1 //===- ConstantRange.h - Represent a range ----------------------*- C++ -*-===//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 // Represent a range of possible values that may occur when the program is run
10 // for an integral value. This keeps track of a lower and upper bound for the
11 // constant, which MAY wrap around the end of the numeric range. To do this, it
12 // keeps track of a [lower, upper) bound, which specifies an interval just like
13 // STL iterators. When used with boolean values, the following are important
16 // [F, F) = {} = Empty set
19 // [T, T) = {F, T} = Full set
21 // The other integral ranges use min/max values for special range values. For
22 // example, for 8-bit types, it uses:
23 // [0, 0) = {} = Empty set
24 // [255, 255) = {0..255} = Full Set
26 // Note that ConstantRange can be used to represent either signed or
29 //===----------------------------------------------------------------------===//
31 #ifndef LLVM_IR_CONSTANTRANGE_H
32 #define LLVM_IR_CONSTANTRANGE_H
34 #include "llvm/ADT/APInt.h"
35 #include "llvm/IR/InstrTypes.h"
36 #include "llvm/IR/Instruction.h"
37 #include "llvm/Support/Compiler.h"
46 /// This class represents a range of values.
47 class LLVM_NODISCARD ConstantRange
{
50 /// Create empty constant range with same bitwidth.
51 ConstantRange
getEmpty() const {
52 return ConstantRange(getBitWidth(), false);
55 /// Create full constant range with same bitwidth.
56 ConstantRange
getFull() const {
57 return ConstantRange(getBitWidth(), true);
61 /// Initialize a full or empty set for the specified bit width.
62 explicit ConstantRange(uint32_t BitWidth
, bool isFullSet
);
64 /// Initialize a range to hold the single specified value.
65 ConstantRange(APInt Value
);
67 /// Initialize a range of values explicitly. This will assert out if
68 /// Lower==Upper and Lower != Min or Max value for its type. It will also
69 /// assert out if the two APInt's are not the same bit width.
70 ConstantRange(APInt Lower
, APInt Upper
);
72 /// Create empty constant range with the given bit width.
73 static ConstantRange
getEmpty(uint32_t BitWidth
) {
74 return ConstantRange(BitWidth
, false);
77 /// Create full constant range with the given bit width.
78 static ConstantRange
getFull(uint32_t BitWidth
) {
79 return ConstantRange(BitWidth
, true);
82 /// Create non-empty constant range with the given bounds. If Lower and
83 /// Upper are the same, a full range is returned.
84 static ConstantRange
getNonEmpty(APInt Lower
, APInt Upper
) {
86 return getFull(Lower
.getBitWidth());
87 return ConstantRange(std::move(Lower
), std::move(Upper
));
90 /// Initialize a range based on a known bits constraint. The IsSigned flag
91 /// indicates whether the constant range should not wrap in the signed or
93 static ConstantRange
fromKnownBits(const KnownBits
&Known
, bool IsSigned
);
95 /// Produce the smallest range such that all values that may satisfy the given
96 /// predicate with any value contained within Other is contained in the
97 /// returned range. Formally, this returns a superset of
98 /// 'union over all y in Other . { x : icmp op x y is true }'. If the exact
99 /// answer is not representable as a ConstantRange, the return value will be a
100 /// proper superset of the above.
102 /// Example: Pred = ult and Other = i8 [2, 5) returns Result = [0, 4)
103 static ConstantRange
makeAllowedICmpRegion(CmpInst::Predicate Pred
,
104 const ConstantRange
&Other
);
106 /// Produce the largest range such that all values in the returned range
107 /// satisfy the given predicate with all values contained within Other.
108 /// Formally, this returns a subset of
109 /// 'intersection over all y in Other . { x : icmp op x y is true }'. If the
110 /// exact answer is not representable as a ConstantRange, the return value
111 /// will be a proper subset of the above.
113 /// Example: Pred = ult and Other = i8 [2, 5) returns [0, 2)
114 static ConstantRange
makeSatisfyingICmpRegion(CmpInst::Predicate Pred
,
115 const ConstantRange
&Other
);
117 /// Produce the exact range such that all values in the returned range satisfy
118 /// the given predicate with any value contained within Other. Formally, this
119 /// returns the exact answer when the superset of 'union over all y in Other
120 /// is exactly same as the subset of intersection over all y in Other.
121 /// { x : icmp op x y is true}'.
123 /// Example: Pred = ult and Other = i8 3 returns [0, 3)
124 static ConstantRange
makeExactICmpRegion(CmpInst::Predicate Pred
,
127 /// Produce the largest range containing all X such that "X BinOp Y" is
128 /// guaranteed not to wrap (overflow) for *all* Y in Other. However, there may
129 /// be *some* Y in Other for which additional X not contained in the result
130 /// also do not overflow.
132 /// NoWrapKind must be one of OBO::NoUnsignedWrap or OBO::NoSignedWrap.
135 /// typedef OverflowingBinaryOperator OBO;
136 /// #define MGNR makeGuaranteedNoWrapRegion
137 /// MGNR(Add, [i8 1, 2), OBO::NoSignedWrap) == [-128, 127)
138 /// MGNR(Add, [i8 1, 2), OBO::NoUnsignedWrap) == [0, -1)
139 /// MGNR(Add, [i8 0, 1), OBO::NoUnsignedWrap) == Full Set
140 /// MGNR(Add, [i8 -1, 6), OBO::NoSignedWrap) == [INT_MIN+1, INT_MAX-4)
141 /// MGNR(Sub, [i8 1, 2), OBO::NoSignedWrap) == [-127, 128)
142 /// MGNR(Sub, [i8 1, 2), OBO::NoUnsignedWrap) == [1, 0)
143 static ConstantRange
makeGuaranteedNoWrapRegion(Instruction::BinaryOps BinOp
,
144 const ConstantRange
&Other
,
145 unsigned NoWrapKind
);
147 /// Produce the range that contains X if and only if "X BinOp Other" does
149 static ConstantRange
makeExactNoWrapRegion(Instruction::BinaryOps BinOp
,
151 unsigned NoWrapKind
);
153 /// Set up \p Pred and \p RHS such that
154 /// ConstantRange::makeExactICmpRegion(Pred, RHS) == *this. Return true if
156 bool getEquivalentICmp(CmpInst::Predicate
&Pred
, APInt
&RHS
) const;
158 /// Return the lower value for this range.
159 const APInt
&getLower() const { return Lower
; }
161 /// Return the upper value for this range.
162 const APInt
&getUpper() const { return Upper
; }
164 /// Get the bit width of this ConstantRange.
165 uint32_t getBitWidth() const { return Lower
.getBitWidth(); }
167 /// Return true if this set contains all of the elements possible
168 /// for this data-type.
169 bool isFullSet() const;
171 /// Return true if this set contains no members.
172 bool isEmptySet() const;
174 /// Return true if this set wraps around the unsigned domain. Special cases:
175 /// * Empty set: Not wrapped.
176 /// * Full set: Not wrapped.
177 /// * [X, 0) == [X, Max]: Not wrapped.
178 bool isWrappedSet() const;
180 /// Return true if the exclusive upper bound wraps around the unsigned
181 /// domain. Special cases:
182 /// * Empty set: Not wrapped.
183 /// * Full set: Not wrapped.
184 /// * [X, 0): Wrapped.
185 bool isUpperWrapped() const;
187 /// Return true if this set wraps around the signed domain. Special cases:
188 /// * Empty set: Not wrapped.
189 /// * Full set: Not wrapped.
190 /// * [X, SignedMin) == [X, SignedMax]: Not wrapped.
191 bool isSignWrappedSet() const;
193 /// Return true if the (exclusive) upper bound wraps around the signed
194 /// domain. Special cases:
195 /// * Empty set: Not wrapped.
196 /// * Full set: Not wrapped.
197 /// * [X, SignedMin): Wrapped.
198 bool isUpperSignWrapped() const;
200 /// Return true if the specified value is in the set.
201 bool contains(const APInt
&Val
) const;
203 /// Return true if the other range is a subset of this one.
204 bool contains(const ConstantRange
&CR
) const;
206 /// If this set contains a single element, return it, otherwise return null.
207 const APInt
*getSingleElement() const {
208 if (Upper
== Lower
+ 1)
213 /// If this set contains all but a single element, return it, otherwise return
215 const APInt
*getSingleMissingElement() const {
216 if (Lower
== Upper
+ 1)
221 /// Return true if this set contains exactly one member.
222 bool isSingleElement() const { return getSingleElement() != nullptr; }
224 /// Compare set size of this range with the range CR.
225 bool isSizeStrictlySmallerThan(const ConstantRange
&CR
) const;
227 /// Compare set size of this range with Value.
228 bool isSizeLargerThan(uint64_t MaxSize
) const;
230 /// Return true if all values in this range are negative.
231 bool isAllNegative() const;
233 /// Return true if all values in this range are non-negative.
234 bool isAllNonNegative() const;
236 /// Return the largest unsigned value contained in the ConstantRange.
237 APInt
getUnsignedMax() const;
239 /// Return the smallest unsigned value contained in the ConstantRange.
240 APInt
getUnsignedMin() const;
242 /// Return the largest signed value contained in the ConstantRange.
243 APInt
getSignedMax() const;
245 /// Return the smallest signed value contained in the ConstantRange.
246 APInt
getSignedMin() const;
248 /// Return true if this range is equal to another range.
249 bool operator==(const ConstantRange
&CR
) const {
250 return Lower
== CR
.Lower
&& Upper
== CR
.Upper
;
252 bool operator!=(const ConstantRange
&CR
) const {
253 return !operator==(CR
);
256 /// Subtract the specified constant from the endpoints of this constant range.
257 ConstantRange
subtract(const APInt
&CI
) const;
259 /// Subtract the specified range from this range (aka relative complement of
261 ConstantRange
difference(const ConstantRange
&CR
) const;
263 /// If represented precisely, the result of some range operations may consist
264 /// of multiple disjoint ranges. As only a single range may be returned, any
265 /// range covering these disjoint ranges constitutes a valid result, but some
266 /// may be more useful than others depending on context. The preferred range
267 /// type specifies whether a range that is non-wrapping in the unsigned or
268 /// signed domain, or has the smallest size, is preferred. If a signedness is
269 /// preferred but all ranges are non-wrapping or all wrapping, then the
270 /// smallest set size is preferred. If there are multiple smallest sets, any
271 /// one of them may be returned.
272 enum PreferredRangeType
{ Smallest
, Unsigned
, Signed
};
274 /// Return the range that results from the intersection of this range with
275 /// another range. If the intersection is disjoint, such that two results
276 /// are possible, the preferred range is determined by the PreferredRangeType.
277 ConstantRange
intersectWith(const ConstantRange
&CR
,
278 PreferredRangeType Type
= Smallest
) const;
280 /// Return the range that results from the union of this range
281 /// with another range. The resultant range is guaranteed to include the
282 /// elements of both sets, but may contain more. For example, [3, 9) union
283 /// [12,15) is [3, 15), which includes 9, 10, and 11, which were not included
284 /// in either set before.
285 ConstantRange
unionWith(const ConstantRange
&CR
,
286 PreferredRangeType Type
= Smallest
) const;
288 /// Return a new range representing the possible values resulting
289 /// from an application of the specified cast operator to this range. \p
290 /// BitWidth is the target bitwidth of the cast. For casts which don't
291 /// change bitwidth, it must be the same as the source bitwidth. For casts
292 /// which do change bitwidth, the bitwidth must be consistent with the
293 /// requested cast and source bitwidth.
294 ConstantRange
castOp(Instruction::CastOps CastOp
,
295 uint32_t BitWidth
) const;
297 /// Return a new range in the specified integer type, which must
298 /// be strictly larger than the current type. The returned range will
299 /// correspond to the possible range of values if the source range had been
300 /// zero extended to BitWidth.
301 ConstantRange
zeroExtend(uint32_t BitWidth
) const;
303 /// Return a new range in the specified integer type, which must
304 /// be strictly larger than the current type. The returned range will
305 /// correspond to the possible range of values if the source range had been
306 /// sign extended to BitWidth.
307 ConstantRange
signExtend(uint32_t BitWidth
) const;
309 /// Return a new range in the specified integer type, which must be
310 /// strictly smaller than the current type. The returned range will
311 /// correspond to the possible range of values if the source range had been
312 /// truncated to the specified type.
313 ConstantRange
truncate(uint32_t BitWidth
) const;
315 /// Make this range have the bit width given by \p BitWidth. The
316 /// value is zero extended, truncated, or left alone to make it that width.
317 ConstantRange
zextOrTrunc(uint32_t BitWidth
) const;
319 /// Make this range have the bit width given by \p BitWidth. The
320 /// value is sign extended, truncated, or left alone to make it that width.
321 ConstantRange
sextOrTrunc(uint32_t BitWidth
) const;
323 /// Return a new range representing the possible values resulting
324 /// from an application of the specified binary operator to an left hand side
325 /// of this range and a right hand side of \p Other.
326 ConstantRange
binaryOp(Instruction::BinaryOps BinOp
,
327 const ConstantRange
&Other
) const;
329 /// Return a new range representing the possible values resulting
330 /// from an addition of a value in this range and a value in \p Other.
331 ConstantRange
add(const ConstantRange
&Other
) const;
333 /// Return a new range representing the possible values resulting
334 /// from an addition with wrap type \p NoWrapKind of a value in this
335 /// range and a value in \p Other.
336 /// If the result range is disjoint, the preferred range is determined by the
337 /// \p PreferredRangeType.
338 ConstantRange
addWithNoWrap(const ConstantRange
&Other
, unsigned NoWrapKind
,
339 PreferredRangeType RangeType
= Smallest
) const;
341 /// Return a new range representing the possible values resulting
342 /// from a subtraction of a value in this range and a value in \p Other.
343 ConstantRange
sub(const ConstantRange
&Other
) const;
345 /// Return a new range representing the possible values resulting
346 /// from a multiplication of a value in this range and a value in \p Other,
347 /// treating both this and \p Other as unsigned ranges.
348 ConstantRange
multiply(const ConstantRange
&Other
) const;
350 /// Return a new range representing the possible values resulting
351 /// from a signed maximum of a value in this range and a value in \p Other.
352 ConstantRange
smax(const ConstantRange
&Other
) const;
354 /// Return a new range representing the possible values resulting
355 /// from an unsigned maximum of a value in this range and a value in \p Other.
356 ConstantRange
umax(const ConstantRange
&Other
) const;
358 /// Return a new range representing the possible values resulting
359 /// from a signed minimum of a value in this range and a value in \p Other.
360 ConstantRange
smin(const ConstantRange
&Other
) const;
362 /// Return a new range representing the possible values resulting
363 /// from an unsigned minimum of a value in this range and a value in \p Other.
364 ConstantRange
umin(const ConstantRange
&Other
) const;
366 /// Return a new range representing the possible values resulting
367 /// from an unsigned division of a value in this range and a value in
369 ConstantRange
udiv(const ConstantRange
&Other
) const;
371 /// Return a new range representing the possible values resulting
372 /// from a signed division of a value in this range and a value in
373 /// \p Other. Division by zero and division of SignedMin by -1 are considered
374 /// undefined behavior, in line with IR, and do not contribute towards the
376 ConstantRange
sdiv(const ConstantRange
&Other
) const;
378 /// Return a new range representing the possible values resulting
379 /// from an unsigned remainder operation of a value in this range and a
380 /// value in \p Other.
381 ConstantRange
urem(const ConstantRange
&Other
) const;
383 /// Return a new range representing the possible values resulting
384 /// from a signed remainder operation of a value in this range and a
385 /// value in \p Other.
386 ConstantRange
srem(const ConstantRange
&Other
) const;
388 /// Return a new range representing the possible values resulting
389 /// from a binary-and of a value in this range by a value in \p Other.
390 ConstantRange
binaryAnd(const ConstantRange
&Other
) const;
392 /// Return a new range representing the possible values resulting
393 /// from a binary-or of a value in this range by a value in \p Other.
394 ConstantRange
binaryOr(const ConstantRange
&Other
) const;
396 /// Return a new range representing the possible values resulting
397 /// from a left shift of a value in this range by a value in \p Other.
398 /// TODO: This isn't fully implemented yet.
399 ConstantRange
shl(const ConstantRange
&Other
) const;
401 /// Return a new range representing the possible values resulting from a
402 /// logical right shift of a value in this range and a value in \p Other.
403 ConstantRange
lshr(const ConstantRange
&Other
) const;
405 /// Return a new range representing the possible values resulting from a
406 /// arithmetic right shift of a value in this range and a value in \p Other.
407 ConstantRange
ashr(const ConstantRange
&Other
) const;
409 /// Perform an unsigned saturating addition of two constant ranges.
410 ConstantRange
uadd_sat(const ConstantRange
&Other
) const;
412 /// Perform a signed saturating addition of two constant ranges.
413 ConstantRange
sadd_sat(const ConstantRange
&Other
) const;
415 /// Perform an unsigned saturating subtraction of two constant ranges.
416 ConstantRange
usub_sat(const ConstantRange
&Other
) const;
418 /// Perform a signed saturating subtraction of two constant ranges.
419 ConstantRange
ssub_sat(const ConstantRange
&Other
) const;
421 /// Return a new range that is the logical not of the current set.
422 ConstantRange
inverse() const;
424 /// Calculate absolute value range. If the original range contains signed
425 /// min, then the resulting range will also contain signed min.
426 ConstantRange
abs() const;
428 /// Represents whether an operation on the given constant range is known to
429 /// always or never overflow.
430 enum class OverflowResult
{
431 /// Always overflows in the direction of signed/unsigned min value.
433 /// Always overflows in the direction of signed/unsigned max value.
435 /// May or may not overflow.
441 /// Return whether unsigned add of the two ranges always/never overflows.
442 OverflowResult
unsignedAddMayOverflow(const ConstantRange
&Other
) const;
444 /// Return whether signed add of the two ranges always/never overflows.
445 OverflowResult
signedAddMayOverflow(const ConstantRange
&Other
) const;
447 /// Return whether unsigned sub of the two ranges always/never overflows.
448 OverflowResult
unsignedSubMayOverflow(const ConstantRange
&Other
) const;
450 /// Return whether signed sub of the two ranges always/never overflows.
451 OverflowResult
signedSubMayOverflow(const ConstantRange
&Other
) const;
453 /// Return whether unsigned mul of the two ranges always/never overflows.
454 OverflowResult
unsignedMulMayOverflow(const ConstantRange
&Other
) const;
456 /// Print out the bounds to a stream.
457 void print(raw_ostream
&OS
) const;
459 /// Allow printing from a debugger easily.
463 inline raw_ostream
&operator<<(raw_ostream
&OS
, const ConstantRange
&CR
) {
468 /// Parse out a conservative ConstantRange from !range metadata.
470 /// E.g. if RangeMD is !{i32 0, i32 10, i32 15, i32 20} then return [0, 20).
471 ConstantRange
getConstantRangeFromMetadata(const MDNode
&RangeMD
);
473 } // end namespace llvm
475 #endif // LLVM_IR_CONSTANTRANGE_H