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

use std::cmp::max;
use std::slice;

pub unsafe trait KernelArg {}

/// A cuda kernel argument builder.
///
/// ```
/// # use kn_cuda_sys::wrapper::rtc::args::KernelArgs;
/// let mut args = KernelArgs::new();
/// args.push(std::ptr::null::<f32>());
/// args.push(16);
/// args.push(32);
/// args.push(1.0f32);
/// let arg_bytes = args.finish();
#[derive(Debug)]
pub struct KernelArgs {
    buffer: Vec<u8>,
    max_alignment: usize,
}

impl KernelArgs {
    pub fn new() -> Self {
        Self {
            buffer: vec![],
            max_alignment: 1,
        }
    }

    pub fn push<T: KernelArg>(&mut self, value: T) {
        // handle alignment
        let alignment = std::mem::align_of::<T>();
        self.max_alignment = max(self.max_alignment, alignment);
        self.pad_to(alignment);

        // append bytes
        unsafe {
            let bytes = slice::from_raw_parts(&value as *const T as *const u8, std::mem::size_of::<T>());
            self.buffer.extend_from_slice(bytes);
        }
    }

    pub fn push_int(&mut self, value: i32) {
        self.push(value)
    }

    pub fn finish(self) -> Vec<u8> {
        // we're not supposed to pad until alignment here
        self.buffer
    }

    fn pad_to(&mut self, alignment: usize) {
        while self.buffer.len() % alignment != 0 {
            self.buffer.push(0);
        }
    }
}

unsafe impl KernelArg for u8 {}

unsafe impl KernelArg for u16 {}

unsafe impl KernelArg for u32 {}

unsafe impl KernelArg for u64 {}

unsafe impl KernelArg for i8 {}

unsafe impl KernelArg for i16 {}

unsafe impl KernelArg for i32 {}

unsafe impl KernelArg for i64 {}

unsafe impl KernelArg for f32 {}

unsafe impl KernelArg for f64 {}

unsafe impl<T> KernelArg for *const T {}

unsafe impl<T> KernelArg for *mut T {}