dfdx 0.13.0

Ergonomic auto differentiation in Rust, with pytorch like apis.
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
160
161
162
163
164
165
166
167
168
169
170

use crate::{
    dtypes::*,
    shapes::Shape,
    tensor::{launch_cfg, Cuda, Tensor},
};
use cudarc::driver::{CudaSlice, LaunchAsync};

use super::{
    CmpKernel, EqKernelOp, GeKernelOp, GtKernelOp, LeKernelOp, LtKernelOp, NeKernelOp,
    ScalarCmpKernel,
};

const PTX_SRC: &str = include_str!(concat!(env!("OUT_DIR"), "/cmp.ptx"));

trait CmpOpCudaKernel<E: Unit> {
    /// Compiled by build.rs
    const PTX_SRC: &'static str;

    /// Unique name for the kernel
    const MODULE_NAME: &'static str;

    /// Name of function in the .cu file
    const FWD_FN_NAME: &'static str;
}

trait ScalarCmpOpCudaKernel<E: Unit> {
    /// Compiled by build.rs
    const PTX_SRC: &'static str;

    /// Unique name for the kernel
    const MODULE_NAME: &'static str;

    /// Name of function in the .cu file
    const FWD_FN_NAME: &'static str;
}

impl<E: Unit, Op: CmpOpCudaKernel<E>> CmpKernel<Op, E> for Cuda {
    fn forward<S: Shape, T>(
        &self,
        lhs: &Tensor<S, E, Self, T>,
        rhs: &Tensor<S, E, Self, T>,
    ) -> Result<Tensor<S, bool, Self>, Self::Err> {
        if !self.dev.has_func(Op::MODULE_NAME, Op::FWD_FN_NAME) {
            self.dev
                .load_ptx(Op::PTX_SRC.into(), Op::MODULE_NAME, &[Op::FWD_FN_NAME])?;
        }

        let shape = lhs.shape;
        let strides = lhs.shape.strides();
        let numel = shape.num_elements();

        let mut storage = unsafe { self.alloc_empty::<bool>(numel) }?;
        self.dev.memset_zeros(&mut storage)?;

        let dims: CudaSlice<usize> = self.dev.htod_copy(shape.concrete().into())?;
        let lhs_strides: CudaSlice<usize> = self.dev.htod_copy(lhs.strides.into())?;
        let rhs_strides: CudaSlice<usize> = self.dev.htod_copy(rhs.strides.into())?;
        let out_strides: CudaSlice<usize> = self.dev.htod_copy(strides.into())?;

        let fwd_fn = self.dev.get_func(Op::MODULE_NAME, Op::FWD_FN_NAME).unwrap();
        let cfg = launch_cfg::<128>(numel as u32);
        let params = (
            numel,             // const size_t numel,
            S::NUM_DIMS,       // const size_t num_dims,
            &dims,             // const size_t *dims,
            lhs.data.as_ref(), // const float *lhs,
            &lhs_strides,      // const size_t *lhs_strides,
            rhs.data.as_ref(), // const float *rhs,
            &rhs_strides,      // const size_t *rhs_strides,
            &mut storage,      // bool *out,
            &out_strides,      // const size_t *out_strides
        );
        unsafe { fwd_fn.launch(cfg, params) }?;
        Ok(self.build_tensor(shape, strides, storage))
    }
}

impl<E: Unit, Op: ScalarCmpOpCudaKernel<E>> ScalarCmpKernel<Op, E> for Cuda {
    fn forward<S: Shape, T>(
        &self,
        lhs: &Tensor<S, E, Self, T>,
        scalar: E,
    ) -> Result<Tensor<S, bool, Self>, Self::Err> {
        if !self.dev.has_func(Op::MODULE_NAME, Op::FWD_FN_NAME) {
            self.dev
                .load_ptx(Op::PTX_SRC.into(), Op::MODULE_NAME, &[Op::FWD_FN_NAME])?;
        }

        let shape = lhs.shape;
        let strides = lhs.shape.strides();
        let numel = shape.num_elements();

        let mut storage = unsafe { self.alloc_empty::<bool>(numel) }?;
        self.dev.memset_zeros(&mut storage)?;

        let dims: CudaSlice<usize> = self.dev.htod_copy(shape.concrete().into())?;
        let lhs_strides: CudaSlice<usize> = self.dev.htod_copy(lhs.strides.into())?;
        let out_strides: CudaSlice<usize> = self.dev.htod_copy(strides.into())?;

        let fwd_fn = self.dev.get_func(Op::MODULE_NAME, Op::FWD_FN_NAME).unwrap();
        let cfg = launch_cfg::<128>(numel as u32);
        let params = (
            numel,             // const size_t numel,
            S::NUM_DIMS,       // const size_t num_dims,
            &dims,             // const size_t *dims,
            lhs.data.as_ref(), // const float *lhs,
            &lhs_strides,      // const size_t *lhs_strides,
            scalar,            // float scalar,
            &mut storage,      // bool *out,
            &out_strides,      // const size_t *out_strides
        );
        unsafe { fwd_fn.launch(cfg, params) }?;
        Ok(self.build_tensor(shape, strides, storage))
    }
}

macro_rules! cmps {
    ($Op:ty, $TypeName:ty, $Fwd:tt, $ScalarFwd:tt) => {
        impl CmpOpCudaKernel<$TypeName> for $Op {
            const PTX_SRC: &'static str = PTX_SRC;
            const MODULE_NAME: &'static str = $Fwd;
            const FWD_FN_NAME: &'static str = $Fwd;
        }
        impl ScalarCmpOpCudaKernel<$TypeName> for $Op {
            const PTX_SRC: &'static str = PTX_SRC;
            const MODULE_NAME: &'static str = $ScalarFwd;
            const FWD_FN_NAME: &'static str = $ScalarFwd;
        }
    };
}

#[cfg(feature = "f16")]
cmps!(EqKernelOp, AMP<f16>, "eq_fwd_f16", "scalar_eq_fwd_f16");
#[cfg(feature = "f16")]
cmps!(NeKernelOp, AMP<f16>, "ne_fwd_f16", "scalar_ne_fwd_f16");
#[cfg(feature = "f16")]
cmps!(GtKernelOp, AMP<f16>, "gt_fwd_f16", "scalar_gt_fwd_f16");
#[cfg(feature = "f16")]
cmps!(GeKernelOp, AMP<f16>, "ge_fwd_f16", "scalar_ge_fwd_f16");
#[cfg(feature = "f16")]
cmps!(LtKernelOp, AMP<f16>, "lt_fwd_f16", "scalar_lt_fwd_f16");
#[cfg(feature = "f16")]
cmps!(LeKernelOp, AMP<f16>, "le_fwd_f16", "scalar_le_fwd_f16");

#[cfg(feature = "f16")]
cmps!(EqKernelOp, f16, "eq_fwd_f16", "scalar_eq_fwd_f16");
#[cfg(feature = "f16")]
cmps!(NeKernelOp, f16, "ne_fwd_f16", "scalar_ne_fwd_f16");
#[cfg(feature = "f16")]
cmps!(GtKernelOp, f16, "gt_fwd_f16", "scalar_gt_fwd_f16");
#[cfg(feature = "f16")]
cmps!(GeKernelOp, f16, "ge_fwd_f16", "scalar_ge_fwd_f16");
#[cfg(feature = "f16")]
cmps!(LtKernelOp, f16, "lt_fwd_f16", "scalar_lt_fwd_f16");
#[cfg(feature = "f16")]
cmps!(LeKernelOp, f16, "le_fwd_f16", "scalar_le_fwd_f16");

cmps!(EqKernelOp, f32, "eq_fwd_f32", "scalar_eq_fwd_f32");
cmps!(NeKernelOp, f32, "ne_fwd_f32", "scalar_ne_fwd_f32");
cmps!(GtKernelOp, f32, "gt_fwd_f32", "scalar_gt_fwd_f32");
cmps!(GeKernelOp, f32, "ge_fwd_f32", "scalar_ge_fwd_f32");
cmps!(LtKernelOp, f32, "lt_fwd_f32", "scalar_lt_fwd_f32");
cmps!(LeKernelOp, f32, "le_fwd_f32", "scalar_le_fwd_f32");

cmps!(EqKernelOp, f64, "eq_fwd_f64", "scalar_eq_fwd_f64");
cmps!(NeKernelOp, f64, "ne_fwd_f64", "scalar_ne_fwd_f64");
cmps!(GtKernelOp, f64, "gt_fwd_f64", "scalar_gt_fwd_f64");
cmps!(GeKernelOp, f64, "ge_fwd_f64", "scalar_ge_fwd_f64");
cmps!(LtKernelOp, f64, "lt_fwd_f64", "scalar_lt_fwd_f64");
cmps!(LeKernelOp, f64, "le_fwd_f64", "scalar_le_fwd_f64");