hpt 0.1.3

High Performance Tensor (HPT) - A fast, efficient, and user-friendly tensor computation library for Rust
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

use hpt_allocator::{
    traits::{Allocator, AllocatorOutputRetrive},
    Cpu,
};
use hpt_common::error::base::TensorError;
use hpt_traits::ops::creation::TensorCreator;
use hpt_traits::{
    ops::windows::WindowOps,
    tensor::{CommonBounds, TensorLike},
};
use hpt_types::{
    dtype::{FloatConst, TypeCommon},
    into_scalar::Cast,
    traits::VecTrait,
    type_promote::{FloatOutBinary, FloatOutUnary, NormalOut},
};
use rayon::{
    iter::{IndexedParallelIterator, IntoParallelRefMutIterator, ParallelIterator},
    slice::ParallelSliceMut,
};

use crate::tensor_base::_Tensor;

pub(crate) type Simd<T> = <T as TypeCommon>::Vec;

impl<T, A2, const DEVICE: usize> WindowOps for _Tensor<T, Cpu, DEVICE, A2>
where
    f64: Cast<T>,
    T: CommonBounds + FloatOutBinary<Output = T> + FloatOutUnary<Output = T> + FloatConst,
    Simd<T>: FloatOutBinary<Simd<T>, Output = Simd<T>> + FloatOutUnary<Output = Simd<T>>,
    usize: Cast<T>,
    i64: Cast<T>,
    A2: Allocator,
    A2::Output: AllocatorOutputRetrive,
{
    type Output = _Tensor<T, Cpu, DEVICE, A2>;
    type Meta = T;

    #[track_caller]
    fn hamming_window(window_length: i64, periodic: bool) -> Result<Self::Output, TensorError> {
        __hamming_window(window_length, (0.54).cast(), (0.46).cast(), periodic)
    }

    #[track_caller]
    fn hann_window(window_length: i64, periodic: bool) -> Result<Self::Output, TensorError> {
        __hamming_window(window_length, (0.5).cast(), (0.5).cast(), periodic)
    }

    #[track_caller]
    fn blackman_window(window_length: i64, periodic: bool) -> Result<Self::Output, TensorError> {
        let a0: T = (0.42).cast();
        let a1: T = (0.5).cast();
        let a2: T = (0.08).cast();
        let two_pi = f64::TWOPI.cast();
        let four_pi = f64::FOURPI.cast();
        let length_usize = if periodic {
            window_length
        } else {
            window_length - 1
        };
        let length: T = length_usize.cast();
        let mut ret = Self::Output::empty(&[length_usize])?;
        ret.as_raw_mut()
            .par_iter_mut()
            .enumerate()
            .for_each(|(idx, x)| {
                let idx: T = idx.cast();
                let a = a1._mul(two_pi._mul(idx)._div(length)._cos());
                let b = a2._mul(four_pi._mul(idx)._div(length)._cos());
                *x = a0._sub(a)._add(b);
            });
        Ok(ret)
    }
}

#[track_caller]
fn __hamming_window<T, A2, const DEVICE: usize>(
    window_length: i64,
    alpha: T,
    beta: T,
    periodic: bool,
) -> Result<_Tensor<T, Cpu, DEVICE, A2>, TensorError>
where
    f64: Cast<T>,
    T: CommonBounds + FloatOutUnary<Output = T> + FloatConst + FloatOutBinary<Output = T>,
    Simd<T>: FloatOutBinary<Simd<T>, Output = Simd<T>> + FloatOutUnary<Output = Simd<T>>,
    usize: Cast<T>,
    i64: Cast<T>,
    A2: Allocator,
    A2::Output: AllocatorOutputRetrive,
{
    let length_usize = (if periodic {
        window_length
    } else {
        window_length - 1
    }) as usize;
    let length: T = length_usize.cast();
    let mut ret = _Tensor::<T, Cpu, DEVICE, A2>::empty(&[length_usize as i64])?;
    let mut chunk_exact = ret.as_raw_mut().par_chunks_exact_mut(Simd::<T>::SIZE);
    let two_pi = Simd::<T>::splat(T::TWOPI);
    let length_vec = Simd::<T>::splat(length);
    let alpha_vec = Simd::<T>::splat(alpha);
    let beta_vec = Simd::<T>::splat(beta._neg());
    let remainder = chunk_exact.remainder();
    remainder.iter_mut().enumerate().for_each(|(idx, x)| {
        let idx: T = idx.cast();
        *x = idx
            ._mul(T::TWOPI._div(length))
            ._cos()
            ._mul_add(beta._neg(), alpha);
    });
    chunk_exact.enumerate().for_each(|(x, vec)| {
        let idx = x * Simd::<T>::SIZE;
        let mut idxes = Simd::<T>::splat(T::ZERO);
        for i in 0..Simd::<T>::SIZE {
            idxes[i] = (idx + i).cast();
        }
        let ptr = vec as *mut _ as *mut Simd<T>;

        let res = hpt_types::traits::VecTrait::mul_add(
            idxes._mul(two_pi._div(length_vec))._cos(),
            beta_vec,
            alpha_vec,
        );
        unsafe {
            ptr.write_unaligned(res);
        }
    });
    Ok(ret)
}