README AFTER YOU READ the README file!!!
This distribution of MOSFLM v6.0 is slightly different from previous editions,
as it includes the build for the Rossmann Fourier Auto-indexing routines, which
are written mainly in C. You should be aware that there are some minor
modifications to the RFA routines, so don't be tempted to install updates
blindly from the Purdue site!
The program has been built successfully on the following machines; the "build
times" refer to elapsed time while the machine has a typical load of users - in
the case of the SGIs and the PC, that's one or two people; for the Alpha, it's
nearer 40.
The Makefile in the top directory has checks for the environment variables
PWD and HOSTTYPE; the different builds are all included in files in
sub-directories according to the host type. You shouldn't need to touch the
Makefiles at all, I hope (problems to the usual address...).
Platform O/S build time (min)
Alpha 275MHz, 1Gb Digital UNIX ~10
SGI 200MHz R4010/4400, 64Mb Irix 5.3 (-Olimit unset) ~8
SGI 175MHz R10000/R10010, 192Mb Irix 6.2 (-Olimit unset) ~4
" " " " " " " (-Olimit 3400) ~18
PC 233MHz pentium, 128Mb Red Hat Linux 5.1 ~7
The build will involve a lot of paging if your machine has limited RAM,
as will running the program.
The executable is ~3.0 - 3.6Mb, and integrating a single 8Mb image uses
up to around 16Mb, so I'd guess that 32Mb RAM is about the minimum you'd want
to use.
LINUX only:
-
The XDL_VIEW libraries from the CCP4 distribution are slightly
broken, so these should be re-made first from the source files included
here in the tar file XDL_VIEW.tar; it assumes that you have already
set up the CCP4 environment variables according to the normal
$CCP4/include/ccp4.setup route. Install the file libxdl_view.a in your
normal $CLIB/ (CCP4 library directory), i.e. the same place as libccp4.a
-
I've assumed that the X11 libraries are in /usr/X11R6/lib. If yours are in
a different place, you'll have to modify the file
mosflm/Makefile.i386-linux to suit.
SGI only:
-
Optimization in mosflm/Makefile.iris4d -Olimit 11400 might increase
performance, but compilation takes forever. The line's there if you want to
uncomment it, but my guess is that you'll lose more in the compilation than
you'll ever gain in performance.
SGI Irix 6.4 and Irix 6.5 only:
-
If you are building on an SGI using either of these versions of Irix,
you will need to do the following two steps prior to building
MOSFLM.
-
untar the file makefiles.tar
after you have extracted all the other files in the distribution; this
will over-write some of the distribution Makefiles.
- Set the environment variable HOSTTYPE to iris4d6.4 (not
iris4d6.5).
Under tcsh or csh, you would use the command
% setenv HOSTTYPE iris4d6.4
Under bash or ksh, you type
%bash HOSTTYPE=iris4d6.4
%bash export HOSTTYPE
There are a few warnings generated during compilation, but these do not appear
serious.
Digital Unix 4.0 and a Powerstorm 4D60T graphics card:
For some reason connected with colour mapping, the xdl_view routines
do not work with this particular type of graphics card.
When an image is to be displayed the following error message occurs:
X Error of failed request: BadMatch (invalid parameter attributes)
Major opcode of failed request: 1 (X_CreateWindow)
Serial number of failed request: 53
Current serial number in output stream: 56
FIX
To get round this problem, replace the file "xdl_view.c" with the new
version called "xdl_view_dec.c" which is in this directory, and remake
the xdl_view library.
The existing source code will be in a directory something like:
/public/ccp4/x-windows/xdl_view/src
This simply has some code commented out compared to the standard
version, but this can lead to a degradation of the performance on some
machines and so is only recommended if you are using this type of
graphics card.
INSTALLATION
-
If you have the old-style autoindexing routines (REFIX), you need to perform
the
following step before building the program. There are two dummy routines
at the end of the file mosflm/mosflm_all_ip_inc.for which need to be removed and the
REFIX file added with (while in the mosflm subdirectory)
sed -e '/SUBROUTINE TOREFIX/,/:$/d' mosflm_all_ip_inc.for > junk
mv junk mosflm_all_ip_inc.for
cat refix_all.for >> mosflm_all_ip_inc.for
This script is included in the file add_refix.com.
-
If you haven't defined CLIB in your normal login setup, you might want to edit
the line in the top-directory Makefile
#define CLIB /public/xtal/ccp4/lib
to suit your local conditions.
-
If you have an old CCP4 distribution (pre- version 3.2??), you may not have the file pack_c.c in
either $CLIB/src/ or in $CCP4/x-windows/ipdisp/src/. If (and only if...) you don't have it, rename pack_c.C to pack_c.c in the
mosflm directory, and edit Mkfile to change the line
OBJS = mosflm_all_ip_inc.o pck.o chelp.o
to read
OBJS = mosflm_all_ip_inc.o pck.o chelp.o pack_c.o
No other changes should be necessary.
-
Some sites have not defined the PWD environment variable, which is
used in the build procedure. You can either define this to be the top directory
of the distribution before you start the build, or you can edit the top-level Makefile
and change the line
DPS = ${PWD}
to suit.
-
In the distribution's top-level directory, type
make
and it should build and install ipmosflm in its bin directory.
The makes in the mosflm and the index subdirectories should build the two
components; I haven't tested rebuilding from the other subdirectories;
my suggestion is not to
do this, as a complete rebuild, even on a PC, takes only a few minutes.
The main impetus behind this new version is the incorporation of the Rossmann/
Steller Fourier auto-indexing routine, which is much more robust than the
previous method (REFIX) and requires only a single image; however, both
methods are retained to enhance choice. The new auto-indexing is the default,
and needs an estimate of the maximum expected cell edge; a value of 200
Angstroms is supplied, which seems to work for most cells in the range
50 - 240 Angstroms.
If you are a new user of MOSFLM and would like the option of using REFIX, you
should e-mail Andrew Leslie and
ask for it.
Harry Powell, MRC-LMB, MRC Centre, Hills Road, Cambridge, U.K. CB2 2QH
harry@mrc-lmb.cam.ac.uk
Last modified: Fri Jul 9 17:09:39 MET DST 1999