FindEM issues
Added by Anand Gupta over 12 years ago
We're trying to run a template picking job on our cluster, having compiled FindEM for our Intel XServes, and running into an issue where FindEM is unable to open the MRC file that Appion feeds it. FindEM complains about not finding a machine stamp in the MRC file before a CCP4 library call fails. Any thoughts on what could be happening/what to try?
| findem.log (2.35 KB) findem.log | |||
| tmplrun3.appionsub.job (619 Bytes) tmplrun3.appionsub.job |
Replies (2)
RE: FindEM issues - Added by Anchi Cheng over 12 years ago
The presumably newer CCP4 library that reads the mrc file in your compiled Fortrain code may recognize something that we are not aware of. Could you find out what version of CCP4 library you have on the cluster?
You can also do a test run of findem in /myami/programs/findem. The two input files (07jan05b_12gr_02en_00.dwn.mrc and grostain3.dwn.mrc) and the values to feed into findem64.exe (feedtest.txt) are all there. If that does not give error, please send us one of the dwn.mrc that does fails. If the feedtest fails with the same error, you might have compiled findem in the way that it looks for CCP4 library not in findem's self-contained directory.
RE: FindEM issues - Added by Neil Voss over 12 years ago
You are asking questions about FORTRAN, of which I know very little. But I am in charge of the FindEM code. Alan Roseman gave me the copyright.
(1) I tried to build a static FindEM this morning, but there are no fortran-77 libraries on CentOS that I could find.
(2a) Different machines can write the MRC float to the file a different way (I am talking about endianness). The machine stamp says what machine it came from it. If your machine is an Intel machine (not SPARC or alpha) then it is probably okay to comment out the line that checks the machine stamp in the FORTRAN code and let it run.
(2b) It looks as though the machine stamp is missing the mrc that was created. Does pyami.mrc write a machine stamp, I remember talking to Jim about this about 3 years ago. From what I can till it does not, so maybe we need to add one.
(3) We have the same problem earlier with getting FREALIGN to read IMAGIC or MRC files, apparently our python code just did not write the machine stamp and FREALIGN rejected it.
(4) I re-wrote the entire FindEM code in python, but since the FFT is slightly different in C than FORTRAN, there produced slightly different results. Anyway, maybe it is time to bring it back to python.
(5) As Anchi pointed out there is a test in /myami/programs/findem, just type "make check" and it runs the test. Assuming you are setup to compile findem yourself.
Neil