GitLab

Adapt library and examples for 64-bit integer option (USE_INT64=ON); adapt library (but not yet examples) for 32-bit float option (USE_FP64=OFF); make gcc happy on library compilation  with -Wall -Wextra -pedantic; replace malloc/realloc/free with new/delete in container.hpp

pool_array no longer depend on allocation/deallocation order; unit test on pool_array; some fancy functions to retrive derivatives for variable/hessian_variable

It messed up adgprs compillation

Fix compilation issue with std::isinf

Finish memory_pool implementation in container.h

Use of matrices allocated in memory pool for intermediate results in
dense linear algebra operations in inmost_dense.h

Use of memory pool for matrices in inmost_autodiff.h

Change error reporting and return type in Invert, Solve,
CholeskyInvert, CholeskySolve, PseudoInvert, PseudoSolve

Add unit tests for linear algebra on dense matrices.

Additional updates to template handling in AbstractEntry::Access, fix a bug in thread_private class with openmp

thread_private class for static OpenMP objects acts as iterator

EndModification do not unnecessarily delete orphan elements

Fixed Face::getNodes behaviour inside BeginModification and
EndModification region (same fix as in andre_brench)

Added template access functions for AbstractEntry class, allows to get
(matrix of) values, unknown indices or unknowns themselves depending on
template parameter

Examples/OldDrawGrid:
* Adjustments for svg output of color gradient in color bar (previously
was not properly working in Inkscape)
* Preparations for isosurface extraction code

otherwise adgprs does not compile with inmost under vs2017

Added linked_array structure that represents linked list with subarray of constant size. This elemenates possible crash with openmp when mesh is simultaneously modified and data is accessed. Previously when array of links was reallocated in chunk_array due to mesh growth, the pointer may be simultaneously changed due to reallocation and be accessed by another thread. Now reallocation adds another segment of the array in linked list making previous records intact. To restore previous behavior undefine NEW_CHUNKS in container.hpp.

Many updates to GRDECL file format:
- Processing more properties fields, such as pressure, sgas, soil, etc.
- Processing MULTIPLY keyword for properties and transmissibilities
- Introduced field TRANM for transmissibility multipliers set with TRANX/TRANY/TRANZ keywords in MULTIPLY keyword, or with EDITNNC keywords or FAULTS/MULTFLT keyword
- Algorithm that separates faces in place of blocks with zero volume and nonzero ACTNUM. Option ECL_DEGENERATE that replaces algorithm with transmissibility multiplier in TRANM.
-Optimization: bypass algorithm that resolves faces out of set of edges when only one face is expected.
- Processing of NTG keyword.

Fix swap algorithm in dynarray container.

Loading of 2d and mixed 2d/3d grids from VTU files.

Fix Visual Studio warnings in sparse.cpp

Creating elements cause reallocation of internal array, pointing to blocks of data. Other processor may use deallocated pointer to the array stored in cache.

Remove dependence on mesh in Automatizator

Fix few warnings on llvm-g++ in container.hpp

Add converter from elasticity tensor into matrix

Added separate hessian_variable that enables hessian calculation

Incorrect order of parameters to UnpackTagData called from
UnpackElementsData.

OldDrawGrid example: mesh slicing algorithm can now correctly resolve
multiple slices on single cell.

USE_AUTODIFF do not require USE_SOLVER and USE_MESH:
If USE_MESH is not activated, class Automatizator and algebra for
abstract variables is not compiled.
If USE_SOLVER is not activated, class Residual is not compiled.

Fixed reverse_iterator::operator- in ElementArray.

class incident_matrix that computes closed loops out of set of edges or
faces can now work on coordinates other then physical coordinates.

Fixed Face::SplitFace() was not connecting newly created faces to cells.

Cell::Volume() now uses algorithm working correctly on non-convex cells.

Face::CheckNormalOrientation() will now use volume calculation
algorithm for non-convex cells to correctly orient faces.

Face::FixNormalOrientation() will now use Face::SwapCells() instead of
Face::ReorderEdges() on internal faces. New version is thread-safe with
respect to Face::CheckNormalOrientation().

Improved algorithm for loading eclipse meshes.

Improvements for eclipse grid reader.

Fixed a bug in ElementArray::iterator.

Including issue #15

Cleaned bugs; added features.

Fixes for gcc and intel compilers.
gcc-4.1 clean with -Wall -Wextra
intel 11.1 clean with -Wcheck -w1
msvc2010 gives one unrelated warning