ashwa 0.1.2

Hardware accelerated routines for single substring search
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
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149

//! Hardware accelerated routines for single substring search
//!
//! ## Example
//!
//! ```
//! use ashwa::search_one;
//!
//! let haystack = b"The quick brown fox jumps over the lazy dog";
//! assert_eq!(search_one(haystack, b'f'), Some(0x10));
//! assert_eq!(search_one(haystack, b'z'), Some(0x25));
//! assert_eq!(search_one(haystack, b'!'), None);
//! ```

#![cfg_attr(not(test), no_std)]
#![allow(unsafe_op_in_unsafe_fn)]
#![deny(
    missing_docs,
    trivial_casts,
    trivial_numeric_casts,
    unused_extern_crates,
    unused_import_braces,
    unused_results
)]

#[cfg(not(any(target_pointer_width = "64", target_pointer_width = "32")))]
compile_error!("ashwa is only supported on 64 and 32 bit targets");

#[cfg(all(target_arch = "x86_64"))]
use core::{arch::x86_64, sync::atomic};

mod one;

pub use one::search_one;

#[repr(u8)]
#[cfg(target_arch = "x86_64")]
pub(crate) enum ISA {
    NONE,
    SWAR,
    SSE2,
    SSSE3,
    SSE4_2,
    AVX2,
    AVX512BW,
}

#[cfg(target_arch = "x86_64")]
impl From<u8> for ISA {
    #[inline(always)]
    fn from(value: u8) -> Self {
        match value {
            0 => ISA::NONE,
            1 => ISA::SWAR,
            2 => ISA::SSE2,
            3 => ISA::SSSE3,
            4 => ISA::SSE4_2,
            5 => ISA::AVX2,
            6 => ISA::AVX512BW,
            _ => unreachable!("invalid ISA {}", value),
        }
    }
}

/// Best available ISA on the target microarchitecture
#[cfg(target_arch = "x86_64")]
static CPU_FEATURE: atomic::AtomicU8 = atomic::AtomicU8::new(ISA::NONE as u8);

#[inline(always)]
#[allow(unreachable_code)]
#[cfg(target_arch = "x86_64")]
pub(crate) fn get_cpu_feature() -> ISA {
    #[cfg(forced_swar_backend)]
    return ISA::SWAR;

    #[cfg(target_feature = "avx512bw")]
    return ISA::AVX512BW;

    #[cfg(target_feature = "avx2")]
    return ISA::AVX2;

    #[cfg(target_feature = "sse4.2")]
    return ISA::SSE4_2;

    #[cfg(target_feature = "ssse3")]
    return ISA::SSSE3;

    #[cfg(target_feature = "sse2")]
    return ISA::SSE2;

    let feature = CPU_FEATURE.load(atomic::Ordering::Relaxed);
    if feature != 0 {
        return feature.into();
    }

    let detected = unsafe { detect_features_x86_64() };
    CPU_FEATURE.store(detected as u8, atomic::Ordering::Relaxed);

    detected.into()
}

#[cold]
#[inline(never)]
#[cfg(target_arch = "x86_64")]
unsafe fn detect_features_x86_64() -> ISA {
    let cpuid1 = x86_64::__cpuid(1);

    let has_sse2 = (cpuid1.edx & (1 << 0x1A)) != 0;
    let has_ssse3 = (cpuid1.ecx & (1 << 9)) != 0;
    let has_sse4_2 = (cpuid1.ecx & (1 << 0x14)) != 0;

    let osxsave = (cpuid1.ecx & (1 << 0x1B)) != 0;
    if osxsave {
        let xcr0 = x86_64::_xgetbv(0);
        let xmm_ymm_enabled = (xcr0 & 0b110) == 0b110;

        if xmm_ymm_enabled {
            let cpuid7 = x86_64::__cpuid_count(7, 0);

            #[cfg(target_feature = "avx512bw")]
            {
                let avx512f_bw = (1 << 16) | (1 << 30);
                let vbmi_vbmi2 = (1 << 1) | (1 << 6);
                if (cpuid7.ebx & avx512f_bw) == avx512f_bw
                    && (cpuid7.ecx & vbmi_vbmi2) == vbmi_vbmi2
                {
                    return ISA::AVX512BW;
                }
            }

            if (cpuid7.ebx & (1 << 5)) != 0 {
                return ISA::AVX2;
            }
        }
    }

    if has_sse4_2 {
        return ISA::SSE4_2;
    }

    if has_ssse3 {
        return ISA::SSSE3;
    }

    if has_sse2 {
        return ISA::SSE2;
    }

    ISA::SWAR
}