In case I forget...

What ever you occupy, occupy it with passion.

2011-10-10T18:44:00.000-07:00

Everyone must do their part.

How to setup addons for Fedora 14 when running inside VirtualBox

2011-05-05T17:38:00.000-07:00

Well, follow the straightforward instructions on this site:

http://www.sysprobs.com/install-fedora-14-virtualbox-working-guest-additions

Compiling LAPACK as a shared library in linux

2010-12-22T23:24:00.001-08:00

So I've finally gotten MDAnalysis (a python module which relies on LAPACK) to work with custom LAPACK shared library. It makes me wonder what other functions are missing from the lapack libs that numpy/scipy compile against by default. But I think numpy/scipy compile lapack functions statically; so they should be optimized if you compile them from source.

In any case, I worked on LAPACK version 3.3.0 and had to modify the following make related files:
(1) {root}/make.inc
(2) {root}/SRC/Makefile
(3) {root}/BLAS/SRC/Makefile
(4) {root}/Makefile

Where {root} is your extraction path.

(1) For {root}/make.inc:
At minimum, add "-fPIC" to the configuration variables listed below.

FORTRAN
OPTS
LOADER

I also appended "-02 -pipe" for optimization and speedier compilation on to FORTRAN and LOADER

For ease of installation, I added a new varaible, PREFIX, which designates the installation path:

#INSTALLATION PATH
PREFIX=/home/group/shared_libraries/centos64/pkgs/lapack/3.3.0/

(2) For {root}/SRC/Makefile:
Update the main target:

all: ../$(LAPACKLIB) liblapack.so

Add the following build target and rule

liblapack.so: $(ALLOBJ)
gfortran -shared -Wl,-soname,$@ -o $@ $(ALLOBJ)

And you'll need to remove some redundancy from the object files list:

#DSLASRC = spotrs.o sgetrs.o spotrf.o sgetrf.o
DSLASRC = spotrs.o sgetrs.o sgetrf.o

#ZCLASRC = cpotrs.o cgetrs.o cpotrf.o cgetrf.o
ZCLASRC =

(3) For {root}/BLAS/SRC/Makefile:
Update the main target to read as follows

all: $(BLASLIB) libblas.so

Of course, you can create an alias in the {root}/make.inc for BLASLIBSO = libblas.so

Now add the new target:

libblas.so: $(ALLOBJ)
gfortran -shared -Wl,-soname,$@ -o $@ $(ALLOBJ)

(4) For {root}/Makefile
Added an install rule which depends on the PREFIX setting from make.inc:

install: all
cp BLAS/SRC/libblas.so $(PREFIX)/lib
cp SRC/liblapack.so $(PREFIX)/lib
cp *.a $(PREFIX)/lib

That's it!

Installing topicmodels R package

2010-12-17T17:26:00.000-08:00

It has been quite annoying trying to install the "topicmodels" package for R. But here's a run down, in case others out there encounter the following error message when installing the package into a non-standard location.

ctm.c:29:25: error: gsl/gsl_rng.h: No such file or directory
ctm.c:30:28: error: gsl/gsl_vector.h: No such file or directory
ctm.c:31:28: error: gsl/gsl_matrix.h: No such file or directory

First, you will need to install the GNU GSL library. Pick it up from here: ftp://ftp.gnu.org/gnu/gsl/. The typical yum or manual compilation should work just fine. If you're installing this library into a non-standard location, take note of the installation path because you will need that below.

Second, even when passing the "--configure-vars" option to point to the location of the GSL include and shared library folder, R CMD INSTALL will fail. The solution?

Here:
1) Download the topicmodel source from here: http://cran.r-project.org/web/packages/topicmodels/index.html
2) Unpack into a working folder.
3) Modify the src/Makevars file to read as follows:

LIB_GSL=/path/to/gsl/installation/
PKG_LIBS=-lgsl -lgslcblas -L${LIB_GSL}/lib
PKG_CPPFLAGS=-I$(LIB_GSL)/include

Of course, modify the value for LIB_GSL according to your installation path. Now re-compress the folder structure like so:
"tar zcvf topicmodels.tar.gz /path/to/working/folder"

Now run "R CMD INSTALL topicmodels.tar.gz ".

Note that you will also encounter an issue with the vignette generation, so you'll need to install the OAIHarvester package in an R session:

install.packages("OAIHarvester")

Goodluck!
Paul

Southern California Volvo dealerships

2009-09-17T16:44:00.001-07:00

So the past few days have been a bit of a commuter's nightmare. Well, at least, for someone who's lost his car keys in the city of angels.

I drive a Volvo. I love 'em. I've owned two and I've been happy with their safety and reliability. The downside of owning one of these is the cost of maintenance and repair.

I was calling around local dealerships for a replacement set of keys. One word of advice, the downtown Los Angeles Volvo dealership rips people off! The guy on the phone sounded brash and unprofessional too. I saved almost $200 dollars towing my car to the Rusnak dealership in Pasadena.

I have no particular ties (advertising, cronyism, etc) to either of these dealerships. In fact, the best car service I've received has been from the Santa Monica and the Power Volvo Irvine dealerships.

Custom PyCUDA installation on Fedora 10

2009-07-11T06:11:00.000-07:00

Below is an email response I posted in the pycuda mailing list. I was running into conflicting versions of the boost library. I had to remove the default version of boost (1.33?) which is part of the fedora 10 distribution.

====

For anyone else running into this issue (and if you're using Fedora/redhat distribution), here's the fix. This assumes that you've compiled and installed CUDA 2.2 and that you've set CUDA_ROOT to your cuda installation path (including updating your LD_LIBRARY_PATH, etc).

i) make sure you have 'chrpath' installed along with rpm tools:
yum install chrpath
yum install rpm-build

1) download the latest (at this time, 1.39) src rpm and spec file
wget http://people.redhat.com/bkoz/boost/boost-1.39.0-0.3.fc11.src.rpm
wget http://people.redhat.com/bkoz/boost/boost.spec-1.39.0-0.2

2) rebuild the rpm
rpmbuild --rebuild boost-1.39.0-0.3.fc11.src.rpm

3) install the rpms
go into rpmbuild/RPMS/{your platform}
rpm -Uvh boost-*

4) obtain pycuda, unpack and run configure
python configure.py

5) edit siteconf.py and remove the compiler version, eg, gcc43, from the following variables

BOOST_PYTHON_LIBNAME = ['boost_python-mt']

BOOST_THREAD_LIBNAME = ['boost_thread-mt']

6) issue make and make install and ensure that the examples run

Goodluck,
Paul

Exploring Stockholm, Sweden virtually

2009-07-10T17:05:00.001-07:00

A colleague recommended this very nice site to explore Sweden virtually using a google-esque street view interface. I highly recommend visiting Stockholm!

A great page to discover and reminisce about Stockholm

http://www.hitta.se/gatubild/

virtual linux window manager: screen

2009-05-27T09:29:00.001-07:00

I've been loving using gnu screen for a while now. It's been really helpful and provides an interactive 'nohup'!

Here is my ".screenrc" courtesy of a friend at UCI's ICS helpdesk.

# Turn off splash screen

startup_message off

# Will always put the list of screen sessions at the bottom

# Green text, time, and date; windows in red:

hardstatus alwayslastline "%{=b}%{G} Screen(s): %{R}%w %=%{kG}%C%A %D,%M/%d/%Y "

# please turn off the bell

vbell off

# Change default scrollback value for new windows

defscrollback 200000 # default: 10

A couple of useful tips for pretty-fying your sessions and windows:

To create a new window: C-a c

To change the alias of a running window instead of "bash" or your respective shell: C-a A

To change the title of a running session: C-a :sessionname

To kill a window (you will be prompted): C-a k

CUDA using BFG NVidia GTX 260

2009-05-14T16:14:00.000-07:00

The past couple of days has been quite frustrating. Initially, I had hoped to dual-boot with my existing Mac Pro (version 3,1; early 2008), but there is no native OS X driver support for the GTX 260 (or any of the latest iteration of the GeForce line). Because the development for my project only lasts until the beginning of June, I figured I could sacrifice not booting into OS X.

I never fully appreciated the flexibility of PCs and I've never fully admired the 'tightness' of Apple hardware. By tightness, I mean how the design of this aluminum beast prevents any sort of customization. I hate it.

In any case, I just decided to use a Dell Vostro which has a single x16 PCIe lane (whew). The form factor was cramped to say the least. I had to make several trips to find 90-degree SATA cables which were still too stubby. The box came with a fairly thin 90-degree cable for the HDD which meant I had to do away with the cable for the optical drive. Luckily, my fumbling from the previous day with the MacPro afforded me with a HDD which had Fedora 10 installed. I had to shuffle around the HDD of course along with the cables. What a nightmare.

My final setup now consisted of a GTX 260 tightly tucked in the belly of the tiniest supercomputing beast: Dell Vostro, 2GB of RAM, Core 2 Duo 2.8GHz, 1 TB disk space.

It's an awesome machine pushing 370-380 GFlops with the CUDA n-body simulation. This much computing for under $300? Well, minus the stress and other MacPro setup (4GB kit and 1TB separate HD) for a total of ~$500.

Right now I'm using CUDA, PyCUDA and should be helping out with creating wrappers for the latest OpenCL standard.

Here are very useful links:

CUDA

PyCUDA (developed by Andreas Klockner)

Welcome the new age of commodity supercomputing.

Apple Bootcamp and Fedora (Part 1: Hardware setup)

2009-05-11T18:48:00.000-07:00

I have a class project in which I'll be implementing a couple of machine learning algorithms using GPGPUs as an exercise in parallel programming and some software engineering. I'll be using CUDA, and to a simplify the glue code, I'll be taking advantage of PyCUDA.

My development hardware is a Mac Pro (release 3.1, late 2008), but I'll need 1) a separate hard-disk because my primary disk is full of projects that I wish to protect, and 2) an upgrade to the default ATI video card with a GeForce 8800 GT. Too bad the latest GTX 285 is not supported by OS X until the end of June! I'll just have to settle for 512MB of DDR3.

I'll basically be following the instructions from this site: http://sowerbutts.com/linux-mac-mini/

Right now I'm in the process of activating Bootcamp to dual boot with Fedora. One necessary piece of software is rEFI (http://refit.sourceforge.net/) which will enable boot selection. I've already downloaded Fedora (http://fedoraproject.org/), my favorite distribution ;-)

[45 mins later]

The bootcamp instructions are a bust, or at least, totally unnecessary for dual-booting with linux. In fact, installing rEFI and having it boot the install CD was sufficient. However, I had a little scare after ejecting the installation CD and performing the first reboot. I received the following error message, "Missing operating system," in plain white text on an ominous black screen. Apparently, rEFI defaults to the previous boot device, which was the CD-ROM in my case.

After pressing the apple (command) button, the rEFI menu finally popped up and the pretty boot device selection finally appeared. Whew.

Tomorrow, I'll be purchasing the GeForce 8800 GT card from an apple reseller. I should have a linux CUDA development environment by early afternoon. I will post screenshots, etc.

Working with pygr

2009-05-04T17:16:00.001-07:00

Okay, I'm receiving this weird error with my script...

# Create annotation database
Traceback (most recent call last):
  File "scripts/sql_build_annotations_sgdfeatures.py", line 119, in
   main()
  File "scripts/sql_build_annotations_sgdfeatures.py", line 73, in main
   stop="Stop_coordinate"))
  File "/home/prigor/projects/genomics/hts/lib/python2.5/site-packages/pygr/annotation.py", line 137, in __init__
   self.get_annot_obj(k, slice)
  File "/home/prigor/projects/genomics/hts/lib/python2.5/site-packages/pygr/annotation.py", line 184, in get_annot_obj
   %(k,start,stop))
IndexError: annotation S000007553 has zero or negative length [-683625:-684113]!

VirtualBox

2009-04-07T18:05:00.000-07:00

I've been using VMWare's Fusion for quiet some time now (even from the early betas), but for the past couple of weeks VirtualBox has replaced Fusion. Sun's VB is free and provides similar functionality and, for my purposes, similar performance.

Check both products out though.

Google's april fool's joke

2009-04-01T01:18:00.000-07:00

Pretty elaborate, though not as good as the snail mail copy of your gmail....

http://mail.google.com/mail/help/autopilot/index.html

Turbogears 2 (release candidate)

2009-03-31T12:20:00.000-07:00

The default (quickstart) TG2 projects are funcational except when you enable authentication. For some reason (still waiting on a thread reply), the default admin/manager user tables are not populated. Or, at least, there is no documentation on the schema in case you decide *not* to use sqlite.

The error below pops up when attempting to use Catwalk (/admin) when serving both development and test configs.

OperationalError: (OperationalError) no such table: tg_user u'SELECT tg_user.password AS tg_user_password, tg_user.user_id AS tg_user_user_id, tg_user.user_name AS tg_user_user_name, tg_user.email_address AS tg_user_email_address, tg_user.display_name AS tg_user_display_name, tg_user.created AS tg_user_created \nFROM tg_user \nWHERE tg_user.user_name = ? \n LIMIT 2 OFFSET 0' ['manager']

I eagerly await developer replies.