Split Tests spgemm and gauss_seidel into multipl object files.

[kokkos]

Split Tests spgemm and gauss_seidel into multipl object files.

[crtrott]

SPGEMM is closing on two hours:

27473 crtrott   20   0 8655616 5.390g  58560 R 100.0  1.1 112:59.11 /home/projects/pwr8-rhel73-lsf/ibm/xl/xlC/14.1.0/exe/ipa -comp -qalias=ansi -qthreaded -qtls -qtls -maltivec -qtls -qlanglvl=extended0x -qarch=pwr8 -qtune=pwr8 -qsmp=omp /tmp/xlcW0wbmEmP /tmp/xlcW13OIGeu Test_OpenMP_Sparse_spgemm.o /tmp/xlcLj2CORsT.lst /tmp/xlcW25m6I68 

gauss_seidel finished at some point after 85 mins.

[mndevec]

These are compilation times, not test times right?

What are the data types that are enabled?

[crtrott]

Just the default double , complex

[crtrott]

It finished!! Took almost 5 hours.

[mhoemmen]

this makes baby Turing unhappy :(

[mndevec]

I finished the eti of spgemm and gauss seidel.
Would this separation be still required? I mean, would this test split speedup the linking of ETI'd kernels?

[mndevec]

I am closing this issue, as it looks like this does not have much effect once the kernels are ETI'ed.

I separated the tests and compiled it for double and complex double. The overall compilation took 8 minutes.
Without the split it was 7m 57 seconds. There can be noises for timings but apparently this does not have much affect on the compilation time.

[srajama1]

Are you saying XL compilation time went from 5 hrs to 8 mins before splitting SGS and SpGEMM tests ?

[mndevec]

No, 8 min is on my workstation.

Basically, XL compiler issue is solved with ETI (5 hours to 15 min). I was experimenting whether splitting test files would further improve the compile time. I was trying this on my workstation with gcc compiler. Basically, once we the eti is done splitting do not have further improvement.