Struct MappedRanges 

pub struct MappedRanges { /* private fields */ }

Byte ranges of a file in array storage: each range memory-mapped on its own (without a copy, on a Mac with a GPU; read elsewhere), so the bytes between them are neither mapped, wired nor counted. For loads that keep some of a file’s tensors. The file must not be modified while it is mapped (see MappedFile).

use nuvec_core::{DType, MappedRanges};

let page = MappedRanges::page_size();
let path = std::env::temp_dir().join("nuvec-mapped-ranges-doc.bin");
let mut bytes = vec![0u8; 3 * page];
bytes[2 * page..2 * page + 8].copy_from_slice(&[1.0f32, 2.0].map(f32::to_le_bytes).concat());
std::fs::write(&path, &bytes).unwrap();
let file = MappedRanges::open(&path, &[(2 * page as u64, page as u64)])?; // the last page only
assert_eq!(file.array(2 * page, DType::F32, &[2])?.to_vec::<f32>()?, vec![1.0, 2.0]);
assert!(file.array(0, DType::F32, &[2]).is_err()); // not mapped

Implementations

impl MappedRanges

pub fn page_size() -> usize

The granularity of mappings: range offsets must be multiples of it.

pub fn open(path: impl AsRef<Path>, ranges: &[(u64, u64)]) -> Result<Self>

Maps each (offset, length) range of the file at path (offsets multiples of Self::page_size, ranges disjoint and increasing).

pub fn mapped_size(&self) -> u64

The bytes mapped, page-rounded (what the GPU allocator’s mapped counts).

pub fn array( &self, offset: usize, dtype: DType, shape: &[usize], ) -> Result<Array>

A row-major array of dtype and shape at byte offset of the file: a view of the range that holds it (it must lie inside one range and be aligned to its element size).

Trait Implementations

impl Clone for MappedRanges

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for MappedRanges

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

Formats the value using the given formatter.