fgf 3.0.2

Faster Galois Fields: SIMD and portable finite-field arithmetic (binary towers and prime fields) and vector kernels
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

//! Implementation details for experimentation, with no compatibility guarantees.
//!
//! Paths, types, signatures, and behavior may change or disappear in any release.
//! Production consumers should use the supported API in [`crate::ops`],
//! [`crate::bits`], [`crate::field`], and [`crate::kernel`].

/// Direct kernel implementations, preparation tables, and capability tokens.
pub mod kernel {
    pub use crate::kernel::bytes::xor;
    #[cfg(all(feature = "simd", target_arch = "x86_64"))]
    pub use crate::kernel::matrix_provider::{FlatMatrix, Matrix};

    /// Byte-field kernels and preparation.
    pub mod gf8 {
        pub use crate::kernel::gf8::*;
    }

    /// Packed-extension kernels over GF(2¹⁶).
    pub mod gf16 {
        pub use crate::kernel::tower::gf16::*;
    }

    /// Packed-extension kernels over GF(2³²).
    pub mod gf32 {
        pub use crate::kernel::tower::gf32::*;
    }

    /// Packed-extension kernels over GF(2⁶⁴).
    pub mod gf64 {
        pub use crate::kernel::tower::gf64::*;
    }

    /// Packed-extension kernel families of the shipped quadratic towers.
    pub mod tower {
        pub use crate::kernel::tower::*;
    }

    /// Portable prime-field kernels.
    pub mod prime {
        pub use crate::kernel::prime::*;
    }

    /// Portable binary-field kernels and tails.
    pub mod scalar {
        pub use crate::kernel::scalar::*;
    }

    /// Residue row routing: the selector functions the production dispatch
    /// runs, for benchmark routing reports that must not copy constants.
    /// Available only where the x86 SIMD dispatch exists; other targets
    /// route every residue shape through the portable path.
    #[cfg(all(feature = "simd", target_arch = "x86_64"))]
    pub mod residue {
        pub use crate::kernel::residue::{
            DELAYED_ADD_MIN_DEPTH, DELAYED_INTO_MIN_DEPTH, DELAYED_MIN_SOURCES, VectorOp,
            avx2_floor_op_len, delay_depth, delayed_matrix_backend, routes_delayed, routes_vector,
            single_row_backend, vector_backend,
        };
    }
    /// Coefficient tables shared by field and architecture kernels.
    pub mod tables {
        pub use crate::kernel::tables::*;
    }

    /// Token-bearing x86 kernels.
    #[cfg(all(feature = "simd", target_arch = "x86_64"))]
    pub mod x86 {
        pub use crate::kernel::x86::*;

        /// Direct residue kernels and reduction parameters.
        pub mod residue {
            #[cfg(feature = "simd512")]
            pub use crate::kernel::x86::residue::avx512::{
                matrix::{mul_add_matrix_u32_avx512_with, mul_into_matrix_u32_avx512_with},
                single::{
                    mul_add_u8_avx512, mul_add_u32_avx512, mul_add_u64_avx512,
                    mul_assign_u8_avx512, mul_assign_u32_avx512, mul_assign_u64_avx512,
                    mul_elementwise_assign_u32_avx512, mul_elementwise_u32_avx512,
                    mul_into_u8_avx512, mul_into_u32_avx512, mul_into_u64_avx512,
                },
            };
            pub use crate::kernel::x86::residue::*;
        }
    }

    #[cfg(feature = "simd")]
    pub use archmage::SimdToken;
    #[cfg(all(feature = "simd", target_arch = "x86_64"))]
    pub use archmage::{
        X64V1Token, X64V2Token, X64V3GfniCryptoToken, X64V3Token, X64V4Token, X64V4xToken,
    };
}

/// Independent scalar field oracles.
pub mod field {
    /// Fan–Paar recurrence for differential checks of the tower family.
    pub mod fan_paar {
        pub use crate::field::binary::tower::{fp_invert, fp_mul_alpha, fp_multiply, fp_square};
    }
}