Struct cuvsCagraSearchParams 

#[repr(C)]
pub struct cuvsCagraSearchParams {

    pub max_queries: usize,
    pub itopk_size: usize,
    pub max_iterations: usize,
    pub algo: cuvsCagraSearchAlgo,
    pub team_size: usize,
    pub search_width: usize,
    pub min_iterations: usize,
    pub thread_block_size: usize,
    pub hashmap_mode: cuvsCagraHashMode,
    pub hashmap_min_bitlen: usize,
    pub hashmap_max_fill_rate: f32,
    pub num_random_samplings: u32,
    pub rand_xor_mask: u64,
    pub persistent: bool,
    pub persistent_lifetime: f32,
    pub persistent_device_usage: f32,
}

@brief Supplemental parameters to search CAGRA index

Fields

max_queries: usize

Maximum number of queries to search at the same time (batch size). Auto select when 0.

itopk_size: usize

Number of intermediate search results retained during the search.

This is the main knob to adjust trade off between accuracy and search speed. Higher values improve the search accuracy.

max_iterations: usize

Upper limit of search iterations. Auto select when 0.

algo: cuvsCagraSearchAlgo

Which search implementation to use.

team_size: usize

Number of threads used to calculate a single distance. 4, 8, 16, or 32.

search_width: usize

Number of graph nodes to select as the starting point for the search in each iteration. aka search width?

min_iterations: usize

Lower limit of search iterations.

thread_block_size: usize

Thread block size. 0, 64, 128, 256, 512, 1024. Auto selection when 0.

hashmap_mode: cuvsCagraHashMode

Hashmap type. Auto selection when AUTO.

hashmap_min_bitlen: usize

Lower limit of hashmap bit length. More than 8.

hashmap_max_fill_rate: f32

Upper limit of hashmap fill rate. More than 0.1, less than 0.9.

num_random_samplings: u32

Number of iterations of initial random seed node selection. 1 or more.

rand_xor_mask: u64

Bit mask used for initial random seed node selection.

persistent: bool

Whether to use the persistent version of the kernel (only SINGLE_CTA is supported a.t.m.)

persistent_lifetime: f32

Persistent kernel: time in seconds before the kernel stops if no requests received.

persistent_device_usage: f32

Set the fraction of maximum grid size used by persistent kernel. Value 1.0 means the kernel grid size is maximum possible for the selected device. The value must be greater than 0.0 and not greater than 1.0.

One may need to run other kernels alongside this persistent kernel. This parameter can be used to reduce the grid size of the persistent kernel to leave a few SMs idle. Note: running any other work on GPU alongside with the persistent kernel makes the setup fragile.

• Running another kernel in another thread usually works, but no progress guaranteed
• Any CUDA allocations block the context (this issue may be obscured by using pools)
• Memory copies to not-pinned host memory may block the context

Even when we know there are no other kernels working at the same time, setting kDeviceUsage to 1.0 surprisingly sometimes hurts performance. Proceed with care. If you suspect this is an issue, you can reduce this number to ~0.9 without a significant impact on the throughput.

Trait Implementations

impl Clone for cuvsCagraSearchParams

fn clone(&self) -> cuvsCagraSearchParams

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for cuvsCagraSearchParams

impl Debug for cuvsCagraSearchParams

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.