murez / auto_MPI_nonblock

just a naïve idea

auto MPI nonblocking

a very naive idea

Run

mkdir build && cd build
cmake ..
make
cd ..
mpicc -S main.cpp -emit-llvm
opt -f -load-pass-plugin=./build/libReplaceMPIColl.so -passes=replace-mpi-coll -S  main.ll -o m.ll

TODO

1. finish all the collective MPI calls
2. try get data flow and rearrange IR order
3. find stencil pattern and acheive inner & outer seperate computation.