singe-kernel 0.1.0-alpha.4

Reusable CPU and GPU 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
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

//! Fill, zeros, ones, arange, and other device initialization kernels.

use std::sync::Arc;

#[cfg(feature = "cutile")]
use ::cutile::cuda_async::device_buffer::DevicePointer;
#[cfg(feature = "dtype-f16")]
use singe_cuda::types::f16;
use singe_cuda::{
    stream::Stream,
    view::{DeviceRepr, DeviceSliceMut},
};

#[cfg(feature = "cutile")]
use crate::cuda::cutile;
use crate::{
    cuda::interop::{borrowed_stream, output_pointer},
    error::Result,
};

macro_rules! fill_fn {
    ($name:ident, $ty:ty) => {
        pub fn $name(
            stream: &Stream,
            out: &mut impl DeviceSliceMut<$ty>,
            value: $ty,
        ) -> Result<()> {
            fill(stream, out, value, cutile::creation::$name)
        }
    };
}

macro_rules! arange_fn {
    ($name:ident, $ty:ty) => {
        pub fn $name(
            stream: &Stream,
            out: &mut impl DeviceSliceMut<$ty>,
            start: $ty,
            step: $ty,
        ) -> Result<()> {
            arange(stream, out, start, step, cutile::creation::$name)
        }
    };
}

macro_rules! constant_fill_fn {
    ($name:ident, $ty:ty) => {
        pub fn $name(stream: &Stream, out: &mut impl DeviceSliceMut<$ty>) -> Result<()> {
            let stream = borrowed_stream(stream)?;
            cutile::creation::$name(&stream, output_pointer(out), out.len())
        }
    };
}

#[cfg(feature = "dtype-f16")]
fill_fn!(fill_f16, f16);
#[cfg(feature = "dtype-f32")]
fill_fn!(fill_f32, f32);
#[cfg(feature = "dtype-f64")]
fill_fn!(fill_f64, f64);
#[cfg(feature = "dtype-u8")]
fill_fn!(fill_u8, u8);
#[cfg(feature = "dtype-i8")]
fill_fn!(fill_i8, i8);
#[cfg(feature = "dtype-u32")]
fill_fn!(fill_u32, u32);
#[cfg(feature = "dtype-u64")]
fill_fn!(fill_u64, u64);
#[cfg(feature = "dtype-i32")]
fill_fn!(fill_i32, i32);
#[cfg(feature = "dtype-i64")]
fill_fn!(fill_i64, i64);

#[cfg(feature = "dtype-f16")]
constant_fill_fn!(zeros_f16, f16);
#[cfg(feature = "dtype-f16")]
constant_fill_fn!(ones_f16, f16);
#[cfg(feature = "dtype-f32")]
constant_fill_fn!(zeros_f32, f32);
#[cfg(feature = "dtype-f32")]
constant_fill_fn!(ones_f32, f32);
#[cfg(feature = "dtype-f64")]
constant_fill_fn!(zeros_f64, f64);
#[cfg(feature = "dtype-f64")]
constant_fill_fn!(ones_f64, f64);
#[cfg(feature = "dtype-u8")]
constant_fill_fn!(zeros_u8, u8);
#[cfg(feature = "dtype-u8")]
constant_fill_fn!(ones_u8, u8);
#[cfg(feature = "dtype-i8")]
constant_fill_fn!(zeros_i8, i8);
#[cfg(feature = "dtype-i8")]
constant_fill_fn!(ones_i8, i8);
#[cfg(feature = "dtype-u32")]
constant_fill_fn!(zeros_u32, u32);
#[cfg(feature = "dtype-u32")]
constant_fill_fn!(ones_u32, u32);
#[cfg(feature = "dtype-i32")]
constant_fill_fn!(zeros_i32, i32);
#[cfg(feature = "dtype-i32")]
constant_fill_fn!(ones_i32, i32);
#[cfg(feature = "dtype-u64")]
constant_fill_fn!(zeros_u64, u64);
#[cfg(feature = "dtype-u64")]
constant_fill_fn!(ones_u64, u64);
#[cfg(feature = "dtype-i64")]
constant_fill_fn!(zeros_i64, i64);
#[cfg(feature = "dtype-i64")]
constant_fill_fn!(ones_i64, i64);

#[cfg(feature = "dtype-f32")]
arange_fn!(arange_f32, f32);
#[cfg(feature = "dtype-f64")]
arange_fn!(arange_f64, f64);
#[cfg(feature = "dtype-f16")]
arange_fn!(arange_f16, f16);
#[cfg(feature = "dtype-u8")]
arange_fn!(arange_u8, u8);
#[cfg(feature = "dtype-i8")]
arange_fn!(arange_i8, i8);
#[cfg(feature = "dtype-i32")]
arange_fn!(arange_i32, i32);
#[cfg(feature = "dtype-i64")]
arange_fn!(arange_i64, i64);
#[cfg(feature = "dtype-u32")]
arange_fn!(arange_u32, u32);
#[cfg(feature = "dtype-u64")]
arange_fn!(arange_u64, u64);

#[cfg(feature = "dtype-f16")]
arange_fn!(linspace_f16, f16);
#[cfg(feature = "dtype-f32")]
arange_fn!(linspace_f32, f32);
#[cfg(feature = "dtype-f64")]
arange_fn!(linspace_f64, f64);

fn fill<T, F>(stream: &Stream, out: &mut impl DeviceSliceMut<T>, value: T, launch: F) -> Result<()>
where
    T: DeviceRepr,
    F: FnOnce(&Arc<::cutile::cuda_core::Stream>, DevicePointer<T>, T, usize) -> Result<()>,
{
    let stream = borrowed_stream(stream)?;
    launch(&stream, output_pointer(out), value, out.len())
}

fn arange<T, F>(
    stream: &Stream,
    out: &mut impl DeviceSliceMut<T>,
    start: T,
    step: T,
    launch: F,
) -> Result<()>
where
    T: DeviceRepr,
    F: FnOnce(&Arc<::cutile::cuda_core::Stream>, DevicePointer<T>, T, T, usize) -> Result<()>,
{
    let stream = borrowed_stream(stream)?;
    launch(&stream, output_pointer(out), start, step, out.len())
}