custom age makefile fails with g++

[joelmartin]

the Makefile for the custom version of age_align has -O3 option added, on my debian systems at least this pretty well breaks age_align. it hangs most of the time, even at -O it hangs on the test.sh script 90%+ of the time.

tried compiling with gcc 4.6.3 and 4.9.2, nearly the same results.

removing the -O option has it working

[marghoob]

Hi Joel,

Thanks for your interest in MetaSV. Do you get the same problems when compiling from the main AGE repository (https://github.com/abyzovlab/AGE)? Can you email me the logs or sample test input which indicate the failure/breaking of age_align?

On Mon, Jun 15, 2015 at 3:00 PM, Joel Martin [email protected] wrote:

the Makefile for the custom version of age_align has -O3 option added, on my debian systems at least this pretty well breaks age_align. it hangs most of the time, even at -O it hangs on the test.sh script 90%+ of the time.

tried compiling with gcc 4.6.3 and 4.9.2, nearly the same results.

removing the -O option has it working

— Reply to this email directly or view it on GitHub https://github.com/bioinform/metasv/issues/44.

http://www.bina.com

*Marghoob Mohiyuddin, PhD *// Sr. Bioinformatics Scientist, Bina Technologies, part of Roche Sequencing E. [email protected]

Learn More // video http://www.bina.com/product.html // news http://www.bina.com/press-coverage.html // posters http://f1000.com/posters/search?query=bina+technologies // science http://www.bina.com/science.html // platform http://www.bina.com/platform.html *Test Drive Bina // Experience NextGen analysis for yourself * http://www.bina.com/test-drive.html

[joelmartin]

I have the problem with either age repository when -O3 is added ( or -O or -O2 ) make clean && make ./test.sh

occasionally it will make it through one or 2 tests, mostly it hangs. make clean && make rm -f obj/AGEaligner.o obj/age_align.o

Compiling with parallel support.

g++ -fopenmp -DOMP -DAGE_VERSION="v0.4" -DAGE_TIME -O3 -c AGEaligner.cpp -o obj/AGEaligner.o g++ -fopenmp -DOMP -DAGE_VERSION="v0.4" -DAGE_TIME -O3 -c age_align.cpp -o obj/age_align.o g++ -fopenmp -DOMP -DAGE_VERSION="v0.4" -DAGE_TIME -O3 -o age_align obj/AGEaligner.o obj/age_align.o j_martin@genepool12:~/gscratch_jm/brachyinsanity/metasv/AGE$ ./test.sh

• Test #1 For alignment with large deletion and micro insersion. Mismatches and small deletions are present in each flanking sequence. Run as: ./age_align test1_*.fa [hangs here]

It possibly indicates an error in AGE but I haven't worked through all the -Wall reports.

[marghoob]

I did a bit of testing with the compiler flags. It seems that compiling without OpenMP supports works fine with -O3 option. Could you try make OMP=no and run the tests again?

[joelmartin]

Yep that works.

wall was giving AGEaligner.cpp:911:0: warning: ignoring #pragma omp shared [-Wunknown-pragmas] AGEaligner.cpp:1027:0: warning: ignoring #pragma omp shared [-Wunknown-pragmas]

[marghoob]

The OpenMP pragmas are fine. However, there may be a concurrency issue which pops up when the -O flags are used-this will require more investigation. However, for the time being, you have the option of either disabling OpenMP or not using -O flags.

[sbandara]

Since we don't require the customized AGE aligner anymore, can we close this issue here and advise reopening on AGE?