Closed jakirkham closed 8 years ago
Hi! This is the friendly automated conda-forge-linting service.
I just wanted to let you know that I linted all conda-recipes in your PR (recipe
) and found it was in an excellent condition.
Would be curious to hear your thoughts on this, @zym1010.
While this works on OS X, it appears we may have some issues to sort out on Linux.
@jakirkham well I have no objection, as I never encountered this problem. But seems that this option is a default setting, from https://github.com/xianyi/OpenBLAS/blob/ec4390a967cb57eddff39f1b0bf6f3e45a5387a0/lapack-netlib/INSTALL/make.inc.gfortran#L20. So maybe it's good to turn it on?
Update: -frecursive
is not in the default setting. See my comment below.
That's interesting. I wonder why it helped then. 😕
Is that Makefile actually used or is it Linux only?
@jakirkham well sorry for my mistake. I think that option is not used by OpenBLAS's make
by default, as I just compiled the master branch of OpenBLAS and searched through the output, finding no frecursive
.
No worries. Thanks for looking into it. If OpenBLAS was setting -frecursive
by default, I was a bit worried that we might be overriding their default options and thus the real problem remained hidden. Good to know that is not the case.
Did you see if -O2
was being set too? If not, maybe we should be setting that as well.
@jakirkham -O2
is all over the place. See https://www.dropbox.com/s/7wr0b96nkp6iuvi/openblas_output_842d842751c98c92d1c32870b10d81bcec9bb232.txt?dl=0 for raw output.
Sure. Just wanted to make sure we weren't suppressing it by setting the other Fortran flags.
Saw some gfortran
options were dropped/overridden by setting FCOMMON_OPT
. Have switched to FFLAGS
, which seems to alleviate this problem and still ensures -frecursive
is passed through.
Any thoughts on this, @jschueller?
No idea, but I found some frecursive flags: https://github.com/xianyi/OpenBLAS/search?utf8=%E2%9C%93&q=frecursive
Yeah, though that flag isn't actually getting added to the build normally FWICT. Take a look at @zym1010's comment for an example. The build log is completely devoid of this flag.
Thanks everyone for your feedback. As there doesn't seem to be any strong opposition and this is a recommended flag by upstream when building OpenBLAS's multithreaded LAPACK, going to go ahead and merge.
Was seeing this error from OpenBLAS in some BLAS intensive programs. This was often followed by a segfault.
After a little digging, I found this suggestion. As we are using pthreads, this is the right way to go. Rebuilt using this locally and the warning and segfaults went away.