moose 0.2.2

Encrypted learning and data processing framework
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
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356

use super::*;
use crate::host::{RawPrfKey, RawSeed, RawShape};
use crate::TensorLike;

pub trait PlacementFill<S: Session, ShapeT, O> {
    fn fill(&self, sess: &S, value: Constant, shape: &ShapeT) -> O;
}

modelled_kernel! {
    PlacementFill::fill, FillOp{value: Constant},
    [
        (HostPlacement, (HostShape) -> HostRing64Tensor => [runtime] custom |op| {
            let value = match op.value {
                Constant::Ring64(v) => v,
                _ => return Err(Error::KernelError(
                    "Cannot fill HostRing64Tensor with non-Ring64 value.".to_string(),
                ))
            };
            Ok(Box::new(move |sess, host, shape| {
                Self::host_ring64_kernel(sess, host, value, shape)
            }))
        }),
        (HostPlacement, (HostShape) -> HostRing128Tensor => [runtime] custom |op| {
            let value = match op.value {
                Constant::Ring128(v) => v,
                _ => return Err(Error::KernelError(
                    "Cannot fill HostRing64Tensor with non-Ring128 value.".to_string(),
                ))
            };
            Ok(Box::new(move |sess, host, shape| {
                Self::host_ring128_kernel(sess, host, value, shape)
            }))
        }),
        (AdditivePlacement, (HostShape) -> AdditiveRing64Tensor => [hybrid] Self::adt_host_kernel),
        (AdditivePlacement, (HostShape) -> AdditiveRing128Tensor => [hybrid] Self::adt_host_kernel),
        (AdditivePlacement, (AdditiveShape) -> AdditiveRing64Tensor => [concrete] Self::adt_adt_kernel),
        (AdditivePlacement, (AdditiveShape) -> AdditiveRing128Tensor => [concrete] Self::adt_adt_kernel),
        (HostPlacement, (HostShape) -> HostBitTensor => [runtime] custom |op| {
            use std::convert::TryInto;
            let value: u8 = match op.value {
                Constant::Bit(v) => v,
                Constant::Ring64(v) => v.try_into().map_err(|_| {
                    Error::KernelError("Cannot fill HostBitTensor with non-binary value.".to_string())
                })?,
                Constant::Ring128(v) => v.try_into().map_err(|_| {
                    Error::KernelError("Cannot fill HostBitTensor with non-binary value.".to_string())
                })?,
                _ => {
                    return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a HostBitTensor", op.value.ty())))
                }
            };
            if !(value == 0 || value == 1) {
                return Err(Error::KernelError(
                    "Cannot fill HostBitTensor with non-binary value.".to_string(),
                ));
            }
            assert!(value == 0 || value == 1);
            Ok(Box::new(move |sess, host, host_shape| {
                Self::host_bit_kernel(sess, host, value, host_shape)
            }))
        }),
        (ReplicatedPlacement, (ReplicatedShape) -> ReplicatedRing64Tensor => [concrete] custom |op| {
                let value: u64 = match op.value {
                    Constant::Bit(v) => v as u64,
                    Constant::Ring64(v) => v,
                    Constant::Float64(v) => v as u64,
                    Constant::Fixed(FixedpointConstant {
                        value, precision
                    }) => {
                        ((value * ((1u64 << precision) as f64)) as i64) as u64
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a ReplicatedRing64Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::rep_ring64_kernel(sess, rep, value, rep_shape)
                }))
            }),
        (Mirrored3Placement, (ReplicatedShape) -> Mirrored3Ring64Tensor => [concrete] custom |op| {
                let value: u64 = match op.value {
                    Constant::Bit(v) => v as u64,
                    Constant::Ring64(v) => v,
                    Constant::Float64(v) => v as u64,
                    Constant::Fixed(FixedpointConstant {
                        value, precision
                    }) => {
                        ((value * ((1u64 << precision) as f64)) as i64) as u64
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a Mirrored3Ring64Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::mir_ring64_kernel(sess, rep, value, rep_shape)
                }))
        }),
        (ReplicatedPlacement, (ReplicatedShape) -> ReplicatedRing128Tensor => [concrete] custom |op| {
                let value: u128 = match op.value {
                    Constant::Bit(v) => v as u128,
                    Constant::Ring64(v) => v as u128,
                    Constant::Ring128(v) => v,
                    Constant::Float64(v) => v as u128,
                    Constant::Fixed(FixedpointConstant{value, precision}) => {
                        ((value * ((1u128 << precision) as f64)) as i128) as u128
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a ReplicatedRing128Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::rep_ring128_kernel(sess, rep, value, rep_shape)
                }))
        }),
        (Mirrored3Placement, (ReplicatedShape) -> Mirrored3Ring128Tensor => [concrete] custom |op| {
                let value: u128 = match op.value {
                    Constant::Bit(v) => v as u128,
                    Constant::Ring64(v) => v as u128,
                    Constant::Ring128(v) => v,
                    Constant::Float64(v) => v as u128,
                    Constant::Fixed(FixedpointConstant{value, precision}) => {
                        ((value * ((1u128 << precision) as f64)) as i128) as u128
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a Mirrored3Ring128Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::mir_ring128_kernel(sess, rep, value, rep_shape)
                }))
        }),
        (ReplicatedPlacement, (ReplicatedShape) -> ReplicatedBitTensor => [concrete] custom |op| {
                let value: u8 = match op.value {
                    Constant::Bit(v) => v,
                    Constant::Ring64(v) => v as u8,
                    Constant::Ring128(v) => v as u8,
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a ReplicatedBitTensor", op.value.ty()))),
                };
                if value != 0 && value != 1 {
                    return Err(Error::InvalidArgument(format!("Could only support 0 and 1 for the bit tensor fill, got {}", value)));
                }
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::rep_bit_kernel(sess, rep, value, rep_shape)
                }))
        }),
        (Mirrored3Placement, (ReplicatedShape) -> Mirrored3BitTensor => [concrete] custom |op| {
                let value: u8 = match op.value {
                    Constant::Bit(v) => v,
                    Constant::Ring64(v) => v as u8,
                    Constant::Ring128(v) => v as u8,
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a Mirrored3BitTensor", op.value.ty()))),
                };
                if value != 0 && value != 1 {
                    return Err(Error::InvalidArgument(format!("Could only support 0 and 1 for the bit tensor fill, got {}", value)));
                }
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::mir_bit_kernel(sess, rep, value, rep_shape)
                }))
        }),
        (Mirrored3Placement, (ReplicatedShape) -> Mirrored3Fixed64Tensor => [hybrid] custom |op| {
                let (ring_value, fractional_precision, integral_precision) = match op.value {
                    Constant::Fixed(FixedpointConstant{value, precision}) => {
                        let ring_value: u64 = ((value * ((1u64 << precision) as f64)) as i64) as u64;
                        let fractional_precision = precision as u32;
                        let integral_precision = value.log2().ceil() as u32;
                        (ring_value, fractional_precision, integral_precision)
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a Mirrored3Fixed64Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::mir_fixed_kernel(sess, rep, Constant::Ring64(ring_value), rep_shape, fractional_precision, integral_precision)
                }))
        }),
        (Mirrored3Placement, (ReplicatedShape) -> Mirrored3Fixed128Tensor => [hybrid] custom |op| {
                let (ring_value, fractional_precision, integral_precision) = match op.value {
                    Constant::Fixed(FixedpointConstant{value, precision}) => {
                        let ring_value: u128 = ((value * ((1u128 << precision) as f64)) as i128) as u128;
                        let fractional_precision = precision as u32;
                        let integral_precision = value.log2().ceil() as u32;
                        (ring_value, fractional_precision, integral_precision)
                    },
                    _ => return Err(Error::UnimplementedOperator(
                        format!("Cannot fill from {:?} into a Mirrored3Fixed128Tensor", op.value.ty()))),
                };
                Ok(Box::new(move |sess, rep, rep_shape| {
                    Self::mir_fixed_kernel(sess, rep, Constant::Ring128(ring_value), rep_shape, fractional_precision, integral_precision)
                }))
        }),
    ]
}

pub trait PlacementZeros<S: Session, ShapeT, O> {
    fn zeros(&self, sess: &S, shape: &ShapeT) -> O;
}

impl<S: Session, ShapeT, O, P> PlacementZeros<S, ShapeT, O> for P
where
    P: PlacementFill<S, ShapeT, O>,
    O: TensorLike,
    O::Scalar: Into<Constant>,
    O::Scalar: From<u8>,
{
    fn zeros(&self, sess: &S, shape: &ShapeT) -> O {
        let value = O::Scalar::from(0).into();
        self.fill(sess, value, shape)
    }
}

pub trait PlacementOnes<S: Session, ShapeT, O> {
    fn ones(&self, sess: &S, shape: &ShapeT) -> O;
}

impl<S: Session, ShapeT, O, P> PlacementOnes<S, ShapeT, O> for P
where
    P: PlacementFill<S, ShapeT, O>,
    O: TensorLike,
    O::Scalar: Into<Constant>,
    O::Scalar: From<u8>,
{
    fn ones(&self, sess: &S, shape: &ShapeT) -> O {
        let value = O::Scalar::from(1).into();
        self.fill(sess, value, shape)
    }
}

modelled_kernel! {
    PlacementOnes::ones, OnesOp,
    [
        (HostPlacement, (Shape) -> Tensor => [concrete] custom |op| {
            use crate::logical::{AbstractTensor, TensorDType};
            match op.sig.ret() {
                Ty::Tensor(TensorDType::Float32) => Ok(Box::new(move |sess, plc, shape| {
                    Self::logical_host_kernel::<_, Float32Tensor, _, _>(sess, plc, shape).map(AbstractTensor::Float32)
                })),
                Ty::Tensor(TensorDType::Float64) => Ok(Box::new(move |sess, plc, shape| {
                    Self::logical_host_kernel::<_, Float64Tensor, _, _>(sess, plc, shape).map(AbstractTensor::Float64)
                })),
                other => {
                    return Err(Error::UnimplementedOperator(
                        format!("Cannot build ones of type {:?}", other)))
                },
            }
        }),
        // We do not support the ReplicatedPlacement: PlacementFill yet, hence we do not support Ones.
        // Also, logical Tensor can only hold Host tensors at the moment.
        // (ReplicatedPlacement, (HostShape) -> Tensor => [hybrid] Self::logical_rep_kernel),
        (HostPlacement, (HostShape) -> Float32Tensor => [hybrid] Self::host_float_kernel),
        (HostPlacement, (HostShape) -> Float64Tensor => [hybrid] Self::host_float_kernel),
        (HostPlacement, (HostShape) -> HostFloat32Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostFloat64Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt8Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt16Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt32Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt64Tensor => [runtime] Self::host_kernel),
    ]
}

modelled_kernel! {
    PlacementZeros::zeros, ZerosOp,
    [
        (HostPlacement, (Shape) -> Tensor => [concrete] custom |op| {
            use crate::logical::{AbstractTensor, TensorDType};
            match op.sig.ret() {
                Ty::Tensor(TensorDType::Float32) => Ok(Box::new(move |sess, plc, shape| {
                    Self::logical_host_kernel::<_, Float32Tensor, _, _>(sess, plc, shape).map(AbstractTensor::Float32)
                })),
                Ty::Tensor(TensorDType::Float64) => Ok(Box::new(move |sess, plc, shape| {
                    Self::logical_host_kernel::<_, Float64Tensor, _, _>(sess, plc, shape).map(AbstractTensor::Float64)
                })),
                other => {
                    return Err(Error::UnimplementedOperator(
                        format!("Cannot build zeros of type {:?}", other)))
                },
            }
        }),
        // We do not support the ReplicatedPlacement: PlacementFill yet, hence we do not support Zeros.
        // Also, logical Tensor can only hold Host tensors at the moment.
        // (ReplicatedPlacement, (HostShape) -> Tensor => [hybrid] Self::logical_rep_kernel),
        (HostPlacement, (HostShape) -> Float32Tensor => [hybrid] Self::host_float_kernel),
        (HostPlacement, (HostShape) -> Float64Tensor => [hybrid] Self::host_float_kernel),
        (HostPlacement, (HostShape) -> HostFloat32Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostFloat64Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt8Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt16Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt32Tensor => [runtime] Self::host_kernel),
        (HostPlacement, (HostShape) -> HostInt64Tensor => [runtime] Self::host_kernel),
    ]
}

pub trait PlacementConstant<S: Session, O> {
    fn constant(&self, sess: &S, value: Constant) -> O;
}

// TODO(Morten) Since other work is updating the kernel macros I did not want to change
// any of these right now to match the need for individual `attributes[value: T]`.
// If we want we can get rid of `unwrapper` in the near future by updating the macros.

macro_rules! unwrapper {
    ($op:ident, $t:ident, $k:path) => {{
        let value: $t = match &$op.value {
            Constant::$t(v) => Ok(v.clone()),
            _ => Err(crate::error::Error::UnimplementedOperator("".to_string())),
        }?;
        Ok(Box::new(move |sess, plc| $k(sess, plc, value.clone())))
    }};
}

modelled_kernel! {
    PlacementConstant::constant, ConstantOp{value: Constant},
    [
        (HostPlacement, () -> HostRing64Tensor => [runtime] custom |op| unwrapper!(op, HostRing64Tensor, Self::kernel)),
        (HostPlacement, () -> HostRing128Tensor => [runtime] custom |op| unwrapper!(op, HostRing128Tensor, Self::kernel)),
        (HostPlacement, () -> HostFloat32Tensor => [runtime] custom |op| unwrapper!(op, HostFloat32Tensor, Self::kernel)),
        (HostPlacement, () -> HostFloat64Tensor => [runtime] custom |op| unwrapper!(op, HostFloat64Tensor, Self::kernel)),
        (HostPlacement, () -> HostInt8Tensor => [runtime] custom |op| unwrapper!(op, HostInt8Tensor, Self::kernel)),
        (HostPlacement, () -> HostInt16Tensor => [runtime] custom |op| unwrapper!(op, HostInt16Tensor, Self::kernel)),
        (HostPlacement, () -> HostInt32Tensor => [runtime] custom |op| unwrapper!(op, HostInt32Tensor, Self::kernel)),
        (HostPlacement, () -> HostInt64Tensor => [runtime] custom |op| unwrapper!(op, HostInt64Tensor, Self::kernel)),
        (HostPlacement, () -> HostUint8Tensor => [runtime] custom |op| unwrapper!(op, HostUint8Tensor, Self::kernel)),
        (HostPlacement, () -> HostUint16Tensor => [runtime] custom |op| unwrapper!(op, HostUint16Tensor, Self::kernel)),
        (HostPlacement, () -> HostUint32Tensor => [runtime] custom |op| unwrapper!(op, HostUint32Tensor, Self::kernel)),
        (HostPlacement, () -> HostUint64Tensor => [runtime] custom |op| unwrapper!(op, HostUint64Tensor, Self::kernel)),
        (HostPlacement, () -> HostBitTensor => [runtime] custom |op| unwrapper!(op, HostBitTensor, Self::kernel)),
        (HostPlacement, () -> HostString => [runtime] custom |op| unwrapper!(op, String, Self::string_kernel)),
        (HostPlacement, () -> HostPrfKey => [runtime] custom |op| unwrapper!(op, RawPrfKey, Self::prf_key_kernel)),
        (HostPlacement, () -> HostSeed => [runtime] custom |op| unwrapper!(op, RawSeed, Self::seed_kernel)),
        (HostPlacement, () -> Tensor => [concrete] custom |op| {
            let sig = op.sig;
            let value = op.value.clone();
            Ok(Box::new(move |sess, plc| {
                Self::logical_kernel(sess, plc, sig, value.clone())
            }))
        }),
        (HostPlacement, () -> Float32Tensor => [concrete] Self::float_kernel),
        (HostPlacement, () -> Float64Tensor => [concrete] Self::float_kernel),
        (HostPlacement, () -> Uint64Tensor => [concrete] Self::u64_kernel),
        (HostPlacement, () -> BooleanTensor => [concrete] Self::bool_kernel),
        (Mirrored3Placement, () -> Tensor => [concrete] custom |op| {
            let sig = op.sig;
            let value = op.value.clone();
            Ok(Box::new(move |sess, plc| {
                Self::mir3_logical_kernel(sess, plc, sig, value.clone())
            }))
        }),
        (Mirrored3Placement, () -> Float32Tensor => [concrete] Self::mir3_float_kernel),
        (Mirrored3Placement, () -> Float64Tensor => [concrete] Self::mir3_float_kernel),
        (HostPlacement, () -> Shape => [concrete] custom |op| {
            let sig = op.sig;
            let value = op.value.clone();
            Ok(Box::new(move |sess, plc| {
                Self::shape_logical_kernel(sess, plc, sig, value.clone())
            }))
        }),
        (HostPlacement, () -> HostShape => [runtime] custom |op| unwrapper!(op, RawShape, Self::shape_kernel)),
    ]
}