high performance computing on graphics processing units: hgpu.org

Code parallelizers are employed these days to reduce the efforts needed in manually parallelizing sequential code. But they are ineffective when it comes to handling programming constructs like pointers. Code parallelizers like Par4all have a limited support for pointers while approaches like the ASET + BONES cannot handle pointers at all. In this thesis we have developed a pointer analysis infrastructure to enable pointer support for the ASET + BONES approach. Our pointer analysis infrastructure is based upon LLVM’s compiler infrastructure and relies on an indigenously developed flow insensitive, context sensitive analysis pass called ex-ptrinfo. The pass is designed to perform static analysis of source code and extract required pointer analysis information regarding pointer constructs that have been employed for performing memory accesses in PAINt and the Data Path algorithms. Our results show that the developed pointer analysis infrastructure can correctly identify such pointer constructs and extracts the required pointer analysis information which can be used for parallelizing the PAINt and the Data Path algorithms with the ASET + BONES approach. At the moment the ASET + BONES approach is under development and has several limitations. In future we intend to overcome these limitations and apply the ASET + BONES approach for parallelizing real world code.