1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
|
//=== ReplaceWithVeclib.cpp - Replace vector intrinsics with veclib calls -===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// Replaces LLVM IR instructions with vector operands (i.e., the frem
// instruction or calls to LLVM intrinsics) with matching calls to functions
// from a vector library (e.g libmvec, SVML) using TargetLibraryInfo interface.
//
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/ReplaceWithVeclib.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Analysis/DemandedBits.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/VFABIDemangler.h"
#include "llvm/Support/TypeSize.h"
#include "llvm/Transforms/Utils/ModuleUtils.h"
using namespace llvm;
#define DEBUG_TYPE "replace-with-veclib"
STATISTIC(NumCallsReplaced,
"Number of calls to intrinsics that have been replaced.");
STATISTIC(NumTLIFuncDeclAdded,
"Number of vector library function declarations added.");
STATISTIC(NumFuncUsedAdded,
"Number of functions added to `llvm.compiler.used`");
/// Returns a vector Function that it adds to the Module \p M. When an \p
/// ScalarFunc is not null, it copies its attributes to the newly created
/// Function.
Function *getTLIFunction(Module *M, FunctionType *VectorFTy,
const StringRef TLIName,
Function *ScalarFunc = nullptr) {
Function *TLIFunc = M->getFunction(TLIName);
if (!TLIFunc) {
TLIFunc =
Function::Create(VectorFTy, Function::ExternalLinkage, TLIName, *M);
if (ScalarFunc)
TLIFunc->copyAttributesFrom(ScalarFunc);
LLVM_DEBUG(dbgs() << DEBUG_TYPE << ": Added vector library function `"
<< TLIName << "` of type `" << *(TLIFunc->getType())
<< "` to module.\n");
++NumTLIFuncDeclAdded;
// Add the freshly created function to llvm.compiler.used, similar to as it
// is done in InjectTLIMappings.
appendToCompilerUsed(*M, {TLIFunc});
LLVM_DEBUG(dbgs() << DEBUG_TYPE << ": Adding `" << TLIName
<< "` to `@llvm.compiler.used`.\n");
++NumFuncUsedAdded;
}
return TLIFunc;
}
/// Replace the instruction \p I with a call to the corresponding function from
/// the vector library (\p TLIVecFunc).
static void replaceWithTLIFunction(Instruction &I, VFInfo &Info,
Function *TLIVecFunc) {
IRBuilder<> IRBuilder(&I);
auto *CI = dyn_cast<CallInst>(&I);
SmallVector<Value *> Args(CI ? CI->args() : I.operands());
if (auto OptMaskpos = Info.getParamIndexForOptionalMask()) {
auto *MaskTy =
VectorType::get(Type::getInt1Ty(I.getContext()), Info.Shape.VF);
Args.insert(Args.begin() + OptMaskpos.value(),
Constant::getAllOnesValue(MaskTy));
}
// If it is a call instruction, preserve the operand bundles.
SmallVector<OperandBundleDef, 1> OpBundles;
if (CI)
CI->getOperandBundlesAsDefs(OpBundles);
auto *Replacement = IRBuilder.CreateCall(TLIVecFunc, Args, OpBundles);
I.replaceAllUsesWith(Replacement);
// Preserve fast math flags for FP math.
if (isa<FPMathOperator>(Replacement))
Replacement->copyFastMathFlags(&I);
}
/// Returns true when successfully replaced \p I with a suitable function taking
/// vector arguments, based on available mappings in the \p TLI. Currently only
/// works when \p I is a call to vectorized intrinsic or the frem instruction.
static bool replaceWithCallToVeclib(const TargetLibraryInfo &TLI,
Instruction &I) {
// At the moment VFABI assumes the return type is always widened unless it is
// a void type.
auto *VTy = dyn_cast<VectorType>(I.getType());
ElementCount EC(VTy ? VTy->getElementCount() : ElementCount::getFixed(0));
// Compute the argument types of the corresponding scalar call and the scalar
// function name. For calls, it additionally finds the function to replace
// and checks that all vector operands match the previously found EC.
SmallVector<Type *, 8> ScalarArgTypes;
std::string ScalarName;
Function *FuncToReplace = nullptr;
auto *CI = dyn_cast<CallInst>(&I);
if (CI) {
FuncToReplace = CI->getCalledFunction();
Intrinsic::ID IID = FuncToReplace->getIntrinsicID();
assert(IID != Intrinsic::not_intrinsic && "Not an intrinsic");
for (auto Arg : enumerate(CI->args())) {
auto *ArgTy = Arg.value()->getType();
if (isVectorIntrinsicWithScalarOpAtArg(IID, Arg.index())) {
ScalarArgTypes.push_back(ArgTy);
} else if (auto *VectorArgTy = dyn_cast<VectorType>(ArgTy)) {
ScalarArgTypes.push_back(VectorArgTy->getElementType());
// When return type is void, set EC to the first vector argument, and
// disallow vector arguments with different ECs.
if (EC.isZero())
EC = VectorArgTy->getElementCount();
else if (EC != VectorArgTy->getElementCount())
return false;
} else
// Exit when it is supposed to be a vector argument but it isn't.
return false;
}
// Try to reconstruct the name for the scalar version of the instruction,
// using scalar argument types.
ScalarName = Intrinsic::isOverloaded(IID)
? Intrinsic::getName(IID, ScalarArgTypes, I.getModule())
: Intrinsic::getName(IID).str();
} else {
assert(VTy && "Return type must be a vector");
auto *ScalarTy = VTy->getScalarType();
LibFunc Func;
if (!TLI.getLibFunc(I.getOpcode(), ScalarTy, Func))
return false;
ScalarName = TLI.getName(Func);
ScalarArgTypes = {ScalarTy, ScalarTy};
}
// Try to find the mapping for the scalar version of this intrinsic and the
// exact vector width of the call operands in the TargetLibraryInfo. First,
// check with a non-masked variant, and if that fails try with a masked one.
const VecDesc *VD =
TLI.getVectorMappingInfo(ScalarName, EC, /*Masked*/ false);
if (!VD && !(VD = TLI.getVectorMappingInfo(ScalarName, EC, /*Masked*/ true)))
return false;
LLVM_DEBUG(dbgs() << DEBUG_TYPE << ": Found TLI mapping from: `" << ScalarName
<< "` and vector width " << EC << " to: `"
<< VD->getVectorFnName() << "`.\n");
// Replace the call to the intrinsic with a call to the vector library
// function.
Type *ScalarRetTy = I.getType()->getScalarType();
FunctionType *ScalarFTy =
FunctionType::get(ScalarRetTy, ScalarArgTypes, /*isVarArg*/ false);
const std::string MangledName = VD->getVectorFunctionABIVariantString();
auto OptInfo = VFABI::tryDemangleForVFABI(MangledName, ScalarFTy);
if (!OptInfo)
return false;
// There is no guarantee that the vectorized instructions followed the VFABI
// specification when being created, this is why we need to add extra check to
// make sure that the operands of the vector function obtained via VFABI match
// the operands of the original vector instruction.
if (CI) {
for (auto &VFParam : OptInfo->Shape.Parameters) {
if (VFParam.ParamKind == VFParamKind::GlobalPredicate)
continue;
// tryDemangleForVFABI must return valid ParamPos, otherwise it could be
// a bug in the VFABI parser.
assert(VFParam.ParamPos < CI->arg_size() &&
"ParamPos has invalid range.");
Type *OrigTy = CI->getArgOperand(VFParam.ParamPos)->getType();
if (OrigTy->isVectorTy() != (VFParam.ParamKind == VFParamKind::Vector)) {
LLVM_DEBUG(dbgs() << DEBUG_TYPE << ": Will not replace: " << ScalarName
<< ". Wrong type at index " << VFParam.ParamPos
<< ": " << *OrigTy << "\n");
return false;
}
}
}
FunctionType *VectorFTy = VFABI::createFunctionType(*OptInfo, ScalarFTy);
if (!VectorFTy)
return false;
Function *TLIFunc = getTLIFunction(I.getModule(), VectorFTy,
VD->getVectorFnName(), FuncToReplace);
replaceWithTLIFunction(I, *OptInfo, TLIFunc);
LLVM_DEBUG(dbgs() << DEBUG_TYPE << ": Replaced call to `" << ScalarName
<< "` with call to `" << TLIFunc->getName() << "`.\n");
++NumCallsReplaced;
return true;
}
/// Supported instruction \p I must be a vectorized frem or a call to an
/// intrinsic that returns either void or a vector.
static bool isSupportedInstruction(Instruction *I) {
Type *Ty = I->getType();
if (auto *CI = dyn_cast<CallInst>(I))
return (Ty->isVectorTy() || Ty->isVoidTy()) && CI->getCalledFunction() &&
CI->getCalledFunction()->getIntrinsicID() !=
Intrinsic::not_intrinsic;
if (I->getOpcode() == Instruction::FRem && Ty->isVectorTy())
return true;
return false;
}
static bool runImpl(const TargetLibraryInfo &TLI, Function &F) {
bool Changed = false;
SmallVector<Instruction *> ReplacedCalls;
for (auto &I : instructions(F)) {
if (!isSupportedInstruction(&I))
continue;
if (replaceWithCallToVeclib(TLI, I)) {
ReplacedCalls.push_back(&I);
Changed = true;
}
}
// Erase the calls to the intrinsics that have been replaced
// with calls to the vector library.
for (auto *CI : ReplacedCalls)
CI->eraseFromParent();
return Changed;
}
////////////////////////////////////////////////////////////////////////////////
// New pass manager implementation.
////////////////////////////////////////////////////////////////////////////////
PreservedAnalyses ReplaceWithVeclib::run(Function &F,
FunctionAnalysisManager &AM) {
const TargetLibraryInfo &TLI = AM.getResult<TargetLibraryAnalysis>(F);
auto Changed = runImpl(TLI, F);
if (Changed) {
LLVM_DEBUG(dbgs() << "Instructions replaced with vector libraries: "
<< NumCallsReplaced << "\n");
PreservedAnalyses PA;
PA.preserveSet<CFGAnalyses>();
PA.preserve<TargetLibraryAnalysis>();
PA.preserve<ScalarEvolutionAnalysis>();
PA.preserve<LoopAccessAnalysis>();
PA.preserve<DemandedBitsAnalysis>();
PA.preserve<OptimizationRemarkEmitterAnalysis>();
return PA;
}
// The pass did not replace any calls, hence it preserves all analyses.
return PreservedAnalyses::all();
}
////////////////////////////////////////////////////////////////////////////////
// Legacy PM Implementation.
////////////////////////////////////////////////////////////////////////////////
bool ReplaceWithVeclibLegacy::runOnFunction(Function &F) {
const TargetLibraryInfo &TLI =
getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F);
return runImpl(TLI, F);
}
void ReplaceWithVeclibLegacy::getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesCFG();
AU.addRequired<TargetLibraryInfoWrapperPass>();
AU.addPreserved<TargetLibraryInfoWrapperPass>();
AU.addPreserved<ScalarEvolutionWrapperPass>();
AU.addPreserved<AAResultsWrapperPass>();
AU.addPreserved<OptimizationRemarkEmitterWrapperPass>();
AU.addPreserved<GlobalsAAWrapperPass>();
}
////////////////////////////////////////////////////////////////////////////////
// Legacy Pass manager initialization
////////////////////////////////////////////////////////////////////////////////
char ReplaceWithVeclibLegacy::ID = 0;
INITIALIZE_PASS_BEGIN(ReplaceWithVeclibLegacy, DEBUG_TYPE,
"Replace intrinsics with calls to vector library", false,
false)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_PASS_END(ReplaceWithVeclibLegacy, DEBUG_TYPE,
"Replace intrinsics with calls to vector library", false,
false)
FunctionPass *llvm::createReplaceWithVeclibLegacyPass() {
return new ReplaceWithVeclibLegacy();
}
|