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
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
use crate::{Matrix, Vector};
use num::Float;
use std::mem::MaybeUninit;
#[derive(Debug, Clone)]
pub struct DynTensor<const D: usize, T = f32> {
shape: Vector<D, usize>,
data: Box<[T]>,
}
impl<T, const D: usize> DynTensor<D, T> {
pub fn len(&self) -> usize { self.shape.0.iter().product() }
pub fn is_empty(&self) -> bool { self.shape.0.iter().any(|&v| v == 0) }
pub fn shape(&self) -> &[usize; D] { &self.shape.0 }
pub fn iter(&self) -> impl Iterator<Item = &T> { self.data.iter() }
pub fn reshape<const D2: usize>(self, new_shape: Vector<D2, usize>) -> DynTensor<D2, T> {
assert_eq!(
new_shape.0.iter().product::<usize>(),
self.shape.0.iter().product::<usize>(),
"Cannot reshape {:?} into {:?}: Unequal number of elements",
self.shape,
new_shape
);
DynTensor {
shape: new_shape,
data: self.data,
}
}
pub fn from_iter(shape: [usize; D], iter: impl Iterator<Item = T>) -> Self {
let size = shape.iter().product();
let mut data = Box::new_uninit_slice(size);
let mut count = 0;
for (i, v) in iter.enumerate() {
data[i] = MaybeUninit::new(v);
count += 1;
}
assert_eq!(count, size, "Iterator passed did not have enough elements");
let data = unsafe { data.assume_init() };
let shape = Vector(shape);
Self { shape, data }
}
}
trait TensorMul<T, const I: usize> {
type Output;
fn dot(&self, rh: DynTensor<I, T>) -> Self::Output;
}
impl<T: Float> TensorMul<T, 1> for DynTensor<1, T> {
type Output = T;
fn dot(&self, rhs: DynTensor<1, T>) -> Self::Output {
assert_eq!(
self.shape, rhs.shape,
"Length Mismatch between dot product of dyn vectors"
);
self
.data
.iter()
.zip(rhs.data.iter())
.map(|(&l, &r)| l * r)
.fold(T::zero(), |acc, n| acc + n)
}
}
impl<T: Float> TensorMul<T, 1> for DynTensor<2, T> {
type Output = T;
fn dot(&self, rhs: DynTensor<1, T>) -> Self::Output {
assert_eq!(
self.shape[1], rhs.shape[0],
"Length Mismatch between {:?} {:?}",
self.shape, rhs.shape
);
todo!();
}
}
impl<T, const N: usize> Vector<N, T> {
pub fn dyn_tensor(self) -> DynTensor<1, T> {
let shape = Vector([N]);
let data = Box::new(self.0);
DynTensor { shape, data }
}
}
impl<T: Copy, const N: usize, const M: usize> Matrix<M, N, T> {
pub fn dyn_tensor(self) -> DynTensor<2, T> {
let shape = Vector([M, N]);
let mut data = Box::new_uninit_slice(M * N);
let data = unsafe {
for i in 0..N {
for j in 0..M {
data[i * M + j] = MaybeUninit::new(self[i][j]);
}
}
data.assume_init()
};
DynTensor { shape, data }
}
}
macro_rules! impl_bin_op {
($(impl $typ: ident, $func: ident, $op: tt;)*) => {
$(
impl<T: $typ + Copy, const D: usize> $typ for DynTensor<D, T> {
type Output = DynTensor<D, T::Output>;
fn $func(self, o: Self) -> Self::Output {
assert_eq!(self.shape, o.shape, "Currently only supports inputs of same size");
let iter = self.iter().zip(o.iter()).map(|(&l, &r)| l $op r);
DynTensor::from_iter(self.shape.0, iter)
}
}
)*
};
}
use std::ops::{Add, Div, Mul, Sub};
impl_bin_op!(
impl Mul, mul, *;
impl Add, add, +;
impl Sub, sub, -;
impl Div, div, /;
);
#[test]
fn example() {
use crate::{Mat3, Vec3};
let v = Vec3::of(3.0);
let m = Mat3::scale(&Vec3::of(3.0));
let t = m.dyn_tensor();
}