refactor[next][dace]: cleanup dace backend module #1811
Conversation
…-gtir-refact-module
…-gtir-refact-module
| def debug_info( | ||
| node: gtir.Node, *, default: Optional[dace.dtypes.DebugInfo] = None | ||
| ) -> Optional[dace.dtypes.DebugInfo]: | ||
| """Include the GT4Py node location as debug information in the corresponding SDFG nodes.""" | ||
| location = node.location | ||
| if location: | ||
| return dace.dtypes.DebugInfo( | ||
| start_line=location.line, | ||
| start_column=location.column if location.column else 0, | ||
| end_line=location.end_line if location.end_line else -1, | ||
| end_column=location.end_column if location.end_column else 0, | ||
| filename=location.filename, | ||
| ) | ||
| return default | ||
|
|
||
|
|
There was a problem hiding this comment.
I can not find the new location of this function.
Also it is still referenced by other parts of the code.
There was a problem hiding this comment.
Good catch, I silently moved it to gtir_sdfg_utils so I could remove the dependency on gtir in the utils module.
There was a problem hiding this comment.
Okay.
I just could not find it that was all.
But I noticed that for some reason Ctrl+F is a bit unreliable, but I do not understand why.
But I very much like your reason.
Sorry for the false alarm.
| @@ -0,0 +1,227 @@ | |||
| # GT4Py - GridTools Framework | |||
There was a problem hiding this comment.
The transformations subpackage has now tow files with the exact same content, once utils.py (which is new) and once util.py (which was there before).
I actually do not care which one survives and in fact I would say that utils.py should survive.
There was a problem hiding this comment.
Good catch. This must be an artifact of VSCode refactoring.
This PR fixes a bug in `DistributedBufferRelocator` that was observed in ICON4Py's `TestUpdateThetaAndExner` test. In essence there was an `assert` that assumed that checked if this temporary was a sink node, but, the code that finds all write backs was never excluding such cases, i.e. the temporaries that were selected might not be sink nodes in the state where they are defined. The `assert` was not part of the original implementation and is not a requirement of the transformation, instead it was introduced by [PR#1799](GridTools#1799), that fixed some issues in the analysis of read write dependencies. There are two solutions for this, either removing the `assert` or prune these kinds of temporaries. After some consideration, it was realized that handling such cases will not lead to invalid SDFG, as long as the other restrictions on the global data are respected. For that reason the `assert` was removed. However, we should thinking of doing something more intelligent in that case.
PR #1753 removed the
dace_iteratorbackend, that before existed by side ofdace_fieldviewbackend. Code common to these two backends was placed indace_commonmodule.The current GTIR-DaCe backend supports both iterator view and field view flavors of the GTIR, so there is no need for a distinction at the top level of the backend module.
This PR applies the following cleanup tasks:
The module
runners/dace/workflowwas also split into sub-modules. A doc string was added torunners/dace/workflow/__init__.py