Files
cfg_if
lib.rs
libm
math
acos.rsacosf.rsacosh.rsacoshf.rsasin.rsasinf.rsasinh.rsasinhf.rsatan.rsatan2.rsatan2f.rsatanf.rsatanh.rsatanhf.rscbrt.rscbrtf.rsceil.rsceilf.rscopysign.rscopysignf.rscos.rscosf.rscosh.rscoshf.rserf.rserff.rsexp.rsexp10.rsexp10f.rsexp2.rsexp2f.rsexpf.rsexpm1.rsexpm1f.rsexpo2.rsfabs.rsfabsf.rsfdim.rsfdimf.rsfenv.rsfloor.rsfloorf.rsfma.rsfmaf.rsfmax.rsfmaxf.rsfmin.rsfminf.rsfmod.rsfmodf.rsfrexp.rsfrexpf.rshypot.rshypotf.rsilogb.rsilogbf.rsj0.rsj0f.rsj1.rsj1f.rsjn.rsjnf.rsk_cos.rsk_cosf.rsk_expo2.rsk_expo2f.rsk_sin.rsk_sinf.rsk_tan.rsk_tanf.rsldexp.rsldexpf.rslgamma.rslgamma_r.rslgammaf.rslgammaf_r.rslog.rslog10.rslog10f.rslog1p.rslog1pf.rslog2.rslog2f.rslogf.rsmod.rsmodf.rsmodff.rspow.rspowf.rsrem_pio2.rsrem_pio2_large.rsrem_pio2f.rsremquo.rsremquof.rsround.rsroundf.rsscalbn.rsscalbnf.rssin.rssincos.rssincosf.rssinf.rssinh.rssinhf.rssqrt.rssqrtf.rstan.rstanf.rstanh.rstanhf.rstgamma.rstgammaf.rstrunc.rstruncf.rs
lib.rs
packed_simd_2
api
cast
macros.rsv128.rsv16.rsv256.rsv32.rsv512.rsv64.rs
cmp
eq.rsord.rspartial_eq.rspartial_ord.rsvertical.rs
fmt
binary.rsdebug.rslower_hex.rsoctal.rsupper_hex.rs
from
from_array.rsfrom_vector.rs
into_bits
arch_specific.rsmacros.rsv128.rsv16.rsv256.rsv32.rsv512.rsv64.rs
math
float
abs.rsconsts.rscos.rsexp.rsln.rsmul_add.rsmul_adde.rspowf.rsrecpre.rsrsqrte.rssin.rssqrt.rssqrte.rstanh.rs
float.rs
minimal
iuf.rsmask.rsptr.rs
ops
scalar_arithmetic.rsscalar_bitwise.rsscalar_mask_bitwise.rsscalar_shifts.rsvector_arithmetic.rsvector_bitwise.rsvector_float_min_max.rsvector_int_min_max.rsvector_mask_bitwise.rsvector_neg.rsvector_rotates.rsvector_shifts.rs
ptr
gather_scatter.rs
reductions
bitwise.rsfloat_arithmetic.rsinteger_arithmetic.rsmask.rsmin_max.rs
slice
from_slice.rswrite_to_slice.rs
bit_manip.rsbitmask.rscast.rscmp.rsdefault.rsfmt.rsfrom.rshash.rsinto_bits.rsmath.rsminimal.rsops.rsptr.rsreductions.rsselect.rsshuffle.rsshuffle1_dyn.rsslice.rsswap_bytes.rs
codegen
math
float
abs.rscos.rscos_pi.rsexp.rsln.rsmacros.rsmul_add.rsmul_adde.rspowf.rssin.rssin_cos_pi.rssin_pi.rssqrt.rssqrte.rstanh.rs
float.rs
reductions
mask
x86
sse.rssse2.rs
fallback_impl.rsx86.rs
mask.rs
bit_manip.rsllvm.rsmath.rspointer_sized_int.rsreductions.rsshuffle.rsshuffle1_dyn.rsswap_bytes.rsv128.rsv16.rsv256.rsv32.rsv512.rsv64.rsvPtr.rsvSize.rs
testing
macros.rs
api.rscodegen.rslib.rsmasks.rssealed.rstesting.rsv128.rsv16.rsv256.rsv32.rsv512.rsv64.rsvPtr.rsvSize.rs
  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

//! Vertical floating-point `tanh`
#![allow(unused)]

// FIXME 64-bit 1 elem vectors tanh

use crate::*;

crate trait Tanh {
    fn tanh(self) -> Self;
}

macro_rules! define_tanh {
    ($name:ident, $basetype:ty, $simdtype:ty, $lanes:expr, $trait:path) => {
        fn $name(x: $simdtype) -> $simdtype {
            use core::intrinsics::transmute;
            let mut buf: [$basetype; $lanes] = unsafe { transmute(x) };
            for elem in &mut buf {
                *elem = <$basetype as $trait>::tanh(*elem);
            }
            unsafe { transmute(buf) }
        }
    };

    (f32 => $name:ident, $type:ty, $lanes:expr) => {
        define_tanh!($name, f32, $type, $lanes, libm::F32Ext);
    };

    (f64 => $name:ident, $type:ty, $lanes:expr) => {
        define_tanh!($name, f64, $type, $lanes, libm::F64Ext);
    };
}

// llvm does not seem to expose the hyperbolic versions of trigonometric
// functions; we thus call the classical rust versions on all of them (which
// stem from cmath).
define_tanh!(f32 => tanh_v2f32, f32x2, 2);
define_tanh!(f32 => tanh_v4f32, f32x4, 4);
define_tanh!(f32 => tanh_v8f32, f32x8, 8);
define_tanh!(f32 => tanh_v16f32, f32x16, 16);

define_tanh!(f64 => tanh_v2f64, f64x2, 2);
define_tanh!(f64 => tanh_v4f64, f64x4, 4);
define_tanh!(f64 => tanh_v8f64, f64x8, 8);

fn tanh_f32(x: f32) -> f32 {
    libm::F32Ext::tanh(x)
}

fn tanh_f64(x: f64) -> f64 {
    libm::F64Ext::tanh(x)
}

gen_unary_impl_table!(Tanh, tanh);

cfg_if! {
    if #[cfg(target_arch = "s390x")] {
        // FIXME: https://github.com/rust-lang-nursery/packed_simd/issues/14
        impl_unary!(f32x2[f32; 2]: tanh_f32);
        impl_unary!(f32x4[f32; 4]: tanh_f32);
        impl_unary!(f32x8[f32; 8]: tanh_f32);
        impl_unary!(f32x16[f32; 16]: tanh_f32);

        impl_unary!(f64x2[f64; 2]: tanh_f64);
        impl_unary!(f64x4[f64; 4]: tanh_f64);
        impl_unary!(f64x8[f64; 8]: tanh_f64);
    } else if #[cfg(all(target_arch = "x86_64", feature = "sleef-sys"))] {
        use sleef_sys::*;
        cfg_if! {
            if #[cfg(target_feature = "avx2")] {
                impl_unary!(f32x2[t => f32x4]: Sleef_tanhf4_u10avx2128);
                impl_unary!(f32x16[h => f32x8]: Sleef_tanhf8_u10avx2);
                impl_unary!(f64x8[h => f64x4]: Sleef_tanhd4_u10avx2);

                impl_unary!(f32x4: Sleef_tanhf4_u10avx2128);
                impl_unary!(f32x8: Sleef_tanhf8_u10avx2);
                impl_unary!(f64x2: Sleef_tanhd2_u10avx2128);
                impl_unary!(f64x4: Sleef_tanhd4_u10avx2);
            } else if #[cfg(target_feature = "avx")] {
                impl_unary!(f32x2[t => f32x4]: Sleef_tanhf4_u10sse4);
                impl_unary!(f32x16[h => f32x8]: Sleef_tanhf8_u10avx);
                impl_unary!(f64x8[h => f64x4]: Sleef_tanhd4_u10avx);

                impl_unary!(f32x4: Sleef_tanhf4_u10sse4);
                impl_unary!(f32x8: Sleef_tanhf8_u10avx);
                impl_unary!(f64x2: Sleef_tanhd2_u10sse4);
                impl_unary!(f64x4: Sleef_tanhd4_u10avx);
            } else if #[cfg(target_feature = "sse4.2")] {
                impl_unary!(f32x2[t => f32x4]: Sleef_tanhf4_u10sse4);
                impl_unary!(f32x16[q => f32x4]: Sleef_tanhf4_u10sse4);
                impl_unary!(f64x8[q => f64x2]: Sleef_tanhd2_u10sse4);

                impl_unary!(f32x4: Sleef_tanhf4_u10sse4);
                impl_unary!(f32x8[h => f32x4]: Sleef_tanhf4_u10sse4);
                impl_unary!(f64x2: Sleef_tanhd2_u10sse4);
                impl_unary!(f64x4[h => f64x2]: Sleef_tanhd2_u10sse4);
            } else {
                impl_unary!(f32x2[f32; 2]: tanh_f32);
                impl_unary!(f32x16: tanh_v16f32);
                impl_unary!(f64x8: tanh_v8f64);

                impl_unary!(f32x4: tanh_v4f32);
                impl_unary!(f32x8: tanh_v8f32);
                impl_unary!(f64x2: tanh_v2f64);
                impl_unary!(f64x4: tanh_v4f64);
            }
        }
    } else {
        impl_unary!(f32x2[f32; 2]: tanh_f32);
        impl_unary!(f32x4: tanh_v4f32);
        impl_unary!(f32x8: tanh_v8f32);
        impl_unary!(f32x16: tanh_v16f32);

        impl_unary!(f64x2: tanh_v2f64);
        impl_unary!(f64x4: tanh_v4f64);
        impl_unary!(f64x8: tanh_v8f64);
    }
}