorn Author Name
--> Elena C.

Email Address
--> elena@cct.lsu.edu

Add another author? Y/N
--> n

Creating new arrangement CactusShrimp
Creating thorn Shrimp in CactusShrimp
All done.
Please remember to fill out the README and doc/documentation.tex files.
________________________________________________________________________

- copy the shrimp files in the "src" directory and rename the file with the code from .f90 to .F90
- finally, these files should be in src:

Bolivar2003.prn fineelmtrim.txt make.code.defn
CBelevations.prn GITemps2003.txt Shrimp_02_08_07_clean_UNIX.F90
EGelevations.prn Grand Isle 2003.txt var1.i

Note:
- the difference between the extensions .f90 and .F90 is about the "no preprocessing" Fortran 90 free form
- the file has to be rename, otherwise later there will be problems

3. Prepare the F90 file for compilation

- we choose to integrate into Cactus the optmized/tuned version of the shrimp code

- here are the changes we made in order to put into Cactus:

-#include "cctk.h"
-#include "cctk_Arguments.h"
!! including the necessary libraries

!!make the main program a subroutine:
-subroutine Shrimp_main(CCTK_ARGUMENTS)
- implicit none
-
- DECLARE_CCTK_ARGUMENTS

!! and don't forget to close the subroutine instead of the program;
-end subroutine Shrimp_main-

!! change all the INCLUDE 'name' to #include "name"
i.e.
-#include "var1.i"

4. Modify make.code.defn from Cactus/arrangements/CactusShrimp/Shrimp/src/

# Main make.code.defn file for thorn Shrimp
# $Header$

# Source files in this directory
SRCS = Shrimp_02_08_07_clean_UNIX.F90

# Subdirectories containing source files
SUBDIRS =

5. Modify the schedule.ccl from Cactus/arrangements/CactusShrimp/Shrimp/

# Schedule definitions for thorn Shrimp
# $Header$

SCHEDULE Shrimp_main AT EVOL
{
LANG: FORTRAN
} "Main function of the Shrimp code"

6. Modify the interface.ccl to

# Interface definition for thorn Shrimp
# $Header$

implements: Shrimp #name of the thorn
inherits:

7. Create a file with the thorns to be used in the compilation

cd ~/Cactus/
vim file

-in our case, the file contains only CactusShrimp/Shrimp
i.e.
less file
CactusShrimp/Shrimp

8. Use an option file for compilation (i.e. the options-numrel-intel file from Enrique)

9. Compile the code

gmake shrimp_conf-config THORNLIST=file options=options-numrel-intel

10. You should have the following message returned at the end :

________________________________________________________________________
All done !
________________________________________________________________________

11. Create a parameter file in Cactus/exe

less shrimp.par
ActiveThorns="Shrimp"
Cactus::cctk_itlast = 80

12. This is important and it might be omitted.

- the external files used by the code should be inside the exe directory as well.
- therefore the following should be there:

Bolivar2003.prn checkhab.out GITemps2003.txt var1.i
cactus_shrimp_conf EGelevations.prn Grand Isle 2003.txt
CBelevations.prn fineelmtrim.txt shrimp.par

13. From inside the exe directory run the code:

- cactus_shrimp_conf should be there

mpirun -np 1 cactus_shrimp_conf shrimp.par

Enjoy!!!

Profiling

2007-03-05T10:51:00.000-08:00

Profiling a Fortran code:

990 ifort -o Shrimp.out Shrimp_02_08_07_clean_UNIX.f90 -O3 -pg
991 ./Shrimp.out
992 gprof Shrimp.out > Shrimp.gprof

Problems with Fortran compiler on Darwin and othe bench stuff...

2006-12-29T12:46:00.000-08:00

Today I had problems with compiling the Cactus benchmark on a Darwin machine.

Something useful I found on the CactusCode Developer's list:
http://www.cactuscode.org/pipermail/users/2005-March/001391.html

-off-topic:Someone sent me today this link:
http://www.dilbert.com/comics/dilbert/archive/images/dilbert2006152711228.gif

Here I have my final gnuplot script:

#This script contains the commands to generate the plot for the BSSN_PUGH benchmark results.

#This is the dataset we used and which goes in BSSN_PUGH_100l.dat
#machine wall time #processors
#Bassi_sp 567.268 1
#numrel02 728.128 1
#catbert 853.416 1
#xws2 965.440 1
#santaka 1140.422 1
#helix2 1436.51 1

#Set the range for the axes
set xrange [-0.5:6]
set yrange [0:1500]

#Set the width of the boxes for the histogram
set boxwidth 0.15 absolute
set style fill solid 1.00 border -1

#Set the labels on the x axis and rotate them
set xtics ("Bassi" 0.00000, "numrel02" 1.0000, "catbert" 2.00000, "xws2" 3.00000, "santaka" 4.00000, "helix2" 5.00000)
#set xtics rotate by -45
plot "BSSN_PUGH_100l.dat" u :2 w boxes

#Set the title and the axes labels
set title "Wall Time Comparison for Different Machines"
set ylabel "Wall Time in Seconds"
set xlabel "Machine and Architecture"

#Plot the graph and save the output to a file
plot "BSSN_PUGH_100l.dat" u :2 w boxes
set terminal png color small
set output "BSSN_PUGH_100l.png"
plot "BSSN_PUGH_100l.dat" u :2 w boxes

...I finished the Presentation web page. Now it looks really nice. I uploaded the .bib, -bib.html and the .html files.
Here is the link... but I don't know if it's working for those that don't have an account:
http://www.cactuscode.org/Documentation/Presentations/

I also made the News item that I owed to Gab and Tom Goodale about the BSSN_PUGH scaling on BGW.
http://www.cactuscode.org/News/bssnpughonbgw
For this I had an Excel file and (Excel) graph, but I wrote a gnuplot script for this too, to plot the scaling results.

Today I also cleaned the KrancLatex CVS. It took me more than one hour... but I'm done for this year ... or at least for this payperiod.

Things to do:
-ask Peter or Yaakoub or the cactusmain group about Nemeaux and/or greengrass
-find if there is any repository for Presentations, if not create one
-convert the sdd into pdf
-eventually upload the ppts and the sdds.
-figure out how to label with bullets instead of numbers
-do the same things for Bench_whisky_Carpet, as for Bench_BSSN_PUGH

Working with bibtex

2006-12-29T12:28:00.000-08:00

Yesterday I learnt how to modify a bibtex style file bst. I didn't change much; I just made the title in italics.
Here is the portion of code that does this:

FUNCTION {format.title}
{ title empty$
{ "" }
{ title "t" change.case$ emphasize}
if$
}

The commands for creating the html file are in my case:

-to sort the entries by year and reverse them:
./bib2bib -r -s year Presentations.bib > Presentations1.bib

-to create the html files:
./bibtex2html -style plain.bst Presentations1.bib

An entry in the Presentations.bib looks like this:

@INPROCEEDINGS{CactusPresentations_Allen01b,
author = {Gabrielle Allen},
title = {{Numerical Relativity with Cactus}},
urlpdf = {http://www.cactuscode.org/Documentation/presentations/gr16July01.pdf
},
note = {Description: Numerical Relativity Workshop, Satellite Meeting of GR16
- Johannesberg, South Africa},
month = {July},
year = {2001}
}

I tried to find a way to have the entries labeled with bullets instead of numbers for the html, but I didn't find anything yet. :(

Yesterday I partially finished running the benchmark. ( I still couldn't get my hands on igertws09 and I couldn't compile the Bench_BSSN_PUGH on Nemeaux or greengrass - arch=Darwin.) In the same time with the runnings I wrote the .bib file and convert the ppts to pdfs. There were some presentations on sdd format and I was told that these could be opened with OpenOffice. Now, the presentation web page looks reasonable. Eventually I would have to delimitate the entries by year with a line - this was the suggestion I received; this would make the page looks nice, but having bullets instead of numbers would make it nicer.

Useful Unix commands

2006-12-25T05:19:00.000-08:00

uname -a --prints all system information

cat /proc/cpuinfo -- tells the processor type, the clockspeed, etc for each processor

top -- shows the top jobs running on the machine

Who says it's not easy to create a PC-Mac network?

2006-12-22T03:03:00.000-08:00

Yesterday I made my first small home-network: my Mac connects to Internet through a PC with a wireless card. I did not have a router, but only two computers, a USB cable and a wireless card. I had to install the wireless card, connect the computers and create a bridge connection between the two. Finally, I had to allow the dial-up connection to share the internet via wireless.
It was fun reading about networks and setting up computers. :)
Who could have ever imagined that I would have to do this on my first day of Christmas break... after 22 hours of flight?...

histograms in gnuplot

2006-12-12T17:13:00.000-08:00

These days I'm doing benchmarking stuff.
Here are a few commands to generate histograms with gnuplot:

set xrange [-0.5:4]
set yrange [0:1100]
set boxwidth 0.5 absolute
set style fill solid 1.00 border -1
set xtics ("Bassi(SP P5-575 1.9Ghz)" 0.00000, "Bassi(DP P5-575 1.9Ghz)" 1.0000,
"Loni(P5-575 1.9Ghz)" 2.00000, "Supermike(Intel Xeon 3.06Ghz)" 3.00000)
set title "Wall Time Comparison for Different Machines"
set ylabel "Wall Time in Seconds"
set xlabel "Machine and Architecture"
plot "BSSN_PUGH_100l.dat" u :2 w boxes
plot "BSSN_PUGH_100l.dat" u :2 w boxes
set terminal postscript
set output "BSSN_PUGH_100l.eps"
plot "BSSN_PUGH_100l.dat" u :2 w boxes

Facts I "discovered" today:
-gnuplot's postscript terminal does not have the option of changing the page size.
-gnuplot writes the entire .eps file from scratch

What's this all about?

2006-12-09T17:14:00.001-08:00

Hi there!

This is my first entry in a blog that is supposed to describe my CCT work. I just hope that I can make time to write on it.
It will be only for my personal reference. This last summer I had an internship at LBL and those guys asked me to have a web-page with the projects I was working on. At the end, when I had to give two presentations it was really helpful. There were lots of details I wrote down at the time I was writing the pages and that I wouldn't have remember otherwise.

More, I think it's good to have something you can come back to at the end of the week and see what your progress of that week was.

I wish y'all the best,
Elena