1 | use core::ops; |
2 | |
3 | /// Minimal integer implementations needed on all integer types, including wide integers. |
4 | #[allow (dead_code)] |
5 | pub trait MinInt: |
6 | Copy |
7 | + core::fmt::Debug |
8 | + ops::BitOr<Output = Self> |
9 | + ops::Not<Output = Self> |
10 | + ops::Shl<u32, Output = Self> |
11 | { |
12 | /// Type with the same width but other signedness |
13 | type OtherSign: MinInt; |
14 | /// Unsigned version of Self |
15 | type UnsignedInt: MinInt; |
16 | |
17 | /// If `Self` is a signed integer |
18 | const SIGNED: bool; |
19 | |
20 | /// The bitwidth of the int type |
21 | const BITS: u32; |
22 | |
23 | const ZERO: Self; |
24 | const ONE: Self; |
25 | const MIN: Self; |
26 | const MAX: Self; |
27 | } |
28 | |
29 | /// Trait for some basic operations on integers |
30 | #[allow (dead_code)] |
31 | pub trait Int: |
32 | MinInt |
33 | + PartialEq |
34 | + PartialOrd |
35 | + ops::AddAssign |
36 | + ops::SubAssign |
37 | + ops::BitAndAssign |
38 | + ops::BitOrAssign |
39 | + ops::BitXorAssign |
40 | + ops::ShlAssign<i32> |
41 | + ops::ShrAssign<u32> |
42 | + ops::Add<Output = Self> |
43 | + ops::Sub<Output = Self> |
44 | + ops::Mul<Output = Self> |
45 | + ops::Div<Output = Self> |
46 | + ops::Shr<u32, Output = Self> |
47 | + ops::BitXor<Output = Self> |
48 | + ops::BitAnd<Output = Self> |
49 | { |
50 | /// LUT used for maximizing the space covered and minimizing the computational cost of fuzzing |
51 | /// in `testcrate`. For example, Self = u128 produces [0,1,2,7,8,15,16,31,32,63,64,95,96,111, |
52 | /// 112,119,120,125,126,127]. |
53 | const FUZZ_LENGTHS: [u8; 20] = make_fuzz_lengths(<Self as MinInt>::BITS); |
54 | |
55 | /// The number of entries of `FUZZ_LENGTHS` actually used. The maximum is 20 for u128. |
56 | const FUZZ_NUM: usize = { |
57 | let log2 = (<Self as MinInt>::BITS - 1).count_ones() as usize; |
58 | if log2 == 3 { |
59 | // case for u8 |
60 | 6 |
61 | } else { |
62 | // 3 entries on each extreme, 2 in the middle, and 4 for each scale of intermediate |
63 | // boundaries. |
64 | 8 + (4 * (log2 - 4)) |
65 | } |
66 | }; |
67 | |
68 | fn unsigned(self) -> Self::UnsignedInt; |
69 | fn from_unsigned(unsigned: Self::UnsignedInt) -> Self; |
70 | fn unsigned_abs(self) -> Self::UnsignedInt; |
71 | |
72 | fn from_bool(b: bool) -> Self; |
73 | |
74 | /// Prevents the need for excessive conversions between signed and unsigned |
75 | fn logical_shr(self, other: u32) -> Self; |
76 | |
77 | /// Absolute difference between two integers. |
78 | fn abs_diff(self, other: Self) -> Self::UnsignedInt; |
79 | |
80 | // copied from primitive integers, but put in a trait |
81 | fn is_zero(self) -> bool; |
82 | fn wrapping_neg(self) -> Self; |
83 | fn wrapping_add(self, other: Self) -> Self; |
84 | fn wrapping_mul(self, other: Self) -> Self; |
85 | fn wrapping_sub(self, other: Self) -> Self; |
86 | fn wrapping_shl(self, other: u32) -> Self; |
87 | fn wrapping_shr(self, other: u32) -> Self; |
88 | fn rotate_left(self, other: u32) -> Self; |
89 | fn overflowing_add(self, other: Self) -> (Self, bool); |
90 | fn leading_zeros(self) -> u32; |
91 | fn ilog2(self) -> u32; |
92 | } |
93 | |
94 | pub(crate) const fn make_fuzz_lengths(bits: u32) -> [u8; 20] { |
95 | let mut v = [0u8; 20]; |
96 | v[0] = 0; |
97 | v[1] = 1; |
98 | v[2] = 2; // important for parity and the iX::MIN case when reversed |
99 | let mut i = 3; |
100 | |
101 | // No need for any more until the byte boundary, because there should be no algorithms |
102 | // that are sensitive to anything not next to byte boundaries after 2. We also scale |
103 | // in powers of two, which is important to prevent u128 corner tests from getting too |
104 | // big. |
105 | let mut l = 8; |
106 | loop { |
107 | if l >= ((bits / 2) as u8) { |
108 | break; |
109 | } |
110 | // get both sides of the byte boundary |
111 | v[i] = l - 1; |
112 | i += 1; |
113 | v[i] = l; |
114 | i += 1; |
115 | l *= 2; |
116 | } |
117 | |
118 | if bits != 8 { |
119 | // add the lower side of the middle boundary |
120 | v[i] = ((bits / 2) - 1) as u8; |
121 | i += 1; |
122 | } |
123 | |
124 | // We do not want to jump directly from the Self::BITS/2 boundary to the Self::BITS |
125 | // boundary because of algorithms that split the high part up. We reverse the scaling |
126 | // as we go to Self::BITS. |
127 | let mid = i; |
128 | let mut j = 1; |
129 | loop { |
130 | v[i] = (bits as u8) - (v[mid - j]) - 1; |
131 | if j == mid { |
132 | break; |
133 | } |
134 | i += 1; |
135 | j += 1; |
136 | } |
137 | v |
138 | } |
139 | |
140 | macro_rules! int_impl_common { |
141 | ($ty:ty) => { |
142 | fn from_bool(b: bool) -> Self { |
143 | b as $ty |
144 | } |
145 | |
146 | fn logical_shr(self, other: u32) -> Self { |
147 | Self::from_unsigned(self.unsigned().wrapping_shr(other)) |
148 | } |
149 | |
150 | fn is_zero(self) -> bool { |
151 | self == Self::ZERO |
152 | } |
153 | |
154 | fn wrapping_neg(self) -> Self { |
155 | <Self>::wrapping_neg(self) |
156 | } |
157 | |
158 | fn wrapping_add(self, other: Self) -> Self { |
159 | <Self>::wrapping_add(self, other) |
160 | } |
161 | |
162 | fn wrapping_mul(self, other: Self) -> Self { |
163 | <Self>::wrapping_mul(self, other) |
164 | } |
165 | fn wrapping_sub(self, other: Self) -> Self { |
166 | <Self>::wrapping_sub(self, other) |
167 | } |
168 | |
169 | fn wrapping_shl(self, other: u32) -> Self { |
170 | <Self>::wrapping_shl(self, other) |
171 | } |
172 | |
173 | fn wrapping_shr(self, other: u32) -> Self { |
174 | <Self>::wrapping_shr(self, other) |
175 | } |
176 | |
177 | fn rotate_left(self, other: u32) -> Self { |
178 | <Self>::rotate_left(self, other) |
179 | } |
180 | |
181 | fn overflowing_add(self, other: Self) -> (Self, bool) { |
182 | <Self>::overflowing_add(self, other) |
183 | } |
184 | |
185 | fn leading_zeros(self) -> u32 { |
186 | <Self>::leading_zeros(self) |
187 | } |
188 | |
189 | fn ilog2(self) -> u32 { |
190 | <Self>::ilog2(self) |
191 | } |
192 | }; |
193 | } |
194 | |
195 | macro_rules! int_impl { |
196 | ($ity:ty, $uty:ty) => { |
197 | impl MinInt for $uty { |
198 | type OtherSign = $ity; |
199 | type UnsignedInt = $uty; |
200 | |
201 | const BITS: u32 = <Self as MinInt>::ZERO.count_zeros(); |
202 | const SIGNED: bool = Self::MIN != Self::ZERO; |
203 | |
204 | const ZERO: Self = 0; |
205 | const ONE: Self = 1; |
206 | const MIN: Self = <Self>::MIN; |
207 | const MAX: Self = <Self>::MAX; |
208 | } |
209 | |
210 | impl Int for $uty { |
211 | fn unsigned(self) -> $uty { |
212 | self |
213 | } |
214 | |
215 | // It makes writing macros easier if this is implemented for both signed and unsigned |
216 | #[allow(clippy::wrong_self_convention)] |
217 | fn from_unsigned(me: $uty) -> Self { |
218 | me |
219 | } |
220 | |
221 | fn unsigned_abs(self) -> Self { |
222 | self |
223 | } |
224 | |
225 | fn abs_diff(self, other: Self) -> Self { |
226 | self.abs_diff(other) |
227 | } |
228 | |
229 | int_impl_common!($uty); |
230 | } |
231 | |
232 | impl MinInt for $ity { |
233 | type OtherSign = $uty; |
234 | type UnsignedInt = $uty; |
235 | |
236 | const BITS: u32 = <Self as MinInt>::ZERO.count_zeros(); |
237 | const SIGNED: bool = Self::MIN != Self::ZERO; |
238 | |
239 | const ZERO: Self = 0; |
240 | const ONE: Self = 1; |
241 | const MIN: Self = <Self>::MIN; |
242 | const MAX: Self = <Self>::MAX; |
243 | } |
244 | |
245 | impl Int for $ity { |
246 | fn unsigned(self) -> $uty { |
247 | self as $uty |
248 | } |
249 | |
250 | fn from_unsigned(me: $uty) -> Self { |
251 | me as $ity |
252 | } |
253 | |
254 | fn unsigned_abs(self) -> Self::UnsignedInt { |
255 | self.unsigned_abs() |
256 | } |
257 | |
258 | fn abs_diff(self, other: Self) -> $uty { |
259 | self.abs_diff(other) |
260 | } |
261 | |
262 | int_impl_common!($ity); |
263 | } |
264 | }; |
265 | } |
266 | |
267 | int_impl!(isize, usize); |
268 | int_impl!(i8, u8); |
269 | int_impl!(i16, u16); |
270 | int_impl!(i32, u32); |
271 | int_impl!(i64, u64); |
272 | int_impl!(i128, u128); |
273 | |
274 | /// Trait for integers twice the bit width of another integer. This is implemented for all |
275 | /// primitives except for `u8`, because there is not a smaller primitive. |
276 | pub trait DInt: MinInt { |
277 | /// Integer that is half the bit width of the integer this trait is implemented for |
278 | type H: HInt<D = Self>; |
279 | |
280 | /// Returns the low half of `self` |
281 | fn lo(self) -> Self::H; |
282 | /// Returns the high half of `self` |
283 | fn hi(self) -> Self::H; |
284 | /// Returns the low and high halves of `self` as a tuple |
285 | fn lo_hi(self) -> (Self::H, Self::H) { |
286 | (self.lo(), self.hi()) |
287 | } |
288 | /// Constructs an integer using lower and higher half parts |
289 | fn from_lo_hi(lo: Self::H, hi: Self::H) -> Self { |
290 | lo.zero_widen() | hi.widen_hi() |
291 | } |
292 | } |
293 | |
294 | /// Trait for integers half the bit width of another integer. This is implemented for all |
295 | /// primitives except for `u128`, because it there is not a larger primitive. |
296 | pub trait HInt: Int { |
297 | /// Integer that is double the bit width of the integer this trait is implemented for |
298 | type D: DInt<H = Self> + MinInt; |
299 | |
300 | // NB: some of the below methods could have default implementations (e.g. `widen_hi`), but for |
301 | // unknown reasons this can cause infinite recursion when optimizations are disabled. See |
302 | // <https://github.com/rust-lang/compiler-builtins/pull/707> for context. |
303 | |
304 | /// Widens (using default extension) the integer to have double bit width |
305 | fn widen(self) -> Self::D; |
306 | /// Widens (zero extension only) the integer to have double bit width. This is needed to get |
307 | /// around problems with associated type bounds (such as `Int<Othersign: DInt>`) being unstable |
308 | fn zero_widen(self) -> Self::D; |
309 | /// Widens the integer to have double bit width and shifts the integer into the higher bits |
310 | fn widen_hi(self) -> Self::D; |
311 | /// Widening multiplication with zero widening. This cannot overflow. |
312 | fn zero_widen_mul(self, rhs: Self) -> Self::D; |
313 | /// Widening multiplication. This cannot overflow. |
314 | fn widen_mul(self, rhs: Self) -> Self::D; |
315 | } |
316 | |
317 | macro_rules! impl_d_int { |
318 | ($($X:ident $D:ident),*) => { |
319 | $( |
320 | impl DInt for $D { |
321 | type H = $X; |
322 | |
323 | fn lo(self) -> Self::H { |
324 | self as $X |
325 | } |
326 | fn hi(self) -> Self::H { |
327 | (self >> <$X as MinInt>::BITS) as $X |
328 | } |
329 | } |
330 | )* |
331 | }; |
332 | } |
333 | |
334 | macro_rules! impl_h_int { |
335 | ($($H:ident $uH:ident $X:ident),*) => { |
336 | $( |
337 | impl HInt for $H { |
338 | type D = $X; |
339 | |
340 | fn widen(self) -> Self::D { |
341 | self as $X |
342 | } |
343 | fn zero_widen(self) -> Self::D { |
344 | (self as $uH) as $X |
345 | } |
346 | fn zero_widen_mul(self, rhs: Self) -> Self::D { |
347 | self.zero_widen().wrapping_mul(rhs.zero_widen()) |
348 | } |
349 | fn widen_mul(self, rhs: Self) -> Self::D { |
350 | self.widen().wrapping_mul(rhs.widen()) |
351 | } |
352 | fn widen_hi(self) -> Self::D { |
353 | (self as $X) << <Self as MinInt>::BITS |
354 | } |
355 | } |
356 | )* |
357 | }; |
358 | } |
359 | |
360 | impl_d_int!(u8 u16, u16 u32, u32 u64, u64 u128, i8 i16, i16 i32, i32 i64, i64 i128); |
361 | impl_h_int!( |
362 | u8 u8 u16, |
363 | u16 u16 u32, |
364 | u32 u32 u64, |
365 | u64 u64 u128, |
366 | i8 u8 i16, |
367 | i16 u16 i32, |
368 | i32 u32 i64, |
369 | i64 u64 i128 |
370 | ); |
371 | |
372 | /// Trait to express (possibly lossy) casting of integers |
373 | pub trait CastInto<T: Copy>: Copy { |
374 | fn cast(self) -> T; |
375 | } |
376 | |
377 | pub trait CastFrom<T: Copy>: Copy { |
378 | fn cast_from(value: T) -> Self; |
379 | } |
380 | |
381 | impl<T: Copy, U: CastInto<T> + Copy> CastFrom<U> for T { |
382 | fn cast_from(value: U) -> Self { |
383 | value.cast() |
384 | } |
385 | } |
386 | |
387 | macro_rules! cast_into { |
388 | ($ty:ty) => { |
389 | cast_into!($ty; usize, isize, u8, i8, u16, i16, u32, i32, u64, i64, u128, i128); |
390 | }; |
391 | ($ty:ty; $($into:ty),*) => {$( |
392 | impl CastInto<$into> for $ty { |
393 | fn cast(self) -> $into { |
394 | self as $into |
395 | } |
396 | } |
397 | )*}; |
398 | } |
399 | |
400 | cast_into!(usize); |
401 | cast_into!(isize); |
402 | cast_into!(u8); |
403 | cast_into!(i8); |
404 | cast_into!(u16); |
405 | cast_into!(i16); |
406 | cast_into!(u32); |
407 | cast_into!(i32); |
408 | cast_into!(u64); |
409 | cast_into!(i64); |
410 | cast_into!(u128); |
411 | cast_into!(i128); |
412 | |