GitLab

Added proper diagonal rescaling and integrated it into contractions so they no longer need to be unpacked for preserved and summed SY groups.

Also made contraction unfold_broken_sym compare symmetries rather than check with respect to only NS

changed the way symmetrization and desymmetrization, mostly for summation.  Now the product can be desymmetrized and operands can be presymmetrized

Changed the way symmetrization/desymmetrization is done, so that SY desymmetrized tensor do not have data from the original SY tensor and an auxillary zero SY tensor is used to symmetrize before adding to the original SY tensor

1. Added capability (via constructor) to repack tensor data into a different packed symmetric layout without doing the normal symmetrization permutations that happen during sum
2. Fixed bug associated with C["ij"]=A["ij"]*B["ijkl"] where C and A are SY and B is fully NS.
3. Getting (2) to work in parallel required resolving a bug associated with desymmetrization (in the two unfold_broken_Sym functions), which were changing symmetry without changing sym_table, and as a result yielding extra mapping restrictions.

The bug fix to (3) might improve performance generally.

corrected support for user-defined types. Changed conversions from Pair<dtype> to char* to be alignment safe

Correct set() to scalar function for tensors, added appropriate test to scalar.cxx, and also fixed a new bug in tensor print()

Fixed a bug involving multipel reads/writes to/from the same (or symmetrically equivallent) index.  Added a bool that allows read_all to unpack tensor (off by default).

added limits include to set.h, changed configure check for C++11, and took out a zero_out_padding call that looks like it was left over from debugging

Added a clean nonsymmetric transpose and separate code into file src/redistribution/nosym_transp.cxx

Conflicts should have already been previously resolved in memeff_redist, so no significant changes made in this merge

Merge branch 'memeff_redist'

Conflicts:
	doc/doxygen.h
	include/ctf.hpp
	src/ctr_comm/strp_tsr.cxx
	src/ctr_seq/sym_seq_sum_ref.hxx
	src/dist_tensor/cyclopstf.hpp
	src/dist_tensor/dist_tensor_fold.cxx
	src/dist_tensor/dist_tensor_map.cxx
	src/dist_tensor/tensor_object.cxx
	src/interface/ctf_term.cxx

switched to redistribution via global ordering, and rewrote global ordering kernel to use instantiated loop structure, which improves performance

Got new assignment of elements to virtual blocks working single-threaded, required change ctr_2d_general, to send one block at a time.

Changed ordering of bucket_offsets, added phys_phase to distribution, made virtual blocks corresponds to tensor elements that are a physical phase away from each other