spcl dace issues - Githubissues

spcl / dace

DaCe - Data Centric Parallel Programming

http://dace.is/fast

BSD 3-Clause "New" or "Revised" License

499 stars 129 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

[Draft] Interstate assignment to map

#1790 ThrudPrimrose opened 23 hours ago
0
[Multi SDFG, Fortran] Drop the "not implemented error" for the `ArrayLoopNodeLister`, and we have support for multiple ranges. Also, rewrote the logic to be more concise and readable.

#1789 pratyai closed 5 hours ago
0
Symbol redeclaration fix

#1788 ThrudPrimrose opened 23 hours ago
0
[Multi SDFG, Fortran] Fix `intrinsic_blas_test.py`: The `simplify()` pass is broken in `multi_sdfg` branch (probably because it didn't get the bug fixes from `main` branch). Most tests (if any) should not rely on the `simplify()` pass anyway.

#1787 pratyai closed 1 day ago
1
[Multi SDFG, Fortran] Drop one of obsolete/redundant tests (that are failing currently).

#1786 pratyai closed 1 day ago
0
[Multi SDFG, Fortran] Fix `parent_test.py`: The first test was already passing, we just applied the "modern treatment". The second test was failing, because variable declarations in the module are now populated on a different member than before.

#1785 pratyai closed 1 day ago
0
[Multi SDFG, Fortran, After PR#1783] Fix `fortran_language_test.py`. Only one test was failing really, but we apply the same treatment to all of them anyway. The failing one was solved by just avoiding the `transient_mode = False` trick and picking the entry point explicitly (and dropping the `program` that should not be needed in the first place).

#1784 pratyai closed 1 day ago
0
[Multi SDFG, Fortran] Fix the `empty_test.py` by: 1) passing a valid test code, 2) removing redundant parts, 3) using a fresh list of functions everytime we need it, 4) avoiding the `transient_mode = False` trick, and 5) ensuring that some internal classes are initialized correctly.

#1783 pratyai closed 1 day ago
0
[Multi SDFG, Fortran] Testing `allocate()` requires Deferred Allocation support on DaCe, which we do not have yet. So, we skip the test for now.

#1782 pratyai closed 1 day ago
0
[Multi SDFG, Fortran] Fix the array attribute tests: 1) Make the test Fortran codes `gfortran -Wall` certified, 2) Remove the unnecessary main program from them since we can set arbitrary entry points now, 3) Pass the extra arguments for array sizes that should have been there in the first place

#1781 pratyai closed 1 day ago
0
[Multi SDFG, Fortran, After PR#1779] Parse the internal subprograms of subprograms correctly. Currently, we do not parse them correctly.

#1780 pratyai closed 1 day ago
0
[Multi SDFG, Fortran] Format the python code in `dace/frontend/fortran` so that the future PRs are not so noisy with formatting diffs.

#1779 pratyai closed 1 day ago
0
Map over trivial tasklet

#1778 ThrudPrimrose opened 2 days ago
3
[Multi SDFG, Fortran, After PR#1774] New tests. In particular, now we can start testing the construction of internal SDFG.

#1777 pratyai closed 2 days ago
0
[Multi SDFG, Fortran] 1) Account for `dep_graph.nodes.get(dep).get('info_list') is None`, since importing _everything_ from an empty module implies just that. 2) Keep a track of all the identifiers from the scope of the "use-site" of a module, then intersect with that list during pruning.

#1776 pratyai closed 2 days ago
0
Size in `new[]` explicitly casted to `size_t`

#1775 philip-paul-mueller opened 3 days ago
0
[Multi SDFG, Fortran] Make it so that after `own_ast, program = create_internal_ast()`, we need to use as little information as possible from `own_ast`.

#1774 pratyai closed 2 days ago
0
Invalid CUDA code: host computation on device buffer after GPU transformations

#1773 edopao opened 3 days ago
0
Iteration Order of `CopyToMap`

#1772 philip-paul-mueller opened 4 days ago
0
Arguments to GPU Kernel Not Correctly Detected

#1771 philip-paul-mueller opened 4 days ago
0
[Multi SDFG, Fortran] Make `functionStatementEliminator()` to modify the `node` itself and return it, so that it cannot forget to account for all the members of `node` when constructing a new one.

#1769 pratyai closed 2 days ago
0
Remove deprecated features

#1767 tbennun opened 6 days ago
0
Make scalar to symbol promotion robust to node order in state

#1766 tbennun closed 6 days ago
0
[Multi SDFG, Fortran] Extract the "internal AST construction -> SDFG construction" into a separate function, that also supports multiple entry points.

#1765 pratyai closed 4 days ago
1
[Multi SDFG, Fortran] Add a test case where "a module contains a subroutine contains another subroutine" etc.

#1764 pratyai closed 6 days ago
0
Redundant Array Removal is too Strict with Storage

#1763 philip-paul-mueller opened 1 week ago
1
[Multi SDFG, Fortran] Remove the partial internal AST creation in topological order, since at this point (i.e., after the recent fixes), we should have full compilable AST after pruning anyway.

#1762 pratyai closed 6 days ago
0
[Multi SDFG branch, Fortran frontend] The `get_used_modules` were not handling multiple invocations of `use lib ...` correctly.

#1761 pratyai closed 1 week ago
0
`list_of_module_vars` is not actually a string list of the variables, but a structure that needs to be unfolded to get the list.

#1760 pratyai closed 1 week ago
0
Fixed `dace::math::pi` and `dace::math::nan` on GPU

#1759 philip-paul-mueller closed 1 week ago
0
[Multi SDFG branch, Fortran frontend] Change the interface of `recursive_ast_improver()` to hide the implementation-specific objects that helps the recursion, by defining a local function to manage the states involved in recursion internally.

#1758 pratyai closed 1 week ago
0
Fix scalar to symbol promotion exception when trying to promote same scalar again

#1757 phschaad closed 1 week ago
0
Fix `pystr_to_symbolic` not correctly interpreting constants as boolean values in boolean comparisons

#1756 phschaad closed 1 week ago
0
[Multi SDFG branch, Fortran frontend] Make `eliminate_dependencies()` support `use lib` without `only`.

#1755 pratyai closed 1 week ago
0
[Multi SDFG branch, Fortran frontend] Take another piece out of `create_sdfg_from_string()` and test it.

#1754 pratyai closed 1 week ago
0
[Multi SDFG branch, Fortran frontend] These tests relied on `normalize_offset=False` which has changed, since `True` is the default now.

#1753 pratyai closed 1 week ago
0
[Multi SDFG branch, Fortran frontend] Several `par_Decl_Range_Finder()` were missing the `structure` argument (therefore passing a wrong boolean argument in its place).

#1752 pratyai closed 1 week ago
3
[Multi SDFG branch, Fortran frontend] Set `normalize_offset=True` for the handful of tests that passes afterward.

#1751 pratyai closed 1 week ago
0
[Multi SDFG Branch, Fortran Frontend] Extract a piece of code block out of `create_sdfg_from_string()`, that simplifies the dependency graph found from `recursive_ast_improver()`.

#1750 pratyai closed 1 week ago
0
Tracking issue for Fortran frontend test failures on `multi_sdfg` branch.

#1749 pratyai opened 1 week ago
3
Fixed `dace::math::pow` for Integer case

#1748 philip-paul-mueller opened 1 week ago
6
Fixing `SDFGState._read_and_write_sets()`

#1747 philip-paul-mueller closed 1 week ago
0
Fix OpenMP dynamic loop bounds that use persistent memory

#1746 tbennun closed 1 week ago
0
Fix temporary transient counter during Python parsing of nested calls

#1745 tbennun closed 1 week ago
0
Instead of compiler specific align hint use C++ standard aligned alloc

#1744 ThrudPrimrose closed 2 days ago
5
Fixed MapExpansion Transformation

#1743 Berke-Ates opened 1 week ago
0
Unskip unit tests and provide reasons for skipped tests

#1742 tbennun closed 1 week ago
0
Do not rename generated code files

#1741 tbennun closed 2 weeks ago
1
SDFG API additions for version 1.0

#1740 tbennun closed 1 week ago
0
Memlets propagated incorrectly out of nested SDFGs

#1739 phschaad opened 2 weeks ago
0