Struct GPUSorter 

pub struct GPUSorter { /* private fields */ }

Sorting pipeline. It can be used to sort key-value pairs stored in SortBuffers

Implementations

impl GPUSorter

pub fn new(device: &Device, subgroup_size: u32) -> Self

pub fn sort( &self, encoder: &mut CommandEncoder, queue: &Queue, sort_buffers: &SortBuffers, sort_first_n: Option<u32>, )

Writes sort commands to command encoder. If sort_first_n is not none one the first n elements are sorted otherwise everything is sorted.

IMPORTANT: if less than the whole buffer is sorted the rest of the keys buffer will be be corrupted

pub fn sort_indirect( &self, encoder: &mut CommandEncoder, sort_buffers: &SortBuffers, dispatch_buffer: &Buffer, )

Initiates sorting with an indirect call. The dispatch buffer must contain the struct wgpu::util::DispatchIndirectArgs.

number of y and z workgroups must be 1

x = (N + HISTO_BLOCK_KVS- 1 )/HISTO_BLOCK_KVS, where N are the first N elements to be sorted

SortBuffers::state_buffer contains the number of keys that will be sorted. This is set to sort the whole buffer by default.

IMPORTANT: if less than the whole buffer is sorted the rest of the keys buffer will most likely be corrupted.

pub fn create_sort_buffers( &self, device: &Device, length: NonZeroU32, ) -> SortBuffers

creates all buffers necessary for sorting