File Manager

003 File Manager

Current Path: /usr/src/contrib/llvm-project/llvm/include/llvm/Transforms/Scalar

usr / src / contrib / llvm-project / llvm / include / llvm / Transforms / Scalar /

📁 ..
📄 ADCE.h(1.38 KB)
📄 AlignmentFromAssumptions.h(1.6 KB)
📄 BDCE.h(1.04 KB)
📄 CallSiteSplitting.h(810 B)
📄 ConstantHoisting.h(8.02 KB)
📄 CorrelatedValuePropagation.h(810 B)
📄 DCE.h(892 B)
📄 DeadStoreElimination.h(1.2 KB)
📄 DivRemPairs.h(1.01 KB)
📄 EarlyCSE.h(1.41 KB)
📄 Float2Int.h(1.74 KB)
📄 GVN.h(12.57 KB)
📄 GVNExpression.h(20.21 KB)
📄 GuardWidening.h(1.18 KB)
📄 IVUsersPrinter.h(960 B)
📄 IndVarSimplify.h(1.03 KB)
📄 InductiveRangeCheckElimination.h(950 B)
📄 InstSimplifyPass.h(1.64 KB)
📄 JumpThreading.h(6.55 KB)
📄 LICM.h(2.78 KB)
📄 LoopAccessAnalysisPrinter.h(999 B)
📄 LoopDataPrefetch.h(1.06 KB)
📄 LoopDeletion.h(1.13 KB)
📄 LoopDistribute.h(1.08 KB)
📄 LoopFuse.h(876 B)
📄 LoopIdiomRecognize.h(1.18 KB)
📄 LoopInstSimplify.h(1.07 KB)
📄 LoopLoadElimination.h(1.1 KB)
📄 LoopPassManager.h(15.85 KB)
📄 LoopPredication.h(1.12 KB)
📄 LoopRotation.h(1.13 KB)
📄 LoopSimplifyCFG.h(1.31 KB)
📄 LoopSink.h(1.54 KB)
📄 LoopStrengthReduce.h(1.52 KB)
📄 LoopUnrollAndJamPass.h(921 B)
📄 LoopUnrollPass.h(4.97 KB)
📄 LowerAtomic.h(974 B)
📄 LowerConstantIntrinsics.h(1.4 KB)
📄 LowerExpectIntrinsic.h(1.25 KB)
📄 LowerGuardIntrinsic.h(992 B)
📄 LowerMatrixIntrinsics.h(868 B)
📄 LowerWidenableCondition.h(937 B)
📄 MakeGuardsExplicit.h(1.85 KB)
📄 MemCpyOptimizer.h(2.71 KB)
📄 MergeICmps.h(714 B)
📄 MergedLoadStoreMotion.h(2.02 KB)
📄 NaryReassociate.h(6.98 KB)
📄 NewGVN.h(935 B)
📄 PartiallyInlineLibCalls.h(1.04 KB)
📄 Reassociate.h(4.6 KB)
📄 RewriteStatepointsForGC.h(1.27 KB)
📄 SCCP.h(1.81 KB)
📄 SROA.h(5.29 KB)
📄 Scalarizer.h(1.18 KB)
📄 SimpleLoopUnswitch.h(3.35 KB)
📄 SimplifyCFG.h(2.01 KB)
📄 Sink.h(998 B)
📄 SpeculateAroundPHIs.h(3.69 KB)
📄 SpeculativeExecution.h(3.04 KB)
📄 TailRecursionElimination.h(3.3 KB)
📄 WarnMissedTransforms.h(1.19 KB)

Editing: TailRecursionElimination.h

//===---- TailRecursionElimination.h ----------------------------*- C++ -*-===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
// This file transforms calls of the current function (self recursion) followed
// by a return instruction with a branch to the entry of the function, creating
// a loop.  This pass also implements the following extensions to the basic
// algorithm:
//
//  1. Trivial instructions between the call and return do not prevent the
//     transformation from taking place, though currently the analysis cannot
//     support moving any really useful instructions (only dead ones).
//  2. This pass transforms functions that are prevented from being tail
//     recursive by an associative and commutative expression to use an
//     accumulator variable, thus compiling the typical naive factorial or
//     'fib' implementation into efficient code.
//  3. TRE is performed if the function returns void, if the return
//     returns the result returned by the call, or if the function returns a
//     run-time constant on all exits from the function.  It is possible, though
//     unlikely, that the return returns something else (like constant 0), and
//     can still be TRE'd.  It can be TRE'd if ALL OTHER return instructions in
//     the function return the exact same value.
//  4. If it can prove that callees do not access their caller stack frame,
//     they are marked as eligible for tail call elimination (by the code
//     generator).
//
// There are several improvements that could be made:
//
//  1. If the function has any alloca instructions, these instructions will be
//     moved out of the entry block of the function, causing them to be
//     evaluated each time through the tail recursion.  Safely keeping allocas
//     in the entry block requires analysis to proves that the tail-called
//     function does not read or write the stack object.
//  2. Tail recursion is only performed if the call immediately precedes the
//     return instruction.  It's possible that there could be a jump between
//     the call and the return.
//  3. There can be intervening operations between the call and the return that
//     prevent the TRE from occurring.  For example, there could be GEP's and
//     stores to memory that will not be read or written by the call.  This
//     requires some substantial analysis (such as with DSA) to prove safe to
//     move ahead of the call, but doing so could allow many more TREs to be
//     performed, for example in TreeAdd/TreeAlloc from the treeadd benchmark.
//  4. The algorithm we use to detect if callees access their caller stack
//     frames is very primitive.
//
//===----------------------------------------------------------------------===//

#ifndef LLVM_TRANSFORMS_SCALAR_TAILRECURSIONELIMINATION_H
#define LLVM_TRANSFORMS_SCALAR_TAILRECURSIONELIMINATION_H

#include "llvm/IR/Function.h"
#include "llvm/IR/PassManager.h"

namespace llvm {

struct TailCallElimPass : PassInfoMixin<TailCallElimPass> {
  PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
};
}

#endif // LLVM_TRANSFORMS_SCALAR_TAILRECURSIONELIMINATION_H

003 File Manager

Editing: TailRecursionElimination.h

Upload File

Create Folder