holographic-memory 0.6.3

Local-first search engine and agent memory: fast approximate nearest-neighbor and BM25 full-text search, encrypted stores, provable deletion, built on vector-symbolic architectures.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

// Copyright 2024-2026 WritersLogic Contributors
// SPDX-License-Identifier: Apache-2.0

use super::types::HardwareCapabilities;

/// Detect CPU features without executing unsupported instructions.
pub fn capabilities() -> HardwareCapabilities {
    #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
    let (avx2, avx512f) = (
        std::is_x86_feature_detected!("avx2"),
        std::is_x86_feature_detected!("avx512f"),
    );
    #[cfg(not(any(target_arch = "x86", target_arch = "x86_64")))]
    let (avx2, avx512f) = (false, false);

    #[cfg(target_arch = "aarch64")]
    let neon = std::arch::is_aarch64_feature_detected!("neon");
    #[cfg(target_arch = "arm")]
    // Runtime detection on 32-bit ARM is unstable; report what the target was compiled with.
    let neon = cfg!(target_feature = "neon");
    #[cfg(not(any(target_arch = "aarch64", target_arch = "arm")))]
    let neon = false;

    HardwareCapabilities {
        architecture: std::env::consts::ARCH.to_string(),
        avx2,
        avx512f,
        neon,
        // Sparse sorted u32 indices favor adaptive merge/galloping over dense SIMD.
        sparse_intersection_kernel: "adaptive_merge_galloping".to_string(),
    }
}