1 //===- PolynomialApproximation.cpp - Approximate math operations ----------===// 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 implements expansion of math operations to fast approximations 10 // that do not rely on any of the library functions. 11 // 12 //===----------------------------------------------------------------------===// 13 #include "mlir/Dialect/LLVMIR/LLVMDialect.h" 14 #include "mlir/Dialect/LLVMIR/LLVMTypes.h" 15 #include "mlir/Dialect/Math/IR/Math.h" 16 #include "mlir/Dialect/Math/Transforms/Passes.h" 17 #include "mlir/Dialect/Vector/VectorOps.h" 18 #include "mlir/IR/Builders.h" 19 #include "mlir/IR/ImplicitLocOpBuilder.h" 20 #include "mlir/Transforms/DialectConversion.h" 21 #include "mlir/Transforms/GreedyPatternRewriteDriver.h" 22 #include <limits.h> 23 24 using namespace mlir; 25 using namespace mlir::vector; 26 27 using TypePredicate = llvm::function_ref<bool(Type)>; 28 29 // Returns vector width if the element type is matching the predicate (scalars 30 // that do match the predicate have width equal to `1`). 31 static Optional<int> vectorWidth(Type type, TypePredicate pred) { 32 // If the type matches the predicate then its width is `1`. 33 if (pred(type)) 34 return 1; 35 36 // Otherwise check if the type is a vector type. 37 auto vectorType = type.dyn_cast<VectorType>(); 38 if (vectorType && pred(vectorType.getElementType())) { 39 assert(vectorType.getRank() == 1 && "only 1d vectors are supported"); 40 return vectorType.getDimSize(0); 41 } 42 43 return llvm::None; 44 } 45 46 // Returns vector width of the type. If the type is a scalar returns `1`. 47 static int vectorWidth(Type type) { 48 auto vectorType = type.dyn_cast<VectorType>(); 49 return vectorType ? vectorType.getDimSize(0) : 1; 50 } 51 52 // Returns vector element type. If the type is a scalar returns the argument. 53 LLVM_ATTRIBUTE_UNUSED static Type elementType(Type type) { 54 auto vectorType = type.dyn_cast<VectorType>(); 55 return vectorType ? vectorType.getElementType() : type; 56 } 57 58 LLVM_ATTRIBUTE_UNUSED static bool isF32(Type type) { return type.isF32(); } 59 60 LLVM_ATTRIBUTE_UNUSED static bool isI32(Type type) { 61 return type.isInteger(32); 62 } 63 64 //----------------------------------------------------------------------------// 65 // Broadcast scalar types and values into vector types and values. 66 //----------------------------------------------------------------------------// 67 68 // Broadcasts scalar type into vector type (iff width is greater then 1). 69 static Type broadcast(Type type, int width) { 70 assert(!type.isa<VectorType>() && "must be scalar type"); 71 return width > 1 ? VectorType::get({width}, type) : type; 72 } 73 74 // Broadcasts scalar value into vector (iff width is greater then 1). 75 static Value broadcast(ImplicitLocOpBuilder &builder, Value value, int width) { 76 assert(!value.getType().isa<VectorType>() && "must be scalar value"); 77 auto type = broadcast(value.getType(), width); 78 return width > 1 ? builder.create<BroadcastOp>(type, value) : value; 79 } 80 81 //----------------------------------------------------------------------------// 82 // Helper functions to create constants. 83 //----------------------------------------------------------------------------// 84 85 static Value f32Cst(ImplicitLocOpBuilder &builder, float value) { 86 return builder.create<ConstantOp>(builder.getF32Type(), 87 builder.getF32FloatAttr(value)); 88 } 89 90 static Value i32Cst(ImplicitLocOpBuilder &builder, int32_t value) { 91 return builder.create<ConstantOp>(builder.getI32Type(), 92 builder.getI32IntegerAttr(value)); 93 } 94 95 static Value f32FromBits(ImplicitLocOpBuilder &builder, uint32_t bits) { 96 Value i32Value = i32Cst(builder, static_cast<int32_t>(bits)); 97 return builder.create<LLVM::BitcastOp>(builder.getF32Type(), i32Value); 98 } 99 100 //----------------------------------------------------------------------------// 101 // Helper functions to build math functions approximations. 102 //----------------------------------------------------------------------------// 103 104 static Value min(ImplicitLocOpBuilder &builder, Value a, Value b) { 105 return builder.create<SelectOp>( 106 builder.create<CmpFOp>(CmpFPredicate::OLT, a, b), a, b); 107 } 108 109 static Value max(ImplicitLocOpBuilder &builder, Value a, Value b) { 110 return builder.create<SelectOp>( 111 builder.create<CmpFOp>(CmpFPredicate::OGT, a, b), a, b); 112 } 113 114 static Value clamp(ImplicitLocOpBuilder &builder, Value value, Value lowerBound, 115 Value upperBound) { 116 return max(builder, min(builder, value, upperBound), lowerBound); 117 } 118 119 // Decomposes given floating point value `arg` into a normalized fraction and 120 // an integral power of two (see std::frexp). Returned values have float type. 121 static std::pair<Value, Value> frexp(ImplicitLocOpBuilder &builder, Value arg, 122 bool is_positive = false) { 123 assert(isF32(elementType(arg.getType())) && "argument must be f32 type"); 124 125 int width = vectorWidth(arg.getType()); 126 127 auto bcast = [&](Value value) -> Value { 128 return broadcast(builder, value, width); 129 }; 130 131 auto i32 = builder.getIntegerType(32); 132 auto i32Vec = broadcast(i32, width); 133 auto f32Vec = broadcast(builder.getF32Type(), width); 134 135 Value cst126f = f32Cst(builder, 126.0f); 136 Value cstHalf = f32Cst(builder, 0.5f); 137 Value cstInvMantMask = f32FromBits(builder, ~0x7f800000u); 138 139 // Bitcast to i32 for bitwise operations. 140 Value i32Half = builder.create<LLVM::BitcastOp>(i32, cstHalf); 141 Value i32InvMantMask = builder.create<LLVM::BitcastOp>(i32, cstInvMantMask); 142 Value i32Arg = builder.create<LLVM::BitcastOp>(i32Vec, arg); 143 144 // Compute normalized fraction. 145 Value tmp0 = builder.create<LLVM::AndOp>(i32Arg, bcast(i32InvMantMask)); 146 Value tmp1 = builder.create<LLVM::OrOp>(tmp0, bcast(i32Half)); 147 Value normalizedFraction = builder.create<LLVM::BitcastOp>(f32Vec, tmp1); 148 149 // Compute exponent. 150 Value arg0 = is_positive ? arg : builder.create<AbsFOp>(arg); 151 Value biasedExponentBits = builder.create<UnsignedShiftRightOp>( 152 builder.create<LLVM::BitcastOp>(i32Vec, arg0), 153 bcast(i32Cst(builder, 23))); 154 Value biasedExponent = builder.create<SIToFPOp>(f32Vec, biasedExponentBits); 155 Value exponent = builder.create<SubFOp>(biasedExponent, bcast(cst126f)); 156 157 return {normalizedFraction, exponent}; 158 } 159 160 // Computes exp2 for an i32 argument. 161 static Value exp2I32(ImplicitLocOpBuilder &builder, Value arg) { 162 assert(isI32(elementType(arg.getType())) && "argument must be i32 type"); 163 164 int width = vectorWidth(arg.getType()); 165 166 auto bcast = [&](Value value) -> Value { 167 return broadcast(builder, value, width); 168 }; 169 170 auto f32Vec = broadcast(builder.getF32Type(), width); 171 // The exponent of f32 located at 23-bit. 172 auto exponetBitLocation = bcast(i32Cst(builder, 23)); 173 // Set the exponent bias to zero. 174 auto bias = bcast(i32Cst(builder, 127)); 175 176 Value biasedArg = builder.create<AddIOp>(arg, bias); 177 Value exp2ValueInt = 178 builder.create<ShiftLeftOp>(biasedArg, exponetBitLocation); 179 Value exp2ValueF32 = builder.create<LLVM::BitcastOp>(f32Vec, exp2ValueInt); 180 181 return exp2ValueF32; 182 } 183 184 //----------------------------------------------------------------------------// 185 // TanhOp approximation. 186 //----------------------------------------------------------------------------// 187 188 namespace { 189 struct TanhApproximation : public OpRewritePattern<math::TanhOp> { 190 public: 191 using OpRewritePattern::OpRewritePattern; 192 193 LogicalResult matchAndRewrite(math::TanhOp op, 194 PatternRewriter &rewriter) const final; 195 }; 196 } // namespace 197 198 LogicalResult 199 TanhApproximation::matchAndRewrite(math::TanhOp op, 200 PatternRewriter &rewriter) const { 201 auto width = vectorWidth(op.operand().getType(), isF32); 202 if (!width.hasValue()) 203 return rewriter.notifyMatchFailure(op, "unsupported operand type"); 204 205 ImplicitLocOpBuilder builder(op->getLoc(), rewriter); 206 auto bcast = [&](Value value) -> Value { 207 return broadcast(builder, value, *width); 208 }; 209 210 // Clamp operand into [plusClamp, minusClamp] range. 211 Value minusClamp = bcast(f32Cst(builder, -7.9053111076354980f)); 212 Value plusClamp = bcast(f32Cst(builder, 7.90531110763549805f)); 213 Value x = clamp(builder, op.operand(), minusClamp, plusClamp); 214 215 // Mask for tiny values that are approximated with `operand`. 216 Value tiny = bcast(f32Cst(builder, 0.0004f)); 217 Value tinyMask = builder.create<CmpFOp>( 218 CmpFPredicate::OLT, builder.create<AbsFOp>(op.operand()), tiny); 219 220 // The monomial coefficients of the numerator polynomial (odd). 221 Value alpha1 = bcast(f32Cst(builder, 4.89352455891786e-03f)); 222 Value alpha3 = bcast(f32Cst(builder, 6.37261928875436e-04f)); 223 Value alpha5 = bcast(f32Cst(builder, 1.48572235717979e-05f)); 224 Value alpha7 = bcast(f32Cst(builder, 5.12229709037114e-08f)); 225 Value alpha9 = bcast(f32Cst(builder, -8.60467152213735e-11f)); 226 Value alpha11 = bcast(f32Cst(builder, 2.00018790482477e-13f)); 227 Value alpha13 = bcast(f32Cst(builder, -2.76076847742355e-16f)); 228 229 // The monomial coefficients of the denominator polynomial (even). 230 Value beta0 = bcast(f32Cst(builder, 4.89352518554385e-03f)); 231 Value beta2 = bcast(f32Cst(builder, 2.26843463243900e-03f)); 232 Value beta4 = bcast(f32Cst(builder, 1.18534705686654e-04f)); 233 Value beta6 = bcast(f32Cst(builder, 1.19825839466702e-06f)); 234 235 // Since the polynomials are odd/even, we need x^2. 236 Value x2 = builder.create<MulFOp>(x, x); 237 238 // Evaluate the numerator polynomial p. 239 Value p = builder.create<FmaFOp>(x2, alpha13, alpha11); 240 p = builder.create<FmaFOp>(x2, p, alpha9); 241 p = builder.create<FmaFOp>(x2, p, alpha7); 242 p = builder.create<FmaFOp>(x2, p, alpha5); 243 p = builder.create<FmaFOp>(x2, p, alpha3); 244 p = builder.create<FmaFOp>(x2, p, alpha1); 245 p = builder.create<MulFOp>(x, p); 246 247 // Evaluate the denominator polynomial q. 248 Value q = builder.create<FmaFOp>(x2, beta6, beta4); 249 q = builder.create<FmaFOp>(x2, q, beta2); 250 q = builder.create<FmaFOp>(x2, q, beta0); 251 252 // Divide the numerator by the denominator. 253 Value res = 254 builder.create<SelectOp>(tinyMask, x, builder.create<DivFOp>(p, q)); 255 256 rewriter.replaceOp(op, res); 257 258 return success(); 259 } 260 261 #define LN2_VALUE \ 262 0.693147180559945309417232121458176568075500134360255254120680009493393621L 263 #define LOG2E_VALUE \ 264 1.442695040888963407359924681001892137426645954152985934135449406931109219L 265 266 //----------------------------------------------------------------------------// 267 // LogOp and Log2Op approximation. 268 //----------------------------------------------------------------------------// 269 270 namespace { 271 template <typename Op> 272 struct LogApproximationBase : public OpRewritePattern<Op> { 273 using OpRewritePattern<Op>::OpRewritePattern; 274 275 /// Base 2 if 'base2' is set; natural logarithm (base e) otherwise. 276 LogicalResult logMatchAndRewrite(Op op, PatternRewriter &rewriter, 277 bool base2) const; 278 }; 279 } // namespace 280 281 // This approximation comes from Julien Pommier's SSE math library. 282 // Link: http://gruntthepeon.free.fr/ssemath 283 template <typename Op> 284 LogicalResult 285 LogApproximationBase<Op>::logMatchAndRewrite(Op op, PatternRewriter &rewriter, 286 bool base2) const { 287 auto width = vectorWidth(op.operand().getType(), isF32); 288 if (!width.hasValue()) 289 return rewriter.notifyMatchFailure(op, "unsupported operand type"); 290 291 ImplicitLocOpBuilder builder(op->getLoc(), rewriter); 292 auto bcast = [&](Value value) -> Value { 293 return broadcast(builder, value, *width); 294 }; 295 296 Value cstZero = bcast(f32Cst(builder, 0.0f)); 297 Value cstOne = bcast(f32Cst(builder, 1.0f)); 298 Value cstNegHalf = bcast(f32Cst(builder, -0.5f)); 299 300 // The smallest non denormalized float number. 301 Value cstMinNormPos = bcast(f32FromBits(builder, 0x00800000u)); 302 Value cstMinusInf = bcast(f32FromBits(builder, 0xff800000u)); 303 Value cstPosInf = bcast(f32FromBits(builder, 0x7f800000u)); 304 Value cstNan = bcast(f32FromBits(builder, 0x7fc00000)); 305 306 // Polynomial coefficients. 307 Value cstCephesSQRTHF = bcast(f32Cst(builder, 0.707106781186547524f)); 308 Value cstCephesLogP0 = bcast(f32Cst(builder, 7.0376836292E-2f)); 309 Value cstCephesLogP1 = bcast(f32Cst(builder, -1.1514610310E-1f)); 310 Value cstCephesLogP2 = bcast(f32Cst(builder, 1.1676998740E-1f)); 311 Value cstCephesLogP3 = bcast(f32Cst(builder, -1.2420140846E-1f)); 312 Value cstCephesLogP4 = bcast(f32Cst(builder, +1.4249322787E-1f)); 313 Value cstCephesLogP5 = bcast(f32Cst(builder, -1.6668057665E-1f)); 314 Value cstCephesLogP6 = bcast(f32Cst(builder, +2.0000714765E-1f)); 315 Value cstCephesLogP7 = bcast(f32Cst(builder, -2.4999993993E-1f)); 316 Value cstCephesLogP8 = bcast(f32Cst(builder, +3.3333331174E-1f)); 317 318 Value x = op.operand(); 319 320 // Truncate input values to the minimum positive normal. 321 x = max(builder, x, cstMinNormPos); 322 323 // Extract significant in the range [0.5,1) and exponent. 324 std::pair<Value, Value> pair = frexp(builder, x, /*is_positive=*/true); 325 x = pair.first; 326 Value e = pair.second; 327 328 // Shift the inputs from the range [0.5,1) to [sqrt(1/2), sqrt(2)) and shift 329 // by -1.0. The values are then centered around 0, which improves the 330 // stability of the polynomial evaluation: 331 // 332 // if( x < SQRTHF ) { 333 // e -= 1; 334 // x = x + x - 1.0; 335 // } else { x = x - 1.0; } 336 Value mask = builder.create<CmpFOp>(CmpFPredicate::OLT, x, cstCephesSQRTHF); 337 Value tmp = builder.create<SelectOp>(mask, x, cstZero); 338 339 x = builder.create<SubFOp>(x, cstOne); 340 e = builder.create<SubFOp>(e, 341 builder.create<SelectOp>(mask, cstOne, cstZero)); 342 x = builder.create<AddFOp>(x, tmp); 343 344 Value x2 = builder.create<MulFOp>(x, x); 345 Value x3 = builder.create<MulFOp>(x2, x); 346 347 // Evaluate the polynomial approximant of degree 8 in three parts. 348 Value y0, y1, y2; 349 y0 = builder.create<FmaFOp>(cstCephesLogP0, x, cstCephesLogP1); 350 y1 = builder.create<FmaFOp>(cstCephesLogP3, x, cstCephesLogP4); 351 y2 = builder.create<FmaFOp>(cstCephesLogP6, x, cstCephesLogP7); 352 y0 = builder.create<FmaFOp>(y0, x, cstCephesLogP2); 353 y1 = builder.create<FmaFOp>(y1, x, cstCephesLogP5); 354 y2 = builder.create<FmaFOp>(y2, x, cstCephesLogP8); 355 y0 = builder.create<FmaFOp>(y0, x3, y1); 356 y0 = builder.create<FmaFOp>(y0, x3, y2); 357 y0 = builder.create<MulFOp>(y0, x3); 358 359 y0 = builder.create<FmaFOp>(cstNegHalf, x2, y0); 360 x = builder.create<AddFOp>(x, y0); 361 362 if (base2) { 363 Value cstLog2e = bcast(f32Cst(builder, static_cast<float>(LOG2E_VALUE))); 364 x = builder.create<FmaFOp>(x, cstLog2e, e); 365 } else { 366 Value cstLn2 = bcast(f32Cst(builder, static_cast<float>(LN2_VALUE))); 367 x = builder.create<FmaFOp>(e, cstLn2, x); 368 } 369 370 Value invalidMask = 371 builder.create<CmpFOp>(CmpFPredicate::ULT, op.operand(), cstZero); 372 Value zeroMask = 373 builder.create<CmpFOp>(CmpFPredicate::OEQ, op.operand(), cstZero); 374 Value posInfMask = 375 builder.create<CmpFOp>(CmpFPredicate::OEQ, op.operand(), cstPosInf); 376 377 // Filter out invalid values: 378 // • x == 0 -> -INF 379 // • x < 0 -> NAN 380 // • x == +INF -> +INF 381 Value aproximation = builder.create<SelectOp>( 382 zeroMask, cstMinusInf, 383 builder.create<SelectOp>( 384 invalidMask, cstNan, 385 builder.create<SelectOp>(posInfMask, cstPosInf, x))); 386 387 rewriter.replaceOp(op, aproximation); 388 389 return success(); 390 } 391 392 namespace { 393 struct LogApproximation : public LogApproximationBase<math::LogOp> { 394 using LogApproximationBase::LogApproximationBase; 395 396 LogicalResult matchAndRewrite(math::LogOp op, 397 PatternRewriter &rewriter) const final { 398 return logMatchAndRewrite(op, rewriter, /*base2=*/false); 399 } 400 }; 401 } // namespace 402 403 namespace { 404 struct Log2Approximation : public LogApproximationBase<math::Log2Op> { 405 using LogApproximationBase::LogApproximationBase; 406 407 LogicalResult matchAndRewrite(math::Log2Op op, 408 PatternRewriter &rewriter) const final { 409 return logMatchAndRewrite(op, rewriter, /*base2=*/true); 410 } 411 }; 412 } // namespace 413 414 //----------------------------------------------------------------------------// 415 // Exp approximation. 416 //----------------------------------------------------------------------------// 417 418 namespace { 419 420 struct ExpApproximation : public OpRewritePattern<math::ExpOp> { 421 public: 422 using OpRewritePattern::OpRewritePattern; 423 424 LogicalResult matchAndRewrite(math::ExpOp op, 425 PatternRewriter &rewriter) const final; 426 }; 427 } // namespace 428 429 // Approximate exp(x) using its reduced range exp(y) where y is in the range 430 // [0, ln(2)], let y = x - floor(x / ln(2)) * ln(2) = x - k * ln(2), exp(x) 431 // = exp(y) * 2^k. exp(y). 432 LogicalResult 433 ExpApproximation::matchAndRewrite(math::ExpOp op, 434 PatternRewriter &rewriter) const { 435 auto width = vectorWidth(op.operand().getType(), isF32); 436 if (!width.hasValue()) 437 return rewriter.notifyMatchFailure(op, "unsupported operand type"); 438 ImplicitLocOpBuilder builder(op->getLoc(), rewriter); 439 440 // TODO: Consider a common pattern rewriter with all methods below to 441 // write the approximations. 442 auto bcast = [&](Value value) -> Value { 443 return broadcast(builder, value, *width); 444 }; 445 auto fmla = [&](Value a, Value b, Value c) { 446 return builder.create<FmaFOp>(a, b, c); 447 }; 448 auto mul = [&](Value a, Value b) -> Value { 449 return builder.create<MulFOp>(a, b); 450 }; 451 auto sub = [&](Value a, Value b) -> Value { 452 return builder.create<SubFOp>(a, b); 453 }; 454 auto floor = [&](Value a) { return builder.create<FloorFOp>(a); }; 455 456 Value cstLn2 = bcast(f32Cst(builder, static_cast<float>(LN2_VALUE))); 457 Value cstLog2E = bcast(f32Cst(builder, static_cast<float>(LOG2E_VALUE))); 458 459 // Polynomial coefficients. 460 Value cstCephesExpP0 = bcast(f32Cst(builder, 1.0)); 461 Value cstCephesExpP1 = bcast(f32Cst(builder, 1.0)); 462 Value cstCephesExpP2 = bcast(f32Cst(builder, 0.49970514590562437052f)); 463 Value cstCephesExpP3 = bcast(f32Cst(builder, 0.16873890085469545053f)); 464 Value cstCephesExpP4 = bcast(f32Cst(builder, 0.03668965196652099192f)); 465 Value cstCephesExpP5 = bcast(f32Cst(builder, 0.01314350012789660196f)); 466 467 Value x = op.operand(); 468 469 // Reduced y = x - floor(x / ln(2)) * ln(2) = x - k * ln(2) 470 Value xL2Inv = mul(x, cstLog2E); 471 Value kF32 = floor(xL2Inv); 472 Value kLn2 = mul(kF32, cstLn2); 473 Value y = sub(x, kLn2); 474 475 // Use Estrin's evaluation scheme with 3 independent parts: 476 // P(y)^y : (c0 + c1 y) + (c2 + c3 y) y^2 + (c4 + c5 y) y^4 477 Value y2 = mul(y, y); 478 Value y4 = mul(y2, y2); 479 480 Value q0 = fmla(cstCephesExpP1, y, cstCephesExpP0); 481 Value q1 = fmla(cstCephesExpP3, y, cstCephesExpP2); 482 Value q2 = fmla(cstCephesExpP5, y, cstCephesExpP4); 483 Value expY = fmla(q1, y2, q0); 484 expY = fmla(q2, y4, expY); 485 486 auto i32Vec = broadcast(builder.getI32Type(), *width); 487 488 // exp2(k) 489 Value k = builder.create<FPToSIOp>(kF32, i32Vec); 490 Value exp2KValue = exp2I32(builder, k); 491 492 // exp(x) = exp(y) * exp2(k) 493 expY = mul(expY, exp2KValue); 494 495 // Handle overflow, inf and underflow of exp(x). exp(x) range is [0, inf], its 496 // partitioned as the following: 497 // exp(x) = 0, x <= -inf 498 // exp(x) = underflow (min_float), x <= -88 499 // exp(x) = inf (min_float), x >= 88 500 // Note: |k| = 127 is the value where the 8-bits exponent saturates. 501 Value zerof32Const = bcast(f32Cst(builder, 0)); 502 auto constPosInfinity = 503 bcast(f32Cst(builder, std::numeric_limits<float>::infinity())); 504 auto constNegIfinity = 505 bcast(f32Cst(builder, -std::numeric_limits<float>::infinity())); 506 auto underflow = bcast(f32Cst(builder, std::numeric_limits<float>::min())); 507 508 Value kMaxConst = bcast(i32Cst(builder, 127)); 509 Value kMaxNegConst = bcast(i32Cst(builder, -127)); 510 Value rightBound = builder.create<CmpIOp>(CmpIPredicate::sle, k, kMaxConst); 511 Value leftBound = builder.create<CmpIOp>(CmpIPredicate::sge, k, kMaxNegConst); 512 513 Value isNegInfinityX = 514 builder.create<CmpFOp>(CmpFPredicate::OEQ, x, constNegIfinity); 515 Value isPostiveX = 516 builder.create<CmpFOp>(CmpFPredicate::OGT, x, zerof32Const); 517 Value isComputable = builder.create<AndOp>(rightBound, leftBound); 518 519 expY = builder.create<SelectOp>( 520 isComputable, expY, 521 builder.create<SelectOp>( 522 isPostiveX, constPosInfinity, 523 builder.create<SelectOp>(isNegInfinityX, zerof32Const, underflow))); 524 525 rewriter.replaceOp(op, expY); 526 527 return success(); 528 } 529 530 //----------------------------------------------------------------------------// 531 532 void mlir::populateMathPolynomialApproximationPatterns( 533 OwningRewritePatternList &patterns, MLIRContext *ctx) { 534 patterns.insert<TanhApproximation, LogApproximation, Log2Approximation, 535 ExpApproximation>(ctx); 536 } 537