GitLab

modifications and corrections updating gpu capabilities. adjustment to interface for matrices, whose symmetry member is now named symm to distinguish from symmetry array sym

Added unnecessary 'leave from home' (copy) being done when taking care of scaling diagonals. Should improve performance, especially sequentially. Incremented version to 1.33

Lowered dimension of btwn_central code in test_suite to make it work faster

permitted handling of userdefined elementwise functions with non-identity alpha, since this is needed when scaling for symmetric permutations

Made ASSERT() print error and try to print stack trace from processor 0. A write to NULL is still done by all processors so that errors can be examined via valgrind. This may mean that the error proc 0 is supposed to print does not get printed due to interrupt from another processor failing

minor corrections in line with clang warnings and change from _C_ back to _CXX_ for MPI collective datatypes

Added a check for std::function parameter matching support, which seems to be broken on Edison's Intel compiler but is fixed by loading gcc, as the configure script automatically suggests.

still need to think about whether this is actually correct and why

switche to using MPI_LONG_DOUBLE as generic 16-byte type and switche to MPI_C_COMPLEX, etc. for default complex MPI types

Added -std=c++0x by default for Apple

put fopenmp in CXXFLAGS vs LDFLAGS, fixed MPI reduction ops in apsp/btwn_central, betweenness centrality now works

Corrected bugs in slicing

and work deep in progress for betweenness centrality

Conflicts:
	examples/btwn_central.cxx

Also made norms (except for the 2-norm) operate on the default ring rather than the specified algebraic structure

Corrections to some elementwise function functionality and new examples, one for custom bivariate functions and a crazy one for bitonic sorting