mircuda-sys 0.1.0

Private native CUDA integration for mircuda
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

use std::{ffi::c_void, ptr::NonNull, sync::Arc};

use cudarc::driver::CudaStream;

use super::{VariableGroupedFp4Spec, check, validate_sizes};
use crate::{
    Error, Result,
    platform::{
        driver::{Context, Stream},
        memory::{DeviceBuffer, ensure_stream},
    },
};

unsafe extern "C" {
    fn mircuda_paired_variable_grouped_fp4_create(
        groups: i32,
        matrices: i32,
        max_m: i32,
        n: i32,
        k: i32,
        stream: *mut c_void,
        output: *mut *mut c_void,
    ) -> i32;
    fn mircuda_paired_variable_grouped_fp4_execute(
        plan: *mut c_void,
        stream: *mut c_void,
        left_a: *const c_void,
        left_a_scales: *const c_void,
        left_b: *const c_void,
        left_b_scales: *const c_void,
        left_alphas: *const c_void,
        right_a: *const c_void,
        right_a_scales: *const c_void,
        right_b: *const c_void,
        right_b_scales: *const c_void,
        right_alphas: *const c_void,
        indices: *const u32,
        rows: *const u32,
        offsets: *const u32,
        left_c: *mut c_void,
        right_c: *mut c_void,
    ) -> i32;
    fn mircuda_paired_variable_grouped_fp4_destroy(plan: *mut c_void);
}

#[derive(Debug)]
pub struct PairedVariableGroupedFp4Plan {
    raw: NonNull<c_void>,
    stream: Arc<CudaStream>,
    spec: VariableGroupedFp4Spec,
}

// SAFETY: the plan retains its CUDA stream, binds the context before native
// use or destruction, and execution requires exclusive mutable access.
unsafe impl Send for PairedVariableGroupedFp4Plan {}

impl Context {
    pub fn create_paired_variable_grouped_fp4_plan(
        &self,
        stream: &Stream,
        spec: VariableGroupedFp4Spec,
    ) -> Result<PairedVariableGroupedFp4Plan> {
        if !Arc::ptr_eq(&self.inner, stream.inner.context()) {
            return Err(Error::ContextMismatch);
        }
        self.inner.bind_to_thread()?;
        let mut raw = std::ptr::null_mut();
        // SAFETY: output is writable and the safe facade validated all dimensions.
        let status = unsafe {
            mircuda_paired_variable_grouped_fp4_create(
                i32::try_from(spec.groups)?,
                i32::try_from(spec.matrices)?,
                i32::try_from(spec.max_m)?,
                i32::try_from(spec.n)?,
                i32::try_from(spec.k)?,
                stream.inner.cu_stream().cast(),
                &raw mut raw,
            )
        };
        check(status)?;
        Ok(PairedVariableGroupedFp4Plan {
            raw: NonNull::new(raw).ok_or(Error::NullAllocation)?,
            stream: stream.inner.clone(),
            spec,
        })
    }
}

impl PairedVariableGroupedFp4Plan {
    #[allow(clippy::similar_names, clippy::too_many_arguments)]
    pub fn execute(
        &mut self,
        stream: &Stream,
        left_a: &DeviceBuffer,
        left_a_scales: &DeviceBuffer,
        left_b: &DeviceBuffer,
        left_b_scales: &DeviceBuffer,
        left_alphas: &DeviceBuffer,
        right_a: &DeviceBuffer,
        right_a_scales: &DeviceBuffer,
        right_b: &DeviceBuffer,
        right_b_scales: &DeviceBuffer,
        right_alphas: &DeviceBuffer,
        indices: &DeviceBuffer,
        rows: &DeviceBuffer,
        offsets: &DeviceBuffer,
        left_c: &DeviceBuffer,
        right_c: &DeviceBuffer,
    ) -> Result<()> {
        self.stream.context().bind_to_thread()?;
        if !Arc::ptr_eq(&self.stream, &stream.inner) {
            return Err(Error::StreamMismatch);
        }
        let buffers = [
            left_a, left_a_scales, left_b, left_b_scales, left_alphas, right_a, right_a_scales,
            right_b, right_b_scales, right_alphas, indices, rows, offsets, left_c, right_c,
        ];
        for buffer in buffers {
            ensure_stream(buffer, stream)?;
        }
        validate_sizes(
            self.spec, left_a, left_a_scales, left_b, left_b_scales, left_alphas, indices, rows,
            offsets, left_c,
        )?;
        validate_sizes(
            self.spec, right_a, right_a_scales, right_b, right_b_scales, right_alphas, indices,
            rows, offsets, right_c,
        )?;
        // SAFETY: validated buffers are stream-bound and outlive asynchronous execution.
        check(unsafe {
            mircuda_paired_variable_grouped_fp4_execute(
                self.raw.as_ptr(),
                stream.inner.cu_stream().cast(),
                left_a.pointer() as *const c_void,
                left_a_scales.pointer() as *const c_void,
                left_b.pointer() as *const c_void,
                left_b_scales.pointer() as *const c_void,
                left_alphas.pointer() as *const c_void,
                right_a.pointer() as *const c_void,
                right_a_scales.pointer() as *const c_void,
                right_b.pointer() as *const c_void,
                right_b_scales.pointer() as *const c_void,
                right_alphas.pointer() as *const c_void,
                indices.pointer() as *const u32,
                rows.pointer() as *const u32,
                offsets.pointer() as *const u32,
                left_c.pointer() as *mut c_void,
                right_c.pointer() as *mut c_void,
            )
        })
    }
}

impl Drop for PairedVariableGroupedFp4Plan {
    fn drop(&mut self) {
        self.stream.context().record_err(self.stream.context().bind_to_thread());
        // SAFETY: this owns the matching native paired plan.
        unsafe { mircuda_paired_variable_grouped_fp4_destroy(self.raw.as_ptr()) };
    }
}