SciLifeLab / facs

Fast and Accurate Classification of Sequences using Bloom filters
http://facs.scilifelab.se/
Other
16 stars 9 forks source link

is 'mpi_bloom.h' mandatory and included? #129

Open dbrami opened 10 years ago

dbrami commented 10 years ago

I am trying to compile for multi-cpu support using openmpi. Here is my command and the result:

dbrami@asm05.c01:/sgi/asmopt/src/facs/facs
cmd-> make mpi
Make sure you have MPI support on your cluster hint: module load openmpi
#mpicc -c *.c -O3 -Wall -g -DNODEBUG -fopenmp -D_FILE_OFFSET_BITS=64 -D_LARGE_FILE -lm -lz
#mpicc -c mpi_decon.c -O3 -D_FILE_OFFSET_BITS=64 -D_LARGE_FILE 
#mpicc -o mpi_decon mpi_decon.o bloom.o suggestions.o lookup3.o  -lm -O3 -Wall -g -DNODEBUG -fopenmp -D_FILE_OFFSET_BITS=64 -D_LARGE_FILE
mpicc -c mpi_bloom.c -O3 -Wall -g -DNODEBUG -fopenmp -D_FILE_OFFSET_BITS=64 -D_LARGE_FILE
mpi_bloom.c:25:23: error: mpi_bloom.h: No such file or directory
mpi_bloom.c:49: warning: return type defaults to ‘int’
mpi_bloom.c: In function ‘main’:
mpi_bloom.c:138: warning: implicit declaration of function ‘struc_init’
mpi_bloom.c:139: warning: implicit declaration of function ‘make_list’
mpi_bloom.c:139: warning: initialization makes pointer from integer without a cast
mpi_bloom.c:153: warning: implicit declaration of function ‘gz_mpi’
mpi_bloom.c:166: warning: ignoring #pragma omp task
mpi_bloom.c:183: warning: implicit declaration of function ‘gather’
mpi_bloom.c:186: error: too few arguments to function ‘report’
mpi_bloom.c:58: warning: unused variable ‘exit_sign’
mpi_bloom.c: At top level:
mpi_bloom.c:194: error: conflicting types for ‘struc_init’
mpi_bloom.c:138: error: previous implicit declaration of ‘struc_init’ was here
mpi_bloom.c:251: error: conflicting types for ‘gz_mpi’
mpi_bloom.c:153: error: previous implicit declaration of ‘gz_mpi’ was here
make: *** [mpi] Error 1

It seems to be complaining about a missing file 'mpi_bloom.h' that is not included in source nor in the repository. Any ideas?

tzcoolman commented 10 years ago

@brainstorm Very strange. I ll take care of that. My local version doesn't have mpi_bloom.h

dbrami commented 10 years ago

Any progress on this please? Otherwise the size of the data that needs to be filtered makes single CPU FACS unusable.

tzcoolman commented 10 years ago

@brainstorm i'm on it now...

guillermo-carrasco commented 10 years ago

hi @dbrami , could you please pull the latest master branch and try to compile again?

Thanks,

dbrami commented 9 years ago

I compiled again and this time successfully. But running mpi_facs does not recognize the 'n' or 'np' parameter flag for number of processors; Also, the mpi_facs does not seem to jive with the usage statement or the documentation of the usage for vanilla facs.

/sgi/asmopt/src/facs/facs/facs_mpi -n 64 -r ecoli.bloom -q 3xtest.fasta
/sgi/asmopt/src/facs/facs/facs_mpi: invalid option -- n
Unknown option: -n

I also went through the mpi_bloom.c file and looked at line 79-111 and did not find a param for number of processors

tzcoolman commented 9 years ago

Hi @dbrami, if you want to run facs_mpi, do it like this: mpirun -n 4 /sgi/asmopt/src/facs/facs/facs_mpi/facs_mpi -r ecoli.bloom -q 3xtest.fasta This is the standard procedure of running mpi and make sure you have either openmpi or mpich on you machine.