[mvapich-discuss] Cannot build MVAPICH 2.3.4 from source: helper_fns.h not found

[Fang, Leo]

Dear Hari,


Thanks (as always :D) for your prompt replies! The reason I used ch3:nemesis is that on this test machine without IB or any other high-speed network, I got the following error if I set to ch3:mrail (or don’t set any device):

configure: error: 'libibverbs not found. Did you specify --with-ib-libpath=?’

But I just realized libibverbs still exists, and this works for me:

./configure --enable-cuda --prefix=/home/leofang/.mvapich2-2.3.4_cuda --with-pm=hydra --with-ibverbs-include=/usr/include/infiniband/ --with-ibverbs-lib=/usr/lib/x86_64-linux-gnu/ --disable-mcast

(Yeah, I think I did see "YACC is not found” error when building, but after I did "sudo apt install byacc" the scanner.c error showed up instead of complaining yacc not found. But it’s not reproducible now...)

As for not using GDR (which is what I would like to really use in a production env), the reason is this machine missed libibmad.so and libibumad.so to which libmpi.so is linked. I’ll try looking into that.

Thanks again.


Sincerely,
Leo

---
Yao-Lung Leo Fang
Assistant Computational Scientist
Computational Science Initiative
Brookhaven National Laboratory
Bldg. 725, Room 2-169
P.O. Box 5000, Upton, NY 11973-5000
Office: (631) 344-3265
Email: leofang at bnl.gov<mailto:leofang at bnl.gov>
Website: https://urldefense.com/v3/__https://leofang.github.io/__;!!KGKeukY!jolzjoerrN5Q_pwO74Bf-23YzTxo5pOH3UV0_pq4d1eNFl9N-SBt4Qp00mDB2OcCec67R2ZqN8hiYjNmn00n$ 

Subramoni, Hari <subramoni.1 at osu.edu<mailto:subramoni.1 at osu.edu>> 於 2020年7月20日 上午9:49 寫道：

Hi, Leo.

The ch3:nemesis branch has been deprecated. Please use the ch3:mrail branch. For running on GPUs, we strongly recommend using MVAPICH2-GDR for best performance and scalability. This is even true if it is a single node system since MVAPICH2-GDR has a lot of advanced support for intra-node point-to-point and collective communication.

The error with scanner.c is because lex/yacc is not present on your system. Another user reported this on discuss and we will be resolving this with the next release.

Best,
Hari.

From: mvapich-discuss-bounces at cse.ohio-state.edu<mailto:mvapich-discuss-bounces at cse.ohio-state.edu> <mvapich-discuss-bounces at mailman.cse.ohio-state.edu<mailto:mvapich-discuss-bounces at mailman.cse.ohio-state.edu>> On Behalf Of Fang, Leo
Sent: Monday, July 20, 2020 12:44 AM
To: mvapich-discuss at cse.ohio-state.edu<mailto:mvapich-discuss at cse.ohio-state.edu> <mvapich-discuss at mailman.cse.ohio-state.edu<mailto:mvapich-discuss at mailman.cse.ohio-state.edu>>
Subject: [mvapich-discuss] Cannot build MVAPICH 2.3.4 from source: helper_fns.h not found

Hello,


I tried to build MVAPICH 2.3.4 from source with the following command:

./configure --enable-cuda --prefix=/home/leofang/.mvapich2-2.3.4_cuda --with-device=ch3:nemesis --with-pm=hydra

But the build failed with "fatal error: helper_fns.h: No such file or directory” when compiling source codes from src/mpi/coll/. Occasionally the error is: "error: src/pm/mpirun/src/hostfile/scanner.c: No such file or directory”, but it’s not always reproducible.

The machine I’m on has no infiniband or any high-speed network or job scheduler, just an isolated test node with GPUs.

I had a vague impression that this worked in the past versions, but I am not 100% sure about the build command I used. Also, with a quick search I found helper_fns.h is newly added in 2.3.4: https://urldefense.com/v3/__https://fossies.org/diffs/mvapich2/2.3.3_vs_2.3.4/index.html__;!!KGKeukY!jolzjoerrN5Q_pwO74Bf-23YzTxo5pOH3UV0_pq4d1eNFl9N-SBt4Qp00mDB2OcCec67R2ZqN8hiYrCq5Ngy$ <https://urldefense.com/v3/__https:/fossies.org/diffs/mvapich2/2.3.3_vs_2.3.4/index.html__;!!KGKeukY!hmScmeImOTuuWL8EW62m9cvsuaf3mYaLrA4dqZr4GdDngv87DPvlXZgSc4YaRRCR33t_vGAmD9KRF4Q$>. Could it be that the configure script failed to copy it to the build path in certain cases?

Thanks.


Sincerely,
Leo

---
Yao-Lung Leo Fang
Assistant Computational Scientist
Computational Science Initiative
Brookhaven National Laboratory
Bldg. 725, Room 2-169
P.O. Box 5000, Upton, NY 11973-5000
Office: (631) 344-3265
Email: leofang at bnl.gov<mailto:leofang at bnl.gov>
Website: https://urldefense.com/v3/__https://leofang.github.io/__;!!KGKeukY!jolzjoerrN5Q_pwO74Bf-23YzTxo5pOH3UV0_pq4d1eNFl9N-SBt4Qp00mDB2OcCec67R2ZqN8hiYjNmn00n$ <https://urldefense.com/v3/__https:/leofang.github.io/__;!!KGKeukY!hmScmeImOTuuWL8EW62m9cvsuaf3mYaLrA4dqZr4GdDngv87DPvlXZgSc4YaRRCR33t_vGAmB3SVCRs$>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20200720/5e2cb3f7/attachment-0001.html>