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
use crate::Matrix;
#[cfg(feature = "opencl")]
use custos::OpenCL;
use custos::{impl_stack, number::Float, Buffer, Device, MainMemory, Shape, CPU};
#[cfg(feature = "stack")]
use custos::Stack;
#[cfg(feature = "opencl")]
use crate::opencl::cl_write;
pub trait RandBuf<T> {
fn rand(&mut self, lo: T, hi: T);
}
impl<T, S: Shape, D: RandOp<T, S>> RandBuf<T> for Buffer<'_, T, D, S> {
#[inline]
fn rand(&mut self, lo: T, hi: T) {
self.device().rand(self, lo, hi)
}
}
impl<'a, T, S: Shape, D: RandOp<T, S>> Matrix<'a, T, D, S> {
#[inline]
pub fn rand(&mut self, lo: T, hi: T) {
self.as_buf_mut().rand(lo, hi);
}
}
pub trait RandOp<T, S: Shape = (), D: Device = Self>: Device {
fn rand(&self, x: &mut Buffer<T, D, S>, lo: T, hi: T);
}
pub fn rand_slice<T: PartialOrd + Copy + Float>(slice: &mut [T], lo: T, hi: T) {
let rng = fastrand::Rng::new();
for value in slice {
*value = T::as_generic(rng.f64()) * (hi - (lo)) + (lo);
}
}
#[impl_stack]
impl<T: Float, D: MainMemory, S: Shape> RandOp<T, S, D> for CPU {
#[inline]
fn rand(&self, x: &mut Buffer<T, D, S>, lo: T, hi: T) {
rand_slice(x, lo, hi)
}
}
#[cfg(feature = "opencl")]
impl<T: Float> RandOp<T> for OpenCL {
fn rand(&self, x: &mut Buffer<T, OpenCL>, lo: T, hi: T) {
#[cfg(unified_cl)]
rand_slice(x, lo, hi);
#[cfg(not(unified_cl))]
{
let mut data = vec![T::default(); x.len()];
rand_slice(&mut data, lo, hi);
cl_write(self, x, &data)
};
}
}
#[cfg(feature = "cuda")]
use custos::{cuda::api::cu_write, CUDA};
#[cfg(feature = "cuda")]
impl<T: Float> RandOp<T> for CUDA {
fn rand(&self, x: &mut Buffer<T, CUDA>, lo: T, hi: T) {
let mut data = vec![T::default(); x.len()];
rand_slice(&mut data, lo, hi);
cu_write(x.ptr.ptr, &data).unwrap();
}
}