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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
use crate::intermediate_representation::*;
use crate::prelude::*;
use gcd::Gcd;
/// A strided interval of values with a fixed byte size.
///
/// The interval bounds are interpreted as signed integers,
/// i.e. `self.start` is not allowed to be greater than `self.end`
/// as signed integers.
///
/// The values represented by the interval are `start, start + stride, start + 2*stride, ... , end`.
/// The following invariants have to hold for a correct interval instance:
/// - `end - start % stride == 0`
/// - if `start == end`, then the stride should always be set to zero.
#[derive(Serialize, Deserialize, Debug, PartialEq, Eq, Hash, Clone)]
pub struct Interval {
/// The start of the interval. The bound is included in the represented interval.
pub start: Bitvector,
/// The end of the interval. The bound is included in the represented interval.
pub end: Bitvector,
/// The stride.
pub stride: u64,
}
impl Interval {
/// Construct a new interval.
///
/// Both `start` and `end` of the interval are inclusive,
/// i.e. contained in the represented interval.
///
/// The function automatically rounds down `end` (if necessary)
/// so that it is contained in the same residue class as the start value modulo the stride.
/// If the stride is 0 then `end` will be set to `start`.
pub fn new(start: Bitvector, end: Bitvector, stride: u64) -> Interval {
assert_eq!(start.width(), end.width());
let mut interval = Interval { start, end, stride };
interval.adjust_end_to_value_in_stride();
interval
}
/// Construct a new unconstrained interval.
pub fn new_top(bytesize: ByteSize) -> Interval {
Interval {
start: Bitvector::signed_min_value(bytesize.into()),
end: Bitvector::signed_max_value(bytesize.into()),
stride: 1,
}
}
/// Returns true if all values representable by bitvectors of the corresponding length are contained in the interval.
pub fn is_top(&self) -> bool {
(self.start.clone() - &Bitvector::one(self.start.width())) == self.end && self.stride == 1
}
/// Get the size in bytes of values contained in the interval.
pub fn bytesize(&self) -> ByteSize {
self.start.width().into()
}
/// Merge two intervals interpreting both as intervals of signed integers.
pub fn signed_merge(&self, other: &Interval) -> Interval {
if self.start.checked_sgt(&self.end).unwrap()
|| other.start.checked_sgt(&other.end).unwrap()
{
// One of the intervals wraps around
return Interval::new_top(self.bytesize());
}
let start = signed_min(&self.start, &other.start);
let end = signed_max(&self.end, &other.end);
let start_diff = if self.start.checked_sgt(&other.start).unwrap() {
self.start.clone() - &other.start
} else {
other.start.clone() - &self.start
};
let stride = if let Ok(start_diff) = start_diff.try_to_u64() {
self.stride.gcd(other.stride).gcd(start_diff)
} else {
1
};
Interval { start, end, stride }
}
/// Compute the intersection of two intervals as intervals of signed integers.
/// Return an error if the intersection is empty.
pub fn signed_intersect(&self, other: &Interval) -> Result<Interval, Error> {
let start = signed_max(&self.start, &other.start);
let end = signed_min(&self.end, &other.end);
if self.stride == 0 && other.stride == 0 {
if start == end {
return Ok(Interval {
start,
end,
stride: 0,
});
} else {
return Err(anyhow!("Empty interval"));
}
}
if self.start.bytesize() > ByteSize::new(8) {
// We ignore the stride for bytesizes larger than 8 bytes
let stride = !(start == end) as u64; // 0 if start == end, else 1
if start.checked_sle(&end).unwrap() {
return Ok(Interval { start, end, stride });
} else {
return Err(anyhow!("Empty interval"));
}
}
if let Some((stride, remainder)) = compute_intersection_residue_class(self, other)? {
Interval { start, end, stride }.adjust_to_stride_and_remainder(stride, remainder)
} else {
Err(anyhow!("Empty interval"))
}
}
/// If `self.start == self.end` set the stride to 0 and to 1 if `self.start < self.end`.
fn set_stride_to_unknown(&mut self) {
if self.start == self.end {
self.stride = 0;
} else {
self.stride = 1;
}
}
/// Round down `self.end` to the nearest value such that `self.end - self.start` is again divisible by the stride.
/// If afterwards `self.start == self.end` holds then set the stride to 0.
pub fn adjust_end_to_value_in_stride(&mut self) {
if self.stride == 0 {
self.end = self.start.clone();
return;
}
if self.stride == 1 && self.start != self.end {
return;
}
if let (Ok(start), Ok(end)) = (self.start.try_to_i64(), self.end.try_to_i64()) {
let diff = (end - start) as u64 % self.stride;
let diff = Bitvector::from_u64(diff).into_resize_unsigned(self.end.bytesize());
self.end.checked_sub_assign(&diff).unwrap();
if self.start == self.end {
self.stride = 0;
}
} else {
self.set_stride_to_unknown();
}
}
/// Round up `self.start` to the nearest value such that `self.end - self.start` is again divisible by the stride.
/// If afterwards `self.start == self.end` holds then set the stride to 0.
pub fn adjust_start_to_value_in_stride(&mut self) {
if self.stride == 0 {
self.start = self.end.clone();
return;
}
if self.stride == 1 && self.start != self.end {
return;
}
if let (Ok(start), Ok(end)) = (self.start.try_to_i64(), self.end.try_to_i64()) {
let diff = (end - start) as u64 % self.stride;
let diff = Bitvector::from_u64(diff).into_resize_unsigned(self.end.bytesize());
self.start.checked_add_assign(&diff).unwrap();
if self.start == self.end {
self.stride = 0;
}
} else {
self.set_stride_to_unknown();
}
}
/// Change the given interval such that it only contains values with the given remainder modulo the given stride.
/// This may round up the start of the interval and may round down the end of the interval.
/// If the resulting interval is empty then an error is returned.
/// This function ignores and replaces the previous stride of the interval.
///
/// For intervals with bytesize greater than 8 this function just returns the unmodified interval.
pub fn adjust_to_stride_and_remainder(
self,
stride: u64,
remainder: u64,
) -> Result<Self, Error> {
if self.bytesize() > ByteSize::new(8) {
return Ok(self);
}
let (mut start, mut end) = (
self.start.try_to_i128().unwrap(),
self.end.try_to_i128().unwrap(),
);
let diff = (remainder as i128 - start) % stride as i128;
let diff = (diff + stride as i128) % stride as i128;
start += diff;
let diff = (end - remainder as i128) % stride as i128;
let diff = (diff + stride as i128) % stride as i128;
end -= diff;
if start > i64::MAX as i128 || end < i64::MIN as i128 || start > end {
return Err(anyhow!("Empty interval"));
}
let start = Bitvector::from_i64(start as i64)
.into_truncate(self.start.bytesize())
.unwrap();
let end = Bitvector::from_i64(end as i64)
.into_truncate(self.end.bytesize())
.unwrap();
let stride = if start == end { 0 } else { stride };
Ok(Interval { start, end, stride })
}
/// Compute the interval represented if the byte size of the value is zero-extended.
pub fn zero_extend(self, width: ByteSize) -> Interval {
assert!(self.bytesize() <= width);
if self.bytesize() == width {
return self;
}
if self.start.sign_bit().to_bool() == self.end.sign_bit().to_bool() {
// Both start and end have the same sign
Interval {
start: self.start.into_zero_extend(width).unwrap(),
end: self.end.into_zero_extend(width).unwrap(),
stride: self.stride,
}
} else {
// The interval either contains both -1 and 0 or wraps around
if let Ok(start) = self.start.try_to_i128() {
let stride = 1 << self.stride.trailing_zeros();
let remainder = (start % stride + stride) % stride;
Interval {
start: Bitvector::zero(width.into()),
end: Bitvector::unsigned_max_value(self.end.width())
.into_zero_extend(width)
.unwrap(),
stride: stride as u64,
}
.adjust_to_stride_and_remainder(stride as u64, remainder as u64)
.unwrap()
} else {
Interval {
start: Bitvector::zero(width.into()),
end: Bitvector::unsigned_max_value(self.end.width())
.into_zero_extend(width)
.unwrap(),
stride: 1,
}
}
}
}
/// Truncate the bitvectors in the interval
/// by removing the least significant bytes lower than the `low_byte` from them.
pub fn subpiece_higher(self, low_byte: ByteSize) -> Self {
let start = self.start.subpiece(low_byte, self.bytesize() - low_byte);
let end = self.end.subpiece(low_byte, self.bytesize() - low_byte);
let stride = !(start == end) as u64;
Interval {
start: self.start.subpiece(low_byte, self.bytesize() - low_byte),
end: self.end.subpiece(low_byte, self.bytesize() - low_byte),
stride,
}
}
/// Truncate the bitvectors in the interval to `size`,
/// i.e. the most significant bytes (higher than `size`) are removed from all values.
pub fn subpiece_lower(self, size: ByteSize) -> Self {
let length = self.end.clone() - &self.start;
if length
.checked_ule(
&Bitvector::unsigned_max_value(size.into())
.into_zero_extend(self.bytesize())
.unwrap(),
)
.unwrap()
{
let start = self.start.into_truncate(size).unwrap();
let end = self.end.into_truncate(size).unwrap();
if start.checked_sle(&end).unwrap() {
return Interval {
start,
end,
stride: self.stride,
};
}
}
Self::new_top(size)
}
/// Take a subpiece of the bitvectors.
pub fn subpiece(mut self, low_byte: ByteSize, size: ByteSize) -> Self {
if low_byte != ByteSize::new(0) {
self = self.subpiece_higher(low_byte);
}
if self.bytesize() > size {
self = self.subpiece_lower(size);
}
self
}
/// Piece two intervals together, where `self` contains the most signifcant bytes
/// and `other` contains the least significant bytes of the resulting values.
pub fn piece(&self, other: &Interval) -> Self {
if other.start.sign_bit().to_bool() && !other.end.sign_bit().to_bool() {
// The `other` interval contains both negative and positive values.
let interval = Interval {
start: self
.start
.bin_op(BinOpType::Piece, &Bitvector::zero(other.start.width()))
.unwrap(),
end: self
.end
.bin_op(BinOpType::Piece, &(-Bitvector::one(other.end.width())))
.unwrap(),
stride: 1,
};
if other.bytesize() > ByteSize::new(8) {
interval
} else {
let stride = 1u64 << other.stride.trailing_zeros();
let remainder = other.start.try_to_i128().unwrap() % stride as i128;
let remainder = ((remainder + stride as i128) % stride as i128) as u64;
interval
.adjust_to_stride_and_remainder(stride, remainder)
.unwrap()
}
} else {
let stride = match (self.stride, other.stride) {
(0, _) => other.stride,
(_, 0) => self.stride << other.bytesize().as_bit_length(),
_ => 1u64 << other.stride.trailing_zeros(),
};
Interval {
start: self.start.bin_op(BinOpType::Piece, &other.start).unwrap(),
end: self.end.bin_op(BinOpType::Piece, &other.end).unwrap(),
stride,
}
}
}
/// Take the 2's complement of values in the interval.
pub fn int_2_comp(self) -> Self {
if self
.start
.checked_sgt(&Bitvector::signed_min_value(self.bytesize().into()))
.unwrap()
{
Interval {
start: -self.end,
end: -self.start,
stride: self.stride,
}
} else {
Interval::new_top(self.bytesize())
}
}
/// Compute the bitwise negation of values in the interval.
/// Only exact if there is exactly one value in the interval.
pub fn bitwise_not(self) -> Self {
if self.start == self.end {
self.start.into_bitnot().into()
} else {
Interval::new_top(self.bytesize())
}
}
/// Compute the interval of possible results
/// if one adds a value from `self` to a value from `rhs`.
pub fn add(&self, rhs: &Interval) -> Interval {
if let (Some(start), Some(end)) = (
self.start.signed_add_overflow_checked(&rhs.start),
self.end.signed_add_overflow_checked(&rhs.end),
) {
Interval {
start,
end,
stride: self.stride.gcd(rhs.stride),
}
} else {
Interval::new_top(self.bytesize())
}
}
/// Compute the interval of possible results
/// if one subtracts a value in `rhs` from a value in `self`.
pub fn sub(&self, rhs: &Interval) -> Interval {
if let (Some(start), Some(end)) = (
self.start.signed_sub_overflow_checked(&rhs.end),
self.end.signed_sub_overflow_checked(&rhs.start),
) {
Interval {
start,
end,
stride: self.stride.gcd(rhs.stride),
}
} else {
Interval::new_top(self.bytesize())
}
}
/// Compute the interval of possible results
/// if one multiplies a value in `self` with a value in `rhs`.
pub fn signed_mul(&self, rhs: &Interval) -> Interval {
if self.bytesize().as_bit_length() > 64 {
return Interval::new_top(self.bytesize());
}
let val1 = self
.start
.signed_mult_with_overflow_flag(&rhs.start)
.unwrap();
let val2 = self.start.signed_mult_with_overflow_flag(&rhs.end).unwrap();
let val3 = self.end.signed_mult_with_overflow_flag(&rhs.start).unwrap();
let val4 = self.end.signed_mult_with_overflow_flag(&rhs.end).unwrap();
if val1.1 || val2.1 || val3.1 || val4.1 {
// (signed) overflow during multiplication
return Interval::new_top(self.bytesize());
}
let min = signed_min(&val1.0, &signed_min(&val2.0, &signed_min(&val3.0, &val4.0)));
let max = signed_max(&val1.0, &signed_max(&val2.0, &signed_max(&val3.0, &val4.0)));
Interval {
start: min,
end: max,
stride: self.stride.gcd(rhs.stride),
}
}
/// Return `true` if `bitvec` is contained in the strided interval.
/// Panics if the interval and `bitvec` have different bytesizes.
pub fn contains(&self, bitvec: &Bitvector) -> bool {
if self.start == *bitvec {
return true;
}
self.start.checked_sle(bitvec).unwrap() && self.end.checked_sge(bitvec).unwrap() && {
if let Ok(diff) = (bitvec - &self.start).try_to_u64() {
self.stride > 0 && diff % self.stride == 0
} else {
true
}
}
}
}
impl From<Bitvector> for Interval {
/// Create an interval that only contains the given bitvector.
fn from(bitvec: Bitvector) -> Self {
Interval {
start: bitvec.clone(),
end: bitvec,
stride: 0,
}
}
}
/// Helper function returning the (signed) minimum of two bitvectors.
fn signed_min(v1: &Bitvector, v2: &Bitvector) -> Bitvector {
if v1.checked_sle(v2).unwrap() {
v1.clone()
} else {
v2.clone()
}
}
/// Helper function returning the (signed) maximum of two bitvectors.
fn signed_max(v1: &Bitvector, v2: &Bitvector) -> Bitvector {
if v1.checked_sge(v2).unwrap() {
v1.clone()
} else {
v2.clone()
}
}
/// The extended Euclidean algorithm.
///
/// Returns a triple `(gcd, x, y)` such that `gcd` is the greatest common divisor of `a` and `b`
/// and the following equation holds:
/// ```txt
/// gcd = x*a + y*b
/// ```
fn extended_gcd(a: i128, b: i128) -> (i128, i128, i128) {
if a == 0 {
(b, 0, 1)
} else {
let (g, left_inverse, right_inverse) = extended_gcd(b % a, a);
(g, right_inverse - (b / a) * left_inverse, left_inverse)
}
}
/// Compute the stride and the residue class of the intersection of the given intervals using the chinese remainder theorem.
/// The inputs are required to have byte sizes not larger than 8 (= 64bit).
///
/// If the intersection is empty, then `Ok(None)` is returned.
/// If an error occured during the computation (e.g. because of an integer overflow), then an error is returned.
/// Note that this also includes the case where the computed stride is larger than [`u64::MAX`].
fn compute_intersection_residue_class(
interval_left: &Interval,
interval_right: &Interval,
) -> Result<Option<(u64, u64)>, Error> {
match (interval_left.stride, interval_right.stride) {
(0, 0) => {
// both intervals contain exactly one value
if interval_left.start == interval_right.start {
return Ok(Some((0, 0)));
} else {
return Ok(None);
}
}
(0, _) => {
if interval_right.contains(&interval_left.start) {
let stride = interval_right.stride as i128;
let remainder = interval_right.start.try_to_i128()? % stride;
let remainder = (remainder + stride) % stride;
return Ok(Some((stride as u64, remainder as u64)));
} else {
return Ok(None);
}
}
(_, 0) => {
if interval_left.contains(&interval_right.start) {
let stride = interval_left.stride as i128;
let remainder = interval_left.start.try_to_i128()? % stride;
let remainder = (remainder + stride) % stride;
return Ok(Some((stride as u64, remainder as u64)));
} else {
return Ok(None);
}
}
_ => (),
}
// We compute everything in i128 to reduce the likelihood of integer overflows.
let (stride_left, stride_right) = (interval_left.stride as i128, interval_right.stride as i128);
let (base_left, base_right) = (
interval_left.start.try_to_i64().unwrap() as i128,
interval_right.start.try_to_i64().unwrap() as i128,
);
// The result of the extended euclidean algorithm satisfies
// `gcd = left_inverse * stride_left + right_inverse * stride_right`.
// For us most important is the equation system
// ```
// left_inverse * stride_left = 0 (modulo stride_left)
// left_inverse * stride_left = gcd (modulo stride_right)
// right_inverse * stride_right = gcd (modulo stride_left)
// right_inverse * stride_right = 0 (modulo stride_right)
// ```
let (gcd, left_inverse, right_inverse) = extended_gcd(stride_left, stride_right);
if base_left % gcd != base_right % gcd {
// The residue classes do not intersect, thus the intersection is empty.
Ok(None)
} else {
let lcm = (stride_left / gcd) * stride_right;
// The residue class of the intersection is computed such that the following equations hold:
// ```
// residue_class = base_right (modulo stride_right)
// residue_class = base_left (modulo stride_left)
// ```
// The `% lcm` operations are there to reduce the risk of integer overflows
let residue_class = ((base_right % lcm) / gcd * (left_inverse * stride_left)) % lcm // = base_right / gcd * gcd (modulo stride_right)
+ ((base_left % lcm) / gcd * (right_inverse * stride_right)) % lcm // = base_left / gcd * gcd (modulo stride_left)
+ base_left % gcd; // = base_left % gcd = base_right % gcd
// Ensure that the residue class is not negative
let residue_class = (residue_class + lcm) % lcm;
// Since we cannot rule out integer overflows for all possible inputs,
// we need to check the correctness of the result.
if lcm <= u64::MAX as i128
&& lcm % stride_left == 0
&& lcm % stride_right == 0
&& (base_left - residue_class) % stride_left == 0
&& (base_right - residue_class) % stride_right == 0
{
Ok(Some((lcm as u64, residue_class as u64)))
} else {
Err(anyhow!(
"Integer overflow during chinese remainder theorem computation."
))
}
}
}
#[cfg(test)]
mod tests;