Boost logo

Boost Users :

Subject: [Boost-users] Compiling Boost with PGI
From: Ruediger Berlich (ruediger.berlich_at_[hidden])
Date: 2009-04-22 07:59:05


Hi there,

I am trying to port my heavily Boost-based application to different
compiler(-versions) . I was successful with Intel and various Linux-based
gcc versions so far. Now I am trying my luck with the Portland group
compiler in version 8. Corresponding compiler support seems to be available
in Boost:

./tools/build/v2/tools/pgi.jam
./boost/config/compiler/pgi.hpp

A "./configure --with-toolset=pgi" succeeds.

The subsequent make, however, results in a plethora of error messages such
as the following:

gcc.compile.c++
bin.v2/libs/math/build/gcc/release/threading-multi/nextafter.o
/bin/sh: line 1: -ftemplate-depth-128: command not found

      -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -pthread -fP
IC -DBOOST_ALL_NO_LIB=1 -DBOOST_MATH_TR1_DYN_LINK=1 -DNDEBUG -I"." -c -o "bin.
v2/libs/math/build/gcc/release/threading-multi/nextafter.o" "libs/math/build/../
src/tr1/nextafter.cpp"

Looking into the pgi configuration files makes it appear as if the pgi
support isn't finished yet. E.g. pgi.jam has a commented-out line

# flags pgi.compile OPTIONS <threading>multi : -mt ;

Indeed, pgCC does not accept this switch:

$ pgCC -mt
pgCC-Error-Unknown switch: -mt

The above error message also makes it appear as if gcc options
(e.g. "-Wall -pthread") are passed to pgCC.

Can anyone comment on the status of PGI support by Boost ?

The above happens on a system whose /etc/redhat-release states "Scientific
Linux SL Release 4.0 (Beryllium)". The kernel is 2.6.9-42 . Scientific
Linux is a Fedora-derivative, being used mostly in particle physics.

Thanks and Best Regards,
Ruediger


Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net