1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
|
TILEDARRAY REVISION HISTORY
Following is a brief summary of changes made in each release of TiledArray.
- 16-November-2020: 1.0.0
- resolved issue 77: negative indices are supported, use signed 1-index type by default (this brings TA::Range
in sync with btas::RangeNd), to revert to legacy behavior configure with TA_SIGNED_1INDEX_TYPE=OFF (PR #214)
- removed interface to Elemental (PR #201)
- added Python interface (PR #198 and #212)
- added interface to ScaLAPACK (PR #173)
- added ability to fuse vector<DistArray> -> DistArray and extract subarray from the fused array (PR #160 and #162)
- resolved boost check issue (PR #161)
- revamped TA::foreach and improved conversions to be able to handle non-standard policies
- C++17 is now prerequisite
- CUDA 11 is now prerequisite for execution on NVIDIA GPUs; CMake 3.18 is required if CUDA is enabled
- 07-June-2019: 1.0.0-alpha.2
- modernized CMake handling of CUDA, CMake 3.10 is now required
- improved CMake handling of MADNESS arguments, e.g. gperftools
- Pmaps no longer have to store O(N) data, and no existing Pmap does
- 09-May-2019: 1.0.0-alpha.1
- first-class support for CUDA; requires Umpire [github.com/LLNL/Umpire] and cuTT [github.com/ValeevGroup/cutt
fork of github.com/ap-hynninen/cutt], both can be installed automatically
- better usability with BTAS tensors and hierarchical tiles (i.e. tensors of tensors)
- disabled automatic conversion of multiply expressions into scalars
- updates to support C++17 (but only C++14 is required)
- resolved issues 145, 157
- bump to CMake 3.8
- 15-Feb-2018: 0.7.0
- MADNESS bump to 0.11.0, requires C++14 as the result
- added interface to latest Elemental ('El')
- resolved issues 96, 100, 101, 102, 104, 106, 107, 108, 110, 124
- significant API changes:
- mixed-expressions branch is incorporated, it makes possible to implement operations on arrays of different types,
e.g. sparse + dense = dense, etc. It also supports asynchronous tile conversion operators (i.e. returning Future<Tile>
rather than Tile) to support expensive/communicating conversions.
- added expr.set_{world,shape,pmap}() to override the computed defaults; key application is to
impose the sparsity on the result of an expression
- TiledRange1::element2tile -> TiledRange1::element_to_tile for consistency with TiledRange
- ArrayIiterator -> ArrayIterator
- BTAS is a required dependence, automatically installed if missing
- 27-Nov-2016: 0.6.0
- block expression code
- DSL extensions to "force" shape, world, and pmap
- minor API improvements and bug fixes.
|
