I struggle with Caffe compilation. Unfortunately I failed to compile it.
Steps I followed:
git clone https://github.com/BVLC/caffe.git
cd caffe
mkdir build
cd build
cmake ..
make all
Running make all fails with the following error message:
[ 2%] Building NVCC (Device) object src/caffe/CMakeFiles/cuda_compile.dir/util/cuda_compile_generated_im2col.cu.o
In file included from /usr/include/cuda_runtime.h:59:0,
from <command-line>:0:
/usr/include/host_config.h:82:2: error: #error -- unsupported GNU version! gcc 4.9 and up are not supported!
#error -- unsupported GNU version! gcc 4.9 and up are not supported!
^
CMake Error at cuda_compile_generated_im2col.cu.o.cmake:207 (message):
Error generating /mydir/caffe/build/src/caffe/CMakeFiles/cuda_compile.dir/util/./cuda_compile_generated_im2col.cu.o
Software version:
OS: Debian.
gcc version: 5.3.1.
nvcc version: 6.5.12.
cat /proc/driver/nvidia/version result:
NVRM version: NVIDIA UNIX x86_64 Kernel Module 352.63 Sat Nov 7 21:25:42 PST 2015
GCC version: gcc version 4.8.5 (Debian 4.8.5-3)
Attempts to solve the problem
1st try
Simple solutions are often best ones, so (as suggested here) I tried to comment out macro checking gcc version from /usr/include/host_config.h (line 82). Unfortunately it doesn't work and compilation fails badly:
1 catastrophic error detected in the compilation of "/tmp/tmpxft_000069c2_00000000-4_im2col.cpp4.ii".
2nd try
I tried to run:
cmake -D CMAKE_CXX_COMPILER=g++-4.8 ..
make
but it fails with exactly the same error message (even though g++-4.8 should be accepted).
3rd try
I've found similar problem (though not related to Caffe) and I tried to solve it as suggested in the accepted answer.
What I did:
I've ran grep -iR "find_package(CUDA" caffe command and found Cuda.cmake file which has find_package(CUDA 5.5 QUIET) in line 225.
I added set(CUDA_HOST_COMPILER /usr/bin/gcc-4.8) to Cuda.cmake, line before line: find_package(CUDA 5.5 QUIET).
I removed everything from build directory and ran cmake and make again - with and without -D CMAKE_CXX_COMPILER=g++-4.8.
Unfortunately result is exactly the same. Caffe probably overwrites it somehow - I didn't figure it out how.
make VERBOSE=1 2>&1 | grep -i compiler-bindir returns nothing.
What's interesting, make VERBOSE=1 prints command that fails, which is:
/usr/bin/nvcc -M -D__CUDACC__ /mydir/caffe/src/caffe/util/im2col.cu -o /mydir/caffe/build/src/caffe/CMakeFiles/cuda_compile.dir/util/cuda_compile_generated_im2col.cu.o.NVCC-depend -ccbin /usr/bin/cc -m64 -DUSE_LMDB -DUSE_LEVELDB -DUSE_OPENCV -DWITH_PYTHON_LAYER -DGTEST_USE_OWN_TR1_TUPLE -Xcompiler ,\"-fPIC\",\"-Wall\",\"-Wno-sign-compare\",\"-Wno-uninitialized\",\"-O3\",\"-DNDEBUG\" -gencode arch=compute_20,code=sm_21 -Xcudafe --diag_suppress=cc_clobber_ignored -Xcudafe --diag_suppress=integer_sign_change -Xcudafe --diag_suppress=useless_using_declaration -Xcudafe --diag_suppress=set_but_not_used -Xcompiler -fPIC -DNVCC -I/usr/include -I/mydir/caffe/src -I/usr/include -I/mydir/caffe/build/include -I/usr/include/hdf5/serial -I/usr/include/opencv -I/usr/include/atlas -I/usr/include/python2.7 -I/usr/lib/python2.7/dist-packages/numpy/core/include -I/mydir/caffe/include -I/mydir/caffe/build
when I add --compiler-bindir /usr/bin/gcc-4.8 flag manually, it prints error:
nvcc fatal : redefinition of argument 'compiler-bindir'
which may be related to this bug report.
Edit: I didn't notice that --compiler-bindir and -ccbin are the same options, and the latter is already set in above command that failed. When I changed -ccbin /usr/bin/cc to -ccbin /usr/bin/gcc-4.8 in above command that failed, it completes successfully. Now I need to find option in Caffe's CMake file that overwrite -ccbin in all subsequent Caffe's CMakes. Looking at cmake/Cuda.cmake:252:list(APPEND CUDA_NVCC_FLAGS ${NVCC_FLAGS_EXTRA} seems to be good way to go.
How can I successfully complete my compilation? Any help is appreciated.
Related SO questions:
host_config.h:unsupported GNU version! gcc versions later than 4.9 are not supported.
CUDA 6.5 complains about not supporting gcc 4.9 - what to do?.
cmake -D CUDA_NVCC_FLAGS="-ccbin gcc-4.8" .. && make causes successful compilation.
Now another problem showed up: linking Google's libgflags or libprotobuf fails probably due to fact that it was compiled with newer gcc version but it's not related to asked question.
My machine runs Ubuntu 15.10, and my default compiler version is gcc 5.2.1 .
Commenting out the #error directive in line 115 of file
/usr/local/cuda-7.5/include/host_config.h
(or whatever the path on your system is) did the trick for me. Caffe compiled fine, all tests ran smoothly.
On the other hand, if one chooses to ignore this and proceed to compile part of the project with one compiler version, part of the project with another (for me it was gcc-4.8 and gcc-5.2.1), linking problems will arise. The linking problems of protobuf and libgflags another answer mentions are not unrelated to this.
Related
In order to make use of newer C++ 17 and 20 features, I want to switch from Apple clang to a homebrew installed vanilla clang. I installed it via brew install llvm.
When trying to build a minimal CMake based test project like
cmake -DCMAKE_BUILD_TYPE=Debug -DCMAKE_C_COMPILER=/usr/local/opt/llvm/bin/clang -DCMAKE_CXX_COMPILER=/usr/local/opt/llvm/bin/clang++ -G Ninja testProject
CMake fails with
-- The C compiler identification is Clang 13.0.0
-- The CXX compiler identification is Clang 13.0.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - failed
-- Check for working C compiler: /usr/local/opt/llvm/bin/clang
-- Check for working C compiler: /usr/local/opt/llvm/bin/clang - broken
CMake Error at /Applications/CLion.app/Contents/bin/cmake/mac/share/cmake-3.20/Modules/CMakeTestCCompiler.cmake:66 (message):
The C compiler
"/usr/local/opt/llvm/bin/clang"
is not able to compile a simple test program.
It fails with the following output:
Change Dir: /Users/me/Projects/testProject/build/CLion/Debug/CMakeFiles/CMakeTmp
Run Build Command(s):/usr/local/bin/ninja cmTC_d2d4b && [1/2] Building C object CMakeFiles/cmTC_d2d4b.dir/testCCompiler.c.o
[2/2] Linking C executable cmTC_d2d4b
FAILED: cmTC_d2d4b
: && /usr/local/opt/llvm/bin/clang -isysroot /Library/Developer/CommandLineTools/SDKs/MacOSX11.0.sdk -Wl,-search_paths_first -Wl,-headerpad_max_install_names CMakeFiles/cmTC_d2d4b.dir/testCCompiler.c.o -o cmTC_d2d4b && :
ld: library not found for -lSystem
clang-13: error: linker command failed with exit code 1 (use -v to see invocation)
ninja: build stopped: subcommand failed.
So it seems that it fails to locate the correct library search path out of the box.
When trying to build a simple hello world program from the command line like
/usr/local/opt/llvm/bin/clang++ main.cpp -isysroot /Library/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX11.0.sdk
It fails with even not finding some headers.
Supplying the header and library search path explicitly to the build command like
/usr/local/opt/llvm/bin/clang++ main.cpp -I"/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include" -L"/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib"
makes it work. However, both are no solutions to make the built-in CMake compiler check work and they don't seem like the correct way to go.
Once this is working correctly, I'm furthermore interested in being able to compile ObjectiveC++ code which currently fails with even more errors.
I think I understand the problem that the compiler does not know the correct search paths, but I'm lacking some understanding about how it would normally figure out those search paths in a CMake project.
I finally managed to get it working by supplying a few linker flags through the LDFLAGS environment variable:
-L/Library/Developer/CommandLineTools/SDKs/MacOSX11.0.sdk/usr/lib
In order to find the system libraries needed by the C compiler check
-L/usr/local/opt/llvm/lib
In order to find the C++ libraries as needed by the C++ compiler check
In order to successfully compile Objective-C code that links to Apple frameworks, I furthermore needed to add
-F/Library/Developer/CommandLineTools/SDKs/MacOSX11.0.sdk/System/Library/Frameworks -lobjc
With those flags, I now can successfully build my projects. Still, I'm not 100% sure if this is the right way, especially if the libraries linked from /usr/local/opt/llvm/lib can be expected to be present on all target systems?
I just moved my PC from Ubuntu 15.10 to Linux Mint 17.3.
Before this shift, this project compiled just fine in CLion. Now, it gives the following error:
c++: error: unrecognized command line option ‘-std=c++14’
This is probably due to this line in my CMake file:
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -W -Wall -Wextra -pedantic")
I suspect this error is due to some outdated library/compiler. I don't know exactly what I need to do.
P.S.: I know that C++14 is not completely supported in CLion, and the editor will complain about syntax errors, but it will compile fine nonetheless.
-std=c++14 is called -std=c++1y in old gcc versions (at least 4.9 and older).
I am working on a high-performance cluster and the "default" version of gcc is 4.8.5, which is why I also ran into the error c++: error: unrecognized command line option ‘-std=c++14’.
Not having sudo rights, I could solve the issue by loading a module (https://modules.readthedocs.io/en/latest/) containing a more recent gcc version (module load compiler/gnu/9.1). So if you are working on a cluster check out module avail (or in general the documentation of the cluster).
Can you (reasonably) get Fedora 21 to where it only has llvm/clang/libc++/libc++abi? (I found some things suggesting no, but they were all about 3 years old, and llvm/clang has come a long way since then.)
With a fresh install, I tried
yum install gcc gcc-c++
(downloaded, built, installed llvm/cfe(clang)/compiler-rt/libcxx/libcxxabi from svn)
yum remove gcc gcc-c++
added to /etc/profile: export CC=/usr/local/bin/clang \ export CXX=/usr/local/bin/clang++
(in case of hard wiring)
ln -s /usr/local/bin/clang /usr/local/bin/gcc
ln -s /usr/local/bin/clang /usr/local/bin/cc
ln -s /usr/local/bin/clang++ /usr/local/bin/g++
ln -s /usr/local/bin/clang++ /usr/local/bin/c++
ldconfig
I was all happy, then went to build something, and I got:
ld: cannot find crtbegin.o
ld: cannot find -lgcc
ld: cannot find -lgcc_s
clang -v includes
Found candidate GCC installation: /usr/lib/gcc/x86_64-redhat-linux/4.9.2
ldconfig && ldconfig -p | grep libgcc does show
libgcc_s.so.1 (libc6,x86-64) => /lib64/libgcc_s.so.1
And /lib64 is a symlink to /usr/lib64. And, /usr/lib64/libgcc_s.so.1 is a symlink to /usr/lib64/libgcc_s-4.9.2-20150212.so.1, which exists as a real file (92816 bytes.)
So, I don't get what ld's problem is on -lgcc_s. crtbegin is nowhere to be found, and gcc (no _s) is nowhere to be found.
yum install libgcc says it's already installed and latest version, nothing to do.
Since I have an installed clang source build, can I re-build clang, this time using clang rather than gcc, to get rid of the dependency? (Maybe then the "candidate GCC installation" bit goes away.)
Can I force -stdlib=c++ and -lc++abi to be default, or at least have libc++ and libc++abi installed without gcc?
Having spent some time trying to get clang to work with libc++ and libc++abi without GCC, I have found that it is indeed possible, even if a bit problematic given the current state of LLVM/clang. In addition to small test programs, I've been able to build CMake and some other software packages written in C++ with no GCC installed, and with the resulting binaries being independent of libstdc++; they only depend on libc++/libc++abi according to ldd output. Unfortunately, I haven't been able to build clang itself with clang that was build using GCC. I've been experimenting on different Linux platforms (Fedora 21 32-bit, Amazon Linux release 2015.3 (RPM-based) 64-bit, CentOS 7.1 64-bit, and Ubuntu 14.04 64-bit).
Even though one can build software with clang using libc++/libc++abi without dependency on libstdc++ and without GCC compiler present, a typical Linux installation is so tied to libgcc and libstdc++ that getting rid of these is not practical. Try removing these two packages and you will see how much of the system depends on them. Even on FreeBSD 10.1, with clang being the default compiler and no GCC installed, libgcc.a, libgcc_s.so, and a few crt*.o files are used when building a program as revealed by the -v option. Also, on FreeBSD 10.1, resulting binaries depend on libgcc according to ldd. On Ubuntu, which has dpkg as the package manager, the files
libgcc.a
libgcc_s.so
crtbegin.o
crtbeginT.o
crtbeginS.o
crtendS.o
crtend.o
are in the libgcc-devel package, while on an RPM-based system, such as Fedora, these are in the gcc package. In addition, you might possibly need these files, even though I didn't need them for the code I tried building:
crtfastmath.o
crtprec32.o
crtprec80.o
crtprec64.o
Thus one might argue that the aforementioned files better belong in libgcc, rather than in gcc. As far as I can tell, the following needs to be done on an RPM-based system before removing the gcc package:
1) Create the symlink
libgcc_s.so -> libgcc_s.so.1
in whatever directory libgcc_s.so.1 is located.
2) Copy the crt*.o files listed above to that directory.
3) In the same directory create the symlink (libstdc++.so.x should already be there; x is a number):
libstdc++.so -> libstdc++.so.x
You only need this if you are going to use libstdc++; this isn't needed if you only plan to use libc++. On some
systems libstdc++.so, which is a symlink to libstdc++.so.x belonging to the libstdc++ package, is placed by the libstdc++-devel package into the GCC library directory, so you can remove that directory after uninstalling GCC and just create the symlink in the same directory where libstdc++.so.x lives.
Now you should be able to do the following:
1) Build a C program:
clang progname.c
2) Build a C++ program using libstdc++ headers/libs:
clang++ -I<location of headers> progname.cpp
On RPM-based systems I've looked at, the libstdc++ headers are part of the libstdc++-devel package and their location can be found from rpm -ql on the package.
3) Build a C++ program using libc++ headers/libs:
clang++ -I/<location of headers> progname.cpp -nodefaultlibs -lc++ -lc++abi -lm -lc -lgcc -lgcc_s
The location of the headers is wherever they were installed when you built LLVM+clang etc.
Please see http://libcxx.llvm.org/ for additional information. When building C++ code using libc++/libc++abi, you may use -stdlib=libc++ instead of the -I flag, but in my testing that only worked with clang built from source, not with clang installed from a repository (you can install clang from repo and use it to build libc++/libc++abi; or you can use gcc to build libc++(abi), then remove gcc and use the libs with the repo-provided clang).
When configuring a software package to build it using clang + libc++, you might need to set the following:
LIBS="-nodefaultlibs -lc++ -lc++abi -lm -lc -lgcc_s -lgcc"
CXX=clang++
CXXFLAGS="-stdlib=libc++"
CC=clang
Please note that to configure CMake source in order to build it I had to use a wrapper script like this:
#!/bin/bash
MYLFLAGS="-nodefaultlibs -lc++ -lc++abi -lm -lc -lgcc_s -lgcc"
# Catch the case when we only want to compile; this helps us avoid some warnings:
if echo "$#" | egrep "(^-c | -c | -c$)" >/dev/null 2>&1; then
MYLFLAGS=""
fi
/usr/local/bin/clang++ -stdlib=libc++ "$#" $MYLFLAGS
It might be useful for other purposes as well.
For more information please see my article at http://www.omniprog.info/clang_no_gcc.html
I'm trying to compile and install the x264 H.264/AVC encoder.
I've got gcc installed. But I get the 'No working C compiler found' error when I run:
./configure --enable-shared --enable-static
What can I do?
The config log said:
/bin/gcc conftest.c -Wall -I. -I$(SRCPATH) -falign-loops=16 -mdynamic-no-pic -o conftest
clang: error: unknown argument: '-falign-loops=16' [-Wunused-command-line-argument-hard-error-in-future]
clang: note: this will be a hard error (cannot be downgraded to a warning) in the future
I encountered the same error and found a simple solution here:
http://www.xin.at/x264/x264-guide-macosx-en.htm
Before actually being able to start the build we will however need to remove a GCC compiler flag from the configure script, that the newer LLVM+CLANG compiler will not be able to handle. For that, please open the file configure in your favorite text editor and look for the following spot:
darwin*)
SYS="MACOSX"
CFLAGS="$CFLAGS -falign-loops=16"
Replace that with the following, effectively removing the -falign-loops=16 option:
darwin*)
SYS="MACOSX"
CFLAGS="$CFLAGS"
After doing the above, libx264 builds just fine :)
The configure script is trying to set a compiler option -falign-loops=16 that the clang compiler (masquerading as gcc) declines to accept.
Either get (compile) your own real GCC and use that (I've done the compilation and installation; it's not very hard, though neither is it trivial), or work out how to stop the configure script from failing simply because it assumes that the -falign-loops=16 option must be supported by all versions of GCC. That is the sort of thing the configure script should be checking so that you don't run into that sort of failure. Ultimately, this is a bug in the configuration for this code.
I am attempting to compile gcc 4.4.0 on opensolaris 2009.6
Currently in the box (which is a AMD 64bit machine), I have the gcc 3.4.6 installed.
I unpacked the gcc 4.4.0 tarball.
I set the following env variables:
export CXX=/usr/local/bin/g++
export CC=/usr/local/bin/gcc
Then I ran "configure && make" and this is the error message that I got:
checking for i386-pc-solaris2.11-gcc... /export/home/me/wd/gcc/gcc-4.4.0/host-i386-pc-solaris2.11/gcc/xgcc -B/export/home/me/wd/gcc/gcc-4.4.0/host-i386-pc-solaris2.11/gcc/ -B/usr/local/i386-pc-solaris2.11/bin/ -B/usr/local/i386-pc-solaris2.11/lib/ -isystem /usr/local/i386-pc-solaris2.11/include -isystem /usr/local/i386-pc-solaris2.11/sys-include -m64
checking for suffix of object files... configure: error: in `/export/home/me/wd/gcc/gcc-4.4.0/i386-pc-solaris2.11/amd64/libgcc':
configure: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
Anyone has any suggestion as to how to work around this error message?
/Edit:
Content of the config.log is posted here: link text
Normally the GCC build is bootstrapped, i.e. first it uses the system compiler to build GCC C compiler, and then it uses the freshly built compiler to recompile GCC once again (and then even once more time again). The configure line shows that it is not the system compiler but the already-built GCC compiler which is used for configure test there.
Since it fails, the problem is that the freshly-built GCC is somehow "stillborn" here. If config.log will not help you, I'd suggest to ask at gcc-help#gcc.gnu.org.
EDIT: Ah-ha, I think it is the assembler. You are using GNU assembler, but the unsupported options look like they were meant for Sun assembler. This should be solved by adding --with-gnu-as configure option (and then possibly having to specify its path explicitly with --with-as=/usr/gnu/bin/as)
You can also take a look at Solaris-specific GCC build instructions.
There's a readily available build for gcc4, which you can try updating. Its current version is 4.3.3. To get started, install pkg-get from OpenCSW and check out the build from the subversion repository:
svn co https://gar.svn.sourceforge.net/svnroot/gar/csw/mgar/pkg/gcc4/trunk/ gcc4
cd gcc4
gmake package