Readme

Installing LABELIT

It is assumed that the user has downloaded a LABELIT bundle. This document explains the CCP4 requirements. Note: LABELIT bundles come in two varieties, either with a prepackaged version of Python (for all users) or without (for developers only).

OpenMP

As of May 2009 LABELIT uses multiple processing cores for the DPS autoindexing step. This is primarily intended to speed up processing at synchrotron sites, where Linux computers are routinely available with multiple cores. A 16-core machine, for example, can speed up autoindexing by a factor of about 10 (The gain is not a full 16-fold because some time is spent dividing up the problem and marshaling the results).
On the download page, binary bundles labeled "OpenMP" are compiled with the multiprocessing option.
It may be necessary to set an environment variable defining the maximum number of threads that the LABELIT process is permitted to use. For a two-processor machine:
```
setenv OMP_NUM_THREADS 2
```
A script can then be run to confirm the use of both threads:
```
> labelit.python labelit_sources/labelit/excercise_omp.py
The parallel region is executed by thread 1
The parallel region is executed by thread 0
e=  2.718282  pi=  3.141568
OK
```
Newer operating systems support the OpenMP specification. On older RedHat systems, it may be necessary to install the libgomp rpm package.

Enabling OpenMP from a LABELIT source build:

Change to a new, empty directory for the build. Install Python with the command:
```
[path to source tree]/labelit/bundle/install_python_here.csh
```

Configure with the command:

base/bin/python [path to source tree]/cctbx_sources/libtbx/configure.py --enable-openmp-if-possible=True labelit
source setpaths.csh
make

Enabling OpenMP from a PHENIX source build:

Download the source installer for the latest developmental version of Phenix, from phenix-online.org. Untar the bundle and go in to the directory.
```
./install --help  #gives all the options

./install --prefix=[] --nproc=[] --openmp  #asks for compile with open MP
```
...fails with warning message about incompatibility with GUI.
```
./install --prefix=[] --nproc=[] --openmp  --no-gui
```
...now it should work in full. Will not be possible to run the Phenix GUI, but everything else OK.
To test the ability to run multiple threads:
```
labelit.python $PHENIX/labelit/excercise_omp.py
```
...will enumerate all the threads being used.

CCP4 and MOSFLM

CCP4 6.xx is supported (5.0.2 is also expected to work).

MOSFLM. Versions 6.2.4 and above will work. The 7.0.7 executable (recommended) may be downloaded from http://www.mrc-lmb.cam.ac.uk/harry/mosflm.

The CCP4 setup file must be sourced, and the command "ipmosflm" aliased to the MOSFLM executable must be in the user's PATH.

For data from very large unit cells, MOSFLM versions prior to 7xx must be recompiled with larger array dimensions. Here is an easily run script to accomplish this. This script also redimensions the program for the unbinned Quantum 315 as noted in the above paragraph. It has been tested under Red Hat 8.0 Linux/ gcc 3.2, but could be modified for other platforms.

Python (for developers only; users read on to see why you shouldn't install python)

For most circumstances it is recommended that the user download a LABELIT bundle containing a pre-built Python executable; in which case this entire section can be skipped. This includes Power Users and System Administrators! Explanation to Power Users: Installing LABELIT with its own self-contained pre-built Python will not affect your shell environment in any way. Specifically, it will not affect the ability to run the system's pre-existing Python version for other applications. If despite this the user wishes to supply a standalone Python, the following steps must be taken:
Python must be installed and in the user's unix PATH. Python 2.6 is recommended.
Third party Python modules:

Numeric-24.2. Download the source tar.gz file & unzip. (Older versions such as 23.1 and 23.3 are also expected to work.)

On the Tru64 platform, the source code must be edited. The following line must be inserted at the very top of the files Src/multiarraymodule.c and Src/ranlib.c:
```
#include "Python.h"
```
Then on all platforms:
```
cd Numeric-24.2
python setup.py install
```

Zlib 1.2.2 (only required on Tru64) needed for compiling the Image Library in the next step:

wget http://www.zlib.net/zlib-1.2.2.tar.gz
gunzip -c zlib-1.2.2.tar.gz | tar xf -
cd zlib-1.2.2
make clean
#previous PIL versions use shared z libary: ./configure -s
./configure
make

Python Image Library Version 1.1.6 (Dec. 3, 2006) is currently supported.

wget http://effbot.org/downloads/Imaging-1.1.6.tar.gz
gunzip -c Imaging-1.1.6.tar.gz | tar xf -
cd Imaging-1.1.6

The following line must be inserted at the very top of the file path.c:
```
#include "pyconfig.h"
```
Edit the file setup.py.
Change the _tkinter import to "import disable_tkinter"
Change the line 'ZLIB_ROOT = None' to 'ZLIB_ROOT = "../zlib-1.2.2"'
On the IRIX platform it may be necessary to disable the linkage to the preinstalled jpeg library. This can be done by deleting the four lines in setup.py beginning with 'if find_library_file(self, "jpeg"):'

Modify the file libImaging/ZipEncode.c so that "Z_DEFAULT_COMPRESSION" is changed to "Z_BEST_SPEED". This will speed up png compression at the price of slightly increased output file size.

python setup.py install

Reportlab 1.17. Newer versions may also work. Download the source tar.gz file & unzip.

The following edit must be performed to make ReportLab 1.17 compatible with Python 2.4 and above (current ReportLab versions have already been fixed). In the files "reportlab/graphics/charts/axes.py" and "reportlab/platypus/doctemplate.py", change instances of "None =" to "unused =".

On the Tru64 & Irix platforms, the source code file lib/_rl_accel.c must be edited. Instead of raising the error "Don't know how to define STRICMP" in line 22, define it as follows:
```
# define STRICMP strcasecmp
```
Important: If you use the most recent ReportLab release (v1.20) you will have to download and install _rl_accel as a separate package.

Then on all platforms:
```
cd reportlab
python setup.py install
```