Repository navigation
Conversation
d4c4686 to
3b0658d
Compare
3b0658d to
788525e
Compare
|
cscs-ci run default |
1 similar comment
|
cscs-ci run default |
1b31304 to
d72f949
Compare
|
cscs-ci run default |
1 similar comment
|
cscs-ci run default |
|
The failure in the RoCM CI ( |
|
cscs-ci run default |
2 similar comments
|
cscs-ci run default |
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
8c085c6 to
02259e7
Compare
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
edopao
changed the title
|
cscs-ci run default |
|
cscs-ci run dace-determinism |
|
cscs-ci run default |
philip-paul-mueller
left a comment
philip-paul-mueller
left a comment
There was a problem hiding this comment.
It generally looks okay, but there are some things that must be changed.
| if isinstance(arg, IteratorExpr) and use_full_shape: | ||
| if isinstance(arg, IteratorExpr): | ||
| return IteratorExpr(inner_node, arg.gt_dtype, arg.field_domain, arg.indices) | ||
| elif isinstance(arg_expr, MemletExpr): |
There was a problem hiding this comment.
I am just wondering why this case was not here before.
| If a transformation connects different data to the connector, it has to restate | ||
| the connector itself, see `restate_nested_sdfg_connector()`, because only the | ||
| transformation knows where the old data is located inside the new one. |
There was a problem hiding this comment.
Maybe I am too tired, but this sounds strange to me.
| passed_symbols.update( | ||
| str(sym) for off in ss_offset if dace_sym.issymbolic(off) for sym in off.free_symbols | ||
| ) | ||
| for sym in sorted(passed_symbols): |
There was a problem hiding this comment.
I looked at the dace source code and to me it looks like as symbol renaming is performed upon validation.
So this whole thing here is technically not needed or do I miss something.
| # A nested SDFG connector has to be equivalent to the data it is connected to. | ||
| # Thus `i0` is still `a`, but `o0` has been restated as `e`, into which the | ||
| # nested SDFG now writes directly, at the offset of the removed copy chain. |
There was a problem hiding this comment.
I think this sentence is used a bit too much here.
| SDFG has to be equivalent to the data it is connected to, so it cannot write | ||
| into an element of the outer array directly. This is also how the lowering | ||
| writes the result of an if-expression. |
There was a problem hiding this comment.
Yes this sentences is written way too much here all over the place.
| assert str(nsdfg.symbol_mapping[exp_stride]) == exp_stride | ||
|
|
||
| # Now we propagate `a` and `b`, but not `c`. | ||
| # Now we change the strides of `a1` and propagate them, but not the ones of `c1`. |
There was a problem hiding this comment.
Why is b1 no longer modified?
|
cscs-ci run default |
Update
daceto v2.0-beta1, which introduces the concept of no-view nested SDFG: the data descriptor for global arrays inside a nested SDFG have to be the same as the outer data descriptors.ICON-blueline performance comparison w.r.t. main branch:
