Currently, a sequential implementation along with examples and
benchmark codes is available. Required are a C++ compiler and the Gnu
m4 macro processor. The programming model provides an implementation
of grids (arrays) and k-ary trees. This demonstrates the programming
model and can be used for program development and debugging. No
adavanced compilers are needed.
Further, some parallel implementations for pthreads [Unix, Windows],
SIMD vector [Pentium/AMD SSE, PowerPC AltiVec], Cell BE [Sony
Playstation3, IBM Blade Server], and Cuda [Nvidia GeForce GPU] are available. This is the code transformation
system . An additional data dependence analysis tool based on Gnu g++
is needed, which is also be provided here. Note that the programming
model is in fact compatible to the sequential model available. In
addition to the Gnu m4 macro processor and perl, a special compiler is
needed, which can be downloaded separately.
The dependence analysis tool is based on Gnu g++ 4 (4.1, 4.2, 4.3). It is provided
as a patch to the standard source distribution. This tool is a
necessary part of the parallel code generation.
Coming soon: An update of the code transformation system for MPI parallel code.