Januari 8, 2001
A few small improvements in Crunchmake.
A revised version of Crunch may be downloaded from the usual site as of
now. Installation creates a directory crunch_1.1, your old and trusted
version will therefore not be destroyed. The minimum required RAM is now
Various improvements have been made. The most important:
The next version will have one very important additional feature: I'm
currently developing and testing a set of routines aimed at determining
'protein' substructures such as Se, S, Cl and P from their anomalous
signal. Determination of heavy atom positions from SIR data will be
supported as well. I think I'll name that version Crunch_2.0.
- An alternative E-value calculation has been implemented. Running
'crunch code blessing' results in E's calculated by the programs Levy
and Eval of the DREAR program suite. This can be very useful
especially if the quality of your data is not all you'd like it to
be. If you do this all subsequent calculations will use the
alternative E's. To use the Normal80 E's again clean up the directory
by typing 'crunch code clear'. See the manual provided for details.
- Crunch determines the resolution of your data and takes this into
account at all stages of the process of structure determination.
- Sometimes structures contain a number of atoms which have
significantly higher B-values than the rest. Autofour used to have
problems with this. The new version automatically switches to
difference maps to locate these atoms, assigning them weaker
scattering and testing each additional atom against R2. This does use
more CPU time. However, the models obtained are much improved.
- The determination of atomic numbers has been stabilized. Note that,
if Autofour cannot find all atoms you've specified, no attempt is
made to discriminate between atoms which are taken to possess the same
scattering power during the extension of the initial model (for
instance C and N).
- There is no longer a separate Linux version. The installation script
determines the type of machine it is running on. Supported and tested
are: Linux on Pentium and Alpha, True64 on Alpha, SGI IRIX and IBM
AIX. Deter now uses Lapack and Blas routines for the matrix and
vector operations. The script Crunchmake supports the possibility of
using locally available Lapack and Blas libraries. Details are in the
manual or you might want to consult the comments in Crunchmake.
- Some large structures caused Deter to crash on systematic overflow.
This has been taken care of by carrying out some critical
calculations in double precision. The way of enforcing positivity of
the matrices in the initial stages of the phase refinement has been
changed as well.
There is no Windows version yet. However, pressure has been mounting. In
my occasional moments of weakness I'm actually considering building
such a thing. However, don't hold your collective breaths.
April 20, 2000:
The routine avrg.f, present in the subdirectory source of the system,
contains an error. If you've downloaded Crunch before April 21, 2000,
please download the corrected routine. Replace the version in the source
directory, compile and link it and replace the excutable in the programs
Before the summer of 2000 version 1.1 will be posted on the CRUNCH site.
Februari 5, 1999: Another update. Some modern compilers do not accept
the use of one memory location for different types of variable. I have
effected a separation between integer, floating point and double
precision variables while keeping the dynamic allocation of memory
As 64bit-word machines are used more and more, I have changed the code
to take care of alignment problems as well.
November 27, 1998: An updated version of Crunch is now available.
The main difference is the improved handling of heavy atom structures.
A few minor bugs have been removed. Linux support had been allowed to
laps a bit. I've now updated the Linux version to the same level as the
general one. Compiler support now includes g77, f2c and cc linked with
f77 - the old way of doing fortran under Linux - and fort77, Red Hat's
pseudo fortran compiler.
Januari 16, 1998: Apart from the usual sources of X-rays, such as Cu,
Mo, etc, in the crysinfile Sy may be specified now. The user is prompted
to specify the wavelength used in the experiment. This enables the crunching
of synchroton datasets.
NB: Some values in the resulting crysdafile are zeroed in this case
such as the linear absorption coefficient and the anomalous dispersion
August 20, 1997: A small error in the script has been repaired. The
effect of the change is largely cosmetic.
June 25, 1997: I've removed a bug from the phase-refinement. It's
surprising how robust this refinement is. Nevertheless, it should work
even better now.
June 1997: Version 1.0 is available now. Changes from 0.9:
While Dirdif is no longer essential for using Crunch, I consider
the programs to be complementary. For compounds containing (many)
heavy atoms Dirdif is usually more powerful, while for organic
molecules, possibly containing a few second row elements, in general
Crunch is more effective. This is the reason why Crunch uses the same
data files as Dirdif does. This is still true in version 1.0
- Crunch no longer needs Dirdif installed on the same machine.
- The scripts have been generalized, there is now one version which
runs on most UNIX systems. I'm still
working on the HP version. Linux is still supported specifically, while
on other platforms such as Alpha, SUN, IBM/AIX, SGI/IRIX the general
version must be used.
- Various larger and smaller bugs have been fixed:
- All elements in the periodic table, excepting the noble gasses,
may be specified now.
- There are small but significant changes in the matrix-construction
- The conhkl-program has been adapted to even more curious inputfiles.
Installing the new version does not overwrite the old one. Users might
consider keeping the old version for some time.