Hi
I'm trying to use pyDardshan on my laptop to analyze logs from Polaris. I've successfully built and installed darshan-util and the pydarshan from a cloned repo.
./prepare.sh
cd darshan-util/
./configure --prefix ~/.local --enable-pydarshan
make
make install
Followed by installing the module into my virtual environment.
cd pydarshon
pip install .
When I import darshan and then try to run "report = darshan.DarshanReport(filename, read_all=True)" where filename is one of my Polaris darshan logs I get an error message about APMPI (one of the modules used to generate the darshan log).
File ~/projects/presto/venv/lib/python3.10/site-packages/darshan/report.py:753, in DarshanReport.mod_read_all_apmpi_records(self, mod, dtype, warnings)
749 self.counters[mod] = {}
751 # fetch records
752 # fetch header record
--> 753 rec = backend.log_get_apmpi_record(self.log, mod, "HEADER", dtype=dtype)
754 while rec != None:
755 self.records[mod].append(rec)
AttributeError: module 'darshan.backend.cffi_backend' has no attribute 'log_get_apmpi_record'
I assume this is because the module wasn't included during the darshan-util build.
I'm trying to build this module into darshan-util and have added the autoperf submodule to the darshan repo. When I add --enable-apmpi-mod to the configure step above i get an error that it can't find the header.
checking for GNU default python prefix... ${prefix}
checking for GNU default python exec_prefix... ${exec_prefix}
checking for python script directory (pythondir)... ${PYTHON_PREFIX}/lib/python3.10/site-packages
checking for python extension module directory (pyexecdir)... ${PYTHON_EXEC_PREFIX}/lib/python3.10/site-packages
checking for /home/jpr/projects/presto/darshan/darshan-util/../modules/autoperf/apmpi/darshan-apmpi-log-format.h... no
configure: error: The autoperf MPI module is not present
Looking at the config.log, it seems the test is just trying to confirm the include of the header using the same path, which does exist on my laptop.
I'm a bit stuck on how to proceed. Any hints appreciated.
Thanks.
