2,652
edits
Docandreas (talk | contribs) No edit summary |
|||
(16 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
The possibility of using external libraries (that are loaded at runtime) has been available in C/C++ | The possibility of using external libraries (that are loaded at runtime) has been available in C/C++ for a long time, but in Fortran became only available as of Fortran2003. | ||
In the case of XDS, frame-reading and computation can be separated starting with version November-2016. This allows users/companies to develop their own specialized frame-reading libraries, and relieves the XDS maintainers from implementing even more file formats. | In the case of XDS, frame-reading and computation can be separated starting with version November-2016. This allows users/companies to develop their own specialized frame-reading libraries, and relieves the XDS maintainers from implementing even more file formats. The feature was developed in order to be able to natively (i.e. without temporary intermediates) read the HDF5 files written for data from the [[Eiger]] detector. | ||
In the following, small examples are given for | In the following, small examples are given for | ||
* how a program ("'''host'''") may use an existing external library, e.g. the dectris-neggia library ([https://github.com/dectris/neggia source]; [https://www.dectris.com/ | * how a program ("'''host'''") may use an existing external library, e.g. the dectris-neggia library ([https://github.com/dectris/neggia source]; [https://www.dectris.com/company/news/newsroom/news-details/process-eiger-data-with-xds-fast pre-compiled]) | ||
* how an external library ("'''client'''") may be implemented that XDS can use | * how an external library ("'''client'''") may be implemented that XDS can use | ||
The glue code between host and client is based on http://cims.nyu.edu/~donev/Fortran/DLL/DLL.Forum.txt . There should be no need to change this, unless the interface design changes. | The [[LIB#Glue_code|glue code]] between host and client is based on http://cims.nyu.edu/~donev/Fortran/DLL/DLL.Forum.txt . There should be no need to change this, unless the interface design changes. | ||
The interface was designed by Markus Mathes (Dectris), Vittorio Boccone (Dectris) and Kay Diederichs. It is supposed to be generic, i.e. useful beyond XDS. In particular, the 4096 bytes of the info_array can be utilized to obtain and use header information (e.g. wavelength, distance, axes specifications and other metadata). | The interface was designed by Markus Mathes (Dectris), Vittorio Boccone (Dectris) and Kay Diederichs. It is supposed to be generic, i.e. useful beyond XDS. In particular, the 4096 bytes of the info_array can be utilized to obtain and use header information (e.g. wavelength, distance, axes specifications and other metadata). | ||
The interface has several [[LIB#Existing_implementations|implementations]]. | |||
== Host code example == | == Host code example == | ||
Line 20: | Line 22: | ||
! ifort -qopenmp generic_data_plugin.f90 test_generic_host.f90 -o test_generic_host | ! ifort -qopenmp generic_data_plugin.f90 test_generic_host.f90 -o test_generic_host | ||
! or | ! or | ||
! gfortran -O -fopenmp -ldl generic_data_plugin.f90 | ! gfortran -O -fopenmp -ldl generic_data_plugin.f90 test_generic_host.f90 -o test_generic_host | ||
! run with | ! run with | ||
! ./test_generic_host < test.in | ! ./test_generic_host < test.in | ||
Line 52: | Line 54: | ||
! set some more module variables | ! set some more module variables | ||
firstqm=INDEX(actnam,'?') | firstqm=INDEX(actnam,'?') ! qm means question mark | ||
lastqm =INDEX(actnam,'?',BACK=.TRUE.) | lastqm =INDEX(actnam,'?',BACK=.TRUE.) | ||
len =LEN_TRIM(actnam) | len =LEN_TRIM(actnam) | ||
Line 125: | Line 127: | ||
! gfortran -c -fpic test_generic_client.f90 && ld -shared test_generic_client.o -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/ -lgfortran -o libtest_generic_client.so | ! gfortran -c -fpic test_generic_client.f90 && ld -shared test_generic_client.o -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/ -lgfortran -o libtest_generic_client.so | ||
! (attention: the above - from "gfortran" to "libtest_generic_client.so" - is one looong line) | ! (attention: the above - from "gfortran" to "libtest_generic_client.so" - is one looong line) | ||
! The resulting file can be used with a LIB=./libtest_generic_client.so line in XDS.INP, and enables | |||
! reading of data files with a 7680 bytes header plus 1024*1024 pixels of integer data, without any record structure. | |||
MODULE plugin_test_mod | MODULE plugin_test_mod | ||
CHARACTER :: fn_template*132='' | CHARACTER :: fn_template*132='',cformat*6='(i4.4)' | ||
INTEGER :: lenfn,firstqm,lastqm | INTEGER :: lenfn,firstqm,lastqm | ||
END MODULE | END MODULE | ||
Line 150: | Line 154: | ||
firstqm=INDEX(fn_template,'?') | firstqm=INDEX(fn_template,'?') | ||
lastqm =INDEX(fn_template,'?',BACK=.TRUE.) | lastqm =INDEX(fn_template,'?',BACK=.TRUE.) | ||
WRITE(cformat(3:5),'(i1,a1,i1)')lastqm-firstqm+1,'.',lastqm-firstqm+1 | |||
END SUBROUTINE plugin_open | END SUBROUTINE plugin_open | ||
! | ! | ||
Line 180: | Line 185: | ||
! local variables | ! local variables | ||
INTEGER k,i,dummy | INTEGER k,i,dummy | ||
CHARACTER :: fn*132 | CHARACTER :: fn*132 | ||
fn=fn_template | fn=fn_template | ||
IF (frame_number>0) WRITE(fn(firstqm:lastqm),cformat) frame_number | IF (frame_number>0) WRITE(fn(firstqm:lastqm),cformat) frame_number | ||
! -qopenmp compile option needs to be used otherwise race in writing fn | ! -qopenmp compile option needs to be used otherwise race in writing fn | ||
Line 222: | Line 226: | ||
! THE USE OR OTHER DEALINGS IN THE SOFTWARE. | ! THE USE OR OTHER DEALINGS IN THE SOFTWARE. | ||
! | ! | ||
! | ! For more information, please refer to <http://unlicense.org/> | ||
! | ! | ||
! | ! | ||
Line 668: | Line 672: | ||
</pre> | </pre> | ||
Google returns https://github.com/apple/cups/issues/4410 and https://bugzilla.redhat.com/show_bug.cgi?id=1065695 when searching for similar problems. Overall, this appears to be harmless and in fact I don't know how to change the code to make the segfault disappear - I'd appreciate a patch! | Google returns https://github.com/apple/cups/issues/4410 and https://bugzilla.redhat.com/show_bug.cgi?id=1065695 when searching for similar problems. Overall, this appears to be harmless and in fact I don't know how to change the code to make the segfault disappear - I'd appreciate a patch! | ||
== Existing implementations == | |||
# [https://github.com/dectris/neggia Dectris Neggia-plugin] to read HDF5 written by Dectris-supplied software of Eiger detectors | |||
# [https://github.com/DiamondLightSource/durin Diamond's Durin-plugin] to read HDF5 written by Eiger detectors at Diamond (and presumably elsewhere); latest binaries for MacOS and Linux (RHEL6) as well as example XDS.INP and source at https://github.com/DiamondLightSource/durin/releases/latest | |||
# [https://git.embl.de/nikolova/xds-zcbf/ EMBL-Hamburg's zcbf-plugin] to read gzip-compressed CBF files without intermediate file | |||
== See also == | |||
# https://rosettacode.org/wiki/Call_a_function_in_a_shared_library#GNU_Fortran_on_Linux |