Change split functions to work on box instead of chunk

[psalz]

chunk is a purely user-facing type (and of questionable usefulness at that), and since all of our internal tracking is built around box, splitting should follow suit.

[github-actions]

Check-perf-impact results: (000e2892abb21ddd1ae413a5c86d7d95)

❓ No new benchmark data submitted. ❓
Please re-run the microbenchmarks and include the results if your commit could potentially affect performance.

[github-actions]

⚠️ Clang-Tidy found issue(s) with the introduced code (1/1)

[github-actions]

⚠️ readability-inconsistent-declaration-parameter-name ⚠️
function celerity::detail::split_1d has a definition with different parameter names

[github-actions]

⚠️ readability-inconsistent-declaration-parameter-name ⚠️
function celerity::detail::split_2d has a definition with different parameter names

[coveralls]

Pull Request Test Coverage Report for Build 12434960318

Details

• 47 of 47 (100.0%) changed or added relevant lines in 3 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.005%) to 95.055%

---

Totals
Change from base Build 12390656669:	0.005%
Covered Lines:	7098
Relevant Lines:	7203

---

💛 - Coveralls

[GagaLP]

Nice

[fknorr]

I like it!

If I'm not mistaken you could also define command_graph_generator::assigned_chunk in terms of boxes and get rid of even more casts without much hassle - buffer_access_map already operates on boxes.

Also, a drive-by suggestion:

diff --git a/src/command_graph_generator.cc b/src/command_graph_generator.cc
index da56682e..3cfb6d22 100644
--- a/src/command_graph_generator.cc
+++ b/src/command_graph_generator.cc
@@ -117,11 +117,9 @@ std::vector<const command*> command_graph_generator::build_task(const task& tsk)
 }

 void command_graph_generator::report_overlapping_writes(const task& tsk, const box_vector<3>& local_chunks) const {
-       const chunk<3> full_chunk{tsk.get_global_offset(), tsk.get_global_size(), tsk.get_global_size()};
-
        // Since this check is run distributed on every node, we avoid quadratic behavior by only checking for conflicts between all local chunks and the
        // region-union of remote chunks. This way, every conflict will be reported by at least one node.
-       const box<3> global_chunk(subrange(full_chunk.offset, full_chunk.range));
+       const box<3> global_chunk(subrange(tsk.get_global_offset(), tsk.get_global_size()));
        auto remote_chunks = region_difference(global_chunk, region(box_vector<3>(local_chunks))).into_boxes();

        // detect_overlapping_writes takes a single box_vector, so we concatenate local and global chunks (the order does not matter)