[TASK][JNI] Investigate train of null_count after explode

[abellina]

While analyzing an nsys trace for a Spark job with deeply nested tables, we see an explode kernel call that is followed by a train of null_count, which end in is_valid.

After we call cudf::explode we build up a table, and construct java ColumnVector objects. I think the construction of these objects is triggering it.

This task is to confirm that the columns with missing a null count are coming from the explode kernels. If they are coming from explode, it would be great if explode could compute null count as part of that kernel.

In this screenshot, it is the ~20ms at the end after explode:

[GregoryKimball]

I'd like to cross-reference this issue with #11968. It's likely that the null_count appearances in profiles will change as we refactor null_count for compatibility with user-provided streams.

[jrhemstad]

it would be great if explode could compute null count as part of that kernel.

I'm not following. explode returns a table:

cudf/cpp/include/cudf/lists/explode.hpp

Line 72 in 7d173c9

std::unique_ptr<table> explode(

Are you then constructing a column_view for each of the lists that were exploded?

If so, then yeah, you're going to have a problem with computing the null count of each of those column_views individually.

To make that efficient, you'd have to do what we do in cudf::split where we compute the individual null counts in bulk with a single segmented_null_count:

cudf/cpp/include/cudf/detail/null_mask.hpp

Lines 186 to 204 in 9c06330

	/**
	* @brief Given a validity bitmask, counts the number of null elements (unset
	* bits) in every range `[indices[2*i], indices[(2*i)+1])` (where 0 <= i <
	* indices.size() / 2).
	*
	* If `bitmask == nullptr`, all elements are assumed to be valid and a vector of
	* length `indices.size()` containing all zeros is returned.
	*
	* @throws cudf::logic_error if `indices.size() % 2 != 0`
	* @throws cudf::logic_error if `indices[2*i] < 0 or indices[2*i] > indices[(2*i)+1]`
	*
	* @param[in] bitmask Validity bitmask residing in device memory.
	* @param[in] indices A host_span of indices specifying ranges to count the number of null elements.
	* @param[in] stream CUDA stream used for device memory operations and kernel launches.
	* @return A vector storing the number of null elements in each specified range.
	*/
	std::vector<size_type> segmented_null_count(bitmask_type const* bitmask,
	host_span<size_type const> indices,
	rmm::cuda_stream_view stream);