| 1 | //===- InferIntRangeCommon.cpp - Inference for common ops ------------===// |
| 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 implementations of range inference for operations that are |
| 10 | // common to both the `arith` and `index` dialects to facilitate reuse. |
| 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
| 14 | #include "mlir/Interfaces/Utils/InferIntRangeCommon.h" |
| 15 | |
| 16 | #include "mlir/Interfaces/InferIntRangeInterface.h" |
| 17 | #include "mlir/Interfaces/ShapedOpInterfaces.h" |
| 18 | |
| 19 | #include "llvm/ADT/ArrayRef.h" |
| 20 | #include "llvm/ADT/STLExtras.h" |
| 21 | |
| 22 | #include "llvm/Support/Debug.h" |
| 23 | |
| 24 | #include <iterator> |
| 25 | #include <optional> |
| 26 | |
| 27 | using namespace mlir; |
| 28 | |
| 29 | #define DEBUG_TYPE "int-range-analysis" |
| 30 | |
| 31 | //===----------------------------------------------------------------------===// |
| 32 | // General utilities |
| 33 | //===----------------------------------------------------------------------===// |
| 34 | |
| 35 | /// Function that evaluates the result of doing something on arithmetic |
| 36 | /// constants and returns std::nullopt on overflow. |
| 37 | using ConstArithFn = |
| 38 | function_ref<std::optional<APInt>(const APInt &, const APInt &)>; |
| 39 | using ConstArithStdFn = |
| 40 | std::function<std::optional<APInt>(const APInt &, const APInt &)>; |
| 41 | |
| 42 | /// Compute op(minLeft, minRight) and op(maxLeft, maxRight) if possible, |
| 43 | /// If either computation overflows, make the result unbounded. |
| 44 | static ConstantIntRanges computeBoundsBy(ConstArithFn op, const APInt &minLeft, |
| 45 | const APInt &minRight, |
| 46 | const APInt &maxLeft, |
| 47 | const APInt &maxRight, bool isSigned) { |
| 48 | std::optional<APInt> maybeMin = op(minLeft, minRight); |
| 49 | std::optional<APInt> maybeMax = op(maxLeft, maxRight); |
| 50 | if (maybeMin && maybeMax) |
| 51 | return ConstantIntRanges::range(min: *maybeMin, max: *maybeMax, isSigned); |
| 52 | return ConstantIntRanges::maxRange(bitwidth: minLeft.getBitWidth()); |
| 53 | } |
| 54 | |
| 55 | /// Compute the minimum and maximum of `(op(l, r) for l in lhs for r in rhs)`, |
| 56 | /// ignoring unbounded values. Returns the maximal range if `op` overflows. |
| 57 | static ConstantIntRanges minMaxBy(ConstArithFn op, ArrayRef<APInt> lhs, |
| 58 | ArrayRef<APInt> rhs, bool isSigned) { |
| 59 | unsigned width = lhs[0].getBitWidth(); |
| 60 | APInt min = |
| 61 | isSigned ? APInt::getSignedMaxValue(numBits: width) : APInt::getMaxValue(numBits: width); |
| 62 | APInt max = |
| 63 | isSigned ? APInt::getSignedMinValue(numBits: width) : APInt::getZero(numBits: width); |
| 64 | for (const APInt &left : lhs) { |
| 65 | for (const APInt &right : rhs) { |
| 66 | std::optional<APInt> maybeThisResult = op(left, right); |
| 67 | if (!maybeThisResult) |
| 68 | return ConstantIntRanges::maxRange(bitwidth: width); |
| 69 | APInt result = std::move(*maybeThisResult); |
| 70 | min = (isSigned ? result.slt(RHS: min) : result.ult(RHS: min)) ? result : min; |
| 71 | max = (isSigned ? result.sgt(RHS: max) : result.ugt(RHS: max)) ? result : max; |
| 72 | } |
| 73 | } |
| 74 | return ConstantIntRanges::range(min, max, isSigned); |
| 75 | } |
| 76 | |
| 77 | //===----------------------------------------------------------------------===// |
| 78 | // Ext, trunc, index op handling |
| 79 | //===----------------------------------------------------------------------===// |
| 80 | |
| 81 | ConstantIntRanges |
| 82 | mlir::intrange::inferIndexOp(const InferRangeFn &inferFn, |
| 83 | ArrayRef<ConstantIntRanges> argRanges, |
| 84 | intrange::CmpMode mode) { |
| 85 | ConstantIntRanges sixtyFour = inferFn(argRanges); |
| 86 | SmallVector<ConstantIntRanges, 2> truncated; |
| 87 | llvm::transform(Range&: argRanges, d_first: std::back_inserter(x&: truncated), |
| 88 | F: [](const ConstantIntRanges &range) { |
| 89 | return truncRange(range, /*destWidth=*/indexMinWidth); |
| 90 | }); |
| 91 | ConstantIntRanges thirtyTwo = inferFn(truncated); |
| 92 | ConstantIntRanges thirtyTwoAsSixtyFour = |
| 93 | extRange(range: thirtyTwo, /*destWidth=*/indexMaxWidth); |
| 94 | ConstantIntRanges sixtyFourAsThirtyTwo = |
| 95 | truncRange(range: sixtyFour, /*destWidth=*/indexMinWidth); |
| 96 | |
| 97 | LLVM_DEBUG(llvm::dbgs() << "Index handling: 64-bit result = " << sixtyFour |
| 98 | << " 32-bit = " << thirtyTwo << "\n" ); |
| 99 | bool truncEqual = false; |
| 100 | switch (mode) { |
| 101 | case intrange::CmpMode::Both: |
| 102 | truncEqual = (thirtyTwo == sixtyFourAsThirtyTwo); |
| 103 | break; |
| 104 | case intrange::CmpMode::Signed: |
| 105 | truncEqual = (thirtyTwo.smin() == sixtyFourAsThirtyTwo.smin() && |
| 106 | thirtyTwo.smax() == sixtyFourAsThirtyTwo.smax()); |
| 107 | break; |
| 108 | case intrange::CmpMode::Unsigned: |
| 109 | truncEqual = (thirtyTwo.umin() == sixtyFourAsThirtyTwo.umin() && |
| 110 | thirtyTwo.umax() == sixtyFourAsThirtyTwo.umax()); |
| 111 | break; |
| 112 | } |
| 113 | if (truncEqual) |
| 114 | // Returing the 64-bit result preserves more information. |
| 115 | return sixtyFour; |
| 116 | ConstantIntRanges merged = sixtyFour.rangeUnion(other: thirtyTwoAsSixtyFour); |
| 117 | return merged; |
| 118 | } |
| 119 | |
| 120 | ConstantIntRanges mlir::intrange::(const ConstantIntRanges &range, |
| 121 | unsigned int destWidth) { |
| 122 | APInt umin = range.umin().zext(width: destWidth); |
| 123 | APInt umax = range.umax().zext(width: destWidth); |
| 124 | APInt smin = range.smin().sext(width: destWidth); |
| 125 | APInt smax = range.smax().sext(width: destWidth); |
| 126 | return {umin, umax, smin, smax}; |
| 127 | } |
| 128 | |
| 129 | ConstantIntRanges mlir::intrange::extUIRange(const ConstantIntRanges &range, |
| 130 | unsigned destWidth) { |
| 131 | APInt umin = range.umin().zext(width: destWidth); |
| 132 | APInt umax = range.umax().zext(width: destWidth); |
| 133 | return ConstantIntRanges::fromUnsigned(umin, umax); |
| 134 | } |
| 135 | |
| 136 | ConstantIntRanges mlir::intrange::extSIRange(const ConstantIntRanges &range, |
| 137 | unsigned destWidth) { |
| 138 | APInt smin = range.smin().sext(width: destWidth); |
| 139 | APInt smax = range.smax().sext(width: destWidth); |
| 140 | return ConstantIntRanges::fromSigned(smin, smax); |
| 141 | } |
| 142 | |
| 143 | ConstantIntRanges mlir::intrange::truncRange(const ConstantIntRanges &range, |
| 144 | unsigned int destWidth) { |
| 145 | // If you truncate the first four bytes in [0xaaaabbbb, 0xccccbbbb], |
| 146 | // the range of the resulting value is not contiguous ind includes 0. |
| 147 | // Ex. If you truncate [256, 258] from i16 to i8, you validly get [0, 2], |
| 148 | // but you can't truncate [255, 257] similarly. |
| 149 | bool hasUnsignedRollover = |
| 150 | range.umin().lshr(shiftAmt: destWidth) != range.umax().lshr(shiftAmt: destWidth); |
| 151 | APInt umin = hasUnsignedRollover ? APInt::getZero(numBits: destWidth) |
| 152 | : range.umin().trunc(width: destWidth); |
| 153 | APInt umax = hasUnsignedRollover ? APInt::getMaxValue(numBits: destWidth) |
| 154 | : range.umax().trunc(width: destWidth); |
| 155 | |
| 156 | // Signed post-truncation rollover will not occur when either: |
| 157 | // - The high parts of the min and max, plus the sign bit, are the same |
| 158 | // - The high halves + sign bit of the min and max are either all 1s or all 0s |
| 159 | // and you won't create a [positive, negative] range by truncating. |
| 160 | // For example, you can truncate the ranges [256, 258]_i16 to [0, 2]_i8 |
| 161 | // but not [255, 257]_i16 to a range of i8s. You can also truncate |
| 162 | // [-256, -256]_i16 to [-2, 0]_i8, but not [-257, -255]_i16. |
| 163 | // You can also truncate [-130, 0]_i16 to i8 because -130_i16 (0xff7e) |
| 164 | // will truncate to 0x7e, which is greater than 0 |
| 165 | APInt sminHighPart = range.smin().ashr(ShiftAmt: destWidth - 1); |
| 166 | APInt smaxHighPart = range.smax().ashr(ShiftAmt: destWidth - 1); |
| 167 | bool hasSignedOverflow = |
| 168 | (sminHighPart != smaxHighPart) && |
| 169 | !(sminHighPart.isAllOnes() && |
| 170 | (smaxHighPart.isAllOnes() || smaxHighPart.isZero())) && |
| 171 | !(sminHighPart.isZero() && smaxHighPart.isZero()); |
| 172 | APInt smin = hasSignedOverflow ? APInt::getSignedMinValue(numBits: destWidth) |
| 173 | : range.smin().trunc(width: destWidth); |
| 174 | APInt smax = hasSignedOverflow ? APInt::getSignedMaxValue(numBits: destWidth) |
| 175 | : range.smax().trunc(width: destWidth); |
| 176 | return {umin, umax, smin, smax}; |
| 177 | } |
| 178 | |
| 179 | //===----------------------------------------------------------------------===// |
| 180 | // Addition |
| 181 | //===----------------------------------------------------------------------===// |
| 182 | |
| 183 | ConstantIntRanges |
| 184 | mlir::intrange::inferAdd(ArrayRef<ConstantIntRanges> argRanges, |
| 185 | OverflowFlags ovfFlags) { |
| 186 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 187 | |
| 188 | ConstArithStdFn uadd = [=](const APInt &a, |
| 189 | const APInt &b) -> std::optional<APInt> { |
| 190 | bool overflowed = false; |
| 191 | APInt result = any(Val: ovfFlags & OverflowFlags::Nuw) |
| 192 | ? a.uadd_sat(RHS: b) |
| 193 | : a.uadd_ov(RHS: b, Overflow&: overflowed); |
| 194 | return overflowed ? std::optional<APInt>() : result; |
| 195 | }; |
| 196 | ConstArithStdFn sadd = [=](const APInt &a, |
| 197 | const APInt &b) -> std::optional<APInt> { |
| 198 | bool overflowed = false; |
| 199 | APInt result = any(Val: ovfFlags & OverflowFlags::Nsw) |
| 200 | ? a.sadd_sat(RHS: b) |
| 201 | : a.sadd_ov(RHS: b, Overflow&: overflowed); |
| 202 | return overflowed ? std::optional<APInt>() : result; |
| 203 | }; |
| 204 | |
| 205 | ConstantIntRanges urange = computeBoundsBy( |
| 206 | op: uadd, minLeft: lhs.umin(), minRight: rhs.umin(), maxLeft: lhs.umax(), maxRight: rhs.umax(), /*isSigned=*/false); |
| 207 | ConstantIntRanges srange = computeBoundsBy( |
| 208 | op: sadd, minLeft: lhs.smin(), minRight: rhs.smin(), maxLeft: lhs.smax(), maxRight: rhs.smax(), /*isSigned=*/true); |
| 209 | return urange.intersection(other: srange); |
| 210 | } |
| 211 | |
| 212 | //===----------------------------------------------------------------------===// |
| 213 | // Subtraction |
| 214 | //===----------------------------------------------------------------------===// |
| 215 | |
| 216 | ConstantIntRanges |
| 217 | mlir::intrange::inferSub(ArrayRef<ConstantIntRanges> argRanges, |
| 218 | OverflowFlags ovfFlags) { |
| 219 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 220 | |
| 221 | ConstArithStdFn usub = [=](const APInt &a, |
| 222 | const APInt &b) -> std::optional<APInt> { |
| 223 | bool overflowed = false; |
| 224 | APInt result = any(Val: ovfFlags & OverflowFlags::Nuw) |
| 225 | ? a.usub_sat(RHS: b) |
| 226 | : a.usub_ov(RHS: b, Overflow&: overflowed); |
| 227 | return overflowed ? std::optional<APInt>() : result; |
| 228 | }; |
| 229 | ConstArithStdFn ssub = [=](const APInt &a, |
| 230 | const APInt &b) -> std::optional<APInt> { |
| 231 | bool overflowed = false; |
| 232 | APInt result = any(Val: ovfFlags & OverflowFlags::Nsw) |
| 233 | ? a.ssub_sat(RHS: b) |
| 234 | : a.ssub_ov(RHS: b, Overflow&: overflowed); |
| 235 | return overflowed ? std::optional<APInt>() : result; |
| 236 | }; |
| 237 | ConstantIntRanges urange = computeBoundsBy( |
| 238 | op: usub, minLeft: lhs.umin(), minRight: rhs.umax(), maxLeft: lhs.umax(), maxRight: rhs.umin(), /*isSigned=*/false); |
| 239 | ConstantIntRanges srange = computeBoundsBy( |
| 240 | op: ssub, minLeft: lhs.smin(), minRight: rhs.smax(), maxLeft: lhs.smax(), maxRight: rhs.smin(), /*isSigned=*/true); |
| 241 | return urange.intersection(other: srange); |
| 242 | } |
| 243 | |
| 244 | //===----------------------------------------------------------------------===// |
| 245 | // Multiplication |
| 246 | //===----------------------------------------------------------------------===// |
| 247 | |
| 248 | ConstantIntRanges |
| 249 | mlir::intrange::inferMul(ArrayRef<ConstantIntRanges> argRanges, |
| 250 | OverflowFlags ovfFlags) { |
| 251 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 252 | |
| 253 | ConstArithStdFn umul = [=](const APInt &a, |
| 254 | const APInt &b) -> std::optional<APInt> { |
| 255 | bool overflowed = false; |
| 256 | APInt result = any(Val: ovfFlags & OverflowFlags::Nuw) |
| 257 | ? a.umul_sat(RHS: b) |
| 258 | : a.umul_ov(RHS: b, Overflow&: overflowed); |
| 259 | return overflowed ? std::optional<APInt>() : result; |
| 260 | }; |
| 261 | ConstArithStdFn smul = [=](const APInt &a, |
| 262 | const APInt &b) -> std::optional<APInt> { |
| 263 | bool overflowed = false; |
| 264 | APInt result = any(Val: ovfFlags & OverflowFlags::Nsw) |
| 265 | ? a.smul_sat(RHS: b) |
| 266 | : a.smul_ov(RHS: b, Overflow&: overflowed); |
| 267 | return overflowed ? std::optional<APInt>() : result; |
| 268 | }; |
| 269 | |
| 270 | ConstantIntRanges urange = |
| 271 | minMaxBy(op: umul, lhs: {lhs.umin(), lhs.umax()}, rhs: {rhs.umin(), rhs.umax()}, |
| 272 | /*isSigned=*/false); |
| 273 | ConstantIntRanges srange = |
| 274 | minMaxBy(op: smul, lhs: {lhs.smin(), lhs.smax()}, rhs: {rhs.smin(), rhs.smax()}, |
| 275 | /*isSigned=*/true); |
| 276 | return urange.intersection(other: srange); |
| 277 | } |
| 278 | |
| 279 | //===----------------------------------------------------------------------===// |
| 280 | // DivU, CeilDivU (Unsigned division) |
| 281 | //===----------------------------------------------------------------------===// |
| 282 | |
| 283 | /// Fix up division results (ex. for ceiling and floor), returning an APInt |
| 284 | /// if there has been no overflow |
| 285 | using DivisionFixupFn = function_ref<std::optional<APInt>( |
| 286 | const APInt &lhs, const APInt &rhs, const APInt &result)>; |
| 287 | |
| 288 | static ConstantIntRanges inferDivURange(const ConstantIntRanges &lhs, |
| 289 | const ConstantIntRanges &rhs, |
| 290 | DivisionFixupFn fixup) { |
| 291 | const APInt &lhsMin = lhs.umin(), &lhsMax = lhs.umax(), &rhsMin = rhs.umin(), |
| 292 | &rhsMax = rhs.umax(); |
| 293 | |
| 294 | if (!rhsMin.isZero()) { |
| 295 | auto udiv = [&fixup](const APInt &a, |
| 296 | const APInt &b) -> std::optional<APInt> { |
| 297 | return fixup(a, b, a.udiv(RHS: b)); |
| 298 | }; |
| 299 | return minMaxBy(op: udiv, lhs: {lhsMin, lhsMax}, rhs: {rhsMin, rhsMax}, |
| 300 | /*isSigned=*/false); |
| 301 | } |
| 302 | |
| 303 | APInt umin = APInt::getZero(numBits: rhsMin.getBitWidth()); |
| 304 | if (lhsMin.uge(RHS: rhsMax) && !rhsMax.isZero()) |
| 305 | umin = lhsMin.udiv(RHS: rhsMax); |
| 306 | |
| 307 | // X u/ Y u<= X. |
| 308 | APInt umax = lhsMax; |
| 309 | return ConstantIntRanges::fromUnsigned(umin, umax); |
| 310 | } |
| 311 | |
| 312 | ConstantIntRanges |
| 313 | mlir::intrange::inferDivU(ArrayRef<ConstantIntRanges> argRanges) { |
| 314 | return inferDivURange(lhs: argRanges[0], rhs: argRanges[1], |
| 315 | fixup: [](const APInt &lhs, const APInt &rhs, |
| 316 | const APInt &result) { return result; }); |
| 317 | } |
| 318 | |
| 319 | ConstantIntRanges |
| 320 | mlir::intrange::inferCeilDivU(ArrayRef<ConstantIntRanges> argRanges) { |
| 321 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 322 | |
| 323 | auto ceilDivUIFix = [](const APInt &lhs, const APInt &rhs, |
| 324 | const APInt &result) -> std::optional<APInt> { |
| 325 | if (!lhs.urem(RHS: rhs).isZero()) { |
| 326 | bool overflowed = false; |
| 327 | APInt corrected = |
| 328 | result.uadd_ov(RHS: APInt(result.getBitWidth(), 1), Overflow&: overflowed); |
| 329 | return overflowed ? std::optional<APInt>() : corrected; |
| 330 | } |
| 331 | return result; |
| 332 | }; |
| 333 | return inferDivURange(lhs, rhs, fixup: ceilDivUIFix); |
| 334 | } |
| 335 | |
| 336 | //===----------------------------------------------------------------------===// |
| 337 | // DivS, CeilDivS, FloorDivS (Signed division) |
| 338 | //===----------------------------------------------------------------------===// |
| 339 | |
| 340 | static ConstantIntRanges inferDivSRange(const ConstantIntRanges &lhs, |
| 341 | const ConstantIntRanges &rhs, |
| 342 | DivisionFixupFn fixup) { |
| 343 | const APInt &lhsMin = lhs.smin(), &lhsMax = lhs.smax(), &rhsMin = rhs.smin(), |
| 344 | &rhsMax = rhs.smax(); |
| 345 | bool canDivide = rhsMin.isStrictlyPositive() || rhsMax.isNegative(); |
| 346 | |
| 347 | if (canDivide) { |
| 348 | auto sdiv = [&fixup](const APInt &a, |
| 349 | const APInt &b) -> std::optional<APInt> { |
| 350 | bool overflowed = false; |
| 351 | APInt result = a.sdiv_ov(RHS: b, Overflow&: overflowed); |
| 352 | return overflowed ? std::optional<APInt>() : fixup(a, b, result); |
| 353 | }; |
| 354 | return minMaxBy(op: sdiv, lhs: {lhsMin, lhsMax}, rhs: {rhsMin, rhsMax}, |
| 355 | /*isSigned=*/true); |
| 356 | } |
| 357 | return ConstantIntRanges::maxRange(bitwidth: rhsMin.getBitWidth()); |
| 358 | } |
| 359 | |
| 360 | ConstantIntRanges |
| 361 | mlir::intrange::inferDivS(ArrayRef<ConstantIntRanges> argRanges) { |
| 362 | return inferDivSRange(lhs: argRanges[0], rhs: argRanges[1], |
| 363 | fixup: [](const APInt &lhs, const APInt &rhs, |
| 364 | const APInt &result) { return result; }); |
| 365 | } |
| 366 | |
| 367 | ConstantIntRanges |
| 368 | mlir::intrange::inferCeilDivS(ArrayRef<ConstantIntRanges> argRanges) { |
| 369 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 370 | |
| 371 | auto ceilDivSIFix = [](const APInt &lhs, const APInt &rhs, |
| 372 | const APInt &result) -> std::optional<APInt> { |
| 373 | if (!lhs.srem(RHS: rhs).isZero() && lhs.isNonNegative() == rhs.isNonNegative()) { |
| 374 | bool overflowed = false; |
| 375 | APInt corrected = |
| 376 | result.sadd_ov(RHS: APInt(result.getBitWidth(), 1), Overflow&: overflowed); |
| 377 | return overflowed ? std::optional<APInt>() : corrected; |
| 378 | } |
| 379 | // Special case where the usual implementation of ceilDiv causes |
| 380 | // INT_MIN / [positive number] to be positive. This doesn't match the |
| 381 | // definition of signed ceiling division mathematically, but it prevents |
| 382 | // inconsistent constant-folding results. This arises because (-int_min) is |
| 383 | // still negative, so -(-int_min / b) is -(int_min / b), which is |
| 384 | // positive See #115293. |
| 385 | if (lhs.isMinSignedValue() && rhs.sgt(RHS: 1)) { |
| 386 | return -result; |
| 387 | } |
| 388 | return result; |
| 389 | }; |
| 390 | ConstantIntRanges result = inferDivSRange(lhs, rhs, fixup: ceilDivSIFix); |
| 391 | if (lhs.smin().isMinSignedValue() && lhs.smax().sgt(RHS: lhs.smin())) { |
| 392 | // If lhs range includes INT_MIN and lhs is not a single value, we can |
| 393 | // suddenly wrap to positive val, skipping entire negative range, add |
| 394 | // [INT_MIN + 1, smax()] range to the result to handle this. |
| 395 | auto newLhs = ConstantIntRanges::fromSigned(smin: lhs.smin() + 1, smax: lhs.smax()); |
| 396 | result = result.rangeUnion(other: inferDivSRange(lhs: newLhs, rhs, fixup: ceilDivSIFix)); |
| 397 | } |
| 398 | return result; |
| 399 | } |
| 400 | |
| 401 | ConstantIntRanges |
| 402 | mlir::intrange::inferFloorDivS(ArrayRef<ConstantIntRanges> argRanges) { |
| 403 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 404 | |
| 405 | auto floorDivSIFix = [](const APInt &lhs, const APInt &rhs, |
| 406 | const APInt &result) -> std::optional<APInt> { |
| 407 | if (!lhs.srem(RHS: rhs).isZero() && lhs.isNonNegative() != rhs.isNonNegative()) { |
| 408 | bool overflowed = false; |
| 409 | APInt corrected = |
| 410 | result.ssub_ov(RHS: APInt(result.getBitWidth(), 1), Overflow&: overflowed); |
| 411 | return overflowed ? std::optional<APInt>() : corrected; |
| 412 | } |
| 413 | return result; |
| 414 | }; |
| 415 | return inferDivSRange(lhs, rhs, fixup: floorDivSIFix); |
| 416 | } |
| 417 | |
| 418 | //===----------------------------------------------------------------------===// |
| 419 | // Signed remainder (RemS) |
| 420 | //===----------------------------------------------------------------------===// |
| 421 | |
| 422 | ConstantIntRanges |
| 423 | mlir::intrange::inferRemS(ArrayRef<ConstantIntRanges> argRanges) { |
| 424 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 425 | const APInt &lhsMin = lhs.smin(), &lhsMax = lhs.smax(), &rhsMin = rhs.smin(), |
| 426 | &rhsMax = rhs.smax(); |
| 427 | |
| 428 | unsigned width = rhsMax.getBitWidth(); |
| 429 | APInt smin = APInt::getSignedMinValue(numBits: width); |
| 430 | APInt smax = APInt::getSignedMaxValue(numBits: width); |
| 431 | // No bounds if zero could be a divisor. |
| 432 | bool canBound = (rhsMin.isStrictlyPositive() || rhsMax.isNegative()); |
| 433 | if (canBound) { |
| 434 | APInt maxDivisor = rhsMin.isStrictlyPositive() ? rhsMax : rhsMin.abs(); |
| 435 | bool canNegativeDividend = lhsMin.isNegative(); |
| 436 | bool canPositiveDividend = lhsMax.isStrictlyPositive(); |
| 437 | APInt zero = APInt::getZero(numBits: maxDivisor.getBitWidth()); |
| 438 | APInt maxPositiveResult = maxDivisor - 1; |
| 439 | APInt minNegativeResult = -maxPositiveResult; |
| 440 | smin = canNegativeDividend ? minNegativeResult : zero; |
| 441 | smax = canPositiveDividend ? maxPositiveResult : zero; |
| 442 | // Special case: sweeping out a contiguous range in N/[modulus]. |
| 443 | if (rhsMin == rhsMax) { |
| 444 | if ((lhsMax - lhsMin).ult(RHS: maxDivisor)) { |
| 445 | APInt minRem = lhsMin.srem(RHS: maxDivisor); |
| 446 | APInt maxRem = lhsMax.srem(RHS: maxDivisor); |
| 447 | if (minRem.sle(RHS: maxRem)) { |
| 448 | smin = minRem; |
| 449 | smax = maxRem; |
| 450 | } |
| 451 | } |
| 452 | } |
| 453 | } |
| 454 | return ConstantIntRanges::fromSigned(smin, smax); |
| 455 | } |
| 456 | |
| 457 | //===----------------------------------------------------------------------===// |
| 458 | // Unsigned remainder (RemU) |
| 459 | //===----------------------------------------------------------------------===// |
| 460 | |
| 461 | ConstantIntRanges |
| 462 | mlir::intrange::inferRemU(ArrayRef<ConstantIntRanges> argRanges) { |
| 463 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 464 | const APInt &rhsMin = rhs.umin(), &rhsMax = rhs.umax(); |
| 465 | |
| 466 | unsigned width = rhsMin.getBitWidth(); |
| 467 | APInt umin = APInt::getZero(numBits: width); |
| 468 | // Remainder can't be larger than either of its arguments. |
| 469 | APInt umax = llvm::APIntOps::umin(A: (rhsMax - 1), B: lhs.umax()); |
| 470 | |
| 471 | if (!rhsMin.isZero()) { |
| 472 | // Special case: sweeping out a contiguous range in N/[modulus] |
| 473 | if (rhsMin == rhsMax) { |
| 474 | const APInt &lhsMin = lhs.umin(), &lhsMax = lhs.umax(); |
| 475 | if ((lhsMax - lhsMin).ult(RHS: rhsMax)) { |
| 476 | APInt minRem = lhsMin.urem(RHS: rhsMax); |
| 477 | APInt maxRem = lhsMax.urem(RHS: rhsMax); |
| 478 | if (minRem.ule(RHS: maxRem)) { |
| 479 | umin = minRem; |
| 480 | umax = maxRem; |
| 481 | } |
| 482 | } |
| 483 | } |
| 484 | } |
| 485 | return ConstantIntRanges::fromUnsigned(umin, umax); |
| 486 | } |
| 487 | |
| 488 | //===----------------------------------------------------------------------===// |
| 489 | // Max and min (MaxS, MaxU, MinS, MinU) |
| 490 | //===----------------------------------------------------------------------===// |
| 491 | |
| 492 | ConstantIntRanges |
| 493 | mlir::intrange::inferMaxS(ArrayRef<ConstantIntRanges> argRanges) { |
| 494 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 495 | |
| 496 | const APInt &smin = lhs.smin().sgt(RHS: rhs.smin()) ? lhs.smin() : rhs.smin(); |
| 497 | const APInt &smax = lhs.smax().sgt(RHS: rhs.smax()) ? lhs.smax() : rhs.smax(); |
| 498 | return ConstantIntRanges::fromSigned(smin, smax); |
| 499 | } |
| 500 | |
| 501 | ConstantIntRanges |
| 502 | mlir::intrange::inferMaxU(ArrayRef<ConstantIntRanges> argRanges) { |
| 503 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 504 | |
| 505 | const APInt &umin = lhs.umin().ugt(RHS: rhs.umin()) ? lhs.umin() : rhs.umin(); |
| 506 | const APInt &umax = lhs.umax().ugt(RHS: rhs.umax()) ? lhs.umax() : rhs.umax(); |
| 507 | return ConstantIntRanges::fromUnsigned(umin, umax); |
| 508 | } |
| 509 | |
| 510 | ConstantIntRanges |
| 511 | mlir::intrange::inferMinS(ArrayRef<ConstantIntRanges> argRanges) { |
| 512 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 513 | |
| 514 | const APInt &smin = lhs.smin().slt(RHS: rhs.smin()) ? lhs.smin() : rhs.smin(); |
| 515 | const APInt &smax = lhs.smax().slt(RHS: rhs.smax()) ? lhs.smax() : rhs.smax(); |
| 516 | return ConstantIntRanges::fromSigned(smin, smax); |
| 517 | } |
| 518 | |
| 519 | ConstantIntRanges |
| 520 | mlir::intrange::inferMinU(ArrayRef<ConstantIntRanges> argRanges) { |
| 521 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 522 | |
| 523 | const APInt &umin = lhs.umin().ult(RHS: rhs.umin()) ? lhs.umin() : rhs.umin(); |
| 524 | const APInt &umax = lhs.umax().ult(RHS: rhs.umax()) ? lhs.umax() : rhs.umax(); |
| 525 | return ConstantIntRanges::fromUnsigned(umin, umax); |
| 526 | } |
| 527 | |
| 528 | //===----------------------------------------------------------------------===// |
| 529 | // Bitwise operators (And, Or, Xor) |
| 530 | //===----------------------------------------------------------------------===// |
| 531 | |
| 532 | /// "Widen" bounds - if 0bvvvvv??? <= a <= 0bvvvvv???, |
| 533 | /// relax the bounds to 0bvvvvv000 <= a <= 0bvvvvv111, where vvvvv are the bits |
| 534 | /// that both bonuds have in common. This gives us a consertive approximation |
| 535 | /// for what values can be passed to bitwise operations. |
| 536 | static std::tuple<APInt, APInt> |
| 537 | widenBitwiseBounds(const ConstantIntRanges &bound) { |
| 538 | APInt leftVal = bound.umin(), rightVal = bound.umax(); |
| 539 | unsigned bitwidth = leftVal.getBitWidth(); |
| 540 | unsigned differingBits = bitwidth - (leftVal ^ rightVal).countl_zero(); |
| 541 | leftVal.clearLowBits(loBits: differingBits); |
| 542 | rightVal.setLowBits(differingBits); |
| 543 | return std::make_tuple(args: std::move(leftVal), args: std::move(rightVal)); |
| 544 | } |
| 545 | |
| 546 | ConstantIntRanges |
| 547 | mlir::intrange::inferAnd(ArrayRef<ConstantIntRanges> argRanges) { |
| 548 | auto [lhsZeros, lhsOnes] = widenBitwiseBounds(bound: argRanges[0]); |
| 549 | auto [rhsZeros, rhsOnes] = widenBitwiseBounds(bound: argRanges[1]); |
| 550 | auto andi = [](const APInt &a, const APInt &b) -> std::optional<APInt> { |
| 551 | return a & b; |
| 552 | }; |
| 553 | return minMaxBy(op: andi, lhs: {lhsZeros, lhsOnes}, rhs: {rhsZeros, rhsOnes}, |
| 554 | /*isSigned=*/false); |
| 555 | } |
| 556 | |
| 557 | ConstantIntRanges |
| 558 | mlir::intrange::inferOr(ArrayRef<ConstantIntRanges> argRanges) { |
| 559 | auto [lhsZeros, lhsOnes] = widenBitwiseBounds(bound: argRanges[0]); |
| 560 | auto [rhsZeros, rhsOnes] = widenBitwiseBounds(bound: argRanges[1]); |
| 561 | auto ori = [](const APInt &a, const APInt &b) -> std::optional<APInt> { |
| 562 | return a | b; |
| 563 | }; |
| 564 | return minMaxBy(op: ori, lhs: {lhsZeros, lhsOnes}, rhs: {rhsZeros, rhsOnes}, |
| 565 | /*isSigned=*/false); |
| 566 | } |
| 567 | |
| 568 | /// Get bitmask of all bits which can change while iterating in |
| 569 | /// [bound.umin(), bound.umax()]. |
| 570 | static APInt getVaryingBitsMask(const ConstantIntRanges &bound) { |
| 571 | APInt leftVal = bound.umin(), rightVal = bound.umax(); |
| 572 | unsigned bitwidth = leftVal.getBitWidth(); |
| 573 | unsigned differingBits = bitwidth - (leftVal ^ rightVal).countl_zero(); |
| 574 | return APInt::getLowBitsSet(numBits: bitwidth, loBitsSet: differingBits); |
| 575 | } |
| 576 | |
| 577 | ConstantIntRanges |
| 578 | mlir::intrange::inferXor(ArrayRef<ConstantIntRanges> argRanges) { |
| 579 | // Construct mask of varying bits for both ranges, xor values and then replace |
| 580 | // masked bits with 0s and 1s to get min and max values respectively. |
| 581 | ConstantIntRanges lhs = argRanges[0], rhs = argRanges[1]; |
| 582 | APInt mask = getVaryingBitsMask(bound: lhs) | getVaryingBitsMask(bound: rhs); |
| 583 | APInt res = lhs.umin() ^ rhs.umin(); |
| 584 | APInt min = res & ~mask; |
| 585 | APInt max = res | mask; |
| 586 | return ConstantIntRanges::fromUnsigned(umin: min, umax: max); |
| 587 | } |
| 588 | |
| 589 | //===----------------------------------------------------------------------===// |
| 590 | // Shifts (Shl, ShrS, ShrU) |
| 591 | //===----------------------------------------------------------------------===// |
| 592 | |
| 593 | ConstantIntRanges |
| 594 | mlir::intrange::inferShl(ArrayRef<ConstantIntRanges> argRanges, |
| 595 | OverflowFlags ovfFlags) { |
| 596 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 597 | const APInt &rhsUMin = rhs.umin(), &rhsUMax = rhs.umax(); |
| 598 | |
| 599 | // The signed/unsigned overflow behavior of shl by `rhs` matches a mul with |
| 600 | // 2^rhs. |
| 601 | ConstArithStdFn ushl = [=](const APInt &l, |
| 602 | const APInt &r) -> std::optional<APInt> { |
| 603 | bool overflowed = false; |
| 604 | APInt result = any(Val: ovfFlags & OverflowFlags::Nuw) |
| 605 | ? l.ushl_sat(RHS: r) |
| 606 | : l.ushl_ov(Amt: r, Overflow&: overflowed); |
| 607 | return overflowed ? std::optional<APInt>() : result; |
| 608 | }; |
| 609 | ConstArithStdFn sshl = [=](const APInt &l, |
| 610 | const APInt &r) -> std::optional<APInt> { |
| 611 | bool overflowed = false; |
| 612 | APInt result = any(Val: ovfFlags & OverflowFlags::Nsw) |
| 613 | ? l.sshl_sat(RHS: r) |
| 614 | : l.sshl_ov(Amt: r, Overflow&: overflowed); |
| 615 | return overflowed ? std::optional<APInt>() : result; |
| 616 | }; |
| 617 | |
| 618 | ConstantIntRanges urange = |
| 619 | minMaxBy(op: ushl, lhs: {lhs.umin(), lhs.umax()}, rhs: {rhsUMin, rhsUMax}, |
| 620 | /*isSigned=*/false); |
| 621 | ConstantIntRanges srange = |
| 622 | minMaxBy(op: sshl, lhs: {lhs.smin(), lhs.smax()}, rhs: {rhsUMin, rhsUMax}, |
| 623 | /*isSigned=*/true); |
| 624 | return urange.intersection(other: srange); |
| 625 | } |
| 626 | |
| 627 | ConstantIntRanges |
| 628 | mlir::intrange::inferShrS(ArrayRef<ConstantIntRanges> argRanges) { |
| 629 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 630 | |
| 631 | auto ashr = [](const APInt &l, const APInt &r) -> std::optional<APInt> { |
| 632 | return r.uge(RHS: r.getBitWidth()) ? std::optional<APInt>() : l.ashr(ShiftAmt: r); |
| 633 | }; |
| 634 | |
| 635 | return minMaxBy(op: ashr, lhs: {lhs.smin(), lhs.smax()}, rhs: {rhs.umin(), rhs.umax()}, |
| 636 | /*isSigned=*/true); |
| 637 | } |
| 638 | |
| 639 | ConstantIntRanges |
| 640 | mlir::intrange::inferShrU(ArrayRef<ConstantIntRanges> argRanges) { |
| 641 | const ConstantIntRanges &lhs = argRanges[0], &rhs = argRanges[1]; |
| 642 | |
| 643 | auto lshr = [](const APInt &l, const APInt &r) -> std::optional<APInt> { |
| 644 | return r.uge(RHS: r.getBitWidth()) ? std::optional<APInt>() : l.lshr(ShiftAmt: r); |
| 645 | }; |
| 646 | return minMaxBy(op: lshr, lhs: {lhs.umin(), lhs.umax()}, rhs: {rhs.umin(), rhs.umax()}, |
| 647 | /*isSigned=*/false); |
| 648 | } |
| 649 | |
| 650 | //===----------------------------------------------------------------------===// |
| 651 | // Comparisons (Cmp) |
| 652 | //===----------------------------------------------------------------------===// |
| 653 | |
| 654 | static intrange::CmpPredicate invertPredicate(intrange::CmpPredicate pred) { |
| 655 | switch (pred) { |
| 656 | case intrange::CmpPredicate::eq: |
| 657 | return intrange::CmpPredicate::ne; |
| 658 | case intrange::CmpPredicate::ne: |
| 659 | return intrange::CmpPredicate::eq; |
| 660 | case intrange::CmpPredicate::slt: |
| 661 | return intrange::CmpPredicate::sge; |
| 662 | case intrange::CmpPredicate::sle: |
| 663 | return intrange::CmpPredicate::sgt; |
| 664 | case intrange::CmpPredicate::sgt: |
| 665 | return intrange::CmpPredicate::sle; |
| 666 | case intrange::CmpPredicate::sge: |
| 667 | return intrange::CmpPredicate::slt; |
| 668 | case intrange::CmpPredicate::ult: |
| 669 | return intrange::CmpPredicate::uge; |
| 670 | case intrange::CmpPredicate::ule: |
| 671 | return intrange::CmpPredicate::ugt; |
| 672 | case intrange::CmpPredicate::ugt: |
| 673 | return intrange::CmpPredicate::ule; |
| 674 | case intrange::CmpPredicate::uge: |
| 675 | return intrange::CmpPredicate::ult; |
| 676 | } |
| 677 | llvm_unreachable("unknown cmp predicate value" ); |
| 678 | } |
| 679 | |
| 680 | static bool isStaticallyTrue(intrange::CmpPredicate pred, |
| 681 | const ConstantIntRanges &lhs, |
| 682 | const ConstantIntRanges &rhs) { |
| 683 | switch (pred) { |
| 684 | case intrange::CmpPredicate::sle: |
| 685 | return lhs.smax().sle(RHS: rhs.smin()); |
| 686 | case intrange::CmpPredicate::slt: |
| 687 | return lhs.smax().slt(RHS: rhs.smin()); |
| 688 | case intrange::CmpPredicate::ule: |
| 689 | return lhs.umax().ule(RHS: rhs.umin()); |
| 690 | case intrange::CmpPredicate::ult: |
| 691 | return lhs.umax().ult(RHS: rhs.umin()); |
| 692 | case intrange::CmpPredicate::sge: |
| 693 | return lhs.smin().sge(RHS: rhs.smax()); |
| 694 | case intrange::CmpPredicate::sgt: |
| 695 | return lhs.smin().sgt(RHS: rhs.smax()); |
| 696 | case intrange::CmpPredicate::uge: |
| 697 | return lhs.umin().uge(RHS: rhs.umax()); |
| 698 | case intrange::CmpPredicate::ugt: |
| 699 | return lhs.umin().ugt(RHS: rhs.umax()); |
| 700 | case intrange::CmpPredicate::eq: { |
| 701 | std::optional<APInt> lhsConst = lhs.getConstantValue(); |
| 702 | std::optional<APInt> rhsConst = rhs.getConstantValue(); |
| 703 | return lhsConst && rhsConst && lhsConst == rhsConst; |
| 704 | } |
| 705 | case intrange::CmpPredicate::ne: { |
| 706 | // While equality requires that there is an interpration of the preceeding |
| 707 | // computations that produces equal constants, whether that be signed or |
| 708 | // unsigned, statically determining inequality requires that neither |
| 709 | // interpretation produce potentially overlapping ranges. |
| 710 | bool sne = isStaticallyTrue(pred: intrange::CmpPredicate::slt, lhs, rhs) || |
| 711 | isStaticallyTrue(pred: intrange::CmpPredicate::sgt, lhs, rhs); |
| 712 | bool une = isStaticallyTrue(pred: intrange::CmpPredicate::ult, lhs, rhs) || |
| 713 | isStaticallyTrue(pred: intrange::CmpPredicate::ugt, lhs, rhs); |
| 714 | return sne && une; |
| 715 | } |
| 716 | } |
| 717 | return false; |
| 718 | } |
| 719 | |
| 720 | std::optional<bool> mlir::intrange::evaluatePred(CmpPredicate pred, |
| 721 | const ConstantIntRanges &lhs, |
| 722 | const ConstantIntRanges &rhs) { |
| 723 | if (isStaticallyTrue(pred, lhs, rhs)) |
| 724 | return true; |
| 725 | if (isStaticallyTrue(pred: invertPredicate(pred), lhs, rhs)) |
| 726 | return false; |
| 727 | return std::nullopt; |
| 728 | } |
| 729 | |
| 730 | //===----------------------------------------------------------------------===// |
| 731 | // Shaped type dimension accessors / ShapedDimOpInterface |
| 732 | //===----------------------------------------------------------------------===// |
| 733 | |
| 734 | ConstantIntRanges |
| 735 | mlir::intrange::inferShapedDimOpInterface(ShapedDimOpInterface op, |
| 736 | const IntegerValueRange &maybeDim) { |
| 737 | unsigned width = |
| 738 | ConstantIntRanges::getStorageBitwidth(type: op->getResult(0).getType()); |
| 739 | APInt zero = APInt::getZero(numBits: width); |
| 740 | APInt typeMax = APInt::getSignedMaxValue(numBits: width); |
| 741 | |
| 742 | auto shapedTy = cast<ShapedType>(op.getShapedValue().getType()); |
| 743 | if (!shapedTy.hasRank()) |
| 744 | return ConstantIntRanges::fromSigned(smin: zero, smax: typeMax); |
| 745 | |
| 746 | int64_t rank = shapedTy.getRank(); |
| 747 | int64_t minDim = 0; |
| 748 | int64_t maxDim = rank - 1; |
| 749 | if (!maybeDim.isUninitialized()) { |
| 750 | const ConstantIntRanges &dim = maybeDim.getValue(); |
| 751 | minDim = std::max(a: minDim, b: dim.smin().getSExtValue()); |
| 752 | maxDim = std::min(a: maxDim, b: dim.smax().getSExtValue()); |
| 753 | } |
| 754 | |
| 755 | std::optional<ConstantIntRanges> result; |
| 756 | auto joinResult = [&](const ConstantIntRanges &thisResult) { |
| 757 | if (!result.has_value()) |
| 758 | result = thisResult; |
| 759 | else |
| 760 | result = result->rangeUnion(other: thisResult); |
| 761 | }; |
| 762 | for (int64_t i = minDim; i <= maxDim; ++i) { |
| 763 | int64_t length = shapedTy.getDimSize(i); |
| 764 | |
| 765 | if (ShapedType::isDynamic(length)) |
| 766 | joinResult(ConstantIntRanges::fromSigned(smin: zero, smax: typeMax)); |
| 767 | else |
| 768 | joinResult(ConstantIntRanges::constant(value: APInt(width, length))); |
| 769 | } |
| 770 | return result.value_or(u: ConstantIntRanges::fromSigned(smin: zero, smax: typeMax)); |
| 771 | } |
| 772 | |