install Parallel-For


sequential
Download the package, un-tar and build and run it. Maybe the location or name of compilers and tools in the makefile need to be changed.
tar xvfz sequential_?.?.tar.gz
cd sequential_?.?
make
./fmm.exe

parallel
Download the package, un-tar and build it. Maybe the location or name of compilers and tools in the makefile need to be changed. The modified compiler needs to be built in advance. Put its path into the makefile. Running the executables depends on the parallel programming model.
Examples: Run the code with 4 threads ./mg.T -t4 ,
run the vectorized code ./mg.SV.
tar xvfz parallel?.?.tar.gz
cd parallel?.?
make

cell
Download the package, un-tar and build it. Maybe the location or name of compilers (gnu g++ or IBM xlC) and tools in the makefile need to be changed. The modified compiler needs to be built in advance. Put its path into the makefile. Run the code on all SPUs ./mg.PPU ,
run the vectorized code ./mg.PPUV.
tar xvfz cell_?.?.tar.gz
cd cell_?.?
make

cuda
Download the package, un-tar and build it. Maybe the location of the Nvidia nvcc compiler (based on gnu g++ or Microsoft Visual C++) and tools in the makefile need to be changed. Use the `-deviceemu' flag to use the Cuda emulator instead of the GPU. The modified parallel-for compiler needs to be built in advance. Put its path into the makefile. Run the code ./mg.CUDA .
tar xvfz cuda_?.?.tar.gz
cd cuda_?.?
make

compiler
Download the Gnu g++ source distribution [core and c++] and un-tar it. Apply the patch file. Configure and build it. This will take some time. Do not use this compiler as default.
tar jxvf gcc-core-4.*.tar.bz2
tar jxvf gcc-g++-4.*.tar.bz2
tar zxvf compiler_?.?.tar.gz
cd gcc-4.*/gcc
patch -p1 < ../../gcc-4.*-patch
cd ../..

mkdir obj
cd obj
../gcc-4.*/configure --enable-languages=c,c++
make bootstrap-lean

cd gcc
mkdir ../../example
cp g++ xgcc ../../example
or
make install
 or whereever you want to put the compiler.
[start] [example] [references] [download]