1 //===-- KnownBits.cpp - Stores known zeros/ones ---------------------------===//
2 //
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
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file contains a class for representing known zeros and ones used by
10 // computeKnownBits.
11 //
12 //===----------------------------------------------------------------------===//
13
14 #include "llvm/Support/KnownBits.h"
15 #include <cassert>
16
17 using namespace llvm;
18
computeForAddCarry(const KnownBits & LHS,const KnownBits & RHS,bool CarryZero,bool CarryOne)19 static KnownBits computeForAddCarry(
20 const KnownBits &LHS, const KnownBits &RHS,
21 bool CarryZero, bool CarryOne) {
22 assert(!(CarryZero && CarryOne) &&
23 "Carry can't be zero and one at the same time");
24
25 APInt PossibleSumZero = LHS.getMaxValue() + RHS.getMaxValue() + !CarryZero;
26 APInt PossibleSumOne = LHS.getMinValue() + RHS.getMinValue() + CarryOne;
27
28 // Compute known bits of the carry.
29 APInt CarryKnownZero = ~(PossibleSumZero ^ LHS.Zero ^ RHS.Zero);
30 APInt CarryKnownOne = PossibleSumOne ^ LHS.One ^ RHS.One;
31
32 // Compute set of known bits (where all three relevant bits are known).
33 APInt LHSKnownUnion = LHS.Zero | LHS.One;
34 APInt RHSKnownUnion = RHS.Zero | RHS.One;
35 APInt CarryKnownUnion = std::move(CarryKnownZero) | CarryKnownOne;
36 APInt Known = std::move(LHSKnownUnion) & RHSKnownUnion & CarryKnownUnion;
37
38 assert((PossibleSumZero & Known) == (PossibleSumOne & Known) &&
39 "known bits of sum differ");
40
41 // Compute known bits of the result.
42 KnownBits KnownOut;
43 KnownOut.Zero = ~std::move(PossibleSumZero) & Known;
44 KnownOut.One = std::move(PossibleSumOne) & Known;
45 return KnownOut;
46 }
47
computeForAddCarry(const KnownBits & LHS,const KnownBits & RHS,const KnownBits & Carry)48 KnownBits KnownBits::computeForAddCarry(
49 const KnownBits &LHS, const KnownBits &RHS, const KnownBits &Carry) {
50 assert(Carry.getBitWidth() == 1 && "Carry must be 1-bit");
51 return ::computeForAddCarry(
52 LHS, RHS, Carry.Zero.getBoolValue(), Carry.One.getBoolValue());
53 }
54
computeForAddSub(bool Add,bool NSW,const KnownBits & LHS,KnownBits RHS)55 KnownBits KnownBits::computeForAddSub(bool Add, bool NSW,
56 const KnownBits &LHS, KnownBits RHS) {
57 KnownBits KnownOut;
58 if (Add) {
59 // Sum = LHS + RHS + 0
60 KnownOut = ::computeForAddCarry(
61 LHS, RHS, /*CarryZero*/true, /*CarryOne*/false);
62 } else {
63 // Sum = LHS + ~RHS + 1
64 std::swap(RHS.Zero, RHS.One);
65 KnownOut = ::computeForAddCarry(
66 LHS, RHS, /*CarryZero*/false, /*CarryOne*/true);
67 }
68
69 // Are we still trying to solve for the sign bit?
70 if (!KnownOut.isNegative() && !KnownOut.isNonNegative()) {
71 if (NSW) {
72 // Adding two non-negative numbers, or subtracting a negative number from
73 // a non-negative one, can't wrap into negative.
74 if (LHS.isNonNegative() && RHS.isNonNegative())
75 KnownOut.makeNonNegative();
76 // Adding two negative numbers, or subtracting a non-negative number from
77 // a negative one, can't wrap into non-negative.
78 else if (LHS.isNegative() && RHS.isNegative())
79 KnownOut.makeNegative();
80 }
81 }
82
83 return KnownOut;
84 }
85
sextInReg(unsigned SrcBitWidth) const86 KnownBits KnownBits::sextInReg(unsigned SrcBitWidth) const {
87 unsigned BitWidth = getBitWidth();
88 assert(BitWidth >= SrcBitWidth && "Illegal sext-in-register");
89
90 // Sign extension. Compute the demanded bits in the result that are not
91 // present in the input.
92 APInt NewBits = APInt::getHighBitsSet(BitWidth, BitWidth - SrcBitWidth);
93
94 // If the sign extended bits are demanded, we know that the sign
95 // bit is demanded.
96 APInt InSignMask = APInt::getSignMask(SrcBitWidth).zext(BitWidth);
97 APInt InDemandedBits = APInt::getLowBitsSet(BitWidth, SrcBitWidth);
98 if (NewBits.getBoolValue())
99 InDemandedBits |= InSignMask;
100
101 KnownBits Result;
102 Result.One = One & InDemandedBits;
103 Result.Zero = Zero & InDemandedBits;
104
105 // If the sign bit of the input is known set or clear, then we know the
106 // top bits of the result.
107 if (Result.Zero.intersects(InSignMask)) { // Input sign bit known clear
108 Result.Zero |= NewBits;
109 Result.One &= ~NewBits;
110 } else if (Result.One.intersects(InSignMask)) { // Input sign bit known set
111 Result.One |= NewBits;
112 Result.Zero &= ~NewBits;
113 } else { // Input sign bit unknown
114 Result.Zero &= ~NewBits;
115 Result.One &= ~NewBits;
116 }
117
118 return Result;
119 }
120
makeGE(const APInt & Val) const121 KnownBits KnownBits::makeGE(const APInt &Val) const {
122 // Count the number of leading bit positions where our underlying value is
123 // known to be less than or equal to Val.
124 unsigned N = (Zero | Val).countLeadingOnes();
125
126 // For each of those bit positions, if Val has a 1 in that bit then our
127 // underlying value must also have a 1.
128 APInt MaskedVal(Val);
129 MaskedVal.clearLowBits(getBitWidth() - N);
130 return KnownBits(Zero, One | MaskedVal);
131 }
132
umax(const KnownBits & LHS,const KnownBits & RHS)133 KnownBits KnownBits::umax(const KnownBits &LHS, const KnownBits &RHS) {
134 // If we can prove that LHS >= RHS then use LHS as the result. Likewise for
135 // RHS. Ideally our caller would already have spotted these cases and
136 // optimized away the umax operation, but we handle them here for
137 // completeness.
138 if (LHS.getMinValue().uge(RHS.getMaxValue()))
139 return LHS;
140 if (RHS.getMinValue().uge(LHS.getMaxValue()))
141 return RHS;
142
143 // If the result of the umax is LHS then it must be greater than or equal to
144 // the minimum possible value of RHS. Likewise for RHS. Any known bits that
145 // are common to these two values are also known in the result.
146 KnownBits L = LHS.makeGE(RHS.getMinValue());
147 KnownBits R = RHS.makeGE(LHS.getMinValue());
148 return KnownBits::commonBits(L, R);
149 }
150
umin(const KnownBits & LHS,const KnownBits & RHS)151 KnownBits KnownBits::umin(const KnownBits &LHS, const KnownBits &RHS) {
152 // Flip the range of values: [0, 0xFFFFFFFF] <-> [0xFFFFFFFF, 0]
153 auto Flip = [](const KnownBits &Val) { return KnownBits(Val.One, Val.Zero); };
154 return Flip(umax(Flip(LHS), Flip(RHS)));
155 }
156
smax(const KnownBits & LHS,const KnownBits & RHS)157 KnownBits KnownBits::smax(const KnownBits &LHS, const KnownBits &RHS) {
158 // Flip the range of values: [-0x80000000, 0x7FFFFFFF] <-> [0, 0xFFFFFFFF]
159 auto Flip = [](const KnownBits &Val) {
160 unsigned SignBitPosition = Val.getBitWidth() - 1;
161 APInt Zero = Val.Zero;
162 APInt One = Val.One;
163 Zero.setBitVal(SignBitPosition, Val.One[SignBitPosition]);
164 One.setBitVal(SignBitPosition, Val.Zero[SignBitPosition]);
165 return KnownBits(Zero, One);
166 };
167 return Flip(umax(Flip(LHS), Flip(RHS)));
168 }
169
smin(const KnownBits & LHS,const KnownBits & RHS)170 KnownBits KnownBits::smin(const KnownBits &LHS, const KnownBits &RHS) {
171 // Flip the range of values: [-0x80000000, 0x7FFFFFFF] <-> [0xFFFFFFFF, 0]
172 auto Flip = [](const KnownBits &Val) {
173 unsigned SignBitPosition = Val.getBitWidth() - 1;
174 APInt Zero = Val.One;
175 APInt One = Val.Zero;
176 Zero.setBitVal(SignBitPosition, Val.Zero[SignBitPosition]);
177 One.setBitVal(SignBitPosition, Val.One[SignBitPosition]);
178 return KnownBits(Zero, One);
179 };
180 return Flip(umax(Flip(LHS), Flip(RHS)));
181 }
182
shl(const KnownBits & LHS,const KnownBits & RHS)183 KnownBits KnownBits::shl(const KnownBits &LHS, const KnownBits &RHS) {
184 unsigned BitWidth = LHS.getBitWidth();
185 KnownBits Known(BitWidth);
186
187 // If the shift amount is a valid constant then transform LHS directly.
188 if (RHS.isConstant() && RHS.getConstant().ult(BitWidth)) {
189 unsigned Shift = RHS.getConstant().getZExtValue();
190 Known = LHS;
191 Known.Zero <<= Shift;
192 Known.One <<= Shift;
193 // Low bits are known zero.
194 Known.Zero.setLowBits(Shift);
195 return Known;
196 }
197
198 // No matter the shift amount, the trailing zeros will stay zero.
199 unsigned MinTrailingZeros = LHS.countMinTrailingZeros();
200
201 // Minimum shift amount low bits are known zero.
202 if (RHS.getMinValue().ult(BitWidth)) {
203 MinTrailingZeros += RHS.getMinValue().getZExtValue();
204 MinTrailingZeros = std::min(MinTrailingZeros, BitWidth);
205 }
206
207 Known.Zero.setLowBits(MinTrailingZeros);
208 return Known;
209 }
210
lshr(const KnownBits & LHS,const KnownBits & RHS)211 KnownBits KnownBits::lshr(const KnownBits &LHS, const KnownBits &RHS) {
212 unsigned BitWidth = LHS.getBitWidth();
213 KnownBits Known(BitWidth);
214
215 if (RHS.isConstant() && RHS.getConstant().ult(BitWidth)) {
216 unsigned Shift = RHS.getConstant().getZExtValue();
217 Known = LHS;
218 Known.Zero.lshrInPlace(Shift);
219 Known.One.lshrInPlace(Shift);
220 // High bits are known zero.
221 Known.Zero.setHighBits(Shift);
222 return Known;
223 }
224
225 // No matter the shift amount, the leading zeros will stay zero.
226 unsigned MinLeadingZeros = LHS.countMinLeadingZeros();
227
228 // Minimum shift amount high bits are known zero.
229 if (RHS.getMinValue().ult(BitWidth)) {
230 MinLeadingZeros += RHS.getMinValue().getZExtValue();
231 MinLeadingZeros = std::min(MinLeadingZeros, BitWidth);
232 }
233
234 Known.Zero.setHighBits(MinLeadingZeros);
235 return Known;
236 }
237
ashr(const KnownBits & LHS,const KnownBits & RHS)238 KnownBits KnownBits::ashr(const KnownBits &LHS, const KnownBits &RHS) {
239 unsigned BitWidth = LHS.getBitWidth();
240 KnownBits Known(BitWidth);
241
242 if (RHS.isConstant() && RHS.getConstant().ult(BitWidth)) {
243 unsigned Shift = RHS.getConstant().getZExtValue();
244 Known = LHS;
245 Known.Zero.ashrInPlace(Shift);
246 Known.One.ashrInPlace(Shift);
247 return Known;
248 }
249
250 // No matter the shift amount, the leading sign bits will stay.
251 unsigned MinLeadingZeros = LHS.countMinLeadingZeros();
252 unsigned MinLeadingOnes = LHS.countMinLeadingOnes();
253
254 // Minimum shift amount high bits are known sign bits.
255 if (RHS.getMinValue().ult(BitWidth)) {
256 if (MinLeadingZeros) {
257 MinLeadingZeros += RHS.getMinValue().getZExtValue();
258 MinLeadingZeros = std::min(MinLeadingZeros, BitWidth);
259 }
260 if (MinLeadingOnes) {
261 MinLeadingOnes += RHS.getMinValue().getZExtValue();
262 MinLeadingOnes = std::min(MinLeadingOnes, BitWidth);
263 }
264 }
265
266 Known.Zero.setHighBits(MinLeadingZeros);
267 Known.One.setHighBits(MinLeadingOnes);
268 return Known;
269 }
270
abs(bool IntMinIsPoison) const271 KnownBits KnownBits::abs(bool IntMinIsPoison) const {
272 // If the source's MSB is zero then we know the rest of the bits already.
273 if (isNonNegative())
274 return *this;
275
276 // Absolute value preserves trailing zero count.
277 KnownBits KnownAbs(getBitWidth());
278 KnownAbs.Zero.setLowBits(countMinTrailingZeros());
279
280 // We only know that the absolute values's MSB will be zero if INT_MIN is
281 // poison, or there is a set bit that isn't the sign bit (otherwise it could
282 // be INT_MIN).
283 if (IntMinIsPoison || (!One.isNullValue() && !One.isMinSignedValue()))
284 KnownAbs.Zero.setSignBit();
285
286 // FIXME: Handle known negative input?
287 // FIXME: Calculate the negated Known bits and combine them?
288 return KnownAbs;
289 }
290
computeForMul(const KnownBits & LHS,const KnownBits & RHS)291 KnownBits KnownBits::computeForMul(const KnownBits &LHS, const KnownBits &RHS) {
292 unsigned BitWidth = LHS.getBitWidth();
293
294 assert(!LHS.hasConflict() && !RHS.hasConflict());
295 // Compute a conservative estimate for high known-0 bits.
296 unsigned LeadZ =
297 std::max(LHS.countMinLeadingZeros() + RHS.countMinLeadingZeros(),
298 BitWidth) -
299 BitWidth;
300 LeadZ = std::min(LeadZ, BitWidth);
301
302 // The result of the bottom bits of an integer multiply can be
303 // inferred by looking at the bottom bits of both operands and
304 // multiplying them together.
305 // We can infer at least the minimum number of known trailing bits
306 // of both operands. Depending on number of trailing zeros, we can
307 // infer more bits, because (a*b) <=> ((a/m) * (b/n)) * (m*n) assuming
308 // a and b are divisible by m and n respectively.
309 // We then calculate how many of those bits are inferrable and set
310 // the output. For example, the i8 mul:
311 // a = XXXX1100 (12)
312 // b = XXXX1110 (14)
313 // We know the bottom 3 bits are zero since the first can be divided by
314 // 4 and the second by 2, thus having ((12/4) * (14/2)) * (2*4).
315 // Applying the multiplication to the trimmed arguments gets:
316 // XX11 (3)
317 // X111 (7)
318 // -------
319 // XX11
320 // XX11
321 // XX11
322 // XX11
323 // -------
324 // XXXXX01
325 // Which allows us to infer the 2 LSBs. Since we're multiplying the result
326 // by 8, the bottom 3 bits will be 0, so we can infer a total of 5 bits.
327 // The proof for this can be described as:
328 // Pre: (C1 >= 0) && (C1 < (1 << C5)) && (C2 >= 0) && (C2 < (1 << C6)) &&
329 // (C7 == (1 << (umin(countTrailingZeros(C1), C5) +
330 // umin(countTrailingZeros(C2), C6) +
331 // umin(C5 - umin(countTrailingZeros(C1), C5),
332 // C6 - umin(countTrailingZeros(C2), C6)))) - 1)
333 // %aa = shl i8 %a, C5
334 // %bb = shl i8 %b, C6
335 // %aaa = or i8 %aa, C1
336 // %bbb = or i8 %bb, C2
337 // %mul = mul i8 %aaa, %bbb
338 // %mask = and i8 %mul, C7
339 // =>
340 // %mask = i8 ((C1*C2)&C7)
341 // Where C5, C6 describe the known bits of %a, %b
342 // C1, C2 describe the known bottom bits of %a, %b.
343 // C7 describes the mask of the known bits of the result.
344 const APInt &Bottom0 = LHS.One;
345 const APInt &Bottom1 = RHS.One;
346
347 // How many times we'd be able to divide each argument by 2 (shr by 1).
348 // This gives us the number of trailing zeros on the multiplication result.
349 unsigned TrailBitsKnown0 = (LHS.Zero | LHS.One).countTrailingOnes();
350 unsigned TrailBitsKnown1 = (RHS.Zero | RHS.One).countTrailingOnes();
351 unsigned TrailZero0 = LHS.countMinTrailingZeros();
352 unsigned TrailZero1 = RHS.countMinTrailingZeros();
353 unsigned TrailZ = TrailZero0 + TrailZero1;
354
355 // Figure out the fewest known-bits operand.
356 unsigned SmallestOperand =
357 std::min(TrailBitsKnown0 - TrailZero0, TrailBitsKnown1 - TrailZero1);
358 unsigned ResultBitsKnown = std::min(SmallestOperand + TrailZ, BitWidth);
359
360 APInt BottomKnown =
361 Bottom0.getLoBits(TrailBitsKnown0) * Bottom1.getLoBits(TrailBitsKnown1);
362
363 KnownBits Res(BitWidth);
364 Res.Zero.setHighBits(LeadZ);
365 Res.Zero |= (~BottomKnown).getLoBits(ResultBitsKnown);
366 Res.One = BottomKnown.getLoBits(ResultBitsKnown);
367 return Res;
368 }
369
udiv(const KnownBits & LHS,const KnownBits & RHS)370 KnownBits KnownBits::udiv(const KnownBits &LHS, const KnownBits &RHS) {
371 unsigned BitWidth = LHS.getBitWidth();
372 assert(!LHS.hasConflict() && !RHS.hasConflict());
373 KnownBits Known(BitWidth);
374
375 // For the purposes of computing leading zeros we can conservatively
376 // treat a udiv as a logical right shift by the power of 2 known to
377 // be less than the denominator.
378 unsigned LeadZ = LHS.countMinLeadingZeros();
379 unsigned RHSMaxLeadingZeros = RHS.countMaxLeadingZeros();
380
381 if (RHSMaxLeadingZeros != BitWidth)
382 LeadZ = std::min(BitWidth, LeadZ + BitWidth - RHSMaxLeadingZeros - 1);
383
384 Known.Zero.setHighBits(LeadZ);
385 return Known;
386 }
387
urem(const KnownBits & LHS,const KnownBits & RHS)388 KnownBits KnownBits::urem(const KnownBits &LHS, const KnownBits &RHS) {
389 unsigned BitWidth = LHS.getBitWidth();
390 assert(!LHS.hasConflict() && !RHS.hasConflict());
391 KnownBits Known(BitWidth);
392
393 if (RHS.isConstant() && RHS.getConstant().isPowerOf2()) {
394 // The upper bits are all zero, the lower ones are unchanged.
395 APInt LowBits = RHS.getConstant() - 1;
396 Known.Zero = LHS.Zero | ~LowBits;
397 Known.One = LHS.One & LowBits;
398 return Known;
399 }
400
401 // Since the result is less than or equal to either operand, any leading
402 // zero bits in either operand must also exist in the result.
403 uint32_t Leaders =
404 std::max(LHS.countMinLeadingZeros(), RHS.countMinLeadingZeros());
405 Known.Zero.setHighBits(Leaders);
406 return Known;
407 }
408
srem(const KnownBits & LHS,const KnownBits & RHS)409 KnownBits KnownBits::srem(const KnownBits &LHS, const KnownBits &RHS) {
410 unsigned BitWidth = LHS.getBitWidth();
411 assert(!LHS.hasConflict() && !RHS.hasConflict());
412 KnownBits Known(BitWidth);
413
414 if (RHS.isConstant() && RHS.getConstant().isPowerOf2()) {
415 // The low bits of the first operand are unchanged by the srem.
416 APInt LowBits = RHS.getConstant() - 1;
417 Known.Zero = LHS.Zero & LowBits;
418 Known.One = LHS.One & LowBits;
419
420 // If the first operand is non-negative or has all low bits zero, then
421 // the upper bits are all zero.
422 if (LHS.isNonNegative() || LowBits.isSubsetOf(LHS.Zero))
423 Known.Zero |= ~LowBits;
424
425 // If the first operand is negative and not all low bits are zero, then
426 // the upper bits are all one.
427 if (LHS.isNegative() && LowBits.intersects(LHS.One))
428 Known.One |= ~LowBits;
429 return Known;
430 }
431
432 // The sign bit is the LHS's sign bit, except when the result of the
433 // remainder is zero. If it's known zero, our sign bit is also zero.
434 if (LHS.isNonNegative())
435 Known.makeNonNegative();
436 return Known;
437 }
438
operator &=(const KnownBits & RHS)439 KnownBits &KnownBits::operator&=(const KnownBits &RHS) {
440 // Result bit is 0 if either operand bit is 0.
441 Zero |= RHS.Zero;
442 // Result bit is 1 if both operand bits are 1.
443 One &= RHS.One;
444 return *this;
445 }
446
operator |=(const KnownBits & RHS)447 KnownBits &KnownBits::operator|=(const KnownBits &RHS) {
448 // Result bit is 0 if both operand bits are 0.
449 Zero &= RHS.Zero;
450 // Result bit is 1 if either operand bit is 1.
451 One |= RHS.One;
452 return *this;
453 }
454
operator ^=(const KnownBits & RHS)455 KnownBits &KnownBits::operator^=(const KnownBits &RHS) {
456 // Result bit is 0 if both operand bits are 0 or both are 1.
457 APInt Z = (Zero & RHS.Zero) | (One & RHS.One);
458 // Result bit is 1 if one operand bit is 0 and the other is 1.
459 One = (Zero & RHS.One) | (One & RHS.Zero);
460 Zero = std::move(Z);
461 return *this;
462 }
463