51
edits
Changes
→Instructions for AMD
3. Yunseon Lee
OpenMP API specification for parallel programming provides an application programming interface (API) that supports multi-platform shared memory multiprocessing programming in C, C++, and Fortran, on most platforms. It consists of a set of compiler directives, library routines, and environment variables that influence run-time behavior.
Benefits of OpenMPI:OpenMP. Why to choose over GPU kernel model?-supports multi-core, vectorization and GPU-allows for "teams of threads"OpenCl is highly -portablebetween various plaforms-heterogeneous memory allocation and custom data mappers [More information (compare OpenMP syntax with CUDA, popular HIP and it's instructions supported on wide verity of devices across all platformsother):https://github.com/ROCm-Developer-Tools/aomp/blob/master/docs/openmp_terms.md]
== Programming GPUs with OpenMP ==
== Instructions for NVIDEA NVIDIA ==
'''How to set up the compiler and target offloading for Linux with a target NVIDIA GPU'''
$ mv cfe-7.0.0.src llvm-7.0.0.src/tools/clang
$ mv openmp-7.0.0.src llvm-7.0.0.src/projects/openmp
$ sudo usermod -a -G video $USER
</pre>
$ clang -fopenmp -fopenmp-targets=nvptx64 -O2 foo.c
</pre>
== Instructions for AMD==
How to set up compiler and target offloading for Linux on AMD GPU: (Elena)
Note: user should be member of 'video' group; if this doesn't help, may add user to 'render' group
[AOMP https://github.com/ROCm-Developer-Tools/aomp] is an open source Clang/LLVM based compiler with added support for the OpenMP® API on Radeon™ GPUs.
'''Hello world compilation example:'''
<pre>
// File helloWorld.c
#include <omp.h>
#include <stdio.h>
int main()
{
#pragma omp parallel
{
printf("Hello world!");
}
}
</pre>
Make sure to export your new AOMP to PATH
<pre>
export AOMP="/usr/lib/aomp"
export PATH=$AOMP/bin:$PATH
clang -fopenmp helloWorld.c -o helloWorld
./helloWorld
</pre>
'''Hello world on GPU example'''
<pre>
// File helloWorld.c
#include <omp.h>
#include <stdio.h>
int main(void)
{
#pragma omp target
#pragma omp parallel
printf("Hello world from GPU! THREAD %d\n", omp_get_thread_num());
}
</pre>
<pre>
export AOMP="/usr/lib/aomp"
export PATH=$AOMP/bin:$PATH
export LIBOMPTARGET_KERNEL_TRACE=1
clang -O2 -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target= Results and Graphs (Nathan/Elena) =amdgcn-amd-amdhsa -march=gfx803 helloWorld.c -0 helloWorld
./helloWorld
</pre>
To see the name of your device for (-march=gfx803) you may run 'rocminfo' tool:
<pre>
$ /opt/rocm/bin/rocminfo
</pre>
== Sources ==
