开发者
资源
旗舰店安装HMPI报错:dangerous relocation: unsupported reocation
旗舰店安装HMPI报错:dangerous relocation: unsupported reocation
发表于2023/07/04
1181

背景:旗舰店上面的HMPI一开始安装的都没有问题,但是后面测试一款新应用需要重装的时候报错了

cke_4469.jpeg

根因分析:

通过查看config.log,看到如下报错:

_CLEAR_CACHE=1 -Wshadow -Isrc/   conftest.c  >&5
conftest.c:82:10: fatal error: 'sys/cpuset.h' file not found
1 error generated.
conftest.c:49:10: fatal error: 'sys/cpuset.h' file not found
1 error generated.
conftest.c:61:24: error: expected expression
1 error generated.
conftest.c:62:13: error: use of undeclared identifier 'cpuset_t'
1 error generated.
conftest.c:58:27: error: expected expression
1 error generated.
conftest.c:59:29: error: expected expression
1 error generated.
conftest.c:87:10: fatal error: 'pthread_np.h' file not found
1 error generated.
conftest.c:54:10: fatal error: 'pthread_np.h' file not found
1 error generated.
/usr/bin/ld: /usr/lib64/libpthread.so.0: error adding symbols: DSO missing from command line
clang-12: error: linker command failed with exit code 1 (use -v to see invocation)
| /* Override any GCC internal prototype to avoid an error.
clang-12: error: linker command failed with exit code 1 (use -v to see invocation)
| /* Override any GCC internal prototype to avoid an error.
clang-12: error: linker command failed with exit code 1 (use -v to see invocation)
| /* Override any GCC internal prototype to avoid an error.
conftest.c:64:10: fatal error: 'sys/platform/ppc.h' file not found
1 error generated.
conftest.c:98:10: fatal error: 'sys/platform/ppc.h' file not found
1 error generated.
conftest.c:65:10: fatal error: 'sys/platform/ppc.h' file not found
1 error generated.
conftest.c:101:10: fatal error: 'sys/capability.h' file not found
1 error generated.
conftest.c:68:10: fatal error: 'sys/capability.h' file not found
1 error generated.
conftest.c:76:13: error: no member named '__u6_addr' in 'struct in6_addr'
1 error generated.
conftest.c:76:20: error: no member named '__u6_addr' in 'struct in6_addr'
1 error generated.
conftest.c:76:18: error: member reference base type '__be32' (aka 'unsigned int') is not a structure or union
1 error generated.
conftest.c:76:25: error: member reference base type '__be32' (aka 'unsigned int') is not a structurga

该报错属于正常的配置错误,关键问题在于/usr/bin/libbfd.a这个库,一般操作系统里面都是集成的libbfd.so,这里突然出现一个静态库,而且是X86上编译得来的(报X86架构识别错误),怀疑是有人拷贝错了,将该库进行备份重命名

mv libbfd.a libbfd.a.bk

重新清理hmpi安装目录

make distclean

然后重新安装HMPI,如下所示:

./autogen.pl

./configure --prefix=$1 --with-platform=contrib/platform/mellanox/optimized --enable-orterun-prefix-by-default --enable-mpi1-compatibility --with-ucx=$1/hucx

make -j64

make install

发现可以安装成功

收藏举报
Level 1
0
帖子
0
粉丝
0
获赞