[BUG] String column copy performs a full gather

[jlowe]

Describe the bug
While analyzing an Nsight trace of a Spark query, I noticed 10ms being spent on a filter. Digging deeper, the filter time was spent in a cudf::table copy constructor, with most of the time being spent copying one string column within the table.

Spark queries often filter input data to remove nulls before further processing, and often nothing is filtered. cudf::apply_boolean_mask, used to implement the Spark filter, has short-circuit logic to avoid a full gather and instead copy-constructs the output table if it realizes nothing will be filtered. However copy-constructing a string column performs a full gather which can be much more expensive than simply copying the input column buffers.

In this specific case, the string column consists of 178200 rows and each row is around 200 characters. The string copy constructor took almost 10 milliseconds on the GPU despite the column containing approximately 30MB of device memory.

Steps/Code to reproduce bug
Perform a copy of a string column where there are many strings that are relatively long (e.g.: 200+ characters per row). I've attached a gzipped Parquet file containing the string column from the specific filter case mentioned above for reference.

filtertest.parquet.gz

Expected behavior
A copy of a strings column view when the view starts at base offset 0 should be performed at device memory speed, copy the underlying buffers rather than performing the more complicated gather computation (which also requires synchronizing and a DtoH transfer).

[jrhemstad]

In addition, even when this operation does need to perform a gather it is inefficient as it materializes the gather map:

cudf/cpp/src/strings/copying/copying.cu

Lines 48 to 60 in 89b802e

	rmm::device_vector<size_type> indices(strings_count);
	thrust::sequence(execpol->on(stream), indices.begin(), indices.end(), start, step);
	// create a column_view as a wrapper of these indices
	column_view indices_view(
	data_type{type_id::INT32}, strings_count, indices.data().get(), nullptr, 0);
	// build a new strings column from the indices
	auto sliced_table = cudf::detail::gather(table_view{{strings.parent()}},
	indices_view,
	cudf::detail::out_of_bounds_policy::NULLIFY,
	cudf::detail::negative_index_policy::NOT_ALLOWED,
	mr,
	stream)
	->release();

This should just use a counting iterator passed to the detail version of gather that uses an iterator for the map.