[REVIEW] Make documentation uniform for stream and memory resource params

[karthikeyann]

Update documentation common params such as mr, stream.

• param mr device memory resource (taken input from "culture of quality" series)
• param stream CUDA stream

[GPUtester]

Please update the changelog in order to start CI tests.

View the gpuCI docs here.

[devavret]

Can you base this on #5069. Much of detail documentation has been removed

[karthikeyann]

cudf::strings::all_float(), cudf::strings::all_integer() accepts device memory resource in arguments. but it is not used inside.

cudf/cpp/include/cudf/strings/char_types/char_types.hpp

Lines 180 to 181 in 839bed5

	bool all_float(strings_column_view const& strings,
	rmm::mr::device_memory_resource* mr = rmm::mr::get_default_resource());

cudf/cpp/include/cudf/strings/char_types/char_types.hpp

Lines 140 to 141 in 839bed5

	bool all_integer(strings_column_view const& strings,
	rmm::mr::device_memory_resource* mr = rmm::mr::get_default_resource());

@davidwendt should this argument be removed?

[davidwendt]

cudf::strings::all_float(), cudf::strings::all_integer() accepts device memory resource in arguments. but it is not used inside.

cudf/cpp/include/cudf/strings/char_types/char_types.hpp

Lines 180 to 181 in 839bed5

	bool all_float(strings_column_view const& strings,
	rmm::mr::device_memory_resource* mr = rmm::mr::get_default_resource());

cudf/cpp/include/cudf/strings/char_types/char_types.hpp

Lines 140 to 141 in 839bed5

	bool all_integer(strings_column_view const& strings,
	rmm::mr::device_memory_resource* mr = rmm::mr::get_default_resource());

@davidwendt should this argument be removed?

Yes, this argument should be removed.

[karthikeyann]

AFAIK, writers don't need device memory resource.
rmm::mr::device_memory_resource _mr is not used in writer::impl of

cudf/cpp/src/io/orc/writer_impl.cu

Line 964 in 4bf064e

_mr(mr)

cudf/cpp/src/io/parquet/writer_impl.cu

Line 433 in 4bf064e

: _mr(mr),

  // TODO : figure out if we want to keep this. It is currently unused.
  rmm::mr::device_memory_resource* _mr = nullptr;

@nvdbaranec @OlivierNV should this argument and private data member (TODO) be removed?

[davidwendt]

Please don't merge this until the other reviewers have approved.

[rgsl888prabhu]

LGTM, minor changes and questions.

[jrhemstad]

Thanks for going through and standardizing all of these. Since we're going through that effort, we should make sure to get it 100% right.

As such, I see a lot of cases where it says "Device memory resource used to allocate the returned column". This isn't quite right. The resource isn't used to allocate the column object, it's used to allocate the device memory contained inside the column. So a more accurate description would be to say: "Device memory resource used to allocate the returned column's device memory."

[OlivierNV]

@nvdbaranec @OlivierNV should this argument and private data member (TODO) be removed?

We've talked about this: the mr is currently unused by io writers, but it could possibly be needed in the future if returning output data in device memory rather than on disk (ok with me either way).