numr 0.5.1

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
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

//! CUDA implementation of logical operations.

use crate::dtype::DType;
use crate::error::{Error, Result};
use crate::ops::LogicalOps;
use crate::runtime::cuda::kernels::{
    launch_logical_and_op, launch_logical_not_op, launch_logical_or_op, launch_logical_xor_op,
};
use crate::runtime::cuda::{CudaClient, CudaRuntime};
use crate::runtime::ensure_contiguous;
use crate::tensor::Tensor;

/// Validate inputs for binary logical operations (U8 dtype, matching shapes).
fn validate_logical_inputs(a: &Tensor<CudaRuntime>, b: &Tensor<CudaRuntime>) -> Result<()> {
    if a.dtype() != DType::U8 {
        return Err(Error::DTypeMismatch {
            lhs: DType::U8,
            rhs: a.dtype(),
        });
    }
    if b.dtype() != DType::U8 {
        return Err(Error::DTypeMismatch {
            lhs: DType::U8,
            rhs: b.dtype(),
        });
    }
    if a.shape() != b.shape() {
        return Err(Error::ShapeMismatch {
            expected: a.shape().to_vec(),
            got: b.shape().to_vec(),
        });
    }
    Ok(())
}

impl LogicalOps<CudaRuntime> for CudaClient {
    fn logical_and(
        &self,
        a: &Tensor<CudaRuntime>,
        b: &Tensor<CudaRuntime>,
    ) -> Result<Tensor<CudaRuntime>> {
        validate_logical_inputs(a, b)?;
        let a_contig = ensure_contiguous(a);
        let b_contig = ensure_contiguous(b);
        let out = Tensor::<CudaRuntime>::empty(a.shape(), DType::U8, &self.device);

        unsafe {
            launch_logical_and_op(
                &self.context,
                &self.stream,
                self.device.index,
                a_contig.ptr(),
                b_contig.ptr(),
                out.ptr(),
                out.numel(),
            )?;
        }

        Ok(out)
    }

    fn logical_or(
        &self,
        a: &Tensor<CudaRuntime>,
        b: &Tensor<CudaRuntime>,
    ) -> Result<Tensor<CudaRuntime>> {
        validate_logical_inputs(a, b)?;
        let a_contig = ensure_contiguous(a);
        let b_contig = ensure_contiguous(b);
        let out = Tensor::<CudaRuntime>::empty(a.shape(), DType::U8, &self.device);

        unsafe {
            launch_logical_or_op(
                &self.context,
                &self.stream,
                self.device.index,
                a_contig.ptr(),
                b_contig.ptr(),
                out.ptr(),
                out.numel(),
            )?;
        }

        Ok(out)
    }

    fn logical_xor(
        &self,
        a: &Tensor<CudaRuntime>,
        b: &Tensor<CudaRuntime>,
    ) -> Result<Tensor<CudaRuntime>> {
        validate_logical_inputs(a, b)?;
        let a_contig = ensure_contiguous(a);
        let b_contig = ensure_contiguous(b);
        let out = Tensor::<CudaRuntime>::empty(a.shape(), DType::U8, &self.device);

        unsafe {
            launch_logical_xor_op(
                &self.context,
                &self.stream,
                self.device.index,
                a_contig.ptr(),
                b_contig.ptr(),
                out.ptr(),
                out.numel(),
            )?;
        }

        Ok(out)
    }

    fn logical_not(&self, a: &Tensor<CudaRuntime>) -> Result<Tensor<CudaRuntime>> {
        if a.dtype() != DType::U8 {
            return Err(Error::DTypeMismatch {
                lhs: DType::U8,
                rhs: a.dtype(),
            });
        }

        let a_contig = ensure_contiguous(a);
        let out = Tensor::<CudaRuntime>::empty(a.shape(), DType::U8, &self.device);

        unsafe {
            launch_logical_not_op(
                &self.context,
                &self.stream,
                self.device.index,
                a_contig.ptr(),
                out.ptr(),
                out.numel(),
            )?;
        }

        Ok(out)
    }
}