[Mvapich-discuss] Build failure with NVHPC compilers

Shineman, Nat shineman.5 at osu.edu
Thu Apr 21 13:49:15 EDT 2022


Hi Martin,

I looked over your patch and it looks like you have edited the configuration file directly. Unfortunately, this file is overwritten anytime we rerun autotools so this fix may work locally, but it won't persist. Fortunately, these flags are only added to a couple files so we can easily patch the source files. I have attached a patch that should resolve the issues. Please try this one and let me know if it resolves the compilation issue.

Thanks,
Nat


________________________________
From: Mvapich-discuss <mvapich-discuss-bounces+shineman.5=osu.edu at lists.osu.edu> on behalf of Martin Cuma via Mvapich-discuss <mvapich-discuss at lists.osu.edu>
Sent: Friday, April 15, 2022 18:02
To: Subramoni, Hari <subramoni.1 at osu.edu>; mvapich-discuss at lists.osu.edu <mvapich-discuss at lists.osu.edu>
Subject: Re: [Mvapich-discuss] Build failure with NVHPC compilers

Hi Hari, sorry for late reply, it took me some time to go over this. The 2.3.7 gave me the same error as 2.3.6 so I went over the configure to add in pieces to make it work. It required a few changes, namely - add the nvc, nvc++ and nvfortran ZjQcmQRYFpfptBannerStart
This Message Is From an External Sender
This message came from outside your organization.
<https://us-phishalarm-ewt.proofpoint.com/EWT/v1/KGKeukY!vwQd8KhtD6YgRRdxXq6lWWC81PlyeW0vBCqp5b8yACiRkosXP6KQQ9YcxsYneVG2H3XnISlmc-Ay31YFp3dWKvJx6l74qWMRYQE-num46xblmIleuiBVdaKisMXe3x78N98ZwfpNHABw4D153bujaA$>
Report Suspicious

ZjQcmQRYFpfptBannerEnd
Hi Hari,

sorry for late reply, it took me some time to go over this.

The 2.3.7 gave me the same error as 2.3.6 so I went over the configure to add in pieces to make it work. It required a few changes, namely
- add the nvc, nvc++ and nvfortran wherever pgcc, pgCC and pgf90 were referenced
- adding an explicit if statement where the -rdynamic is set to not include it if nvc is used (it detects system based gcc which is present and does not check for what CC is set to)
- nvc++ adds the "standard" objects such as crtn.o or crti.o, but pre/postdep_objects were being set so I explicitly don't set it through another if $CXX==nvc++

Attached is a patch that does all this, though, I have to admit that I only tested this  through Spack so I am not sure if the checks for CC or CXX are appropriate outside of Spack. But the patch file may be a good start to get this fixed for you.

Thanks
MC
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osu.edu/pipermail/mvapich-discuss/attachments/20220421/c817b3d0/attachment-0019.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgi-nordynamic.patch
Type: text/x-patch
Size: 891 bytes
Desc: pgi-nordynamic.patch
URL: <http://lists.osu.edu/pipermail/mvapich-discuss/attachments/20220421/c817b3d0/attachment-0019.bin>


More information about the Mvapich-discuss mailing list