- Simplify group_indices
- Only for C ordered, remove complex compile time.
- Could be generalized to non C ordered, but no need.
- Fix slice for custom orders.
- Generalize the group_indices for the custom order.
- Add c_ordered_transposed_view (useful ?)
- Improve slice, special for ellipsis (quicker).
- Simplify TraversalOrder
- Assignement. Specialize one case for speed.
- use FORCEINLINE in foreach, according to speed test for clang
- add one speed test
- Modify iterators for better speed.
- along the lines decided for the foreach
- update doc.
- change the underlying data : do not flatten the linear indices of the
mesh into a single index, keep a higher dim array.
- easier for various places, and necessary for g(nu,nu').
- work on several 2 part. containers.
- add default target (may not be always matrix_valued)
Changed operator()(int, int...) && for array, and views.
- For const_view and regular type, returns value_type (i.e. a copy).
NB : does make a copy, not a move. Ok for scalar type. TODO: think for complicated types.
This allows codes like :
f(x)(0,0) where f : x-> matrix or const_view
to be correct in clef expression evaluation.
- For _view : return a value_type &, as before to allow :
A(....)(0,0) = rhs;
It is not possible to detect dangling refs in that case at compile time.
Added a security in TRIQS_ARRAYS_DEBUG mode to detect a dangling ref at run time,
i.e. the case where the view is "unique" (ref count ==1).
This would be a quite bad design anyway ...
- also :
- clean operator[] for vector (old workaround for old gcc...)
- add IsView flag in ISP impl class, for the impl. of operator() &&
- The previous version of the * operator for matrix was too clever.
It was giving a lazy object and then rewriting C = A *B into gemm (a,A,B,0,C).
The pb was in case of aliasing : when e.g. C = A, or is a part of A.
gemm is not correct that case, and as a result generic code like
a = a *b
may not be correct in matrix case, which is unacceptable.
- So we revert to a simple * operator for matrix
that does immediate computation.
Same thing for matrix* vector
- we also suppress a_x_ty class.
-> for M = a * b,
when M is a matrix, there is no overhead due to move assignment
-> however, when M is a view, there is an additionnal copy.
-Correctness comes first, hence the fix.
However, if one wants more speed and one can guarantee that
there is no aliasing possible, then one has to write a direct gemm call.
-> det_manip class was adapted, since in that case, we can show there
no alias, and we want the speed gain, so the * ops where replaced
by direct blas call (using the array blas interface).
-> also gemm, gemv, ger were overloaded in the case the return
matrix/vector (i.e. last parameter of the function) is not an lvalue,
but a temporary view created on the fly.
- has_view_type_tag mechanism was replaced by a decltype long ago.
Cleaning since it seems to work on all supported compiler now.
- removed add_views : not used any more.
- improve the mem_block and shared_block.
- the reference counting is now done in the mem_block and shared_block, removing the need of shared_ptr.
- speed tests shows that shared_ptr is very slow (due to thread safety?)
the new version is much better, though not perfect.
- Hence introducing weak views.
- also :
-- clean the guard mechanism for python (to allow returning from python without any python ref left).
-- clean code, add documentation for mem_block
-- remove nan init, which was not working, and corresponding test
-- serialisation of view still unchanged (need to forbid serialization of view ??).
- tests ok, incl. valgrind tests.
