Skip to content

Commit

Permalink
Update arpack-ng to 3.2.
Browse files Browse the repository at this point in the history
  • Loading branch information
Viral B. Shah committed Dec 9, 2014
1 parent 181a3ab commit 68e5cc7
Show file tree
Hide file tree
Showing 7 changed files with 3 additions and 998 deletions.
1 change: 0 additions & 1 deletion deps/Makefile
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1139,7 +1139,6 @@ arpack-ng-$(ARPACK_VER).tar.gz:
arpack-ng-$(ARPACK_VER)/configure: arpack-ng-$(ARPACK_VER).tar.gz
$(JLCHECKSUM) $<
$(TAR) zxf $<
( cd arpack-ng-$(ARPACK_VER) && patch -p1 < ../arpack.xlahqr2.patch )
touch -c $@

ifeq ($(USE_ATLAS), 1)
Expand Down
2 changes: 1 addition & 1 deletion deps/Versions.make
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ PCRE_VER = 8.36
DSFMT_VER = 2.2
OPENBLAS_VER = v0.2.12
LAPACK_VER = 3.5.0
ARPACK_VER = 3.1.5
ARPACK_VER = 3.2.0
FFTW_VER = 3.3.4
SUITESPARSE_VER = 4.3.1
UNWIND_VER = 1.1
Expand Down
Loading

18 comments on commit 68e5cc7

@JeffBezanson
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm getting this:

checking for dummy main to link with Fortran 77 libraries... none
checking for Fortran 77 name-mangling scheme... lower case, underscore, no extra underscore
checking for sgemm_ in -L/home/bezanson/src/julia/usr/lib -lopenblas... no
checking if sgemm_ is being linked in already... no
checking for ATL_xerbla in -latlas... no
checking for sgemm_ in -lblas... no
checking for sgemm_ in -lmkl... no
checking for sgemm_ in -framework vecLib... no
checking for sgemm_ in -lcxml... no
checking for sgemm_ in -ldxml... no
checking for sgemm_ in -lscs... no
checking for sgemm_ in -lcomplib.sgimath... no
checking for sgemm_ in -lblas... (cached) no
checking for sgemm_ in -lblas... (cached) no
configure: error: Cannot find BLAS libraries

usr/lib/libopenblas.so exists and does contain sgemm_.

@staticfloat
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you post the full config.log so we can see what the linker invocation for the test program is?

@ViralBShah
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We may want to revert this commit in that case.

Perhaps the same as this: opencollab/arpack-ng#1
@turboencabulator there suggests that a newer ax_blas.m4 may be needed.

http:https://git.savannah.gnu.org/gitweb/?p=autoconf-archive.git;a=history;f=m4/ax_blas.m4

@JeffBezanson
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here's my full config.log: https://gist.github.com/JeffBezanson/9315c50ed19f3c7e5cde

@staticfloat
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks like you're getting bitten by our symbol renaming:

configure:15391: checking for sgemm_ in -L/home/bezanson/src/julia/usr/lib -lopenblas
configure:15419: gcc -m64 -o conftest  -Dsgemm_=sgemm_64_ -Dcheev_=cheev_64_   -Wl,-rpath,'/home/bezanson/src/julia/usr/lib' conftest.c -L/home/bezanson/src/julia/usr/lib -lopenblas   -L/usr/lib/gcc/x86_64-linux-gnu/4.6 -L/usr/lib/gcc/x86_64-linux-gnu/4.6/../../../x86_64-linux-gnu -L/usr/lib/gcc/x86_64-linux-gnu/4.6/../../../../lib -L/lib/x86_64-linux-gnu -L/lib/../lib -L/usr/lib/x86_64-linux-gnu -L/usr/lib/../lib -L/usr/lib/gcc/x86_64-linux-gnu/4.6/../../.. -lgfortran -lm -lquadmath >&5
/tmp/cc3OTmcJ.o: In function `main':
conftest.c:(.text+0xa): undefined reference to `sgemm_64_'
collect2: ld returned 1 exit status
configure:15419: $? = 1

@tkelman

@tkelman
Copy link
Contributor

@tkelman tkelman commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

make -C deps distclean-openblas distclean-suitesparse

@JeffBezanson
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Trying it now. (rebuilding openblas... cries)

@tkelman
Copy link
Contributor

@tkelman tkelman commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Heh, sorry, if you're not building binaries for anyone else then override OPENBLAS_DYNAMIC_ARCH = 0 will speed that up. I also have a way of turning the symbol renaming off in my "Bump OpenBLAS" PR

@JeffBezanson
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok, works now :)

@ivarne
Copy link
Member

@ivarne ivarne commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we post this as a useful hint on the mailing lists?

@tkelman
Copy link
Contributor

@tkelman tkelman commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ivarne I think I did shortly after merging #8734 ?

@ivarne
Copy link
Member

@ivarne ivarne commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry, didn't realize this was such a old issue. I definitely have not added override OPENBLAS_DYNAMIC_ARCH = 0 to my Make.user to speed up openblas compiles, so I must have missed part of it.

@tkelman
Copy link
Contributor

@tkelman tkelman commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh that's technically separate from the symbol renaming and has been the case for even longer. We probably want to open a tracking issue to remind either me or @staticfloat to make fast-compile-mode the default, and group together "portable Julia" mode settings together under a single flag.

Also while I'm pinging @staticfloat, status.julialang.org is really spotty again. Any idea why?

@staticfloat
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you go directly to http:https://128.52.175.131, is it still spotty? Weirdly enough, I seem to be having DNS issues, but using the IP address directly works fine.

@staticfloat
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In fact, DNS for julialang.org seems to be down as well.

@StefanKarpinski So..... you were complaining about 1&1? ;)

@tkelman
Copy link
Contributor

@tkelman tkelman commented on 68e5cc7 Dec 9, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, good call, looks like a DNS problem, direct to the IP works fine. Should we tweak travis-ci/travis-build#318 to refer to the IP or is that not long-term static?

@staticfloat
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is, unfortuantely, not long-term static. I'm doing what I can to make it so we rely more on AWS and less on status.julialang.org though. I'm pretty sure Amazon has got their collective stuff together when it comes to reliability. ;)

@ViralBShah
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JeffBezanson You got by a long time without having to recompile openblas!

Please sign in to comment.