Vukasin Milovanovic

Fix `is_device_write_preferred` in `void_sink` and `user_sink_wrapper`

1

## Description Addresses a few issues in `data_sink` classes to avoid D2H copies in writers when using a `void_sink`. Provide an `is_device_write_preferred` implementation to always prefer device writes. Implement `is_device_write_preferred`...

Align integral types in ORC to specs

2

## Description Use `uint64_t` where specified by the ORC specs: - `PostScript::compressionBlockSize` - `StripeInformation::footerLength` - `StripeInformation::numberOfRows` Using the same type for the derived values. Other changes: - Changed the num_rows...

[C++] Order of row index streams does not match the order of streams in the file footer

10

When writing a file with a string column and multiple row groups, the resulting file has incorrect row index streams. The string column is encoded using direct encoding. The file...

Once https://github.com/rapidsai/cudf/issues/15612 is implemented, libcudf can make further use of the pool to optimize data transfers. Types like `device_scalar` can use a small pinned buffer, since pool makes the allocation...

[FEA] Have a global pinned memory pool by default

2

Users outside of Spark-RAPIDS still use the default, non-pooled, host memory resource and thus have the overhead of pinned memory allocations in `hostdevice_vector`, and any other places where pinned memory...

## Description ## Checklist - [ ] I am familiar with the [Contributing Guidelines](https://github.com/rapidsai/cudf/blob/HEAD/CONTRIBUTING.md). - [ ] New or existing tests cover these changes. - [ ] The documentation is...

Limit pinned memory pool chunk size

1

The size of chunks in the pinned memory pool should be limited by the slice size, since each chunk is always used by a single thread.

## Description Issue #15612 Adds a pooled pinned memory resource that is created on first call to `get_host_memory_resource` or `set_host_memory_resource`. The pool has a fixed size: 0.5% of the device...

Add an option to run cuIO benchmarks with pinned buffers as input

1

## Description Adds `io_type::PINNED_BUFFER`, which allows cuIO benchmarks to use a pinned buffer as an input. The output is still a `std::vector` in this case, same as with `io_type::HOST_BUFFER`. Also...

Try to read timezone file from `$CONDA_PREFIX` dir before trying the system dir

8

## Description Closes https://github.com/rapidsai/cudf/issues/12926 Modified the logic when opening TZ files to try multiple paths before giving up. Without user-specified path, we try `$CONDA_PREFIX/share/zoneinfo` and then `/usr/share/zoneinfo/`. Testing is limited...