Fragmenter_Namespace::anonymous_namespace{InsertOrderFragmenter.cpp} Namespace Reference

Functions
int	compute_device_for_fragment (const int table_id, const int fragment_id, const int num_devices)

Function Documentation

int Fragmenter_Namespace::anonymous_namespace{InsertOrderFragmenter.cpp}::compute_device_for_fragment	(	const int	table_id,
		const int	fragment_id,
		const int	num_devices
	)

Offset the fragment ID by the table ID, meaning single fragment tables end up balanced across multiple GPUs instead of all falling to GPU 0.

Definition at line 103 of file InsertOrderFragmenter.cpp.

References g_use_table_device_offset.

Referenced by Fragmenter_Namespace::InsertOrderFragmenter::createNewFragment(), and Fragmenter_Namespace::InsertOrderFragmenter::getChunkMetadata().

                                                       {
  if (g_use_table_device_offset) {
    return (table_id + fragment_id) % num_devices;
  } else {
    return fragment_id % num_devices;
  }
}

Here is the caller graph for this function: