Using Openmp with Clang

Using OpenMP with clang

Update

Building the latest trunk of LLVM/Clang (clang-3.8), installing libiomp5, and specifying the location of the gomp omp header files worked. Note that the Ubuntu package for libiomp5 isn't quite correct, so you will need to add a symlink in /usr/lib from /usr/lib/libiomp5.so to /usr/lib/libiomp5.so.5.

./clang++ -I/usr/lib/gcc/x86_64-linux-gnu/4.9/include -fopenmp=libiomp5 -o test test.cpp

---

I'm using g++-5.1 and clang++-3.6 on Linux Mint 17.2 (essentially Ubuntu trusty) and I see the same results with the following code.

#include <iostream>
#include <omp.h>
int main() {
    #pragma omp parallel num_threads(4)
    {
        #pragma omp critical
        std::cout << "tid = " << omp_get_thread_num() << std::endl;
    }
}

Running this under ltrace reveals the issue:

g++

$ g++ -fopenmp -o test test.cpp
$ ./test
tid = 0
tid = 3
tid = 2
tid = 1
$ ltrace ./test
__libc_start_main(0x400af6, 1, 0x7ffc937b8198, 0x400bc0 <unfinished ...>
_ZNSt8ios_base4InitC1Ev(0x6021b1, 0xffff, 0x7ffc937b81a8, 5)   = 0
__cxa_atexit(0x4009f0, 0x6021b1, 0x602090, 0x7ffc937b7f70)     = 0
GOMP_parallel(0x400b6d, 0, 4, 0 <unfinished ...>
GOMP_critical_start(0, 128, 0, 0)                              = 0
tid = 3
tid = 2
omp_get_thread_num(0x7f9fe13894a8, 1, 0, 0x493e0)              = 0
_ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc(0x6020a0, 0x400c44, 0, 0x493e0) = 0x6020a0
_ZNSolsEi(0x6020a0, 0, 0x7f9fe1a03988, 0x203d2064)             = 0x6020a0
_ZNSolsEPFRSoS_E(0x6020a0, 0x400920, 0x7f9fe1a03988, 0 <unfinished ...>
_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_(0x6020a0, 0x400920, 0x7f9fe1a03988, 0) = 0x6020a0
<... _ZNSolsEPFRSoS_E resumed> )                               = 0x6020a0
GOMP_critical_end(0x7f9fe0d2d400, 0x7f9fe0d2e9e0, 0, -1)       = 0
tid = 1
tid = 0
<... GOMP_parallel resumed> )                                  = 0
_ZNSt8ios_base4InitD1Ev(0x6021b1, 0, 224, 0x7f9fe0d2df50)      = 0x7f9fe1a08940
+++ exited (status 0) +++

clang

$ clang++ -fopenmp -o test test.cpp
$ ./test
tid = 0
$ ltrace ./test
__libc_start_main(0x4009a0, 1, 0x7ffde4782538, 0x400a00 <unfinished ...>
_ZNSt8ios_base4InitC1Ev(0x6013f4, 0x7ffde4782538, 0x7ffde4782548, 5) = 0
__cxa_atexit(0x400830, 0x6013f4, 0x6012c8, 0x7ffde4782310)     = 0
_ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc(0x6012e0, 0x400a84, 0x7ffde4782548, 6) = 0x6012e0
omp_get_thread_num(0x7f3e4698c006, 0x7f3e4698c000, 0x7f3e46764988, 1024) = 0
_ZNSolsEi(0x6012e0, 0, 0x7f3e46764988, 1024)                   = 0x6012e0
_ZNSolsEPFRSoS_E(0x6012e0, 0x4007a0, 0x7f3e46764988, 0 <unfinished ...>
_ZSt4endlIcSt11char_traitsIcEERSt13basic_ostreamIT_T0_ES6_(0x6012e0, 0x4007a0, 0x7f3e46764988, 0) = 0x6012e0
tid = 0
<... _ZNSolsEPFRSoS_E resumed> )                               = 0x6012e0
_ZNSt8ios_base4InitD1Ev(0x6013f4, 0, 224, 0x7f3e45886f50)      = 0x7f3e46769940
+++ exited (status 0) +++

You can immediately see the problem: clang++ never calls GOMP_parallel, so you always get one thread. This is crazy behavior on the part of clang. Have you tried building and using the "special" OpenMP version of clang?

Using OpenMP with llvm-clang

Despite the fact that its kinda late regarding the time-stamp of my original question, but I would like to mention the answer here so at-least it saves people's time facing similar issue.

LLVM itself currently doesn't support Openmp right out-of-the-box. You can make it compile and run the omp tagged code with Intel Runtime Support. However, if you want to have a clean clang supporting OpenMP, there is a trunk of the project at OpenMP-Clang which you can clone and build. The current support is OpenMP 3.1 specification and they will reach to support OpenMP 4.0 specification soon:

$ git clone https://github.com/clang-omp/llvm_trunk llvm
$ git clone https://github.com/clang-omp/compiler-rt_trunk llvm/projects/compiler-rt
$ git clone https://github.com/clang-omp/clang_trunk llvm/tools/clang

Don't forget to build the Intel® OpenMP* Runtime Library after this as you need omp.h and /path/to/llvm/projects/openmp/runtime/lin_32e/lib/libomp.so

LLVM / Clang 8 Compilation of OpenMP Code in Windows

To compile and link OpenMP code with clang on Windows, you will have to pass -fopenmp to both the compiler and the linker:

clang -fopenmp -o bla.obj -c bla.c
clang -fopenmp -o bla.exe bla.obj

clang 10 & OpenMP on range-based for error (docs say should be ok)

Support for range-based for loops was added to OpenMP 5.0, and, as is also described in the Clang 10 Release Notes that you link to, you need to explicitly use the -fopenmp-version=50 option to activate support for it:

OpenMP Support in Clang

Use -fopenmp-version=50 option to activate support for OpenMP 5.0.

Thus, if we expand you compilation command to clang++-10 -fopenmp -fopenmp-version=50 -std=c++17, the OMP pragma accepts the range based for loop that follows it.

DEMO.

---

Related Topics