1 // Licensed to the .NET Foundation under one or more agreements.
2 // The .NET Foundation licenses this file to you under the MIT license.
3 // See the LICENSE file in the project root for more information.
6 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
7 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
9 Linear Scan Register Allocation
14 - All register requirements are expressed in the code stream, either as destination
15 registers of tree nodes, or as internal registers. These requirements are
16 expressed in the TreeNodeInfo (gtLsraInfo) on each node, which includes:
17 - The number of register sources and destinations.
18 - The register restrictions (candidates) of the target register, both from itself,
19 as producer of the value (dstCandidates), and from its consuming node (srcCandidates).
20 Note that the srcCandidates field of TreeNodeInfo refers to the destination register
21 (not any of its sources).
22 - The number (internalCount) of registers required, and their register restrictions (internalCandidates).
23 These are neither inputs nor outputs of the node, but used in the sequence of code generated for the tree.
24 "Internal registers" are registers used during the code sequence generated for the node.
25 The register lifetimes must obey the following lifetime model:
26 - First, any internal registers are defined.
27 - Next, any source registers are used (and are then freed if they are last use and are not identified as
29 - Next, the internal registers are used (and are then freed).
30 - Next, any registers in the kill set for the instruction are killed.
31 - Next, the destination register(s) are defined (multiple destination registers are only supported on ARM)
32 - Finally, any "delayRegFree" source registers are freed.
33 There are several things to note about this order:
34 - The internal registers will never overlap any use, but they may overlap a destination register.
35 - Internal registers are never live beyond the node.
36 - The "delayRegFree" annotation is used for instructions that are only available in a Read-Modify-Write form.
37 That is, the destination register is one of the sources. In this case, we must not use the same register for
38 the non-RMW operand as for the destination.
40 Overview (doLinearScan):
41 - Walk all blocks, building intervals and RefPositions (buildIntervals)
42 - Allocate registers (allocateRegisters)
43 - Annotate nodes with register assignments (resolveRegisters)
44 - Add move nodes as needed to resolve conflicting register
45 assignments across non-adjacent edges. (resolveEdges, called from resolveRegisters)
50 - GenTree::gtRegNum (and gtRegPair for ARM) is annotated with the register
51 assignment for a node. If the node does not require a register, it is
52 annotated as such (for single registers, gtRegNum = REG_NA; for register
53 pair type, gtRegPair = REG_PAIR_NONE). For a variable definition or interior
54 tree node (an "implicit" definition), this is the register to put the result.
55 For an expression use, this is the place to find the value that has previously
57 - In most cases, this register must satisfy the constraints specified by the TreeNodeInfo.
58 - In some cases, this is difficult:
59 - If a lclVar node currently lives in some register, it may not be desirable to move it
60 (i.e. its current location may be desirable for future uses, e.g. if it's a callee save register,
61 but needs to be in a specific arg register for a call).
62 - In other cases there may be conflicts on the restrictions placed by the defining node and the node which
64 - If such a node is constrained to a single fixed register (e.g. an arg register, or a return from a call),
65 then LSRA is free to annotate the node with a different register. The code generator must issue the appropriate
67 - However, if such a node is constrained to a set of registers, and its current location does not satisfy that
68 requirement, LSRA must insert a GT_COPY node between the node and its parent. The gtRegNum on the GT_COPY node
69 must satisfy the register requirement of the parent.
70 - GenTree::gtRsvdRegs has a set of registers used for internal temps.
71 - A tree node is marked GTF_SPILL if the tree node must be spilled by the code generator after it has been
73 - LSRA currently does not set GTF_SPILLED on such nodes, because it caused problems in the old code generator.
74 In the new backend perhaps this should change (see also the note below under CodeGen).
75 - A tree node is marked GTF_SPILLED if it is a lclVar that must be reloaded prior to use.
76 - The register (gtRegNum) on the node indicates the register to which it must be reloaded.
77 - For lclVar nodes, since the uses and defs are distinct tree nodes, it is always possible to annotate the node
78 with the register to which the variable must be reloaded.
79 - For other nodes, since they represent both the def and use, if the value must be reloaded to a different
80 register, LSRA must insert a GT_RELOAD node in order to specify the register to which it should be reloaded.
82 Local variable table (LclVarDsc):
83 - LclVarDsc::lvRegister is set to true if a local variable has the
84 same register assignment for its entire lifetime.
85 - LclVarDsc::lvRegNum / lvOtherReg: these are initialized to their
86 first value at the end of LSRA (it looks like lvOtherReg isn't?
87 This is probably a bug (ARM)). Codegen will set them to their current value
88 as it processes the trees, since a variable can (now) be assigned different
89 registers over its lifetimes.
91 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
92 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
100 #ifndef LEGACY_BACKEND // This file is ONLY used for the RyuJIT backend that uses the linear scan register allocator
105 const char* LinearScan::resolveTypeName[] = {"Split", "Join", "Critical", "SharedCritical"};
108 /*XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
109 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
111 XX Small Helper functions XX
114 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
115 XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
118 //--------------------------------------------------------------
119 // lsraAssignRegToTree: Assign the given reg to tree node.
122 // tree - Gentree node
123 // reg - register to be assigned
124 // regIdx - register idx, if tree is a multi-reg call node.
125 // regIdx will be zero for single-reg result producing tree nodes.
130 void lsraAssignRegToTree(GenTreePtr tree, regNumber reg, unsigned regIdx)
134 tree->gtRegNum = reg;
136 #if defined(_TARGET_ARM_)
137 else if (tree->OperGet() == GT_MUL_LONG || tree->OperGet() == GT_PUTARG_REG)
140 GenTreeMultiRegOp* mul = tree->AsMultiRegOp();
141 mul->gtOtherReg = reg;
143 else if (tree->OperGet() == GT_COPY)
146 GenTreeCopyOrReload* copy = tree->AsCopyOrReload();
147 copy->gtOtherRegs[0] = (regNumberSmall)reg;
149 else if (tree->OperGet() == GT_PUTARG_SPLIT)
151 GenTreePutArgSplit* putArg = tree->AsPutArgSplit();
152 putArg->SetRegNumByIdx(reg, regIdx);
154 #endif // _TARGET_ARM_
157 assert(tree->IsMultiRegCall());
158 GenTreeCall* call = tree->AsCall();
159 call->SetRegNumByIdx(reg, regIdx);
163 //-------------------------------------------------------------
164 // getWeight: Returns the weight of the RefPosition.
167 // refPos - ref position
170 // Weight of ref position.
171 unsigned LinearScan::getWeight(RefPosition* refPos)
174 GenTreePtr treeNode = refPos->treeNode;
176 if (treeNode != nullptr)
178 if (isCandidateLocalRef(treeNode))
180 // Tracked locals: use weighted ref cnt as the weight of the
182 GenTreeLclVarCommon* lclCommon = treeNode->AsLclVarCommon();
183 LclVarDsc* varDsc = &(compiler->lvaTable[lclCommon->gtLclNum]);
184 weight = varDsc->lvRefCntWtd;
188 // Non-candidate local ref or non-lcl tree node.
189 // These are considered to have two references in the basic block:
190 // a def and a use and hence weighted ref count is 2 times
191 // the basic block weight in which they appear.
192 weight = 2 * this->blockInfo[refPos->bbNum].weight;
197 // Non-tree node ref positions. These will have a single
198 // reference in the basic block and hence their weighted
199 // refcount is equal to the block weight in which they
201 weight = this->blockInfo[refPos->bbNum].weight;
207 // allRegs represents a set of registers that can
208 // be used to allocate the specified type in any point
209 // in time (more of a 'bank' of registers).
210 regMaskTP LinearScan::allRegs(RegisterType rt)
214 return availableFloatRegs;
216 else if (rt == TYP_DOUBLE)
218 return availableDoubleRegs;
220 // TODO-Cleanup: Add an RBM_ALLSIMD
222 else if (varTypeIsSIMD(rt))
224 return availableDoubleRegs;
225 #endif // FEATURE_SIMD
229 return availableIntRegs;
233 //--------------------------------------------------------------------------
234 // allMultiRegCallNodeRegs: represents a set of registers that can be used
235 // to allocate a multi-reg call node.
238 // call - Multi-reg call node
241 // Mask representing the set of available registers for multi-reg call
245 // Multi-reg call node available regs = Bitwise-OR(allregs(GetReturnRegType(i)))
246 // for all i=0..RetRegCount-1.
247 regMaskTP LinearScan::allMultiRegCallNodeRegs(GenTreeCall* call)
249 assert(call->HasMultiRegRetVal());
251 ReturnTypeDesc* retTypeDesc = call->GetReturnTypeDesc();
252 regMaskTP resultMask = allRegs(retTypeDesc->GetReturnRegType(0));
254 unsigned count = retTypeDesc->GetReturnRegCount();
255 for (unsigned i = 1; i < count; ++i)
257 resultMask |= allRegs(retTypeDesc->GetReturnRegType(i));
263 //--------------------------------------------------------------------------
264 // allRegs: returns the set of registers that can accomodate the type of
268 // tree - GenTree node
271 // Mask representing the set of available registers for given tree
273 // Note: In case of multi-reg call node, the full set of registers must be
274 // determined by looking at types of individual return register types.
275 // In this case, the registers may include registers from different register
276 // sets and will not be limited to the actual ABI return registers.
277 regMaskTP LinearScan::allRegs(GenTree* tree)
279 regMaskTP resultMask;
281 // In case of multi-reg calls, allRegs is defined as
282 // Bitwise-Or(allRegs(GetReturnRegType(i)) for i=0..ReturnRegCount-1
283 if (tree->IsMultiRegCall())
285 resultMask = allMultiRegCallNodeRegs(tree->AsCall());
289 resultMask = allRegs(tree->TypeGet());
295 regMaskTP LinearScan::allSIMDRegs()
297 return availableFloatRegs;
300 //------------------------------------------------------------------------
301 // internalFloatRegCandidates: Return the set of registers that are appropriate
302 // for use as internal float registers.
305 // The set of registers (as a regMaskTP).
308 // compFloatingPointUsed is only required to be set if it is possible that we
309 // will use floating point callee-save registers.
310 // It is unlikely, if an internal register is the only use of floating point,
311 // that it will select a callee-save register. But to be safe, we restrict
312 // the set of candidates if compFloatingPointUsed is not already set.
314 regMaskTP LinearScan::internalFloatRegCandidates()
316 if (compiler->compFloatingPointUsed)
318 return allRegs(TYP_FLOAT);
322 return RBM_FLT_CALLEE_TRASH;
326 /*****************************************************************************
328 *****************************************************************************/
330 RegisterType regType(T type)
333 if (varTypeIsSIMD(type))
335 return FloatRegisterType;
337 #endif // FEATURE_SIMD
338 return varTypeIsFloating(TypeGet(type)) ? FloatRegisterType : IntRegisterType;
341 bool useFloatReg(var_types type)
343 return (regType(type) == FloatRegisterType);
346 bool registerTypesEquivalent(RegisterType a, RegisterType b)
348 return varTypeIsIntegralOrI(a) == varTypeIsIntegralOrI(b);
351 bool isSingleRegister(regMaskTP regMask)
353 return (regMask != RBM_NONE && genMaxOneBit(regMask));
356 /*****************************************************************************
357 * Inline functions for RegRecord
358 *****************************************************************************/
360 bool RegRecord::isFree()
362 return ((assignedInterval == nullptr || !assignedInterval->isActive) && !isBusyUntilNextKill);
365 /*****************************************************************************
366 * Inline functions for LinearScan
367 *****************************************************************************/
368 RegRecord* LinearScan::getRegisterRecord(regNumber regNum)
370 return &physRegs[regNum];
375 //----------------------------------------------------------------------------
376 // getConstrainedRegMask: Returns new regMask which is the intersection of
377 // regMaskActual and regMaskConstraint if the new regMask has at least
378 // minRegCount registers, otherwise returns regMaskActual.
381 // regMaskActual - regMask that needs to be constrained
382 // regMaskConstraint - regMask constraint that needs to be
383 // applied to regMaskActual
384 // minRegCount - Minimum number of regs that should be
385 // be present in new regMask.
388 // New regMask that has minRegCount registers after instersection.
389 // Otherwise returns regMaskActual.
390 regMaskTP LinearScan::getConstrainedRegMask(regMaskTP regMaskActual, regMaskTP regMaskConstraint, unsigned minRegCount)
392 regMaskTP newMask = regMaskActual & regMaskConstraint;
393 if (genCountBits(newMask) >= minRegCount)
398 return regMaskActual;
401 //------------------------------------------------------------------------
402 // stressLimitRegs: Given a set of registers, expressed as a register mask, reduce
403 // them based on the current stress options.
406 // mask - The current mask of register candidates for a node
409 // A possibly-modified mask, based on the value of COMPlus_JitStressRegs.
412 // This is the method used to implement the stress options that limit
413 // the set of registers considered for allocation.
415 regMaskTP LinearScan::stressLimitRegs(RefPosition* refPosition, regMaskTP mask)
417 if (getStressLimitRegs() != LSRA_LIMIT_NONE)
419 // The refPosition could be null, for example when called
420 // by getTempRegForResolution().
421 int minRegCount = (refPosition != nullptr) ? refPosition->minRegCandidateCount : 1;
423 switch (getStressLimitRegs())
425 case LSRA_LIMIT_CALLEE:
426 if (!compiler->opts.compDbgEnC)
428 mask = getConstrainedRegMask(mask, RBM_CALLEE_SAVED, minRegCount);
432 case LSRA_LIMIT_CALLER:
434 mask = getConstrainedRegMask(mask, RBM_CALLEE_TRASH, minRegCount);
438 case LSRA_LIMIT_SMALL_SET:
439 if ((mask & LsraLimitSmallIntSet) != RBM_NONE)
441 mask = getConstrainedRegMask(mask, LsraLimitSmallIntSet, minRegCount);
443 else if ((mask & LsraLimitSmallFPSet) != RBM_NONE)
445 mask = getConstrainedRegMask(mask, LsraLimitSmallFPSet, minRegCount);
453 if (refPosition != nullptr && refPosition->isFixedRegRef)
455 mask |= refPosition->registerAssignment;
463 // TODO-Cleanup: Consider adding an overload that takes a varDsc, and can appropriately
464 // set such fields as isStructField
466 Interval* LinearScan::newInterval(RegisterType theRegisterType)
468 intervals.emplace_back(theRegisterType, allRegs(theRegisterType));
469 Interval* newInt = &intervals.back();
472 newInt->intervalIndex = static_cast<unsigned>(intervals.size() - 1);
475 DBEXEC(VERBOSE, newInt->dump());
479 RefPosition* LinearScan::newRefPositionRaw(LsraLocation nodeLocation, GenTree* treeNode, RefType refType)
481 refPositions.emplace_back(curBBNum, nodeLocation, treeNode, refType);
482 RefPosition* newRP = &refPositions.back();
484 newRP->rpNum = static_cast<unsigned>(refPositions.size() - 1);
489 //------------------------------------------------------------------------
490 // resolveConflictingDefAndUse: Resolve the situation where we have conflicting def and use
491 // register requirements on a single-def, single-use interval.
494 // defRefPosition - The interval definition
495 // useRefPosition - The (sole) interval use
501 // The two RefPositions are for the same interval, which is a tree-temp.
504 // We require some special handling for the case where the use is a "delayRegFree" case of a fixedReg.
505 // In that case, if we change the registerAssignment on the useRefPosition, we will lose the fact that,
506 // even if we assign a different register (and rely on codegen to do the copy), that fixedReg also needs
507 // to remain busy until the Def register has been allocated. In that case, we don't allow Case 1 or Case 4
509 // Here are the cases we consider (in this order):
510 // 1. If The defRefPosition specifies a single register, and there are no conflicting
511 // FixedReg uses of it between the def and use, we use that register, and the code generator
512 // will insert the copy. Note that it cannot be in use because there is a FixedRegRef for the def.
513 // 2. If the useRefPosition specifies a single register, and it is not in use, and there are no
514 // conflicting FixedReg uses of it between the def and use, we use that register, and the code generator
515 // will insert the copy.
516 // 3. If the defRefPosition specifies a single register (but there are conflicts, as determined
517 // in 1.), and there are no conflicts with the useRefPosition register (if it's a single register),
518 /// we set the register requirements on the defRefPosition to the use registers, and the
519 // code generator will insert a copy on the def. We can't rely on the code generator to put a copy
520 // on the use if it has multiple possible candidates, as it won't know which one has been allocated.
521 // 4. If the useRefPosition specifies a single register, and there are no conflicts with the register
522 // on the defRefPosition, we leave the register requirements on the defRefPosition as-is, and set
523 // the useRefPosition to the def registers, for similar reasons to case #3.
524 // 5. If both the defRefPosition and the useRefPosition specify single registers, but both have conflicts,
525 // We set the candiates on defRefPosition to be all regs of the appropriate type, and since they are
526 // single registers, codegen can insert the copy.
527 // 6. Finally, if the RefPositions specify disjoint subsets of the registers (or the use is fixed but
528 // has a conflict), we must insert a copy. The copy will be inserted before the use if the
529 // use is not fixed (in the fixed case, the code generator will insert the use).
531 // TODO-CQ: We get bad register allocation in case #3 in the situation where no register is
532 // available for the lifetime. We end up allocating a register that must be spilled, and it probably
533 // won't be the register that is actually defined by the target instruction. So, we have to copy it
534 // and THEN spill it. In this case, we should be using the def requirement. But we need to change
535 // the interface to this method a bit to make that work (e.g. returning a candidate set to use, but
536 // leaving the registerAssignment as-is on the def, so that if we find that we need to spill anyway
537 // we can use the fixed-reg on the def.
540 void LinearScan::resolveConflictingDefAndUse(Interval* interval, RefPosition* defRefPosition)
542 assert(!interval->isLocalVar);
544 RefPosition* useRefPosition = defRefPosition->nextRefPosition;
545 regMaskTP defRegAssignment = defRefPosition->registerAssignment;
546 regMaskTP useRegAssignment = useRefPosition->registerAssignment;
547 RegRecord* defRegRecord = nullptr;
548 RegRecord* useRegRecord = nullptr;
549 regNumber defReg = REG_NA;
550 regNumber useReg = REG_NA;
551 bool defRegConflict = false;
552 bool useRegConflict = false;
554 // If the useRefPosition is a "delayRegFree", we can't change the registerAssignment
555 // on it, or we will fail to ensure that the fixedReg is busy at the time the target
556 // (of the node that uses this interval) is allocated.
557 bool canChangeUseAssignment = !useRefPosition->isFixedRegRef || !useRefPosition->delayRegFree;
559 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CONFLICT));
560 if (!canChangeUseAssignment)
562 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_FIXED_DELAY_USE));
564 if (defRefPosition->isFixedRegRef)
566 defReg = defRefPosition->assignedReg();
567 defRegRecord = getRegisterRecord(defReg);
568 if (canChangeUseAssignment)
570 RefPosition* currFixedRegRefPosition = defRegRecord->recentRefPosition;
571 assert(currFixedRegRefPosition != nullptr &&
572 currFixedRegRefPosition->nodeLocation == defRefPosition->nodeLocation);
574 if (currFixedRegRefPosition->nextRefPosition == nullptr ||
575 currFixedRegRefPosition->nextRefPosition->nodeLocation > useRefPosition->getRefEndLocation())
577 // This is case #1. Use the defRegAssignment
578 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE1));
579 useRefPosition->registerAssignment = defRegAssignment;
584 defRegConflict = true;
588 if (useRefPosition->isFixedRegRef)
590 useReg = useRefPosition->assignedReg();
591 useRegRecord = getRegisterRecord(useReg);
592 RefPosition* currFixedRegRefPosition = useRegRecord->recentRefPosition;
594 // We know that useRefPosition is a fixed use, so the nextRefPosition must not be null.
595 RefPosition* nextFixedRegRefPosition = useRegRecord->getNextRefPosition();
596 assert(nextFixedRegRefPosition != nullptr &&
597 nextFixedRegRefPosition->nodeLocation <= useRefPosition->nodeLocation);
599 // First, check to see if there are any conflicting FixedReg references between the def and use.
600 if (nextFixedRegRefPosition->nodeLocation == useRefPosition->nodeLocation)
602 // OK, no conflicting FixedReg references.
603 // Now, check to see whether it is currently in use.
604 if (useRegRecord->assignedInterval != nullptr)
606 RefPosition* possiblyConflictingRef = useRegRecord->assignedInterval->recentRefPosition;
607 LsraLocation possiblyConflictingRefLocation = possiblyConflictingRef->getRefEndLocation();
608 if (possiblyConflictingRefLocation >= defRefPosition->nodeLocation)
610 useRegConflict = true;
615 // This is case #2. Use the useRegAssignment
616 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE2));
617 defRefPosition->registerAssignment = useRegAssignment;
623 useRegConflict = true;
626 if (defRegRecord != nullptr && !useRegConflict)
629 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE3));
630 defRefPosition->registerAssignment = useRegAssignment;
633 if (useRegRecord != nullptr && !defRegConflict && canChangeUseAssignment)
636 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE4));
637 useRefPosition->registerAssignment = defRegAssignment;
640 if (defRegRecord != nullptr && useRegRecord != nullptr)
643 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE5));
644 RegisterType regType = interval->registerType;
645 assert((getRegisterType(interval, defRefPosition) == regType) &&
646 (getRegisterType(interval, useRefPosition) == regType));
647 regMaskTP candidates = allRegs(regType);
648 defRefPosition->registerAssignment = candidates;
651 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DEFUSE_CASE6));
655 //------------------------------------------------------------------------
656 // conflictingFixedRegReference: Determine whether the current RegRecord has a
657 // fixed register use that conflicts with 'refPosition'
660 // refPosition - The RefPosition of interest
663 // Returns true iff the given RefPosition is NOT a fixed use of this register,
665 // - there is a RefPosition on this RegRecord at the nodeLocation of the given RefPosition, or
666 // - the given RefPosition has a delayRegFree, and there is a RefPosition on this RegRecord at
667 // the nodeLocation just past the given RefPosition.
670 // 'refPosition is non-null.
672 bool RegRecord::conflictingFixedRegReference(RefPosition* refPosition)
674 // Is this a fixed reference of this register? If so, there is no conflict.
675 if (refPosition->isFixedRefOfRegMask(genRegMask(regNum)))
679 // Otherwise, check for conflicts.
680 // There is a conflict if:
681 // 1. There is a recent RefPosition on this RegRecord that is at this location,
682 // except in the case where it is a special "putarg" that is associated with this interval, OR
683 // 2. There is an upcoming RefPosition at this location, or at the next location
684 // if refPosition is a delayed use (i.e. must be kept live through the next/def location).
686 LsraLocation refLocation = refPosition->nodeLocation;
687 if (recentRefPosition != nullptr && recentRefPosition->refType != RefTypeKill &&
688 recentRefPosition->nodeLocation == refLocation &&
689 (!isBusyUntilNextKill || assignedInterval != refPosition->getInterval()))
693 LsraLocation nextPhysRefLocation = getNextRefLocation();
694 if (nextPhysRefLocation == refLocation || (refPosition->delayRegFree && nextPhysRefLocation == (refLocation + 1)))
701 void LinearScan::applyCalleeSaveHeuristics(RefPosition* rp)
703 #ifdef _TARGET_AMD64_
704 if (compiler->opts.compDbgEnC)
706 // We only use RSI and RDI for EnC code, so we don't want to favor callee-save regs.
709 #endif // _TARGET_AMD64_
711 Interval* theInterval = rp->getInterval();
714 regMaskTP calleeSaveMask = calleeSaveRegs(getRegisterType(theInterval, rp));
715 if (doReverseCallerCallee())
717 rp->registerAssignment =
718 getConstrainedRegMask(rp->registerAssignment, calleeSaveMask, rp->minRegCandidateCount);
723 // Set preferences so that this register set will be preferred for earlier refs
724 theInterval->updateRegisterPreferences(rp->registerAssignment);
728 void LinearScan::associateRefPosWithInterval(RefPosition* rp)
730 Referenceable* theReferent = rp->referent;
732 if (theReferent != nullptr)
734 // All RefPositions except the dummy ones at the beginning of blocks
736 if (rp->isIntervalRef())
738 Interval* theInterval = rp->getInterval();
740 applyCalleeSaveHeuristics(rp);
742 if (theInterval->isLocalVar)
744 if (RefTypeIsUse(rp->refType))
746 RefPosition* const prevRP = theInterval->recentRefPosition;
747 if ((prevRP != nullptr) && (prevRP->bbNum == rp->bbNum))
749 prevRP->lastUse = false;
753 rp->lastUse = (rp->refType != RefTypeExpUse) && (rp->refType != RefTypeParamDef) &&
754 (rp->refType != RefTypeZeroInit) && !extendLifetimes();
756 else if (rp->refType == RefTypeUse)
758 // Ensure that we have consistent def/use on SDSU temps.
759 // However, there are a couple of cases where this may over-constrain allocation:
760 // 1. In the case of a non-commutative rmw def (in which the rmw source must be delay-free), or
761 // 2. In the case where the defining node requires a temp distinct from the target (also a
763 // In those cases, if we propagate a single-register restriction from the consumer to the producer
764 // the delayed uses will not see a fixed reference in the PhysReg at that position, and may
765 // incorrectly allocate that register.
766 // TODO-CQ: This means that we may often require a copy at the use of this node's result.
767 // This case could be moved to BuildRefPositionsForNode, at the point where the def RefPosition is
768 // created, causing a RefTypeFixedRef to be added at that location. This, however, results in
769 // more PhysReg RefPositions (a throughput impact), and a large number of diffs that require
770 // further analysis to determine benefit.
772 RefPosition* prevRefPosition = theInterval->recentRefPosition;
773 assert(prevRefPosition != nullptr && theInterval->firstRefPosition == prevRefPosition);
774 // All defs must have a valid treeNode, but we check it below to be conservative.
775 assert(prevRefPosition->treeNode != nullptr);
776 regMaskTP prevAssignment = prevRefPosition->registerAssignment;
777 regMaskTP newAssignment = (prevAssignment & rp->registerAssignment);
778 if (newAssignment != RBM_NONE)
780 if (!isSingleRegister(newAssignment) ||
781 (!theInterval->hasNonCommutativeRMWDef && (prevRefPosition->treeNode != nullptr) &&
782 !prevRefPosition->treeNode->gtLsraInfo.isInternalRegDelayFree))
784 prevRefPosition->registerAssignment = newAssignment;
789 theInterval->hasConflictingDefUse = true;
796 RefPosition* prevRP = theReferent->recentRefPosition;
797 if (prevRP != nullptr)
799 prevRP->nextRefPosition = rp;
803 theReferent->firstRefPosition = rp;
805 theReferent->recentRefPosition = rp;
806 theReferent->lastRefPosition = rp;
810 assert((rp->refType == RefTypeBB) || (rp->refType == RefTypeKillGCRefs));
814 //---------------------------------------------------------------------------
815 // newRefPosition: allocate and initialize a new RefPosition.
818 // reg - reg number that identifies RegRecord to be associated
819 // with this RefPosition
820 // theLocation - LSRA location of RefPosition
821 // theRefType - RefPosition type
822 // theTreeNode - GenTree node for which this RefPosition is created
823 // mask - Set of valid registers for this RefPosition
824 // multiRegIdx - register position if this RefPosition corresponds to a
825 // multi-reg call node.
830 RefPosition* LinearScan::newRefPosition(
831 regNumber reg, LsraLocation theLocation, RefType theRefType, GenTree* theTreeNode, regMaskTP mask)
833 RefPosition* newRP = newRefPositionRaw(theLocation, theTreeNode, theRefType);
835 newRP->setReg(getRegisterRecord(reg));
836 newRP->registerAssignment = mask;
838 newRP->setMultiRegIdx(0);
839 newRP->setAllocateIfProfitable(false);
841 associateRefPosWithInterval(newRP);
843 DBEXEC(VERBOSE, newRP->dump());
847 //---------------------------------------------------------------------------
848 // newRefPosition: allocate and initialize a new RefPosition.
851 // theInterval - interval to which RefPosition is associated with.
852 // theLocation - LSRA location of RefPosition
853 // theRefType - RefPosition type
854 // theTreeNode - GenTree node for which this RefPosition is created
855 // mask - Set of valid registers for this RefPosition
856 // multiRegIdx - register position if this RefPosition corresponds to a
857 // multi-reg call node.
858 // minRegCount - Minimum number registers that needs to be ensured while
859 // constraining candidates for this ref position under
860 // LSRA stress. This is a DEBUG only arg.
865 RefPosition* LinearScan::newRefPosition(Interval* theInterval,
866 LsraLocation theLocation,
868 GenTree* theTreeNode,
870 unsigned multiRegIdx /* = 0 */
871 DEBUGARG(unsigned minRegCandidateCount /* = 1 */))
874 if (theInterval != nullptr && regType(theInterval->registerType) == FloatRegisterType)
876 // In the case we're using floating point registers we must make sure
877 // this flag was set previously in the compiler since this will mandate
878 // whether LSRA will take into consideration FP reg killsets.
879 assert(compiler->compFloatingPointUsed || ((mask & RBM_FLT_CALLEE_SAVED) == 0));
883 // If this reference is constrained to a single register (and it's not a dummy
884 // or Kill reftype already), add a RefTypeFixedReg at this location so that its
885 // availability can be more accurately determined
887 bool isFixedRegister = isSingleRegister(mask);
888 bool insertFixedRef = false;
891 // Insert a RefTypeFixedReg for any normal def or use (not ParamDef or BB)
892 if (theRefType == RefTypeUse || theRefType == RefTypeDef)
894 insertFixedRef = true;
900 regNumber physicalReg = genRegNumFromMask(mask);
901 RefPosition* pos = newRefPosition(physicalReg, theLocation, RefTypeFixedReg, nullptr, mask);
902 assert(theInterval != nullptr);
903 assert((allRegs(theInterval->registerType) & mask) != 0);
906 RefPosition* newRP = newRefPositionRaw(theLocation, theTreeNode, theRefType);
908 newRP->setInterval(theInterval);
911 newRP->isFixedRegRef = isFixedRegister;
913 #ifndef _TARGET_AMD64_
914 // We don't need this for AMD because the PInvoke method epilog code is explicit
915 // at register allocation time.
916 if (theInterval != nullptr && theInterval->isLocalVar && compiler->info.compCallUnmanaged &&
917 theInterval->varNum == compiler->genReturnLocal)
919 mask &= ~(RBM_PINVOKE_TCB | RBM_PINVOKE_FRAME);
920 noway_assert(mask != RBM_NONE);
922 #endif // !_TARGET_AMD64_
923 newRP->registerAssignment = mask;
925 newRP->setMultiRegIdx(multiRegIdx);
926 newRP->setAllocateIfProfitable(false);
929 newRP->minRegCandidateCount = minRegCandidateCount;
932 associateRefPosWithInterval(newRP);
934 DBEXEC(VERBOSE, newRP->dump());
938 /*****************************************************************************
939 * Inline functions for Interval
940 *****************************************************************************/
941 RefPosition* Referenceable::getNextRefPosition()
943 if (recentRefPosition == nullptr)
945 return firstRefPosition;
949 return recentRefPosition->nextRefPosition;
953 LsraLocation Referenceable::getNextRefLocation()
955 RefPosition* nextRefPosition = getNextRefPosition();
956 if (nextRefPosition == nullptr)
962 return nextRefPosition->nodeLocation;
966 // Iterate through all the registers of the given type
967 class RegisterIterator
969 friend class Registers;
972 RegisterIterator(RegisterType type) : regType(type)
974 if (useFloatReg(regType))
976 currentRegNum = REG_FP_FIRST;
980 currentRegNum = REG_INT_FIRST;
985 static RegisterIterator Begin(RegisterType regType)
987 return RegisterIterator(regType);
989 static RegisterIterator End(RegisterType regType)
991 RegisterIterator endIter = RegisterIterator(regType);
992 // This assumes only integer and floating point register types
993 // if we target a processor with additional register types,
994 // this would have to change
995 if (useFloatReg(regType))
997 // This just happens to work for both double & float
998 endIter.currentRegNum = REG_NEXT(REG_FP_LAST);
1002 endIter.currentRegNum = REG_NEXT(REG_INT_LAST);
1008 void operator++(int dummy) // int dummy is c++ for "this is postfix ++"
1010 currentRegNum = REG_NEXT(currentRegNum);
1012 if (regType == TYP_DOUBLE)
1013 currentRegNum = REG_NEXT(currentRegNum);
1016 void operator++() // prefix operator++
1018 currentRegNum = REG_NEXT(currentRegNum);
1020 if (regType == TYP_DOUBLE)
1021 currentRegNum = REG_NEXT(currentRegNum);
1024 regNumber operator*()
1026 return currentRegNum;
1028 bool operator!=(const RegisterIterator& other)
1030 return other.currentRegNum != currentRegNum;
1034 regNumber currentRegNum;
1035 RegisterType regType;
1041 friend class RegisterIterator;
1043 Registers(RegisterType t)
1047 RegisterIterator begin()
1049 return RegisterIterator::Begin(type);
1051 RegisterIterator end()
1053 return RegisterIterator::End(type);
1058 void LinearScan::dumpVarToRegMap(VarToRegMap map)
1060 bool anyPrinted = false;
1061 for (unsigned varIndex = 0; varIndex < compiler->lvaTrackedCount; varIndex++)
1063 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
1064 if (map[varIndex] != REG_STK)
1066 printf("V%02u=%s ", varNum, getRegName(map[varIndex]));
1077 void LinearScan::dumpInVarToRegMap(BasicBlock* block)
1079 printf("Var=Reg beg of BB%02u: ", block->bbNum);
1080 VarToRegMap map = getInVarToRegMap(block->bbNum);
1081 dumpVarToRegMap(map);
1084 void LinearScan::dumpOutVarToRegMap(BasicBlock* block)
1086 printf("Var=Reg end of BB%02u: ", block->bbNum);
1087 VarToRegMap map = getOutVarToRegMap(block->bbNum);
1088 dumpVarToRegMap(map);
1093 LinearScanInterface* getLinearScanAllocator(Compiler* comp)
1095 return new (comp, CMK_LSRA) LinearScan(comp);
1098 //------------------------------------------------------------------------
1105 // The constructor takes care of initializing the data structures that are used
1106 // during Lowering, including (in DEBUG) getting the stress environment variables,
1107 // as they may affect the block ordering.
1109 LinearScan::LinearScan(Compiler* theCompiler)
1110 : compiler(theCompiler)
1111 #if MEASURE_MEM_ALLOC
1112 , lsraIAllocator(nullptr)
1113 #endif // MEASURE_MEM_ALLOC
1114 , intervals(LinearScanMemoryAllocatorInterval(theCompiler))
1115 , refPositions(LinearScanMemoryAllocatorRefPosition(theCompiler))
1118 maxNodeLocation = 0;
1119 activeRefPosition = nullptr;
1121 // Get the value of the environment variable that controls stress for register allocation
1122 lsraStressMask = JitConfig.JitStressRegs();
1125 if (lsraStressMask != 0)
1127 // The code in this #if can be used to debug JitStressRegs issues according to
1128 // method hash. To use, simply set environment variables JitStressRegsHashLo and JitStressRegsHashHi
1129 unsigned methHash = compiler->info.compMethodHash();
1130 char* lostr = getenv("JitStressRegsHashLo");
1131 unsigned methHashLo = 0;
1133 if (lostr != nullptr)
1135 sscanf_s(lostr, "%x", &methHashLo);
1138 char* histr = getenv("JitStressRegsHashHi");
1139 unsigned methHashHi = UINT32_MAX;
1140 if (histr != nullptr)
1142 sscanf_s(histr, "%x", &methHashHi);
1145 if (methHash < methHashLo || methHash > methHashHi)
1149 else if (dump == true)
1151 printf("JitStressRegs = %x for method %s, hash = 0x%x.\n",
1152 lsraStressMask, compiler->info.compFullName, compiler->info.compMethodHash());
1153 printf(""); // in our logic this causes a flush
1159 dumpTerse = (JitConfig.JitDumpTerseLsra() != 0);
1162 enregisterLocalVars = ((compiler->opts.compFlags & CLFLG_REGVAR) != 0) && compiler->lvaTrackedCount > 0;
1163 availableIntRegs = (RBM_ALLINT & ~compiler->codeGen->regSet.rsMaskResvd);
1166 availableIntRegs &= ~RBM_FPBASE;
1167 #endif // ETW_EBP_FRAMED
1169 availableFloatRegs = RBM_ALLFLOAT;
1170 availableDoubleRegs = RBM_ALLDOUBLE;
1172 #ifdef _TARGET_AMD64_
1173 if (compiler->opts.compDbgEnC)
1175 // On x64 when the EnC option is set, we always save exactly RBP, RSI and RDI.
1176 // RBP is not available to the register allocator, so RSI and RDI are the only
1177 // callee-save registers available.
1178 availableIntRegs &= ~RBM_CALLEE_SAVED | RBM_RSI | RBM_RDI;
1179 availableFloatRegs &= ~RBM_CALLEE_SAVED;
1180 availableDoubleRegs &= ~RBM_CALLEE_SAVED;
1182 #endif // _TARGET_AMD64_
1183 compiler->rpFrameType = FT_NOT_SET;
1184 compiler->rpMustCreateEBPCalled = false;
1186 compiler->codeGen->intRegState.rsIsFloat = false;
1187 compiler->codeGen->floatRegState.rsIsFloat = true;
1189 // Block sequencing (the order in which we schedule).
1190 // Note that we don't initialize the bbVisitedSet until we do the first traversal
1191 // (currently during Lowering's second phase, where it sets the TreeNodeInfo).
1192 // This is so that any blocks that are added during the first phase of Lowering
1193 // are accounted for (and we don't have BasicBlockEpoch issues).
1194 blockSequencingDone = false;
1195 blockSequence = nullptr;
1196 blockSequenceWorkList = nullptr;
1200 // Information about each block, including predecessor blocks used for variable locations at block entry.
1201 blockInfo = nullptr;
1203 // Populate the register mask table.
1204 // The first two masks in the table are allint/allfloat
1205 // The next N are the masks for each single register.
1206 // After that are the dynamically added ones.
1207 regMaskTable = new (compiler, CMK_LSRA) regMaskTP[numMasks];
1208 regMaskTable[ALLINT_IDX] = allRegs(TYP_INT);
1209 regMaskTable[ALLFLOAT_IDX] = allRegs(TYP_DOUBLE);
1212 for (reg = REG_FIRST; reg < REG_COUNT; reg = REG_NEXT(reg))
1214 regMaskTable[FIRST_SINGLE_REG_IDX + reg - REG_FIRST] = (reg == REG_STK) ? RBM_NONE : genRegMask(reg);
1216 nextFreeMask = FIRST_SINGLE_REG_IDX + REG_COUNT;
1217 noway_assert(nextFreeMask <= numMasks);
1220 // Return the reg mask corresponding to the given index.
1221 regMaskTP LinearScan::GetRegMaskForIndex(RegMaskIndex index)
1223 assert(index < numMasks);
1224 assert(index < nextFreeMask);
1225 return regMaskTable[index];
1228 // Given a reg mask, return the index it corresponds to. If it is not a 'well known' reg mask,
1229 // add it at the end. This method has linear behavior in the worst cases but that is fairly rare.
1230 // Most methods never use any but the well-known masks, and when they do use more
1231 // it is only one or two more.
1232 LinearScan::RegMaskIndex LinearScan::GetIndexForRegMask(regMaskTP mask)
1234 RegMaskIndex result;
1235 if (isSingleRegister(mask))
1237 result = genRegNumFromMask(mask) + FIRST_SINGLE_REG_IDX;
1239 else if (mask == allRegs(TYP_INT))
1241 result = ALLINT_IDX;
1243 else if (mask == allRegs(TYP_DOUBLE))
1245 result = ALLFLOAT_IDX;
1249 for (int i = FIRST_SINGLE_REG_IDX + REG_COUNT; i < nextFreeMask; i++)
1251 if (regMaskTable[i] == mask)
1257 // We only allocate a fixed number of masks. Since we don't reallocate, we will throw a
1258 // noway_assert if we exceed this limit.
1259 noway_assert(nextFreeMask < numMasks);
1261 regMaskTable[nextFreeMask] = mask;
1262 result = nextFreeMask;
1265 assert(mask == regMaskTable[result]);
1269 // We've decided that we can't use a register during register allocation (probably FPBASE),
1270 // but we've already added it to the register masks. Go through the masks and remove it.
1271 void LinearScan::RemoveRegisterFromMasks(regNumber reg)
1273 JITDUMP("Removing register %s from LSRA register masks\n", getRegName(reg));
1275 regMaskTP mask = ~genRegMask(reg);
1276 for (int i = 0; i < nextFreeMask; i++)
1278 regMaskTable[i] &= mask;
1281 JITDUMP("After removing register:\n");
1282 DBEXEC(VERBOSE, dspRegisterMaskTable());
1286 void LinearScan::dspRegisterMaskTable()
1288 printf("LSRA register masks. Total allocated: %d, total used: %d\n", numMasks, nextFreeMask);
1289 for (int i = 0; i < nextFreeMask; i++)
1292 dspRegMask(regMaskTable[i]);
1298 //------------------------------------------------------------------------
1299 // getNextCandidateFromWorkList: Get the next candidate for block sequencing
1305 // The next block to be placed in the sequence.
1308 // This method currently always returns the next block in the list, and relies on having
1309 // blocks added to the list only when they are "ready", and on the
1310 // addToBlockSequenceWorkList() method to insert them in the proper order.
1311 // However, a block may be in the list and already selected, if it was subsequently
1312 // encountered as both a flow and layout successor of the most recently selected
1315 BasicBlock* LinearScan::getNextCandidateFromWorkList()
1317 BasicBlockList* nextWorkList = nullptr;
1318 for (BasicBlockList* workList = blockSequenceWorkList; workList != nullptr; workList = nextWorkList)
1320 nextWorkList = workList->next;
1321 BasicBlock* candBlock = workList->block;
1322 removeFromBlockSequenceWorkList(workList, nullptr);
1323 if (!isBlockVisited(candBlock))
1331 //------------------------------------------------------------------------
1332 // setBlockSequence:Determine the block order for register allocation.
1341 // On return, the blockSequence array contains the blocks, in the order in which they
1342 // will be allocated.
1343 // This method clears the bbVisitedSet on LinearScan, and when it returns the set
1344 // contains all the bbNums for the block.
1345 // This requires a traversal of the BasicBlocks, and could potentially be
1346 // combined with the first traversal (currently the one in Lowering that sets the
1349 void LinearScan::setBlockSequence()
1351 // Reset the "visited" flag on each block.
1352 compiler->EnsureBasicBlockEpoch();
1353 bbVisitedSet = BlockSetOps::MakeEmpty(compiler);
1354 BlockSet readySet(BlockSetOps::MakeEmpty(compiler));
1355 BlockSet predSet(BlockSetOps::MakeEmpty(compiler));
1357 assert(blockSequence == nullptr && bbSeqCount == 0);
1358 blockSequence = new (compiler, CMK_LSRA) BasicBlock*[compiler->fgBBcount];
1359 bbNumMaxBeforeResolution = compiler->fgBBNumMax;
1360 blockInfo = new (compiler, CMK_LSRA) LsraBlockInfo[bbNumMaxBeforeResolution + 1];
1362 assert(blockSequenceWorkList == nullptr);
1364 bool addedInternalBlocks = false;
1365 verifiedAllBBs = false;
1366 hasCriticalEdges = false;
1367 BasicBlock* nextBlock;
1368 for (BasicBlock* block = compiler->fgFirstBB; block != nullptr; block = nextBlock)
1370 blockSequence[bbSeqCount] = block;
1371 markBlockVisited(block);
1373 nextBlock = nullptr;
1375 // Initialize the blockInfo.
1376 // predBBNum will be set later. 0 is never used as a bbNum.
1377 blockInfo[block->bbNum].predBBNum = 0;
1378 // We check for critical edges below, but initialize to false.
1379 blockInfo[block->bbNum].hasCriticalInEdge = false;
1380 blockInfo[block->bbNum].hasCriticalOutEdge = false;
1381 blockInfo[block->bbNum].weight = block->bbWeight;
1383 #if TRACK_LSRA_STATS
1384 blockInfo[block->bbNum].spillCount = 0;
1385 blockInfo[block->bbNum].copyRegCount = 0;
1386 blockInfo[block->bbNum].resolutionMovCount = 0;
1387 blockInfo[block->bbNum].splitEdgeCount = 0;
1388 #endif // TRACK_LSRA_STATS
1390 if (block->GetUniquePred(compiler) == nullptr)
1392 for (flowList* pred = block->bbPreds; pred != nullptr; pred = pred->flNext)
1394 BasicBlock* predBlock = pred->flBlock;
1395 if (predBlock->NumSucc(compiler) > 1)
1397 blockInfo[block->bbNum].hasCriticalInEdge = true;
1398 hasCriticalEdges = true;
1401 else if (predBlock->bbJumpKind == BBJ_SWITCH)
1403 assert(!"Switch with single successor");
1408 // Determine which block to schedule next.
1410 // First, update the NORMAL successors of the current block, adding them to the worklist
1411 // according to the desired order. We will handle the EH successors below.
1412 bool checkForCriticalOutEdge = (block->NumSucc(compiler) > 1);
1413 if (!checkForCriticalOutEdge && block->bbJumpKind == BBJ_SWITCH)
1415 assert(!"Switch with single successor");
1418 const unsigned numSuccs = block->NumSucc(compiler);
1419 for (unsigned succIndex = 0; succIndex < numSuccs; succIndex++)
1421 BasicBlock* succ = block->GetSucc(succIndex, compiler);
1422 if (checkForCriticalOutEdge && succ->GetUniquePred(compiler) == nullptr)
1424 blockInfo[block->bbNum].hasCriticalOutEdge = true;
1425 hasCriticalEdges = true;
1426 // We can stop checking now.
1427 checkForCriticalOutEdge = false;
1430 if (isTraversalLayoutOrder() || isBlockVisited(succ))
1435 // We've now seen a predecessor, so add it to the work list and the "readySet".
1436 // It will be inserted in the worklist according to the specified traversal order
1437 // (i.e. pred-first or random, since layout order is handled above).
1438 if (!BlockSetOps::IsMember(compiler, readySet, succ->bbNum))
1440 addToBlockSequenceWorkList(readySet, succ, predSet);
1441 BlockSetOps::AddElemD(compiler, readySet, succ->bbNum);
1445 // For layout order, simply use bbNext
1446 if (isTraversalLayoutOrder())
1448 nextBlock = block->bbNext;
1452 while (nextBlock == nullptr)
1454 nextBlock = getNextCandidateFromWorkList();
1456 // TODO-Throughput: We would like to bypass this traversal if we know we've handled all
1457 // the blocks - but fgBBcount does not appear to be updated when blocks are removed.
1458 if (nextBlock == nullptr /* && bbSeqCount != compiler->fgBBcount*/ && !verifiedAllBBs)
1460 // If we don't encounter all blocks by traversing the regular sucessor links, do a full
1461 // traversal of all the blocks, and add them in layout order.
1462 // This may include:
1463 // - internal-only blocks (in the fgAddCodeList) which may not be in the flow graph
1464 // (these are not even in the bbNext links).
1465 // - blocks that have become unreachable due to optimizations, but that are strongly
1466 // connected (these are not removed)
1469 for (Compiler::AddCodeDsc* desc = compiler->fgAddCodeList; desc != nullptr; desc = desc->acdNext)
1471 if (!isBlockVisited(block))
1473 addToBlockSequenceWorkList(readySet, block, predSet);
1474 BlockSetOps::AddElemD(compiler, readySet, block->bbNum);
1478 for (BasicBlock* block = compiler->fgFirstBB; block; block = block->bbNext)
1480 if (!isBlockVisited(block))
1482 addToBlockSequenceWorkList(readySet, block, predSet);
1483 BlockSetOps::AddElemD(compiler, readySet, block->bbNum);
1486 verifiedAllBBs = true;
1494 blockSequencingDone = true;
1497 // Make sure that we've visited all the blocks.
1498 for (BasicBlock* block = compiler->fgFirstBB; block != nullptr; block = block->bbNext)
1500 assert(isBlockVisited(block));
1503 JITDUMP("LSRA Block Sequence: ");
1505 for (BasicBlock *block = startBlockSequence(); block != nullptr; ++i, block = moveToNextBlock())
1507 JITDUMP("BB%02u", block->bbNum);
1509 if (block->isMaxBBWeight())
1515 JITDUMP("(%6s) ", refCntWtd2str(block->getBBWeight(compiler)));
1527 //------------------------------------------------------------------------
1528 // compareBlocksForSequencing: Compare two basic blocks for sequencing order.
1531 // block1 - the first block for comparison
1532 // block2 - the second block for comparison
1533 // useBlockWeights - whether to use block weights for comparison
1536 // -1 if block1 is preferred.
1537 // 0 if the blocks are equivalent.
1538 // 1 if block2 is preferred.
1541 // See addToBlockSequenceWorkList.
1542 int LinearScan::compareBlocksForSequencing(BasicBlock* block1, BasicBlock* block2, bool useBlockWeights)
1544 if (useBlockWeights)
1546 unsigned weight1 = block1->getBBWeight(compiler);
1547 unsigned weight2 = block2->getBBWeight(compiler);
1549 if (weight1 > weight2)
1553 else if (weight1 < weight2)
1559 // If weights are the same prefer LOWER bbnum
1560 if (block1->bbNum < block2->bbNum)
1564 else if (block1->bbNum == block2->bbNum)
1574 //------------------------------------------------------------------------
1575 // addToBlockSequenceWorkList: Add a BasicBlock to the work list for sequencing.
1578 // sequencedBlockSet - the set of blocks that are already sequenced
1579 // block - the new block to be added
1580 // predSet - the buffer to save predecessors set. A block set allocated by the caller used here as a
1581 // temporary block set for constructing a predecessor set. Allocated by the caller to avoid reallocating a new block
1582 // set with every call to this function
1588 // The first block in the list will be the next one to be sequenced, as soon
1589 // as we encounter a block whose successors have all been sequenced, in pred-first
1590 // order, or the very next block if we are traversing in random order (once implemented).
1591 // This method uses a comparison method to determine the order in which to place
1592 // the blocks in the list. This method queries whether all predecessors of the
1593 // block are sequenced at the time it is added to the list and if so uses block weights
1594 // for inserting the block. A block is never inserted ahead of its predecessors.
1595 // A block at the time of insertion may not have all its predecessors sequenced, in
1596 // which case it will be sequenced based on its block number. Once a block is inserted,
1597 // its priority\order will not be changed later once its remaining predecessors are
1598 // sequenced. This would mean that work list may not be sorted entirely based on
1599 // block weights alone.
1601 // Note also that, when random traversal order is implemented, this method
1602 // should insert the blocks into the list in random order, so that we can always
1603 // simply select the first block in the list.
1604 void LinearScan::addToBlockSequenceWorkList(BlockSet sequencedBlockSet, BasicBlock* block, BlockSet& predSet)
1606 // The block that is being added is not already sequenced
1607 assert(!BlockSetOps::IsMember(compiler, sequencedBlockSet, block->bbNum));
1609 // Get predSet of block
1610 BlockSetOps::ClearD(compiler, predSet);
1612 for (pred = block->bbPreds; pred != nullptr; pred = pred->flNext)
1614 BlockSetOps::AddElemD(compiler, predSet, pred->flBlock->bbNum);
1617 // If either a rarely run block or all its preds are already sequenced, use block's weight to sequence
1618 bool useBlockWeight = block->isRunRarely() || BlockSetOps::IsSubset(compiler, sequencedBlockSet, predSet);
1620 BasicBlockList* prevNode = nullptr;
1621 BasicBlockList* nextNode = blockSequenceWorkList;
1623 while (nextNode != nullptr)
1627 if (nextNode->block->isRunRarely())
1629 // If the block that is yet to be sequenced is a rarely run block, always use block weights for sequencing
1630 seqResult = compareBlocksForSequencing(nextNode->block, block, true);
1632 else if (BlockSetOps::IsMember(compiler, predSet, nextNode->block->bbNum))
1634 // always prefer unsequenced pred blocks
1639 seqResult = compareBlocksForSequencing(nextNode->block, block, useBlockWeight);
1647 prevNode = nextNode;
1648 nextNode = nextNode->next;
1651 BasicBlockList* newListNode = new (compiler, CMK_LSRA) BasicBlockList(block, nextNode);
1652 if (prevNode == nullptr)
1654 blockSequenceWorkList = newListNode;
1658 prevNode->next = newListNode;
1662 void LinearScan::removeFromBlockSequenceWorkList(BasicBlockList* listNode, BasicBlockList* prevNode)
1664 if (listNode == blockSequenceWorkList)
1666 assert(prevNode == nullptr);
1667 blockSequenceWorkList = listNode->next;
1671 assert(prevNode != nullptr && prevNode->next == listNode);
1672 prevNode->next = listNode->next;
1674 // TODO-Cleanup: consider merging Compiler::BlockListNode and BasicBlockList
1675 // compiler->FreeBlockListNode(listNode);
1678 // Initialize the block order for allocation (called each time a new traversal begins).
1679 BasicBlock* LinearScan::startBlockSequence()
1681 if (!blockSequencingDone)
1685 BasicBlock* curBB = compiler->fgFirstBB;
1687 curBBNum = curBB->bbNum;
1688 clearVisitedBlocks();
1689 assert(blockSequence[0] == compiler->fgFirstBB);
1690 markBlockVisited(curBB);
1694 //------------------------------------------------------------------------
1695 // moveToNextBlock: Move to the next block in order for allocation or resolution.
1704 // This method is used when the next block is actually going to be handled.
1705 // It changes curBBNum.
1707 BasicBlock* LinearScan::moveToNextBlock()
1709 BasicBlock* nextBlock = getNextBlock();
1711 if (nextBlock != nullptr)
1713 curBBNum = nextBlock->bbNum;
1718 //------------------------------------------------------------------------
1719 // getNextBlock: Get the next block in order for allocation or resolution.
1728 // This method does not actually change the current block - it is used simply
1729 // to determine which block will be next.
1731 BasicBlock* LinearScan::getNextBlock()
1733 assert(blockSequencingDone);
1734 unsigned int nextBBSeqNum = curBBSeqNum + 1;
1735 if (nextBBSeqNum < bbSeqCount)
1737 return blockSequence[nextBBSeqNum];
1742 //------------------------------------------------------------------------
1743 // doLinearScan: The main method for register allocation.
1752 // Lowering must have set the NodeInfo (gtLsraInfo) on each node to communicate
1753 // the register requirements.
1755 void LinearScan::doLinearScan()
1757 unsigned lsraBlockEpoch = compiler->GetCurBasicBlockEpoch();
1759 splitBBNumToTargetBBNumMap = nullptr;
1761 // This is complicated by the fact that physical registers have refs associated
1762 // with locations where they are killed (e.g. calls), but we don't want to
1763 // count these as being touched.
1765 compiler->codeGen->regSet.rsClearRegsModified();
1769 DBEXEC(VERBOSE, TupleStyleDump(LSRA_DUMP_REFPOS));
1770 compiler->EndPhase(PHASE_LINEAR_SCAN_BUILD);
1772 DBEXEC(VERBOSE, lsraDumpIntervals("after buildIntervals"));
1774 clearVisitedBlocks();
1776 allocateRegisters();
1777 compiler->EndPhase(PHASE_LINEAR_SCAN_ALLOC);
1779 compiler->EndPhase(PHASE_LINEAR_SCAN_RESOLVE);
1781 #if TRACK_LSRA_STATS
1782 if ((JitConfig.DisplayLsraStats() != 0)
1788 dumpLsraStats(jitstdout);
1790 #endif // TRACK_LSRA_STATS
1792 DBEXEC(VERBOSE, TupleStyleDump(LSRA_DUMP_POST));
1794 compiler->compLSRADone = true;
1795 noway_assert(lsraBlockEpoch = compiler->GetCurBasicBlockEpoch());
1798 //------------------------------------------------------------------------
1799 // recordVarLocationsAtStartOfBB: Update live-in LclVarDscs with the appropriate
1800 // register location at the start of a block, during codegen.
1803 // bb - the block for which code is about to be generated.
1809 // CodeGen will take care of updating the reg masks and the current var liveness,
1810 // after calling this method.
1811 // This is because we need to kill off the dead registers before setting the newly live ones.
1813 void LinearScan::recordVarLocationsAtStartOfBB(BasicBlock* bb)
1815 if (!enregisterLocalVars)
1819 JITDUMP("Recording Var Locations at start of BB%02u\n", bb->bbNum);
1820 VarToRegMap map = getInVarToRegMap(bb->bbNum);
1823 VarSetOps::AssignNoCopy(compiler, currentLiveVars,
1824 VarSetOps::Intersection(compiler, registerCandidateVars, bb->bbLiveIn));
1825 VarSetOps::Iter iter(compiler, currentLiveVars);
1826 unsigned varIndex = 0;
1827 while (iter.NextElem(&varIndex))
1829 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
1830 LclVarDsc* varDsc = &(compiler->lvaTable[varNum]);
1831 regNumber regNum = getVarReg(map, varIndex);
1833 regNumber oldRegNum = varDsc->lvRegNum;
1834 regNumber newRegNum = regNum;
1836 if (oldRegNum != newRegNum)
1838 JITDUMP(" V%02u(%s->%s)", varNum, compiler->compRegVarName(oldRegNum),
1839 compiler->compRegVarName(newRegNum));
1840 varDsc->lvRegNum = newRegNum;
1843 else if (newRegNum != REG_STK)
1845 JITDUMP(" V%02u(%s)", varNum, compiler->compRegVarName(newRegNum));
1852 JITDUMP(" <none>\n");
1858 void Interval::setLocalNumber(Compiler* compiler, unsigned lclNum, LinearScan* linScan)
1860 LclVarDsc* varDsc = &compiler->lvaTable[lclNum];
1861 assert(varDsc->lvTracked);
1862 assert(varDsc->lvVarIndex < compiler->lvaTrackedCount);
1864 linScan->localVarIntervals[varDsc->lvVarIndex] = this;
1866 assert(linScan->getIntervalForLocalVar(varDsc->lvVarIndex) == this);
1867 this->isLocalVar = true;
1868 this->varNum = lclNum;
1871 // identify the candidates which we are not going to enregister due to
1872 // being used in EH in a way we don't want to deal with
1873 // this logic cloned from fgInterBlockLocalVarLiveness
1874 void LinearScan::identifyCandidatesExceptionDataflow()
1876 VARSET_TP exceptVars(VarSetOps::MakeEmpty(compiler));
1877 VARSET_TP filterVars(VarSetOps::MakeEmpty(compiler));
1878 VARSET_TP finallyVars(VarSetOps::MakeEmpty(compiler));
1881 foreach_block(compiler, block)
1883 if (block->bbCatchTyp != BBCT_NONE)
1885 // live on entry to handler
1886 VarSetOps::UnionD(compiler, exceptVars, block->bbLiveIn);
1889 if (block->bbJumpKind == BBJ_EHFILTERRET)
1891 // live on exit from filter
1892 VarSetOps::UnionD(compiler, filterVars, block->bbLiveOut);
1894 else if (block->bbJumpKind == BBJ_EHFINALLYRET)
1896 // live on exit from finally
1897 VarSetOps::UnionD(compiler, finallyVars, block->bbLiveOut);
1899 #if FEATURE_EH_FUNCLETS
1900 // Funclets are called and returned from, as such we can only count on the frame
1901 // pointer being restored, and thus everything live in or live out must be on the
1903 if (block->bbFlags & BBF_FUNCLET_BEG)
1905 VarSetOps::UnionD(compiler, exceptVars, block->bbLiveIn);
1907 if ((block->bbJumpKind == BBJ_EHFINALLYRET) || (block->bbJumpKind == BBJ_EHFILTERRET) ||
1908 (block->bbJumpKind == BBJ_EHCATCHRET))
1910 VarSetOps::UnionD(compiler, exceptVars, block->bbLiveOut);
1912 #endif // FEATURE_EH_FUNCLETS
1915 // slam them all together (there was really no need to use more than 2 bitvectors here)
1916 VarSetOps::UnionD(compiler, exceptVars, filterVars);
1917 VarSetOps::UnionD(compiler, exceptVars, finallyVars);
1919 /* Mark all pointer variables live on exit from a 'finally'
1920 block as either volatile for non-GC ref types or as
1921 'explicitly initialized' (volatile and must-init) for GC-ref types */
1923 VarSetOps::Iter iter(compiler, exceptVars);
1924 unsigned varIndex = 0;
1925 while (iter.NextElem(&varIndex))
1927 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
1928 LclVarDsc* varDsc = compiler->lvaTable + varNum;
1930 compiler->lvaSetVarDoNotEnregister(varNum DEBUGARG(Compiler::DNER_LiveInOutOfHandler));
1932 if (varTypeIsGC(varDsc))
1934 if (VarSetOps::IsMember(compiler, finallyVars, varIndex) && !varDsc->lvIsParam)
1936 varDsc->lvMustInit = true;
1942 bool LinearScan::isRegCandidate(LclVarDsc* varDsc)
1944 // We shouldn't be called if opt settings do not permit register variables.
1945 assert((compiler->opts.compFlags & CLFLG_REGVAR) != 0);
1947 if (!varDsc->lvTracked)
1952 #if !defined(_TARGET_64BIT_)
1953 if (varDsc->lvType == TYP_LONG)
1955 // Long variables should not be register candidates.
1956 // Lowering will have split any candidate lclVars into lo/hi vars.
1959 #endif // !defined(_TARGET_64BIT)
1961 // If we have JMP, reg args must be put on the stack
1963 if (compiler->compJmpOpUsed && varDsc->lvIsRegArg)
1968 // Don't allocate registers for dependently promoted struct fields
1969 if (compiler->lvaIsFieldOfDependentlyPromotedStruct(varDsc))
1976 // Identify locals & compiler temps that are register candidates
1977 // TODO-Cleanup: This was cloned from Compiler::lvaSortByRefCount() in lclvars.cpp in order
1978 // to avoid perturbation, but should be merged.
1980 void LinearScan::identifyCandidates()
1982 if (enregisterLocalVars)
1984 // Initialize the set of lclVars that are candidates for register allocation.
1985 VarSetOps::AssignNoCopy(compiler, registerCandidateVars, VarSetOps::MakeEmpty(compiler));
1987 // Initialize the sets of lclVars that are used to determine whether, and for which lclVars,
1988 // we need to perform resolution across basic blocks.
1989 // Note that we can't do this in the constructor because the number of tracked lclVars may
1990 // change between the constructor and the actual allocation.
1991 VarSetOps::AssignNoCopy(compiler, resolutionCandidateVars, VarSetOps::MakeEmpty(compiler));
1992 VarSetOps::AssignNoCopy(compiler, splitOrSpilledVars, VarSetOps::MakeEmpty(compiler));
1994 // We set enregisterLocalVars to true only if there are tracked lclVars
1995 assert(compiler->lvaCount != 0);
1997 else if (compiler->lvaCount == 0)
1999 // Nothing to do. Note that even if enregisterLocalVars is false, we still need to set the
2000 // lvLRACandidate field on all the lclVars to false if we have any.
2004 if (compiler->compHndBBtabCount > 0)
2006 identifyCandidatesExceptionDataflow();
2012 // While we build intervals for the candidate lclVars, we will determine the floating point
2013 // lclVars, if any, to consider for callee-save register preferencing.
2014 // We maintain two sets of FP vars - those that meet the first threshold of weighted ref Count,
2015 // and those that meet the second.
2016 // The first threshold is used for methods that are heuristically deemed either to have light
2017 // fp usage, or other factors that encourage conservative use of callee-save registers, such
2018 // as multiple exits (where there might be an early exit that woudl be excessively penalized by
2019 // lots of prolog/epilog saves & restores).
2020 // The second threshold is used where there are factors deemed to make it more likely that fp
2021 // fp callee save registers will be needed, such as loops or many fp vars.
2022 // We keep two sets of vars, since we collect some of the information to determine which set to
2023 // use as we iterate over the vars.
2024 // When we are generating AVX code on non-Unix (FEATURE_PARTIAL_SIMD_CALLEE_SAVE), we maintain an
2025 // additional set of LargeVectorType vars, and there is a separate threshold defined for those.
2026 // It is assumed that if we encounter these, that we should consider this a "high use" scenario,
2027 // so we don't maintain two sets of these vars.
2028 // This is defined as thresholdLargeVectorRefCntWtd, as we are likely to use the same mechanism
2029 // for vectors on Arm64, though the actual value may differ.
2031 unsigned int floatVarCount = 0;
2032 unsigned int thresholdFPRefCntWtd = 4 * BB_UNITY_WEIGHT;
2033 unsigned int maybeFPRefCntWtd = 2 * BB_UNITY_WEIGHT;
2034 VARSET_TP fpMaybeCandidateVars(VarSetOps::UninitVal());
2035 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2036 unsigned int largeVectorVarCount = 0;
2037 unsigned int thresholdLargeVectorRefCntWtd = 4 * BB_UNITY_WEIGHT;
2038 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2039 if (enregisterLocalVars)
2041 VarSetOps::AssignNoCopy(compiler, fpCalleeSaveCandidateVars, VarSetOps::MakeEmpty(compiler));
2042 VarSetOps::AssignNoCopy(compiler, fpMaybeCandidateVars, VarSetOps::MakeEmpty(compiler));
2043 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2044 VarSetOps::AssignNoCopy(compiler, largeVectorVars, VarSetOps::MakeEmpty(compiler));
2045 VarSetOps::AssignNoCopy(compiler, largeVectorCalleeSaveCandidateVars, VarSetOps::MakeEmpty(compiler));
2046 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2049 unsigned refCntStk = 0;
2050 unsigned refCntReg = 0;
2051 unsigned refCntWtdReg = 0;
2052 unsigned refCntStkParam = 0; // sum of ref counts for all stack based parameters
2053 unsigned refCntWtdStkDbl = 0; // sum of wtd ref counts for stack based doubles
2054 doDoubleAlign = false;
2055 bool checkDoubleAlign = true;
2056 if (compiler->codeGen->isFramePointerRequired() || compiler->opts.MinOpts())
2058 checkDoubleAlign = false;
2062 switch (compiler->getCanDoubleAlign())
2064 case MUST_DOUBLE_ALIGN:
2065 doDoubleAlign = true;
2066 checkDoubleAlign = false;
2068 case CAN_DOUBLE_ALIGN:
2070 case CANT_DOUBLE_ALIGN:
2071 doDoubleAlign = false;
2072 checkDoubleAlign = false;
2078 #endif // DOUBLE_ALIGN
2080 // Check whether register variables are permitted.
2081 if (!enregisterLocalVars)
2083 localVarIntervals = nullptr;
2085 else if (compiler->lvaTrackedCount > 0)
2087 // initialize mapping from tracked local to interval
2088 localVarIntervals = new (compiler, CMK_LSRA) Interval*[compiler->lvaTrackedCount];
2091 INTRACK_STATS(regCandidateVarCount = 0);
2092 for (lclNum = 0, varDsc = compiler->lvaTable; lclNum < compiler->lvaCount; lclNum++, varDsc++)
2094 // Initialize all variables to REG_STK
2095 varDsc->lvRegNum = REG_STK;
2096 #ifndef _TARGET_64BIT_
2097 varDsc->lvOtherReg = REG_STK;
2098 #endif // _TARGET_64BIT_
2100 if (!enregisterLocalVars)
2102 varDsc->lvLRACandidate = false;
2107 if (checkDoubleAlign)
2109 if (varDsc->lvIsParam && !varDsc->lvIsRegArg)
2111 refCntStkParam += varDsc->lvRefCnt;
2113 else if (!isRegCandidate(varDsc) || varDsc->lvDoNotEnregister)
2115 refCntStk += varDsc->lvRefCnt;
2116 if ((varDsc->lvType == TYP_DOUBLE) ||
2117 ((varTypeIsStruct(varDsc) && varDsc->lvStructDoubleAlign &&
2118 (compiler->lvaGetPromotionType(varDsc) != Compiler::PROMOTION_TYPE_INDEPENDENT))))
2120 refCntWtdStkDbl += varDsc->lvRefCntWtd;
2125 refCntReg += varDsc->lvRefCnt;
2126 refCntWtdReg += varDsc->lvRefCntWtd;
2129 #endif // DOUBLE_ALIGN
2131 /* Track all locals that can be enregistered */
2133 if (!isRegCandidate(varDsc))
2135 varDsc->lvLRACandidate = 0;
2136 if (varDsc->lvTracked)
2138 localVarIntervals[varDsc->lvVarIndex] = nullptr;
2143 assert(varDsc->lvTracked);
2145 varDsc->lvLRACandidate = 1;
2147 // Start with lvRegister as false - set it true only if the variable gets
2148 // the same register assignment throughout
2149 varDsc->lvRegister = false;
2151 /* If the ref count is zero */
2152 if (varDsc->lvRefCnt == 0)
2154 /* Zero ref count, make this untracked */
2155 varDsc->lvRefCntWtd = 0;
2156 varDsc->lvLRACandidate = 0;
2159 // Variables that are address-exposed are never enregistered, or tracked.
2160 // A struct may be promoted, and a struct that fits in a register may be fully enregistered.
2161 // Pinned variables may not be tracked (a condition of the GCInfo representation)
2162 // or enregistered, on x86 -- it is believed that we can enregister pinned (more properly, "pinning")
2163 // references when using the general GC encoding.
2165 if (varDsc->lvAddrExposed || !varTypeIsEnregisterableStruct(varDsc))
2167 varDsc->lvLRACandidate = 0;
2169 Compiler::DoNotEnregisterReason dner = Compiler::DNER_AddrExposed;
2170 if (!varDsc->lvAddrExposed)
2172 dner = Compiler::DNER_IsStruct;
2175 compiler->lvaSetVarDoNotEnregister(lclNum DEBUGARG(dner));
2177 else if (varDsc->lvPinned)
2179 varDsc->lvTracked = 0;
2180 #ifdef JIT32_GCENCODER
2181 compiler->lvaSetVarDoNotEnregister(lclNum DEBUGARG(Compiler::DNER_PinningRef));
2182 #endif // JIT32_GCENCODER
2185 // Are we not optimizing and we have exception handlers?
2186 // if so mark all args and locals as volatile, so that they
2187 // won't ever get enregistered.
2189 if (compiler->opts.MinOpts() && compiler->compHndBBtabCount > 0)
2191 compiler->lvaSetVarDoNotEnregister(lclNum DEBUGARG(Compiler::DNER_LiveInOutOfHandler));
2194 if (varDsc->lvDoNotEnregister)
2196 varDsc->lvLRACandidate = 0;
2197 localVarIntervals[varDsc->lvVarIndex] = nullptr;
2201 var_types type = genActualType(varDsc->TypeGet());
2205 #if CPU_HAS_FP_SUPPORT
2208 if (compiler->opts.compDbgCode)
2210 varDsc->lvLRACandidate = 0;
2213 if (varDsc->lvIsParam && varDsc->lvIsRegArg)
2215 type = (type == TYP_DOUBLE) ? TYP_LONG : TYP_INT;
2217 #endif // ARM_SOFTFP
2219 #endif // CPU_HAS_FP_SUPPORT
2231 if (varDsc->lvPromoted)
2233 varDsc->lvLRACandidate = 0;
2237 // TODO-1stClassStructs: Move TYP_SIMD8 up with the other SIMD types, after handling the param issue
2238 // (passing & returning as TYP_LONG).
2240 #endif // FEATURE_SIMD
2244 varDsc->lvLRACandidate = 0;
2250 noway_assert(!"lvType not set correctly");
2251 varDsc->lvType = TYP_INT;
2256 varDsc->lvLRACandidate = 0;
2259 if (varDsc->lvLRACandidate)
2261 Interval* newInt = newInterval(type);
2262 newInt->setLocalNumber(compiler, lclNum, this);
2263 VarSetOps::AddElemD(compiler, registerCandidateVars, varDsc->lvVarIndex);
2265 // we will set this later when we have determined liveness
2266 varDsc->lvMustInit = false;
2268 if (varDsc->lvIsStructField)
2270 newInt->isStructField = true;
2273 INTRACK_STATS(regCandidateVarCount++);
2275 // We maintain two sets of FP vars - those that meet the first threshold of weighted ref Count,
2276 // and those that meet the second (see the definitions of thresholdFPRefCntWtd and maybeFPRefCntWtd
2278 CLANG_FORMAT_COMMENT_ANCHOR;
2280 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2281 // Additionally, when we are generating AVX on non-UNIX amd64, we keep a separate set of the LargeVectorType
2283 if (varDsc->lvType == LargeVectorType)
2285 largeVectorVarCount++;
2286 VarSetOps::AddElemD(compiler, largeVectorVars, varDsc->lvVarIndex);
2287 unsigned refCntWtd = varDsc->lvRefCntWtd;
2288 if (refCntWtd >= thresholdLargeVectorRefCntWtd)
2290 VarSetOps::AddElemD(compiler, largeVectorCalleeSaveCandidateVars, varDsc->lvVarIndex);
2294 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2295 if (regType(type) == FloatRegisterType)
2298 unsigned refCntWtd = varDsc->lvRefCntWtd;
2299 if (varDsc->lvIsRegArg)
2301 // Don't count the initial reference for register params. In those cases,
2302 // using a callee-save causes an extra copy.
2303 refCntWtd -= BB_UNITY_WEIGHT;
2305 if (refCntWtd >= thresholdFPRefCntWtd)
2307 VarSetOps::AddElemD(compiler, fpCalleeSaveCandidateVars, varDsc->lvVarIndex);
2309 else if (refCntWtd >= maybeFPRefCntWtd)
2311 VarSetOps::AddElemD(compiler, fpMaybeCandidateVars, varDsc->lvVarIndex);
2317 localVarIntervals[varDsc->lvVarIndex] = nullptr;
2322 if (checkDoubleAlign)
2324 // TODO-CQ: Fine-tune this:
2325 // In the legacy reg predictor, this runs after allocation, and then demotes any lclVars
2326 // allocated to the frame pointer, which is probably the wrong order.
2327 // However, because it runs after allocation, it can determine the impact of demoting
2328 // the lclVars allocated to the frame pointer.
2329 // => Here, estimate of the EBP refCnt and weighted refCnt is a wild guess.
2331 unsigned refCntEBP = refCntReg / 8;
2332 unsigned refCntWtdEBP = refCntWtdReg / 8;
2335 compiler->shouldDoubleAlign(refCntStk, refCntEBP, refCntWtdEBP, refCntStkParam, refCntWtdStkDbl);
2337 #endif // DOUBLE_ALIGN
2339 // The factors we consider to determine which set of fp vars to use as candidates for callee save
2340 // registers current include the number of fp vars, whether there are loops, and whether there are
2341 // multiple exits. These have been selected somewhat empirically, but there is probably room for
2343 CLANG_FORMAT_COMMENT_ANCHOR;
2348 printf("\nFP callee save candidate vars: ");
2349 if (enregisterLocalVars && !VarSetOps::IsEmpty(compiler, fpCalleeSaveCandidateVars))
2351 dumpConvertedVarSet(compiler, fpCalleeSaveCandidateVars);
2361 JITDUMP("floatVarCount = %d; hasLoops = %d, singleExit = %d\n", floatVarCount, compiler->fgHasLoops,
2362 (compiler->fgReturnBlocks == nullptr || compiler->fgReturnBlocks->next == nullptr));
2364 // Determine whether to use the 2nd, more aggressive, threshold for fp callee saves.
2365 if (floatVarCount > 6 && compiler->fgHasLoops &&
2366 (compiler->fgReturnBlocks == nullptr || compiler->fgReturnBlocks->next == nullptr))
2368 assert(enregisterLocalVars);
2372 printf("Adding additional fp callee save candidates: \n");
2373 if (!VarSetOps::IsEmpty(compiler, fpMaybeCandidateVars))
2375 dumpConvertedVarSet(compiler, fpMaybeCandidateVars);
2384 VarSetOps::UnionD(compiler, fpCalleeSaveCandidateVars, fpMaybeCandidateVars);
2391 // Frame layout is only pre-computed for ARM
2392 printf("\nlvaTable after IdentifyCandidates\n");
2393 compiler->lvaTableDump();
2396 #endif // _TARGET_ARM_
2399 // TODO-Throughput: This mapping can surely be more efficiently done
2400 void LinearScan::initVarRegMaps()
2402 if (!enregisterLocalVars)
2404 inVarToRegMaps = nullptr;
2405 outVarToRegMaps = nullptr;
2408 assert(compiler->lvaTrackedFixed); // We should have already set this to prevent us from adding any new tracked
2411 // The compiler memory allocator requires that the allocation be an
2412 // even multiple of int-sized objects
2413 unsigned int varCount = compiler->lvaTrackedCount;
2414 regMapCount = (unsigned int)roundUp(varCount, sizeof(int));
2416 // Not sure why blocks aren't numbered from zero, but they don't appear to be.
2417 // So, if we want to index by bbNum we have to know the maximum value.
2418 unsigned int bbCount = compiler->fgBBNumMax + 1;
2420 inVarToRegMaps = new (compiler, CMK_LSRA) regNumberSmall*[bbCount];
2421 outVarToRegMaps = new (compiler, CMK_LSRA) regNumberSmall*[bbCount];
2425 // This VarToRegMap is used during the resolution of critical edges.
2426 sharedCriticalVarToRegMap = new (compiler, CMK_LSRA) regNumberSmall[regMapCount];
2428 for (unsigned int i = 0; i < bbCount; i++)
2430 VarToRegMap inVarToRegMap = new (compiler, CMK_LSRA) regNumberSmall[regMapCount];
2431 VarToRegMap outVarToRegMap = new (compiler, CMK_LSRA) regNumberSmall[regMapCount];
2433 for (unsigned int j = 0; j < regMapCount; j++)
2435 inVarToRegMap[j] = REG_STK;
2436 outVarToRegMap[j] = REG_STK;
2438 inVarToRegMaps[i] = inVarToRegMap;
2439 outVarToRegMaps[i] = outVarToRegMap;
2444 sharedCriticalVarToRegMap = nullptr;
2445 for (unsigned int i = 0; i < bbCount; i++)
2447 inVarToRegMaps[i] = nullptr;
2448 outVarToRegMaps[i] = nullptr;
2453 void LinearScan::setInVarRegForBB(unsigned int bbNum, unsigned int varNum, regNumber reg)
2455 assert(enregisterLocalVars);
2456 assert(reg < UCHAR_MAX && varNum < compiler->lvaCount);
2457 inVarToRegMaps[bbNum][compiler->lvaTable[varNum].lvVarIndex] = (regNumberSmall)reg;
2460 void LinearScan::setOutVarRegForBB(unsigned int bbNum, unsigned int varNum, regNumber reg)
2462 assert(enregisterLocalVars);
2463 assert(reg < UCHAR_MAX && varNum < compiler->lvaCount);
2464 outVarToRegMaps[bbNum][compiler->lvaTable[varNum].lvVarIndex] = (regNumberSmall)reg;
2467 LinearScan::SplitEdgeInfo LinearScan::getSplitEdgeInfo(unsigned int bbNum)
2469 assert(enregisterLocalVars);
2470 SplitEdgeInfo splitEdgeInfo;
2471 assert(bbNum <= compiler->fgBBNumMax);
2472 assert(bbNum > bbNumMaxBeforeResolution);
2473 assert(splitBBNumToTargetBBNumMap != nullptr);
2474 splitBBNumToTargetBBNumMap->Lookup(bbNum, &splitEdgeInfo);
2475 assert(splitEdgeInfo.toBBNum <= bbNumMaxBeforeResolution);
2476 assert(splitEdgeInfo.fromBBNum <= bbNumMaxBeforeResolution);
2477 return splitEdgeInfo;
2480 VarToRegMap LinearScan::getInVarToRegMap(unsigned int bbNum)
2482 assert(enregisterLocalVars);
2483 assert(bbNum <= compiler->fgBBNumMax);
2484 // For the blocks inserted to split critical edges, the inVarToRegMap is
2485 // equal to the outVarToRegMap at the "from" block.
2486 if (bbNum > bbNumMaxBeforeResolution)
2488 SplitEdgeInfo splitEdgeInfo = getSplitEdgeInfo(bbNum);
2489 unsigned fromBBNum = splitEdgeInfo.fromBBNum;
2492 assert(splitEdgeInfo.toBBNum != 0);
2493 return inVarToRegMaps[splitEdgeInfo.toBBNum];
2497 return outVarToRegMaps[fromBBNum];
2501 return inVarToRegMaps[bbNum];
2504 VarToRegMap LinearScan::getOutVarToRegMap(unsigned int bbNum)
2506 assert(enregisterLocalVars);
2507 assert(bbNum <= compiler->fgBBNumMax);
2508 // For the blocks inserted to split critical edges, the outVarToRegMap is
2509 // equal to the inVarToRegMap at the target.
2510 if (bbNum > bbNumMaxBeforeResolution)
2512 // If this is an empty block, its in and out maps are both the same.
2513 // We identify this case by setting fromBBNum or toBBNum to 0, and using only the other.
2514 SplitEdgeInfo splitEdgeInfo = getSplitEdgeInfo(bbNum);
2515 unsigned toBBNum = splitEdgeInfo.toBBNum;
2518 assert(splitEdgeInfo.fromBBNum != 0);
2519 return outVarToRegMaps[splitEdgeInfo.fromBBNum];
2523 return inVarToRegMaps[toBBNum];
2526 return outVarToRegMaps[bbNum];
2529 //------------------------------------------------------------------------
2530 // setVarReg: Set the register associated with a variable in the given 'bbVarToRegMap'.
2533 // bbVarToRegMap - the map of interest
2534 // trackedVarIndex - the lvVarIndex for the variable
2535 // reg - the register to which it is being mapped
2540 void LinearScan::setVarReg(VarToRegMap bbVarToRegMap, unsigned int trackedVarIndex, regNumber reg)
2542 assert(trackedVarIndex < compiler->lvaTrackedCount);
2543 regNumberSmall regSmall = (regNumberSmall)reg;
2544 assert((regNumber)regSmall == reg);
2545 bbVarToRegMap[trackedVarIndex] = regSmall;
2548 //------------------------------------------------------------------------
2549 // getVarReg: Get the register associated with a variable in the given 'bbVarToRegMap'.
2552 // bbVarToRegMap - the map of interest
2553 // trackedVarIndex - the lvVarIndex for the variable
2556 // The register to which 'trackedVarIndex' is mapped
2558 regNumber LinearScan::getVarReg(VarToRegMap bbVarToRegMap, unsigned int trackedVarIndex)
2560 assert(enregisterLocalVars);
2561 assert(trackedVarIndex < compiler->lvaTrackedCount);
2562 return (regNumber)bbVarToRegMap[trackedVarIndex];
2565 // Initialize the incoming VarToRegMap to the given map values (generally a predecessor of
2567 VarToRegMap LinearScan::setInVarToRegMap(unsigned int bbNum, VarToRegMap srcVarToRegMap)
2569 assert(enregisterLocalVars);
2570 VarToRegMap inVarToRegMap = inVarToRegMaps[bbNum];
2571 memcpy(inVarToRegMap, srcVarToRegMap, (regMapCount * sizeof(regNumber)));
2572 return inVarToRegMap;
2575 // given a tree node
2576 RefType refTypeForLocalRefNode(GenTree* node)
2578 assert(node->IsLocal());
2580 // We don't support updates
2581 assert((node->gtFlags & GTF_VAR_USEASG) == 0);
2583 if (node->gtFlags & GTF_VAR_DEF)
2593 //------------------------------------------------------------------------
2594 // checkLastUses: Check correctness of last use flags
2597 // The block for which we are checking last uses.
2600 // This does a backward walk of the RefPositions, starting from the liveOut set.
2601 // This method was previously used to set the last uses, which were computed by
2602 // liveness, but were not create in some cases of multiple lclVar references in the
2603 // same tree. However, now that last uses are computed as RefPositions are created,
2604 // that is no longer necessary, and this method is simply retained as a check.
2605 // The exception to the check-only behavior is when LSRA_EXTEND_LIFETIMES if set via
2606 // COMPlus_JitStressRegs. In that case, this method is required, because even though
2607 // the RefPositions will not be marked lastUse in that case, we still need to correclty
2608 // mark the last uses on the tree nodes, which is done by this method.
2611 void LinearScan::checkLastUses(BasicBlock* block)
2615 JITDUMP("\n\nCHECKING LAST USES for block %u, liveout=", block->bbNum);
2616 dumpConvertedVarSet(compiler, block->bbLiveOut);
2617 JITDUMP("\n==============================\n");
2620 unsigned keepAliveVarNum = BAD_VAR_NUM;
2621 if (compiler->lvaKeepAliveAndReportThis())
2623 keepAliveVarNum = compiler->info.compThisArg;
2624 assert(compiler->info.compIsStatic == false);
2627 // find which uses are lastUses
2629 // Work backwards starting with live out.
2630 // 'computedLive' is updated to include any exposed use (including those in this
2631 // block that we've already seen). When we encounter a use, if it's
2632 // not in that set, then it's a last use.
2634 VARSET_TP computedLive(VarSetOps::MakeCopy(compiler, block->bbLiveOut));
2636 bool foundDiff = false;
2637 auto currentRefPosition = refPositions.rbegin();
2638 while (currentRefPosition->refType != RefTypeBB)
2640 // We should never see ParamDefs or ZeroInits within a basic block.
2641 assert(currentRefPosition->refType != RefTypeParamDef && currentRefPosition->refType != RefTypeZeroInit);
2642 if (currentRefPosition->isIntervalRef() && currentRefPosition->getInterval()->isLocalVar)
2644 unsigned varNum = currentRefPosition->getInterval()->varNum;
2645 unsigned varIndex = currentRefPosition->getInterval()->getVarIndex(compiler);
2647 LsraLocation loc = currentRefPosition->nodeLocation;
2649 // We should always have a tree node for a localVar, except for the "special" RefPositions.
2650 GenTreePtr tree = currentRefPosition->treeNode;
2651 assert(tree != nullptr || currentRefPosition->refType == RefTypeExpUse ||
2652 currentRefPosition->refType == RefTypeDummyDef);
2654 if (!VarSetOps::IsMember(compiler, computedLive, varIndex) && varNum != keepAliveVarNum)
2656 // There was no exposed use, so this is a "last use" (and we mark it thus even if it's a def)
2658 if (extendLifetimes())
2660 // NOTE: this is a bit of a hack. When extending lifetimes, the "last use" bit will be clear.
2661 // This bit, however, would normally be used during resolveLocalRef to set the value of
2662 // GTF_VAR_DEATH on the node for a ref position. If this bit is not set correctly even when
2663 // extending lifetimes, the code generator will assert as it expects to have accurate last
2664 // use information. To avoid these asserts, set the GTF_VAR_DEATH bit here.
2665 // Note also that extendLifetimes() is an LSRA stress mode, so it will only be true for
2666 // Checked or Debug builds, for which this method will be executed.
2667 if (tree != nullptr)
2669 tree->gtFlags |= GTF_VAR_DEATH;
2672 else if (!currentRefPosition->lastUse)
2674 JITDUMP("missing expected last use of V%02u @%u\n", compiler->lvaTrackedToVarNum[varIndex], loc);
2677 VarSetOps::AddElemD(compiler, computedLive, varIndex);
2679 else if (currentRefPosition->lastUse)
2681 JITDUMP("unexpected last use of V%02u @%u\n", compiler->lvaTrackedToVarNum[varIndex], loc);
2684 else if (extendLifetimes() && tree != nullptr)
2686 // NOTE: see the comment above re: the extendLifetimes hack.
2687 tree->gtFlags &= ~GTF_VAR_DEATH;
2690 if (currentRefPosition->refType == RefTypeDef || currentRefPosition->refType == RefTypeDummyDef)
2692 VarSetOps::RemoveElemD(compiler, computedLive, varIndex);
2696 assert(currentRefPosition != refPositions.rend());
2697 ++currentRefPosition;
2700 VARSET_TP liveInNotComputedLive(VarSetOps::Diff(compiler, block->bbLiveIn, computedLive));
2702 VarSetOps::Iter liveInNotComputedLiveIter(compiler, liveInNotComputedLive);
2703 unsigned liveInNotComputedLiveIndex = 0;
2704 while (liveInNotComputedLiveIter.NextElem(&liveInNotComputedLiveIndex))
2706 unsigned varNum = compiler->lvaTrackedToVarNum[liveInNotComputedLiveIndex];
2707 if (compiler->lvaTable[varNum].lvLRACandidate)
2709 JITDUMP("BB%02u: V%02u is in LiveIn set, but not computed live.\n", block->bbNum, varNum);
2714 VarSetOps::DiffD(compiler, computedLive, block->bbLiveIn);
2715 const VARSET_TP& computedLiveNotLiveIn(computedLive); // reuse the buffer.
2716 VarSetOps::Iter computedLiveNotLiveInIter(compiler, computedLiveNotLiveIn);
2717 unsigned computedLiveNotLiveInIndex = 0;
2718 while (computedLiveNotLiveInIter.NextElem(&computedLiveNotLiveInIndex))
2720 unsigned varNum = compiler->lvaTrackedToVarNum[computedLiveNotLiveInIndex];
2721 if (compiler->lvaTable[varNum].lvLRACandidate)
2723 JITDUMP("BB%02u: V%02u is computed live, but not in LiveIn set.\n", block->bbNum, varNum);
2732 void LinearScan::addRefsForPhysRegMask(regMaskTP mask, LsraLocation currentLoc, RefType refType, bool isLastUse)
2734 for (regNumber reg = REG_FIRST; mask; reg = REG_NEXT(reg), mask >>= 1)
2738 // This assumes that these are all "special" RefTypes that
2739 // don't need to be recorded on the tree (hence treeNode is nullptr)
2740 RefPosition* pos = newRefPosition(reg, currentLoc, refType, nullptr,
2741 genRegMask(reg)); // This MUST occupy the physical register (obviously)
2745 pos->lastUse = true;
2751 //------------------------------------------------------------------------
2752 // getKillSetForNode: Return the registers killed by the given tree node.
2755 // compiler - the compiler context to use
2756 // tree - the tree for which the kill set is needed.
2758 // Return Value: a register mask of the registers killed
2760 regMaskTP LinearScan::getKillSetForNode(GenTree* tree)
2762 regMaskTP killMask = RBM_NONE;
2763 switch (tree->OperGet())
2765 #ifdef _TARGET_XARCH_
2767 // We use the 128-bit multiply when performing an overflow checking unsigned multiply
2769 if (((tree->gtFlags & GTF_UNSIGNED) != 0) && tree->gtOverflowEx())
2771 // Both RAX and RDX are killed by the operation
2772 killMask = RBM_RAX | RBM_RDX;
2777 #if defined(_TARGET_X86_) && !defined(LEGACY_BACKEND)
2780 killMask = RBM_RAX | RBM_RDX;
2787 if (!varTypeIsFloating(tree->TypeGet()))
2789 // RDX needs to be killed early, because it must not be used as a source register
2790 // (unlike most cases, where the kill happens AFTER the uses). So for this kill,
2791 // we add the RefPosition at the tree loc (where the uses are located) instead of the
2792 // usual kill location which is the same as the defs at tree loc+1.
2793 // Note that we don't have to add interference for the live vars, because that
2794 // will be done below, and is not sensitive to the precise location.
2795 LsraLocation currentLoc = tree->gtLsraInfo.loc;
2796 assert(currentLoc != 0);
2797 addRefsForPhysRegMask(RBM_RDX, currentLoc, RefTypeKill, true);
2798 // Both RAX and RDX are killed by the operation
2799 killMask = RBM_RAX | RBM_RDX;
2802 #endif // _TARGET_XARCH_
2805 if (tree->OperIsCopyBlkOp())
2807 assert(tree->AsObj()->gtGcPtrCount != 0);
2808 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_ASSIGN_BYREF);
2814 case GT_STORE_DYN_BLK:
2816 GenTreeBlk* blkNode = tree->AsBlk();
2817 bool isCopyBlk = varTypeIsStruct(blkNode->Data());
2818 switch (blkNode->gtBlkOpKind)
2820 case GenTreeBlk::BlkOpKindHelper:
2823 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_MEMCPY);
2827 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_MEMSET);
2831 #ifdef _TARGET_XARCH_
2832 case GenTreeBlk::BlkOpKindRepInstr:
2835 // rep movs kills RCX, RDI and RSI
2836 killMask = RBM_RCX | RBM_RDI | RBM_RSI;
2840 // rep stos kills RCX and RDI.
2841 // (Note that the Data() node, if not constant, will be assigned to
2842 // RCX, but it's find that this kills it, as the value is not available
2843 // after this node in any case.)
2844 killMask = RBM_RDI | RBM_RCX;
2848 case GenTreeBlk::BlkOpKindRepInstr:
2850 case GenTreeBlk::BlkOpKindUnroll:
2851 case GenTreeBlk::BlkOpKindInvalid:
2852 // for these 'gtBlkOpKind' kinds, we leave 'killMask' = RBM_NONE
2859 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_STOP_FOR_GC);
2863 if (compiler->compFloatingPointUsed)
2865 if (tree->TypeGet() == TYP_DOUBLE)
2867 needDoubleTmpForFPCall = true;
2869 else if (tree->TypeGet() == TYP_FLOAT)
2871 needFloatTmpForFPCall = true;
2874 #endif // _TARGET_X86_
2875 #if defined(_TARGET_X86_) || defined(_TARGET_ARM_)
2876 if (tree->IsHelperCall())
2878 GenTreeCall* call = tree->AsCall();
2879 CorInfoHelpFunc helpFunc = compiler->eeGetHelperNum(call->gtCallMethHnd);
2880 killMask = compiler->compHelperCallKillSet(helpFunc);
2883 #endif // defined(_TARGET_X86_) || defined(_TARGET_ARM_)
2885 // if there is no FP used, we can ignore the FP kills
2886 if (compiler->compFloatingPointUsed)
2888 killMask = RBM_CALLEE_TRASH;
2892 killMask = RBM_INT_CALLEE_TRASH;
2895 if (tree->AsCall()->IsVirtualStub())
2897 killMask |= compiler->virtualStubParamInfo->GetRegMask();
2899 #else // !_TARGET_ARM_
2900 // Verify that the special virtual stub call registers are in the kill mask.
2901 // We don't just add them unconditionally to the killMask because for most architectures
2902 // they are already in the RBM_CALLEE_TRASH set,
2903 // and we don't want to introduce extra checks and calls in this hot function.
2904 assert(!tree->AsCall()->IsVirtualStub() || ((killMask & compiler->virtualStubParamInfo->GetRegMask()) ==
2905 compiler->virtualStubParamInfo->GetRegMask()));
2910 if (compiler->codeGen->gcInfo.gcIsWriteBarrierAsgNode(tree))
2912 killMask = RBM_CALLEE_TRASH_NOGC;
2916 #if defined(PROFILING_SUPPORTED)
2917 // If this method requires profiler ELT hook then mark these nodes as killing
2918 // callee trash registers (excluding RAX and XMM0). The reason for this is that
2919 // profiler callback would trash these registers. See vm\amd64\asmhelpers.asm for
2922 if (compiler->compIsProfilerHookNeeded())
2924 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_PROF_FCN_LEAVE);
2929 if (compiler->compIsProfilerHookNeeded())
2931 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_PROF_FCN_TAILCALL);
2934 #endif // PROFILING_SUPPORTED
2937 // for all other 'tree->OperGet()' kinds, leave 'killMask' = RBM_NONE
2943 //------------------------------------------------------------------------
2944 // buildKillPositionsForNode:
2945 // Given some tree node add refpositions for all the registers this node kills
2948 // tree - the tree for which kill positions should be generated
2949 // currentLoc - the location at which the kills should be added
2952 // true - kills were inserted
2953 // false - no kills were inserted
2956 // The return value is needed because if we have any kills, we need to make sure that
2957 // all defs are located AFTER the kills. On the other hand, if there aren't kills,
2958 // the multiple defs for a regPair are in different locations.
2959 // If we generate any kills, we will mark all currentLiveVars as being preferenced
2960 // to avoid the killed registers. This is somewhat conservative.
2962 bool LinearScan::buildKillPositionsForNode(GenTree* tree, LsraLocation currentLoc)
2964 regMaskTP killMask = getKillSetForNode(tree);
2965 bool isCallKill = ((killMask == RBM_INT_CALLEE_TRASH) || (killMask == RBM_CALLEE_TRASH));
2966 if (killMask != RBM_NONE)
2968 // The killMask identifies a set of registers that will be used during codegen.
2969 // Mark these as modified here, so when we do final frame layout, we'll know about
2970 // all these registers. This is especially important if killMask contains
2971 // callee-saved registers, which affect the frame size since we need to save/restore them.
2972 // In the case where we have a copyBlk with GC pointers, can need to call the
2973 // CORINFO_HELP_ASSIGN_BYREF helper, which kills callee-saved RSI and RDI, if
2974 // LSRA doesn't assign RSI/RDI, they wouldn't get marked as modified until codegen,
2975 // which is too late.
2976 compiler->codeGen->regSet.rsSetRegsModified(killMask DEBUGARG(dumpTerse));
2978 addRefsForPhysRegMask(killMask, currentLoc, RefTypeKill, true);
2980 // TODO-CQ: It appears to be valuable for both fp and int registers to avoid killing the callee
2981 // save regs on infrequently exectued paths. However, it results in a large number of asmDiffs,
2982 // many of which appear to be regressions (because there is more spill on the infrequently path),
2983 // but are not really because the frequent path becomes smaller. Validating these diffs will need
2984 // to be done before making this change.
2985 // if (!blockSequence[curBBSeqNum]->isRunRarely())
2986 if (enregisterLocalVars)
2988 VarSetOps::Iter iter(compiler, currentLiveVars);
2989 unsigned varIndex = 0;
2990 while (iter.NextElem(&varIndex))
2992 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
2993 LclVarDsc* varDsc = compiler->lvaTable + varNum;
2994 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
2995 if (varDsc->lvType == LargeVectorType)
2997 if (!VarSetOps::IsMember(compiler, largeVectorCalleeSaveCandidateVars, varIndex))
3003 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
3004 if (varTypeIsFloating(varDsc) &&
3005 !VarSetOps::IsMember(compiler, fpCalleeSaveCandidateVars, varIndex))
3009 Interval* interval = getIntervalForLocalVar(varIndex);
3012 interval->preferCalleeSave = true;
3014 regMaskTP newPreferences = allRegs(interval->registerType) & (~killMask);
3016 if (newPreferences != RBM_NONE)
3018 interval->updateRegisterPreferences(newPreferences);
3022 // If there are no callee-saved registers, the call could kill all the registers.
3023 // This is a valid state, so in that case assert should not trigger. The RA will spill in order to
3024 // free a register later.
3025 assert(compiler->opts.compDbgEnC || (calleeSaveRegs(varDsc->lvType)) == RBM_NONE);
3030 if (tree->IsCall() && (tree->gtFlags & GTF_CALL_UNMANAGED) != 0)
3032 RefPosition* pos = newRefPosition((Interval*)nullptr, currentLoc, RefTypeKillGCRefs, tree,
3033 (allRegs(TYP_REF) & ~RBM_ARG_REGS));
3041 //----------------------------------------------------------------------------
3042 // defineNewInternalTemp: Defines a ref position for an internal temp.
3045 // tree - Gentree node requiring an internal register
3046 // regType - Register type
3047 // currentLoc - Location of the temp Def position
3048 // regMask - register mask of candidates for temp
3049 // minRegCandidateCount - Minimum registers to be ensured in candidate
3050 // set under LSRA stress mode. This is a
3052 RefPosition* LinearScan::defineNewInternalTemp(GenTree* tree,
3053 RegisterType regType,
3054 LsraLocation currentLoc,
3055 regMaskTP regMask DEBUGARG(unsigned minRegCandidateCount))
3057 Interval* current = newInterval(regType);
3058 current->isInternal = true;
3059 return newRefPosition(current, currentLoc, RefTypeDef, tree, regMask, 0 DEBUG_ARG(minRegCandidateCount));
3062 //------------------------------------------------------------------------
3063 // buildInternalRegisterDefsForNode - build Def positions for internal
3064 // registers required for tree node.
3067 // tree - Gentree node that needs internal registers
3068 // currentLoc - Location at which Def positions need to be defined
3069 // temps - in-out array which is populated with ref positions
3070 // created for Def of internal registers
3071 // minRegCandidateCount - Minimum registers to be ensured in candidate
3072 // set of ref positions under LSRA stress. This is
3073 // a DEBUG only arg.
3076 // The total number of Def positions created for internal registers of tree node.
3077 int LinearScan::buildInternalRegisterDefsForNode(GenTree* tree,
3078 LsraLocation currentLoc,
3079 RefPosition* temps[] // populates
3080 DEBUGARG(unsigned minRegCandidateCount))
3083 int internalIntCount = tree->gtLsraInfo.internalIntCount;
3084 regMaskTP internalCands = tree->gtLsraInfo.getInternalCandidates(this);
3086 // If the number of internal integer registers required is the same as the number of candidate integer registers in
3087 // the candidate set, then they must be handled as fixed registers.
3088 // (E.g. for the integer registers that floating point arguments must be copied into for a varargs call.)
3089 bool fixedRegs = false;
3090 regMaskTP internalIntCandidates = (internalCands & allRegs(TYP_INT));
3091 if (((int)genCountBits(internalIntCandidates)) == internalIntCount)
3096 for (count = 0; count < internalIntCount; count++)
3098 regMaskTP internalIntCands = (internalCands & allRegs(TYP_INT));
3101 internalIntCands = genFindLowestBit(internalIntCands);
3102 internalCands &= ~internalIntCands;
3105 defineNewInternalTemp(tree, IntRegisterType, currentLoc, internalIntCands DEBUG_ARG(minRegCandidateCount));
3108 int internalFloatCount = tree->gtLsraInfo.internalFloatCount;
3109 for (int i = 0; i < internalFloatCount; i++)
3111 regMaskTP internalFPCands = (internalCands & internalFloatRegCandidates());
3113 defineNewInternalTemp(tree, FloatRegisterType, currentLoc, internalFPCands DEBUG_ARG(minRegCandidateCount));
3116 assert(count < MaxInternalRegisters);
3117 assert(count == (internalIntCount + internalFloatCount));
3121 //------------------------------------------------------------------------
3122 // buildInternalRegisterUsesForNode - adds Use positions for internal
3123 // registers required for tree node.
3126 // tree - Gentree node that needs internal registers
3127 // currentLoc - Location at which Use positions need to be defined
3128 // defs - int array containing Def positions of internal
3130 // total - Total number of Def positions in 'defs' array.
3131 // minRegCandidateCount - Minimum registers to be ensured in candidate
3132 // set of ref positions under LSRA stress. This is
3133 // a DEBUG only arg.
3137 void LinearScan::buildInternalRegisterUsesForNode(GenTree* tree,
3138 LsraLocation currentLoc,
3139 RefPosition* defs[],
3140 int total DEBUGARG(unsigned minRegCandidateCount))
3142 assert(total < MaxInternalRegisters);
3144 // defs[] has been populated by buildInternalRegisterDefsForNode
3145 // now just add uses to the defs previously added.
3146 for (int i = 0; i < total; i++)
3148 RefPosition* prevRefPosition = defs[i];
3149 assert(prevRefPosition != nullptr);
3150 regMaskTP mask = prevRefPosition->registerAssignment;
3151 if (prevRefPosition->isPhysRegRef)
3153 newRefPosition(defs[i]->getReg()->regNum, currentLoc, RefTypeUse, tree, mask);
3157 RefPosition* newest = newRefPosition(defs[i]->getInterval(), currentLoc, RefTypeUse, tree, mask,
3158 0 DEBUG_ARG(minRegCandidateCount));
3160 if (tree->gtLsraInfo.isInternalRegDelayFree)
3162 newest->delayRegFree = true;
3168 regMaskTP LinearScan::getUseCandidates(GenTree* useNode)
3170 TreeNodeInfo info = useNode->gtLsraInfo;
3171 return info.getSrcCandidates(this);
3174 regMaskTP LinearScan::getDefCandidates(GenTree* tree)
3176 TreeNodeInfo info = tree->gtLsraInfo;
3177 return info.getDstCandidates(this);
3180 RegisterType LinearScan::getDefType(GenTree* tree)
3182 return tree->TypeGet();
3185 //------------------------------------------------------------------------
3186 // LocationInfoListNode: used to store a single `LocationInfo` value for a
3187 // node during `buildIntervals`.
3189 // This is the node type for `LocationInfoList` below.
3191 class LocationInfoListNode final : public LocationInfo
3193 friend class LocationInfoList;
3194 friend class LocationInfoListNodePool;
3196 LocationInfoListNode* m_next; // The next node in the list
3199 LocationInfoListNode(LsraLocation l, Interval* i, GenTree* t, unsigned regIdx = 0) : LocationInfo(l, i, t, regIdx)
3203 //------------------------------------------------------------------------
3204 // LocationInfoListNode::Next: Returns the next node in the list.
3205 LocationInfoListNode* Next() const
3211 //------------------------------------------------------------------------
3212 // LocationInfoList: used to store a list of `LocationInfo` values for a
3213 // node during `buildIntervals`.
3215 // Given an IR node that either directly defines N registers or that is a
3216 // contained node with uses that define a total of N registers, that node
3217 // will map to N `LocationInfo` values. These values are stored as a
3218 // linked list of `LocationInfoListNode` values.
3220 class LocationInfoList final
3222 friend class LocationInfoListNodePool;
3224 LocationInfoListNode* m_head; // The head of the list
3225 LocationInfoListNode* m_tail; // The tail of the list
3228 LocationInfoList() : m_head(nullptr), m_tail(nullptr)
3232 LocationInfoList(LocationInfoListNode* node) : m_head(node), m_tail(node)
3234 assert(m_head->m_next == nullptr);
3237 //------------------------------------------------------------------------
3238 // LocationInfoList::IsEmpty: Returns true if the list is empty.
3240 bool IsEmpty() const
3242 return m_head == nullptr;
3245 //------------------------------------------------------------------------
3246 // LocationInfoList::Begin: Returns the first node in the list.
3248 LocationInfoListNode* Begin() const
3253 //------------------------------------------------------------------------
3254 // LocationInfoList::End: Returns the position after the last node in the
3255 // list. The returned value is suitable for use as
3256 // a sentinel for iteration.
3258 LocationInfoListNode* End() const
3263 //------------------------------------------------------------------------
3264 // LocationInfoList::Append: Appends a node to the list.
3267 // node - The node to append. Must not be part of an existing list.
3269 void Append(LocationInfoListNode* node)
3271 assert(node->m_next == nullptr);
3273 if (m_tail == nullptr)
3275 assert(m_head == nullptr);
3280 m_tail->m_next = node;
3286 //------------------------------------------------------------------------
3287 // LocationInfoList::Append: Appends another list to this list.
3290 // other - The list to append.
3292 void Append(LocationInfoList other)
3294 if (m_tail == nullptr)
3296 assert(m_head == nullptr);
3297 m_head = other.m_head;
3301 m_tail->m_next = other.m_head;
3304 m_tail = other.m_tail;
3308 //------------------------------------------------------------------------
3309 // LocationInfoListNodePool: manages a pool of `LocationInfoListNode`
3310 // values to decrease overall memory usage
3311 // during `buildIntervals`.
3313 // `buildIntervals` involves creating a list of location info values per
3314 // node that either directly produces a set of registers or that is a
3315 // contained node with register-producing sources. However, these lists
3316 // are short-lived: they are destroyed once the use of the corresponding
3317 // node is processed. As such, there is typically only a small number of
3318 // `LocationInfoListNode` values in use at any given time. Pooling these
3319 // values avoids otherwise frequent allocations.
3320 class LocationInfoListNodePool final
3322 LocationInfoListNode* m_freeList;
3323 Compiler* m_compiler;
3326 //------------------------------------------------------------------------
3327 // LocationInfoListNodePool::LocationInfoListNodePool:
3328 // Creates a pool of `LocationInfoListNode` values.
3331 // compiler - The compiler context.
3332 // preallocate - The number of nodes to preallocate.
3334 LocationInfoListNodePool(Compiler* compiler, unsigned preallocate = 0) : m_compiler(compiler)
3336 if (preallocate > 0)
3338 size_t preallocateSize = sizeof(LocationInfoListNode) * preallocate;
3339 auto* preallocatedNodes = reinterpret_cast<LocationInfoListNode*>(compiler->compGetMem(preallocateSize));
3341 LocationInfoListNode* head = preallocatedNodes;
3342 head->m_next = nullptr;
3344 for (unsigned i = 1; i < preallocate; i++)
3346 LocationInfoListNode* node = &preallocatedNodes[i];
3347 node->m_next = head;
3355 //------------------------------------------------------------------------
3356 // LocationInfoListNodePool::GetNode: Fetches an unused node from the
3360 // l - - The `LsraLocation` for the `LocationInfo` value.
3361 // i - The interval for the `LocationInfo` value.
3362 // t - The IR node for the `LocationInfo` value
3363 // regIdx - The register index for the `LocationInfo` value.
3366 // A pooled or newly-allocated `LocationInfoListNode`, depending on the
3367 // contents of the pool.
3368 LocationInfoListNode* GetNode(LsraLocation l, Interval* i, GenTree* t, unsigned regIdx = 0)
3370 LocationInfoListNode* head = m_freeList;
3371 if (head == nullptr)
3373 head = reinterpret_cast<LocationInfoListNode*>(m_compiler->compGetMem(sizeof(LocationInfoListNode)));
3377 m_freeList = head->m_next;
3383 head->multiRegIdx = regIdx;
3384 head->m_next = nullptr;
3389 //------------------------------------------------------------------------
3390 // LocationInfoListNodePool::ReturnNodes: Returns a list of nodes to the
3394 // list - The list to return.
3396 void ReturnNodes(LocationInfoList& list)
3398 assert(list.m_head != nullptr);
3399 assert(list.m_tail != nullptr);
3401 LocationInfoListNode* head = m_freeList;
3402 list.m_tail->m_next = head;
3403 m_freeList = list.m_head;
3407 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
3409 LinearScan::buildUpperVectorSaveRefPositions(GenTree* tree, LsraLocation currentLoc)
3411 assert(enregisterLocalVars);
3412 VARSET_TP liveLargeVectors(VarSetOps::MakeEmpty(compiler));
3413 regMaskTP fpCalleeKillSet = RBM_NONE;
3414 if (!VarSetOps::IsEmpty(compiler, largeVectorVars))
3416 // We actually need to find any calls that kill the upper-half of the callee-save vector registers.
3417 // But we will use as a proxy any node that kills floating point registers.
3418 // (Note that some calls are masquerading as other nodes at this point so we can't just check for calls.)
3419 fpCalleeKillSet = getKillSetForNode(tree);
3420 if ((fpCalleeKillSet & RBM_FLT_CALLEE_TRASH) != RBM_NONE)
3422 VarSetOps::AssignNoCopy(compiler, liveLargeVectors,
3423 VarSetOps::Intersection(compiler, currentLiveVars, largeVectorVars));
3424 VarSetOps::Iter iter(compiler, liveLargeVectors);
3425 unsigned varIndex = 0;
3426 while (iter.NextElem(&varIndex))
3428 Interval* varInterval = getIntervalForLocalVar(varIndex);
3429 Interval* tempInterval = newInterval(LargeVectorType);
3430 tempInterval->isInternal = true;
3432 newRefPosition(tempInterval, currentLoc, RefTypeUpperVectorSaveDef, tree, RBM_FLT_CALLEE_SAVED);
3433 // We are going to save the existing relatedInterval of varInterval on tempInterval, so that we can set
3434 // the tempInterval as the relatedInterval of varInterval, so that we can build the corresponding
3435 // RefTypeUpperVectorSaveUse RefPosition. We will then restore the relatedInterval onto varInterval,
3436 // and set varInterval as the relatedInterval of tempInterval.
3437 tempInterval->relatedInterval = varInterval->relatedInterval;
3438 varInterval->relatedInterval = tempInterval;
3442 return liveLargeVectors;
3445 void LinearScan::buildUpperVectorRestoreRefPositions(GenTree* tree,
3446 LsraLocation currentLoc,
3447 VARSET_VALARG_TP liveLargeVectors)
3449 assert(enregisterLocalVars);
3450 if (!VarSetOps::IsEmpty(compiler, liveLargeVectors))
3452 VarSetOps::Iter iter(compiler, liveLargeVectors);
3453 unsigned varIndex = 0;
3454 while (iter.NextElem(&varIndex))
3456 Interval* varInterval = getIntervalForLocalVar(varIndex);
3457 Interval* tempInterval = varInterval->relatedInterval;
3458 assert(tempInterval->isInternal == true);
3460 newRefPosition(tempInterval, currentLoc, RefTypeUpperVectorSaveUse, tree, RBM_FLT_CALLEE_SAVED);
3461 // Restore the relatedInterval onto varInterval, and set varInterval as the relatedInterval
3463 varInterval->relatedInterval = tempInterval->relatedInterval;
3464 tempInterval->relatedInterval = varInterval;
3468 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
3471 //------------------------------------------------------------------------
3472 // ComputeOperandDstCount: computes the number of registers defined by a
3475 // For most nodes, this is simple:
3476 // - Nodes that do not produce values (e.g. stores and other void-typed
3477 // nodes) and nodes that immediately use the registers they define
3478 // produce no registers
3479 // - Nodes that are marked as defining N registers define N registers.
3481 // For contained nodes, however, things are more complicated: for purposes
3482 // of bookkeeping, a contained node is treated as producing the transitive
3483 // closure of the registers produced by its sources.
3486 // operand - The operand for which to compute a register count.
3489 // The number of registers defined by `operand`.
3491 static int ComputeOperandDstCount(GenTree* operand)
3493 TreeNodeInfo& operandInfo = operand->gtLsraInfo;
3495 if (operandInfo.isLocalDefUse)
3497 // Operands that define an unused value do not produce any registers.
3500 else if (operandInfo.dstCount != 0)
3502 // Operands that have a specified number of destination registers consume all of their operands
3503 // and therefore produce exactly that number of registers.
3504 return operandInfo.dstCount;
3506 else if (operandInfo.srcCount != 0)
3508 // If an operand has no destination registers but does have source registers, it must be a store
3510 assert(operand->OperIsStore() || operand->OperIsBlkOp() || operand->OperIsPutArgStk() ||
3511 operand->OperIsCompare() || operand->OperIs(GT_CMP) || operand->IsSIMDEqualityOrInequality());
3514 else if (!operand->OperIsFieldListHead() && (operand->OperIsStore() || operand->TypeGet() == TYP_VOID))
3516 // Stores and void-typed operands may be encountered when processing call nodes, which contain
3517 // pointers to argument setup stores.
3522 // If a field list or non-void-typed operand is not an unused value and does not have source registers,
3523 // that argument is contained within its parent and produces `sum(operand_dst_count)` registers.
3525 for (GenTree* op : operand->Operands())
3527 dstCount += ComputeOperandDstCount(op);
3534 //------------------------------------------------------------------------
3535 // ComputeAvailableSrcCount: computes the number of registers available as
3536 // sources for a node.
3538 // This is simply the sum of the number of registers produced by each
3539 // operand to the node.
3542 // node - The node for which to compute a source count.
3545 // The number of registers available as sources for `node`.
3547 static int ComputeAvailableSrcCount(GenTree* node)
3550 for (GenTree* operand : node->Operands())
3552 numSources += ComputeOperandDstCount(operand);
3559 static GenTree* GetFirstOperand(GenTree* node)
3561 GenTree* firstOperand = nullptr;
3562 node->VisitOperands([&firstOperand](GenTree* operand) -> GenTree::VisitResult {
3563 firstOperand = operand;
3564 return GenTree::VisitResult::Abort;
3566 return firstOperand;
3569 void LinearScan::buildRefPositionsForNode(GenTree* tree,
3571 LocationInfoListNodePool& listNodePool,
3572 HashTableBase<GenTree*, LocationInfoList>& operandToLocationInfoMap,
3573 LsraLocation currentLoc)
3576 assert(!isRegPairType(tree->TypeGet()));
3577 #endif // _TARGET_ARM_
3579 // The LIR traversal doesn't visit GT_LIST or GT_ARGPLACE nodes.
3580 // GT_CLS_VAR nodes should have been eliminated by rationalizer.
3581 assert(tree->OperGet() != GT_ARGPLACE);
3582 assert(tree->OperGet() != GT_LIST);
3583 assert(tree->OperGet() != GT_CLS_VAR);
3585 // The LIR traversal visits only the first node in a GT_FIELD_LIST.
3586 assert((tree->OperGet() != GT_FIELD_LIST) || tree->AsFieldList()->IsFieldListHead());
3588 // The set of internal temporary registers used by this node are stored in the
3589 // gtRsvdRegs register mask. Clear it out.
3590 tree->gtRsvdRegs = RBM_NONE;
3592 TreeNodeInfo info = tree->gtLsraInfo;
3593 assert(info.IsValid(this));
3594 int consume = info.srcCount;
3595 int produce = info.dstCount;
3600 lsraDispNode(tree, LSRA_DUMP_REFPOS, (produce != 0));
3602 if (tree->isContained())
3604 JITDUMP("Contained\n");
3606 else if (tree->OperIs(GT_LCL_VAR, GT_LCL_FLD) && info.isLocalDefUse)
3608 JITDUMP("Unused\n");
3612 JITDUMP(" consume=%d produce=%d\n", consume, produce);
3617 JITDUMP("at start of tree, map contains: { ");
3619 for (auto kvp : operandToLocationInfoMap)
3621 GenTree* node = kvp.Key();
3622 LocationInfoList defList = kvp.Value();
3624 JITDUMP("%sN%03u. %s -> (", first ? "" : "; ", node->gtSeqNum, GenTree::OpName(node->OperGet()));
3625 for (LocationInfoListNode *def = defList.Begin(), *end = defList.End(); def != end; def = def->Next())
3627 JITDUMP("%s%d.N%03u", def == defList.Begin() ? "" : ", ", def->loc, def->treeNode->gtSeqNum);
3638 assert(((consume == 0) && (produce == 0)) || (ComputeAvailableSrcCount(tree) == consume));
3640 if (tree->OperIs(GT_LCL_VAR, GT_LCL_FLD))
3642 LclVarDsc* const varDsc = &compiler->lvaTable[tree->AsLclVarCommon()->gtLclNum];
3643 if (isCandidateVar(varDsc))
3645 assert(consume == 0);
3647 // We handle tracked variables differently from non-tracked ones. If it is tracked,
3648 // we simply add a use or def of the tracked variable. Otherwise, for a use we need
3649 // to actually add the appropriate references for loading or storing the variable.
3651 // It won't actually get used or defined until the appropriate ancestor tree node
3652 // is processed, unless this is marked "isLocalDefUse" because it is a stack-based argument
3655 assert(varDsc->lvTracked);
3656 unsigned varIndex = varDsc->lvVarIndex;
3658 // We have only approximate last-use information at this point. This is because the
3659 // execution order doesn't actually reflect the true order in which the localVars
3660 // are referenced - but the order of the RefPositions will, so we recompute it after
3661 // RefPositions are built.
3662 // Use the old value for setting currentLiveVars - note that we do this with the
3663 // not-quite-correct setting of lastUse. However, this is OK because
3664 // 1) this is only for preferencing, which doesn't require strict correctness, and
3665 // 2) the cases where these out-of-order uses occur should not overlap a kill.
3666 // TODO-Throughput: clean this up once we have the execution order correct. At that point
3667 // we can update currentLiveVars at the same place that we create the RefPosition.
3668 if ((tree->gtFlags & GTF_VAR_DEATH) != 0)
3670 VarSetOps::RemoveElemD(compiler, currentLiveVars, varIndex);
3673 if (!info.isLocalDefUse && !tree->isContained())
3675 assert(produce != 0);
3677 LocationInfoList list(listNodePool.GetNode(currentLoc, getIntervalForLocalVar(varIndex), tree));
3678 bool added = operandToLocationInfoMap.AddOrUpdate(tree, list);
3681 tree->gtLsraInfo.definesAnyRegisters = true;
3687 if (tree->isContained())
3689 assert(!info.isLocalDefUse);
3690 assert(consume == 0);
3691 assert(produce == 0);
3692 assert(info.internalIntCount == 0);
3693 assert(info.internalFloatCount == 0);
3695 // Contained nodes map to the concatenated lists of their operands.
3696 LocationInfoList locationInfoList;
3697 tree->VisitOperands([&](GenTree* op) -> GenTree::VisitResult {
3698 if (!op->gtLsraInfo.definesAnyRegisters)
3700 assert(ComputeOperandDstCount(op) == 0);
3701 return GenTree::VisitResult::Continue;
3704 LocationInfoList operandList;
3705 bool removed = operandToLocationInfoMap.TryRemove(op, &operandList);
3708 locationInfoList.Append(operandList);
3709 return GenTree::VisitResult::Continue;
3712 if (!locationInfoList.IsEmpty())
3714 bool added = operandToLocationInfoMap.AddOrUpdate(tree, locationInfoList);
3716 tree->gtLsraInfo.definesAnyRegisters = true;
3722 // Handle the case of local variable assignment
3723 Interval* varDefInterval = nullptr;
3724 RefType defRefType = RefTypeDef;
3726 GenTree* defNode = tree;
3728 // noAdd means the node creates a def but for purposes of map
3729 // management do not add it because data is not flowing up the
3730 // tree but over (as in ASG nodes)
3732 bool noAdd = info.isLocalDefUse;
3733 RefPosition* prevPos = nullptr;
3735 bool isSpecialPutArg = false;
3737 assert(!tree->OperIsAssignment());
3738 if (tree->OperIsLocalStore())
3740 GenTreeLclVarCommon* const store = tree->AsLclVarCommon();
3741 assert((consume > 1) || (regType(store->gtOp1->TypeGet()) == regType(store->TypeGet())));
3743 LclVarDsc* varDsc = &compiler->lvaTable[store->gtLclNum];
3744 if (isCandidateVar(varDsc))
3746 // We always push the tracked lclVar intervals
3747 assert(varDsc->lvTracked);
3748 unsigned varIndex = varDsc->lvVarIndex;
3749 varDefInterval = getIntervalForLocalVar(varIndex);
3750 defRefType = refTypeForLocalRefNode(tree);
3758 assert(consume <= MAX_RET_REG_COUNT);
3761 // Get the location info for the register defined by the first operand.
3762 LocationInfoList operandDefs;
3763 bool found = operandToLocationInfoMap.TryGetValue(GetFirstOperand(tree), &operandDefs);
3766 // Since we only expect to consume one register, we should only have a single register to
3768 assert(operandDefs.Begin()->Next() == operandDefs.End());
3770 LocationInfo& operandInfo = *static_cast<LocationInfo*>(operandDefs.Begin());
3772 Interval* srcInterval = operandInfo.interval;
3773 if (srcInterval->relatedInterval == nullptr)
3775 // Preference the source to the dest, unless this is a non-last-use localVar.
3776 // Note that the last-use info is not correct, but it is a better approximation than preferencing
3777 // the source to the dest, if the source's lifetime extends beyond the dest.
3778 if (!srcInterval->isLocalVar || (operandInfo.treeNode->gtFlags & GTF_VAR_DEATH) != 0)
3780 srcInterval->assignRelatedInterval(varDefInterval);
3783 else if (!srcInterval->isLocalVar)
3785 // Preference the source to dest, if src is not a local var.
3786 srcInterval->assignRelatedInterval(varDefInterval);
3790 if ((tree->gtFlags & GTF_VAR_DEATH) == 0)
3792 VarSetOps::AddElemD(compiler, currentLiveVars, varIndex);
3795 else if (store->gtOp1->OperIs(GT_BITCAST))
3797 store->gtType = store->gtOp1->gtType = store->gtOp1->AsUnOp()->gtOp1->TypeGet();
3799 // Get the location info for the register defined by the first operand.
3800 LocationInfoList operandDefs;
3801 bool found = operandToLocationInfoMap.TryGetValue(GetFirstOperand(store), &operandDefs);
3804 // Since we only expect to consume one register, we should only have a single register to consume.
3805 assert(operandDefs.Begin()->Next() == operandDefs.End());
3807 LocationInfo& operandInfo = *static_cast<LocationInfo*>(operandDefs.Begin());
3809 Interval* srcInterval = operandInfo.interval;
3810 srcInterval->registerType = regType(store->TypeGet());
3812 RefPosition* srcDefPosition = srcInterval->firstRefPosition;
3813 assert(srcDefPosition != nullptr);
3814 assert(srcDefPosition->refType == RefTypeDef);
3815 assert(srcDefPosition->treeNode == store->gtOp1);
3817 srcDefPosition->registerAssignment = allRegs(store->TypeGet());
3818 store->gtOp1->gtLsraInfo.setSrcCandidates(this, allRegs(store->TypeGet()));
3821 else if (noAdd && produce == 0)
3823 // This is the case for dead nodes that occur after
3824 // tree rationalization
3825 // TODO-Cleanup: Identify and remove these dead nodes prior to register allocation.
3826 if (tree->IsMultiRegCall())
3828 // In case of multi-reg call node, produce = number of return registers
3829 produce = tree->AsCall()->GetReturnTypeDesc()->GetReturnRegCount();
3837 Interval* prefSrcInterval = nullptr;
3839 // If this is a binary operator that will be encoded with 2 operand fields
3840 // (i.e. the target is read-modify-write), preference the dst to op1.
3842 bool hasDelayFreeSrc = tree->gtLsraInfo.hasDelayFreeSrc;
3844 #if defined(DEBUG) && defined(_TARGET_X86_)
3845 // On x86, `LSRA_LIMIT_CALLER` is too restrictive to allow the use of special put args: this stress mode
3846 // leaves only three registers allocatable--eax, ecx, and edx--of which the latter two are also used for the
3847 // first two integral arguments to a call. This can leave us with too few registers to succesfully allocate in
3848 // situations like the following:
3850 // t1026 = lclVar ref V52 tmp35 u:3 REG NA <l:$3a1, c:$98d>
3853 // t1352 = * putarg_reg ref REG NA
3855 // t342 = lclVar int V14 loc6 u:4 REG NA $50c
3857 // t343 = const int 1 REG NA $41
3861 // t344 = * + int REG NA $495
3863 // t345 = lclVar int V04 arg4 u:2 REG NA $100
3867 // t346 = * % int REG NA $496
3870 // t1353 = * putarg_reg int REG NA
3872 // t1354 = lclVar ref V52 tmp35 (last use) REG NA
3875 // t1355 = * lea(b+0) byref REG NA
3877 // Here, the first `putarg_reg` would normally be considered a special put arg, which would remove `ecx` from the
3878 // set of allocatable registers, leaving only `eax` and `edx`. The allocator will then fail to allocate a register
3879 // for the def of `t345` if arg4 is not a register candidate: the corresponding ref position will be constrained to
3880 // { `ecx`, `ebx`, `esi`, `edi` }, which `LSRA_LIMIT_CALLER` will further constrain to `ecx`, which will not be
3881 // available due to the special put arg.
3882 const bool supportsSpecialPutArg = getStressLimitRegs() != LSRA_LIMIT_CALLER;
3884 const bool supportsSpecialPutArg = true;
3887 if (supportsSpecialPutArg && tree->OperGet() == GT_PUTARG_REG && isCandidateLocalRef(tree->gtGetOp1()) &&
3888 (tree->gtGetOp1()->gtFlags & GTF_VAR_DEATH) == 0)
3890 // This is the case for a "pass-through" copy of a lclVar. In the case where it is a non-last-use,
3891 // we don't want the def of the copy to kill the lclVar register, if it is assigned the same register
3892 // (which is actually what we hope will happen).
3893 JITDUMP("Setting putarg_reg as a pass-through of a non-last use lclVar\n");
3895 // Get the register information for the first operand of the node.
3896 LocationInfoList operandDefs;
3897 bool found = operandToLocationInfoMap.TryGetValue(GetFirstOperand(tree), &operandDefs);
3900 // Preference the destination to the interval of the first register defined by the first operand.
3901 Interval* srcInterval = operandDefs.Begin()->interval;
3902 assert(srcInterval->isLocalVar);
3903 prefSrcInterval = srcInterval;
3904 isSpecialPutArg = true;
3907 RefPosition* internalRefs[MaxInternalRegisters];
3910 // Number of registers required for tree node is the sum of
3911 // consume + produce + internalCount. This is the minimum
3912 // set of registers that needs to be ensured in candidate
3913 // set of ref positions created.
3914 unsigned minRegCount = consume + produce + info.internalIntCount + info.internalFloatCount;
3917 // make intervals for all the 'internal' register requirements for this node
3918 // where internal means additional registers required temporarily
3919 int internalCount = buildInternalRegisterDefsForNode(tree, currentLoc, internalRefs DEBUG_ARG(minRegCount));
3921 // pop all ref'd tree temps
3922 tree->VisitOperands([&](GenTree* operand) -> GenTree::VisitResult {
3923 // Skip operands that do not define any registers, whether directly or indirectly.
3924 if (!operand->gtLsraInfo.definesAnyRegisters)
3926 return GenTree::VisitResult::Continue;
3929 // Remove the list of registers defined by the current operand from the map. Note that this
3930 // is only correct because tree nodes are singly-used: if this property ever changes (e.g.
3931 // if tree nodes are eventually allowed to be multiply-used), then the removal is only
3932 // correct at the last use.
3933 LocationInfoList operandDefs;
3934 bool removed = operandToLocationInfoMap.TryRemove(operand, &operandDefs);
3936 assert(!operandDefs.IsEmpty());
3938 LocationInfoListNode* const operandDefsEnd = operandDefs.End();
3939 for (LocationInfoListNode* operandDefsIterator = operandDefs.Begin(); operandDefsIterator != operandDefsEnd;
3940 operandDefsIterator = operandDefsIterator->Next())
3942 LocationInfo& locInfo = *static_cast<LocationInfo*>(operandDefsIterator);
3944 // for interstitial tree temps, a use is always last and end; this is set by default in newRefPosition
3945 GenTree* const useNode = locInfo.treeNode;
3946 assert(useNode != nullptr);
3948 Interval* const i = locInfo.interval;
3949 if (useNode->gtLsraInfo.isTgtPref)
3951 prefSrcInterval = i;
3954 const bool delayRegFree = (hasDelayFreeSrc && useNode->gtLsraInfo.isDelayFree);
3957 // If delayRegFree, then Use will interfere with the destination of
3958 // the consuming node. Therefore, we also need add the kill set of
3959 // consuming node to minRegCount.
3961 // For example consider the following IR on x86, where v01 and v02
3962 // are method args coming in ecx and edx respectively.
3965 // For GT_DIV minRegCount will be 3 without adding kill set
3968 // Assume further JitStressRegs=2, which would constrain
3969 // candidates to callee trashable regs { eax, ecx, edx } on
3970 // use positions of v01 and v02. LSRA allocates ecx for v01.
3971 // Use position of v02 cannot be allocated a regs since it
3972 // is marked delay-reg free and {eax,edx} are getting killed
3973 // before the def of GT_DIV. For this reason, minRegCount
3974 // for Use position of v02 also needs to take into account
3975 // of kill set of its consuming node.
3976 unsigned minRegCountForUsePos = minRegCount;
3979 regMaskTP killMask = getKillSetForNode(tree);
3980 if (killMask != RBM_NONE)
3982 minRegCountForUsePos += genCountBits(killMask);
3987 regMaskTP candidates = getUseCandidates(useNode);
3989 // If oper is GT_PUTARG_REG, set bits in useCandidates must be in sequential order.
3990 if (useNode->OperGet() == GT_PUTARG_REG || useNode->OperGet() == GT_COPY)
3992 regMaskTP candidate = genFindLowestReg(candidates);
3993 useNode->gtLsraInfo.setSrcCandidates(this, candidates & ~candidate);
3994 candidates = candidate;
3996 #endif // ARM_SOFTFP
3997 assert((candidates & allRegs(i->registerType)) != 0);
3999 // For non-localVar uses we record nothing, as nothing needs to be written back to the tree.
4000 GenTree* const refPosNode = i->isLocalVar ? useNode : nullptr;
4001 RefPosition* pos = newRefPosition(i, currentLoc, RefTypeUse, refPosNode, candidates,
4002 locInfo.multiRegIdx DEBUG_ARG(minRegCountForUsePos));
4006 pos->delayRegFree = true;
4009 if (useNode->IsRegOptional())
4011 pos->setAllocateIfProfitable(true);
4015 listNodePool.ReturnNodes(operandDefs);
4017 return GenTree::VisitResult::Continue;
4020 buildInternalRegisterUsesForNode(tree, currentLoc, internalRefs, internalCount DEBUG_ARG(minRegCount));
4022 RegisterType registerType = getDefType(tree);
4023 regMaskTP candidates = getDefCandidates(tree);
4024 regMaskTP useCandidates = getUseCandidates(tree);
4027 if (VERBOSE && produce)
4029 printf("Def candidates ");
4030 dumpRegMask(candidates);
4031 printf(", Use candidates ");
4032 dumpRegMask(useCandidates);
4037 #if defined(_TARGET_AMD64_)
4038 // Multi-reg call node is the only node that could produce multi-reg value
4039 assert(produce <= 1 || (tree->IsMultiRegCall() && produce == MAX_RET_REG_COUNT));
4040 #endif // _TARGET_xxx_
4042 // Add kill positions before adding def positions
4043 buildKillPositionsForNode(tree, currentLoc + 1);
4045 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
4046 VARSET_TP liveLargeVectors(VarSetOps::UninitVal());
4047 if (enregisterLocalVars && (RBM_FLT_CALLEE_SAVED != RBM_NONE))
4049 // Build RefPositions for saving any live large vectors.
4050 // This must be done after the kills, so that we know which large vectors are still live.
4051 VarSetOps::AssignNoCopy(compiler, liveLargeVectors, buildUpperVectorSaveRefPositions(tree, currentLoc + 1));
4053 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
4055 ReturnTypeDesc* retTypeDesc = nullptr;
4056 bool isMultiRegCall = tree->IsMultiRegCall();
4059 retTypeDesc = tree->AsCall()->GetReturnTypeDesc();
4060 assert((int)genCountBits(candidates) == produce);
4061 assert(candidates == retTypeDesc->GetABIReturnRegs());
4065 LocationInfoList locationInfoList;
4066 LsraLocation defLocation = currentLoc + 1;
4068 regMaskTP remainingUseCandidates = useCandidates;
4070 for (int i = 0; i < produce; i++)
4072 regMaskTP currCandidates = candidates;
4073 Interval* interval = varDefInterval;
4075 // In case of multi-reg call node, registerType is given by
4076 // the type of ith position return register.
4079 registerType = retTypeDesc->GetReturnRegType((unsigned)i);
4080 currCandidates = genRegMask(retTypeDesc->GetABIReturnReg(i));
4081 useCandidates = allRegs(registerType);
4085 if (tree->OperIsPutArgSplit())
4087 // get i-th candidate
4088 currCandidates = genFindLowestReg(candidates);
4089 candidates &= ~currCandidates;
4092 // If oper is GT_PUTARG_REG, set bits in useCandidates must be in sequential order.
4093 else if (tree->OperGet() == GT_PUTARG_REG || tree->OperGet() == GT_COPY)
4095 useCandidates = genFindLowestReg(remainingUseCandidates);
4096 remainingUseCandidates &= ~useCandidates;
4098 #endif // ARM_SOFTFP
4099 #endif // _TARGET_ARM_
4101 if (interval == nullptr)
4103 // Make a new interval
4104 interval = newInterval(registerType);
4105 if (hasDelayFreeSrc)
4107 interval->hasNonCommutativeRMWDef = true;
4109 else if (tree->OperIsConst())
4111 assert(!tree->IsReuseRegVal());
4112 interval->isConstant = true;
4115 if ((currCandidates & useCandidates) != RBM_NONE)
4117 interval->updateRegisterPreferences(currCandidates & useCandidates);
4120 if (isSpecialPutArg)
4122 interval->isSpecialPutArg = true;
4127 assert(registerTypesEquivalent(interval->registerType, registerType));
4130 if (prefSrcInterval != nullptr)
4132 interval->assignRelatedIntervalIfUnassigned(prefSrcInterval);
4135 // for assignments, we want to create a refposition for the def
4139 locationInfoList.Append(listNodePool.GetNode(defLocation, interval, tree, (unsigned)i));
4142 RefPosition* pos = newRefPosition(interval, defLocation, defRefType, defNode, currCandidates,
4143 (unsigned)i DEBUG_ARG(minRegCount));
4144 if (info.isLocalDefUse)
4146 pos->isLocalDefUse = true;
4147 pos->lastUse = true;
4149 interval->updateRegisterPreferences(currCandidates);
4150 interval->updateRegisterPreferences(useCandidates);
4153 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
4154 // SaveDef position must be at the same location as Def position of call node.
4155 if (enregisterLocalVars)
4157 buildUpperVectorRestoreRefPositions(tree, defLocation, liveLargeVectors);
4159 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
4161 if (!locationInfoList.IsEmpty())
4163 bool added = operandToLocationInfoMap.AddOrUpdate(tree, locationInfoList);
4165 tree->gtLsraInfo.definesAnyRegisters = true;
4170 // make an interval for each physical register
4171 void LinearScan::buildPhysRegRecords()
4173 RegisterType regType = IntRegisterType;
4174 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
4176 RegRecord* curr = &physRegs[reg];
4181 BasicBlock* getNonEmptyBlock(BasicBlock* block)
4183 while (block != nullptr && block->bbTreeList == nullptr)
4185 BasicBlock* nextBlock = block->bbNext;
4186 // Note that here we use the version of NumSucc that does not take a compiler.
4187 // That way this doesn't have to take a compiler, or be an instance method, e.g. of LinearScan.
4188 // If we have an empty block, it must have jump type BBJ_NONE or BBJ_ALWAYS, in which
4189 // case we don't need the version that takes a compiler.
4190 assert(block->NumSucc() == 1 && ((block->bbJumpKind == BBJ_ALWAYS) || (block->bbJumpKind == BBJ_NONE)));
4191 // sometimes the first block is empty and ends with an uncond branch
4192 // assert( block->GetSucc(0) == nextBlock);
4195 assert(block != nullptr && block->bbTreeList != nullptr);
4199 //------------------------------------------------------------------------
4200 // insertZeroInitRefPositions: Handle lclVars that are live-in to the first block
4203 // Prior to calling this method, 'currentLiveVars' must be set to the set of register
4204 // candidate variables that are liveIn to the first block.
4205 // For each register candidate that is live-in to the first block:
4206 // - If it is a GC ref, or if compInitMem is set, a ZeroInit RefPosition will be created.
4207 // - Otherwise, it will be marked as spilled, since it will not be assigned a register
4208 // on entry and will be loaded from memory on the undefined path.
4209 // Note that, when the compInitMem option is not set, we may encounter these on
4210 // paths that are protected by the same condition as an earlier def. However, since
4211 // we don't do the analysis to determine this - and couldn't rely on always identifying
4212 // such cases even if we tried - we must conservatively treat the undefined path as
4213 // being possible. This is a relatively rare case, so the introduced conservatism is
4214 // not expected to warrant the analysis required to determine the best placement of
4215 // an initialization.
4217 void LinearScan::insertZeroInitRefPositions()
4219 assert(enregisterLocalVars);
4221 VARSET_TP expectedLiveVars(VarSetOps::Intersection(compiler, registerCandidateVars, compiler->fgFirstBB->bbLiveIn));
4222 assert(VarSetOps::Equal(compiler, currentLiveVars, expectedLiveVars));
4225 // insert defs for this, then a block boundary
4227 VarSetOps::Iter iter(compiler, currentLiveVars);
4228 unsigned varIndex = 0;
4229 while (iter.NextElem(&varIndex))
4231 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
4232 LclVarDsc* varDsc = compiler->lvaTable + varNum;
4233 if (!varDsc->lvIsParam && isCandidateVar(varDsc))
4235 JITDUMP("V%02u was live in to first block:", varNum);
4236 Interval* interval = getIntervalForLocalVar(varIndex);
4237 if (compiler->info.compInitMem || varTypeIsGC(varDsc->TypeGet()))
4239 JITDUMP(" creating ZeroInit\n");
4240 GenTree* firstNode = getNonEmptyBlock(compiler->fgFirstBB)->firstNode();
4242 newRefPosition(interval, MinLocation, RefTypeZeroInit, firstNode, allRegs(interval->registerType));
4243 varDsc->lvMustInit = true;
4247 setIntervalAsSpilled(interval);
4248 JITDUMP(" marking as spilled\n");
4254 #if defined(FEATURE_UNIX_AMD64_STRUCT_PASSING)
4255 // -----------------------------------------------------------------------
4256 // Sets the register state for an argument of type STRUCT for System V systems.
4257 // See Compiler::raUpdateRegStateForArg(RegState *regState, LclVarDsc *argDsc) in regalloc.cpp
4258 // for how state for argument is updated for unix non-structs and Windows AMD64 structs.
4259 void LinearScan::unixAmd64UpdateRegStateForArg(LclVarDsc* argDsc)
4261 assert(varTypeIsStruct(argDsc));
4262 RegState* intRegState = &compiler->codeGen->intRegState;
4263 RegState* floatRegState = &compiler->codeGen->floatRegState;
4265 if ((argDsc->lvArgReg != REG_STK) && (argDsc->lvArgReg != REG_NA))
4267 if (genRegMask(argDsc->lvArgReg) & (RBM_ALLFLOAT))
4269 assert(genRegMask(argDsc->lvArgReg) & (RBM_FLTARG_REGS));
4270 floatRegState->rsCalleeRegArgMaskLiveIn |= genRegMask(argDsc->lvArgReg);
4274 assert(genRegMask(argDsc->lvArgReg) & (RBM_ARG_REGS));
4275 intRegState->rsCalleeRegArgMaskLiveIn |= genRegMask(argDsc->lvArgReg);
4279 if ((argDsc->lvOtherArgReg != REG_STK) && (argDsc->lvOtherArgReg != REG_NA))
4281 if (genRegMask(argDsc->lvOtherArgReg) & (RBM_ALLFLOAT))
4283 assert(genRegMask(argDsc->lvOtherArgReg) & (RBM_FLTARG_REGS));
4284 floatRegState->rsCalleeRegArgMaskLiveIn |= genRegMask(argDsc->lvOtherArgReg);
4288 assert(genRegMask(argDsc->lvOtherArgReg) & (RBM_ARG_REGS));
4289 intRegState->rsCalleeRegArgMaskLiveIn |= genRegMask(argDsc->lvOtherArgReg);
4294 #endif // defined(FEATURE_UNIX_AMD64_STRUCT_PASSING)
4296 //------------------------------------------------------------------------
4297 // updateRegStateForArg: Updates rsCalleeRegArgMaskLiveIn for the appropriate
4298 // regState (either compiler->intRegState or compiler->floatRegState),
4299 // with the lvArgReg on "argDsc"
4302 // argDsc - the argument for which the state is to be updated.
4304 // Return Value: None
4307 // The argument is live on entry to the function
4308 // (or is untracked and therefore assumed live)
4311 // This relies on a method in regAlloc.cpp that is shared between LSRA
4312 // and regAlloc. It is further abstracted here because regState is updated
4313 // separately for tracked and untracked variables in LSRA.
4315 void LinearScan::updateRegStateForArg(LclVarDsc* argDsc)
4317 #if defined(FEATURE_UNIX_AMD64_STRUCT_PASSING)
4318 // For System V AMD64 calls the argDsc can have 2 registers (for structs.)
4319 // Handle them here.
4320 if (varTypeIsStruct(argDsc))
4322 unixAmd64UpdateRegStateForArg(argDsc);
4325 #endif // defined(FEATURE_UNIX_AMD64_STRUCT_PASSING)
4327 RegState* intRegState = &compiler->codeGen->intRegState;
4328 RegState* floatRegState = &compiler->codeGen->floatRegState;
4329 // In the case of AMD64 we'll still use the floating point registers
4330 // to model the register usage for argument on vararg calls, so
4331 // we will ignore the varargs condition to determine whether we use
4332 // XMM registers or not for setting up the call.
4333 bool isFloat = (isFloatRegType(argDsc->lvType)
4334 #ifndef _TARGET_AMD64_
4335 && !compiler->info.compIsVarArgs
4337 && !compiler->opts.compUseSoftFP);
4339 if (argDsc->lvIsHfaRegArg())
4346 JITDUMP("Float arg V%02u in reg %s\n", (argDsc - compiler->lvaTable), getRegName(argDsc->lvArgReg));
4347 compiler->raUpdateRegStateForArg(floatRegState, argDsc);
4351 JITDUMP("Int arg V%02u in reg %s\n", (argDsc - compiler->lvaTable), getRegName(argDsc->lvArgReg));
4352 #if FEATURE_MULTIREG_ARGS
4353 if (argDsc->lvOtherArgReg != REG_NA)
4355 JITDUMP("(second half) in reg %s\n", getRegName(argDsc->lvOtherArgReg));
4357 #endif // FEATURE_MULTIREG_ARGS
4358 compiler->raUpdateRegStateForArg(intRegState, argDsc);
4363 //------------------------------------------------------------------------
4364 // findPredBlockForLiveIn: Determine which block should be used for the register locations of the live-in variables.
4367 // block - The block for which we're selecting a predecesor.
4368 // prevBlock - The previous block in in allocation order.
4369 // pPredBlockIsAllocated - A debug-only argument that indicates whether any of the predecessors have been seen
4370 // in allocation order.
4373 // The selected predecessor.
4376 // in DEBUG, caller initializes *pPredBlockIsAllocated to false, and it will be set to true if the block
4377 // returned is in fact a predecessor.
4380 // This will select a predecessor based on the heuristics obtained by getLsraBlockBoundaryLocations(), which can be
4382 // LSRA_BLOCK_BOUNDARY_PRED - Use the register locations of a predecessor block (default)
4383 // LSRA_BLOCK_BOUNDARY_LAYOUT - Use the register locations of the previous block in layout order.
4384 // This is the only case where this actually returns a different block.
4385 // LSRA_BLOCK_BOUNDARY_ROTATE - Rotate the register locations from a predecessor.
4386 // For this case, the block returned is the same as for LSRA_BLOCK_BOUNDARY_PRED, but
4387 // the register locations will be "rotated" to stress the resolution and allocation
4390 BasicBlock* LinearScan::findPredBlockForLiveIn(BasicBlock* block,
4391 BasicBlock* prevBlock DEBUGARG(bool* pPredBlockIsAllocated))
4393 BasicBlock* predBlock = nullptr;
4395 assert(*pPredBlockIsAllocated == false);
4396 if (getLsraBlockBoundaryLocations() == LSRA_BLOCK_BOUNDARY_LAYOUT)
4398 if (prevBlock != nullptr)
4400 predBlock = prevBlock;
4405 if (block != compiler->fgFirstBB)
4407 predBlock = block->GetUniquePred(compiler);
4408 if (predBlock != nullptr)
4410 if (isBlockVisited(predBlock))
4412 if (predBlock->bbJumpKind == BBJ_COND)
4414 // Special handling to improve matching on backedges.
4415 BasicBlock* otherBlock = (block == predBlock->bbNext) ? predBlock->bbJumpDest : predBlock->bbNext;
4416 noway_assert(otherBlock != nullptr);
4417 if (isBlockVisited(otherBlock))
4419 // This is the case when we have a conditional branch where one target has already
4420 // been visited. It would be best to use the same incoming regs as that block,
4421 // so that we have less likelihood of having to move registers.
4422 // For example, in determining the block to use for the starting register locations for
4423 // "block" in the following example, we'd like to use the same predecessor for "block"
4424 // as for "otherBlock", so that both successors of predBlock have the same locations, reducing
4425 // the likelihood of needing a split block on a backedge:
4436 for (flowList* pred = otherBlock->bbPreds; pred != nullptr; pred = pred->flNext)
4438 BasicBlock* otherPred = pred->flBlock;
4439 if (otherPred->bbNum == blockInfo[otherBlock->bbNum].predBBNum)
4441 predBlock = otherPred;
4450 predBlock = nullptr;
4455 for (flowList* pred = block->bbPreds; pred != nullptr; pred = pred->flNext)
4457 BasicBlock* candidatePredBlock = pred->flBlock;
4458 if (isBlockVisited(candidatePredBlock))
4460 if (predBlock == nullptr || predBlock->bbWeight < candidatePredBlock->bbWeight)
4462 predBlock = candidatePredBlock;
4463 INDEBUG(*pPredBlockIsAllocated = true;)
4468 if (predBlock == nullptr)
4470 predBlock = prevBlock;
4471 assert(predBlock != nullptr);
4472 JITDUMP("\n\nNo allocated predecessor; ");
4478 void LinearScan::buildIntervals()
4482 // start numbering at 1; 0 is the entry
4483 LsraLocation currentLoc = 1;
4485 JITDUMP("\nbuildIntervals ========\n");
4487 // Now build (empty) records for all of the physical registers
4488 buildPhysRegRecords();
4493 printf("\n-----------------\n");
4494 printf("LIVENESS:\n");
4495 printf("-----------------\n");
4496 foreach_block(compiler, block)
4498 printf("BB%02u use def in out\n", block->bbNum);
4499 dumpConvertedVarSet(compiler, block->bbVarUse);
4501 dumpConvertedVarSet(compiler, block->bbVarDef);
4503 dumpConvertedVarSet(compiler, block->bbLiveIn);
4505 dumpConvertedVarSet(compiler, block->bbLiveOut);
4512 // We will determine whether we should double align the frame during
4513 // identifyCandidates(), but we initially assume that we will not.
4514 doDoubleAlign = false;
4517 identifyCandidates();
4519 // Figure out if we're going to use a frame pointer. We need to do this before building
4520 // the ref positions, because those objects will embed the frame register in various register masks
4521 // if the frame pointer is not reserved. If we decide to have a frame pointer, setFrameType() will
4522 // remove the frame pointer from the masks.
4525 DBEXEC(VERBOSE, TupleStyleDump(LSRA_DUMP_PRE));
4528 JITDUMP("\nbuildIntervals second part ========\n");
4531 // Next, create ParamDef RefPositions for all the tracked parameters,
4532 // in order of their varIndex
4535 unsigned int lclNum;
4537 RegState* intRegState = &compiler->codeGen->intRegState;
4538 RegState* floatRegState = &compiler->codeGen->floatRegState;
4539 intRegState->rsCalleeRegArgMaskLiveIn = RBM_NONE;
4540 floatRegState->rsCalleeRegArgMaskLiveIn = RBM_NONE;
4542 for (unsigned int varIndex = 0; varIndex < compiler->lvaTrackedCount; varIndex++)
4544 lclNum = compiler->lvaTrackedToVarNum[varIndex];
4545 argDsc = &(compiler->lvaTable[lclNum]);
4547 if (!argDsc->lvIsParam)
4552 // Only reserve a register if the argument is actually used.
4553 // Is it dead on entry? If compJmpOpUsed is true, then the arguments
4554 // have to be kept alive, so we have to consider it as live on entry.
4555 // Use lvRefCnt instead of checking bbLiveIn because if it's volatile we
4556 // won't have done dataflow on it, but it needs to be marked as live-in so
4557 // it will get saved in the prolog.
4558 if (!compiler->compJmpOpUsed && argDsc->lvRefCnt == 0 && !compiler->opts.compDbgCode)
4563 if (argDsc->lvIsRegArg)
4565 updateRegStateForArg(argDsc);
4568 if (isCandidateVar(argDsc))
4570 Interval* interval = getIntervalForLocalVar(varIndex);
4571 regMaskTP mask = allRegs(TypeGet(argDsc));
4572 if (argDsc->lvIsRegArg)
4574 // Set this interval as currently assigned to that register
4575 regNumber inArgReg = argDsc->lvArgReg;
4576 assert(inArgReg < REG_COUNT);
4577 mask = genRegMask(inArgReg);
4578 assignPhysReg(inArgReg, interval);
4580 RefPosition* pos = newRefPosition(interval, MinLocation, RefTypeParamDef, nullptr, mask);
4582 else if (varTypeIsStruct(argDsc->lvType))
4584 for (unsigned fieldVarNum = argDsc->lvFieldLclStart;
4585 fieldVarNum < argDsc->lvFieldLclStart + argDsc->lvFieldCnt; ++fieldVarNum)
4587 LclVarDsc* fieldVarDsc = &(compiler->lvaTable[fieldVarNum]);
4588 if (fieldVarDsc->lvLRACandidate)
4590 assert(fieldVarDsc->lvTracked);
4591 Interval* interval = getIntervalForLocalVar(fieldVarDsc->lvVarIndex);
4593 newRefPosition(interval, MinLocation, RefTypeParamDef, nullptr, allRegs(TypeGet(fieldVarDsc)));
4599 // We can overwrite the register (i.e. codegen saves it on entry)
4600 assert(argDsc->lvRefCnt == 0 || !argDsc->lvIsRegArg || argDsc->lvDoNotEnregister ||
4601 !argDsc->lvLRACandidate || (varTypeIsFloating(argDsc->TypeGet()) && compiler->opts.compDbgCode));
4605 // Now set up the reg state for the non-tracked args
4606 // (We do this here because we want to generate the ParamDef RefPositions in tracked
4607 // order, so that loop doesn't hit the non-tracked args)
4609 for (unsigned argNum = 0; argNum < compiler->info.compArgsCount; argNum++, argDsc++)
4611 argDsc = &(compiler->lvaTable[argNum]);
4613 if (argDsc->lvPromotedStruct())
4615 noway_assert(argDsc->lvFieldCnt == 1); // We only handle one field here
4617 unsigned fieldVarNum = argDsc->lvFieldLclStart;
4618 argDsc = &(compiler->lvaTable[fieldVarNum]);
4620 noway_assert(argDsc->lvIsParam);
4621 if (!argDsc->lvTracked && argDsc->lvIsRegArg)
4623 updateRegStateForArg(argDsc);
4627 // If there is a secret stub param, it is also live in
4628 if (compiler->info.compPublishStubParam)
4630 intRegState->rsCalleeRegArgMaskLiveIn |= RBM_SECRET_STUB_PARAM;
4633 LocationInfoListNodePool listNodePool(compiler, 8);
4634 SmallHashTable<GenTree*, LocationInfoList, 32> operandToLocationInfoMap(compiler);
4636 BasicBlock* predBlock = nullptr;
4637 BasicBlock* prevBlock = nullptr;
4639 // Initialize currentLiveVars to the empty set. We will set it to the current
4640 // live-in at the entry to each block (this will include the incoming args on
4641 // the first block).
4642 VarSetOps::AssignNoCopy(compiler, currentLiveVars, VarSetOps::MakeEmpty(compiler));
4644 for (block = startBlockSequence(); block != nullptr; block = moveToNextBlock())
4646 JITDUMP("\nNEW BLOCK BB%02u\n", block->bbNum);
4648 bool predBlockIsAllocated = false;
4649 predBlock = findPredBlockForLiveIn(block, prevBlock DEBUGARG(&predBlockIsAllocated));
4652 JITDUMP("\n\nSetting BB%02u as the predecessor for determining incoming variable registers of BB%02u\n",
4653 block->bbNum, predBlock->bbNum);
4654 assert(predBlock->bbNum <= bbNumMaxBeforeResolution);
4655 blockInfo[block->bbNum].predBBNum = predBlock->bbNum;
4658 if (enregisterLocalVars)
4660 VarSetOps::AssignNoCopy(compiler, currentLiveVars,
4661 VarSetOps::Intersection(compiler, registerCandidateVars, block->bbLiveIn));
4663 if (block == compiler->fgFirstBB)
4665 insertZeroInitRefPositions();
4668 // Any lclVars live-in to a block are resolution candidates.
4669 VarSetOps::UnionD(compiler, resolutionCandidateVars, currentLiveVars);
4671 // Determine if we need any DummyDefs.
4672 // We need DummyDefs for cases where "predBlock" isn't really a predecessor.
4673 // Note that it's possible to have uses of unitialized variables, in which case even the first
4674 // block may require DummyDefs, which we are not currently adding - this means that these variables
4675 // will always be considered to be in memory on entry (and reloaded when the use is encountered).
4676 // TODO-CQ: Consider how best to tune this. Currently, if we create DummyDefs for uninitialized
4677 // variables (which may actually be initialized along the dynamically executed paths, but not
4678 // on all static paths), we wind up with excessive liveranges for some of these variables.
4679 VARSET_TP newLiveIn(VarSetOps::MakeCopy(compiler, currentLiveVars));
4682 // Compute set difference: newLiveIn = currentLiveVars - predBlock->bbLiveOut
4683 VarSetOps::DiffD(compiler, newLiveIn, predBlock->bbLiveOut);
4685 bool needsDummyDefs = (!VarSetOps::IsEmpty(compiler, newLiveIn) && block != compiler->fgFirstBB);
4687 // Create dummy def RefPositions
4691 // If we are using locations from a predecessor, we should never require DummyDefs.
4692 assert(!predBlockIsAllocated);
4694 JITDUMP("Creating dummy definitions\n");
4695 VarSetOps::Iter iter(compiler, newLiveIn);
4696 unsigned varIndex = 0;
4697 while (iter.NextElem(&varIndex))
4699 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
4700 LclVarDsc* varDsc = compiler->lvaTable + varNum;
4701 // Add a dummyDef for any candidate vars that are in the "newLiveIn" set.
4702 // If this is the entry block, don't add any incoming parameters (they're handled with ParamDefs).
4703 if (isCandidateVar(varDsc) && (predBlock != nullptr || !varDsc->lvIsParam))
4705 Interval* interval = getIntervalForLocalVar(varIndex);
4706 RefPosition* pos = newRefPosition(interval, currentLoc, RefTypeDummyDef, nullptr,
4707 allRegs(interval->registerType));
4710 JITDUMP("Finished creating dummy definitions\n\n");
4714 // Add a dummy RefPosition to mark the block boundary.
4715 // Note that we do this AFTER adding the exposed uses above, because the
4716 // register positions for those exposed uses need to be recorded at
4719 RefPosition* pos = newRefPosition((Interval*)nullptr, currentLoc, RefTypeBB, nullptr, RBM_NONE);
4722 LIR::Range& blockRange = LIR::AsRange(block);
4723 for (GenTree* node : blockRange.NonPhiNodes())
4725 assert(node->gtLsraInfo.loc >= currentLoc);
4726 assert(!node->IsValue() || !node->IsUnusedValue() || node->gtLsraInfo.isLocalDefUse);
4728 currentLoc = node->gtLsraInfo.loc;
4729 buildRefPositionsForNode(node, block, listNodePool, operandToLocationInfoMap, currentLoc);
4732 if (currentLoc > maxNodeLocation)
4734 maxNodeLocation = currentLoc;
4739 // Increment the LsraLocation at this point, so that the dummy RefPositions
4740 // will not have the same LsraLocation as any "real" RefPosition.
4743 // Note: the visited set is cleared in LinearScan::doLinearScan()
4744 markBlockVisited(block);
4746 if (enregisterLocalVars)
4748 // Insert exposed uses for a lclVar that is live-out of 'block' but not live-in to the
4749 // next block, or any unvisited successors.
4750 // This will address lclVars that are live on a backedge, as well as those that are kept
4751 // live at a GT_JMP.
4753 // Blocks ending with "jmp method" are marked as BBJ_HAS_JMP,
4754 // and jmp call is represented using GT_JMP node which is a leaf node.
4755 // Liveness phase keeps all the arguments of the method live till the end of
4756 // block by adding them to liveout set of the block containing GT_JMP.
4758 // The target of a GT_JMP implicitly uses all the current method arguments, however
4759 // there are no actual references to them. This can cause LSRA to assert, because
4760 // the variables are live but it sees no references. In order to correctly model the
4761 // liveness of these arguments, we add dummy exposed uses, in the same manner as for
4762 // backward branches. This will happen automatically via expUseSet.
4764 // Note that a block ending with GT_JMP has no successors and hence the variables
4765 // for which dummy use ref positions are added are arguments of the method.
4767 VARSET_TP expUseSet(VarSetOps::MakeCopy(compiler, block->bbLiveOut));
4768 VarSetOps::IntersectionD(compiler, expUseSet, registerCandidateVars);
4769 BasicBlock* nextBlock = getNextBlock();
4770 if (nextBlock != nullptr)
4772 VarSetOps::DiffD(compiler, expUseSet, nextBlock->bbLiveIn);
4774 for (BasicBlock* succ : block->GetAllSuccs(compiler))
4776 if (VarSetOps::IsEmpty(compiler, expUseSet))
4781 if (isBlockVisited(succ))
4785 VarSetOps::DiffD(compiler, expUseSet, succ->bbLiveIn);
4788 if (!VarSetOps::IsEmpty(compiler, expUseSet))
4790 JITDUMP("Exposed uses:");
4791 VarSetOps::Iter iter(compiler, expUseSet);
4792 unsigned varIndex = 0;
4793 while (iter.NextElem(&varIndex))
4795 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
4796 LclVarDsc* varDsc = compiler->lvaTable + varNum;
4797 assert(isCandidateVar(varDsc));
4798 Interval* interval = getIntervalForLocalVar(varIndex);
4800 newRefPosition(interval, currentLoc, RefTypeExpUse, nullptr, allRegs(interval->registerType));
4801 JITDUMP(" V%02u", varNum);
4806 // Clear the "last use" flag on any vars that are live-out from this block.
4808 VarSetOps::Iter iter(compiler, block->bbLiveOut);
4809 unsigned varIndex = 0;
4810 while (iter.NextElem(&varIndex))
4812 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
4813 LclVarDsc* const varDsc = &compiler->lvaTable[varNum];
4814 if (isCandidateVar(varDsc))
4816 RefPosition* const lastRP = getIntervalForLocalVar(varIndex)->lastRefPosition;
4817 if ((lastRP != nullptr) && (lastRP->bbNum == block->bbNum))
4819 lastRP->lastUse = false;
4826 checkLastUses(block);
4831 dumpConvertedVarSet(compiler, block->bbVarUse);
4833 dumpConvertedVarSet(compiler, block->bbVarDef);
4842 if (enregisterLocalVars)
4844 if (compiler->lvaKeepAliveAndReportThis())
4846 // If we need to KeepAliveAndReportThis, add a dummy exposed use of it at the end
4847 unsigned keepAliveVarNum = compiler->info.compThisArg;
4848 assert(compiler->info.compIsStatic == false);
4849 LclVarDsc* varDsc = compiler->lvaTable + keepAliveVarNum;
4850 if (isCandidateVar(varDsc))
4852 JITDUMP("Adding exposed use of this, for lvaKeepAliveAndReportThis\n");
4853 Interval* interval = getIntervalForLocalVar(varDsc->lvVarIndex);
4855 newRefPosition(interval, currentLoc, RefTypeExpUse, nullptr, allRegs(interval->registerType));
4860 if (getLsraExtendLifeTimes())
4863 for (lclNum = 0, varDsc = compiler->lvaTable; lclNum < compiler->lvaCount; lclNum++, varDsc++)
4865 if (varDsc->lvLRACandidate)
4867 JITDUMP("Adding exposed use of V%02u for LsraExtendLifetimes\n", lclNum);
4868 Interval* interval = getIntervalForLocalVar(varDsc->lvVarIndex);
4870 newRefPosition(interval, currentLoc, RefTypeExpUse, nullptr, allRegs(interval->registerType));
4877 // If the last block has successors, create a RefTypeBB to record
4880 if (prevBlock->NumSucc(compiler) > 0)
4882 RefPosition* pos = newRefPosition((Interval*)nullptr, currentLoc, RefTypeBB, nullptr, RBM_NONE);
4886 // Make sure we don't have any blocks that were not visited
4887 foreach_block(compiler, block)
4889 assert(isBlockVisited(block));
4894 lsraDumpIntervals("BEFORE VALIDATING INTERVALS");
4895 dumpRefPositions("BEFORE VALIDATING INTERVALS");
4896 validateIntervals();
4902 void LinearScan::dumpVarRefPositions(const char* title)
4904 if (enregisterLocalVars)
4906 printf("\nVAR REFPOSITIONS %s\n", title);
4908 for (unsigned i = 0; i < compiler->lvaCount; i++)
4910 printf("--- V%02u\n", i);
4912 LclVarDsc* varDsc = compiler->lvaTable + i;
4913 if (varDsc->lvIsRegCandidate())
4915 Interval* interval = getIntervalForLocalVar(varDsc->lvVarIndex);
4916 for (RefPosition* ref = interval->firstRefPosition; ref != nullptr; ref = ref->nextRefPosition)
4926 void LinearScan::validateIntervals()
4928 if (enregisterLocalVars)
4930 for (unsigned i = 0; i < compiler->lvaTrackedCount; i++)
4932 if (!compiler->lvaTable[compiler->lvaTrackedToVarNum[i]].lvLRACandidate)
4936 Interval* interval = getIntervalForLocalVar(i);
4938 bool defined = false;
4939 printf("-----------------\n");
4940 for (RefPosition* ref = interval->firstRefPosition; ref != nullptr; ref = ref->nextRefPosition)
4943 RefType refType = ref->refType;
4944 if (!defined && RefTypeIsUse(refType))
4946 if (compiler->info.compMethodName != nullptr)
4948 printf("%s: ", compiler->info.compMethodName);
4950 printf("LocalVar V%02u: undefined use at %u\n", interval->varNum, ref->nodeLocation);
4952 // Note that there can be multiple last uses if they are on disjoint paths,
4953 // so we can't really check the lastUse flag
4958 if (RefTypeIsDef(refType))
4968 // Set the default rpFrameType based upon codeGen->isFramePointerRequired()
4969 // This was lifted from the register predictor
4971 void LinearScan::setFrameType()
4973 FrameType frameType = FT_NOT_SET;
4975 compiler->codeGen->setDoubleAlign(false);
4978 frameType = FT_DOUBLE_ALIGN_FRAME;
4979 compiler->codeGen->setDoubleAlign(true);
4982 #endif // DOUBLE_ALIGN
4983 if (compiler->codeGen->isFramePointerRequired())
4985 frameType = FT_EBP_FRAME;
4989 if (compiler->rpMustCreateEBPCalled == false)
4994 compiler->rpMustCreateEBPCalled = true;
4995 if (compiler->rpMustCreateEBPFrame(INDEBUG(&reason)))
4997 JITDUMP("; Decided to create an EBP based frame for ETW stackwalking (%s)\n", reason);
4998 compiler->codeGen->setFrameRequired(true);
5002 if (compiler->codeGen->isFrameRequired())
5004 frameType = FT_EBP_FRAME;
5008 frameType = FT_ESP_FRAME;
5015 noway_assert(!compiler->codeGen->isFramePointerRequired());
5016 noway_assert(!compiler->codeGen->isFrameRequired());
5017 compiler->codeGen->setFramePointerUsed(false);
5020 compiler->codeGen->setFramePointerUsed(true);
5023 case FT_DOUBLE_ALIGN_FRAME:
5024 noway_assert(!compiler->codeGen->isFramePointerRequired());
5025 compiler->codeGen->setFramePointerUsed(false);
5027 #endif // DOUBLE_ALIGN
5029 noway_assert(!"rpFrameType not set correctly!");
5033 // If we are using FPBASE as the frame register, we cannot also use it for
5034 // a local var. Note that we may have already added it to the register masks,
5035 // which are computed when the LinearScan class constructor is created, and
5036 // used during lowering. Luckily, the TreeNodeInfo only stores an index to
5037 // the masks stored in the LinearScan class, so we only need to walk the
5038 // unique masks and remove FPBASE.
5039 if (frameType == FT_EBP_FRAME)
5041 if ((availableIntRegs & RBM_FPBASE) != 0)
5043 RemoveRegisterFromMasks(REG_FPBASE);
5045 // We know that we're already in "read mode" for availableIntRegs. However,
5046 // we need to remove the FPBASE register, so subsequent users (like callers
5047 // to allRegs()) get the right thing. The RemoveRegisterFromMasks() code
5048 // fixes up everything that already took a dependency on the value that was
5049 // previously read, so this completes the picture.
5050 availableIntRegs.OverrideAssign(availableIntRegs & ~RBM_FPBASE);
5054 compiler->rpFrameType = frameType;
5057 // Is the copyReg/moveReg given by this RefPosition still busy at the
5059 bool copyOrMoveRegInUse(RefPosition* ref, LsraLocation loc)
5061 assert(ref->copyReg || ref->moveReg);
5062 if (ref->getRefEndLocation() >= loc)
5066 Interval* interval = ref->getInterval();
5067 RefPosition* nextRef = interval->getNextRefPosition();
5068 if (nextRef != nullptr && nextRef->treeNode == ref->treeNode && nextRef->getRefEndLocation() >= loc)
5075 // Determine whether the register represented by "physRegRecord" is available at least
5076 // at the "currentLoc", and if so, return the next location at which it is in use in
5077 // "nextRefLocationPtr"
5079 bool LinearScan::registerIsAvailable(RegRecord* physRegRecord,
5080 LsraLocation currentLoc,
5081 LsraLocation* nextRefLocationPtr,
5082 RegisterType regType)
5084 *nextRefLocationPtr = MaxLocation;
5085 LsraLocation nextRefLocation = MaxLocation;
5086 regMaskTP regMask = genRegMask(physRegRecord->regNum);
5087 if (physRegRecord->isBusyUntilNextKill)
5092 RefPosition* nextPhysReference = physRegRecord->getNextRefPosition();
5093 if (nextPhysReference != nullptr)
5095 nextRefLocation = nextPhysReference->nodeLocation;
5096 // if (nextPhysReference->refType == RefTypeFixedReg) nextRefLocation--;
5098 else if (!physRegRecord->isCalleeSave)
5100 nextRefLocation = MaxLocation - 1;
5103 Interval* assignedInterval = physRegRecord->assignedInterval;
5105 if (assignedInterval != nullptr)
5107 RefPosition* recentReference = assignedInterval->recentRefPosition;
5109 // The only case where we have an assignedInterval, but recentReference is null
5110 // is where this interval is live at procedure entry (i.e. an arg register), in which
5111 // case it's still live and its assigned register is not available
5112 // (Note that the ParamDef will be recorded as a recentReference when we encounter
5113 // it, but we will be allocating registers, potentially to other incoming parameters,
5114 // as we process the ParamDefs.)
5116 if (recentReference == nullptr)
5121 // Is this a copyReg/moveReg? It is if the register assignment doesn't match.
5122 // (the recentReference may not be a copyReg/moveReg, because we could have seen another
5123 // reference since the copyReg/moveReg)
5125 if (!assignedInterval->isAssignedTo(physRegRecord->regNum))
5127 // Don't reassign it if it's still in use
5128 if ((recentReference->copyReg || recentReference->moveReg) &&
5129 copyOrMoveRegInUse(recentReference, currentLoc))
5134 else if (!assignedInterval->isActive && assignedInterval->isConstant)
5136 // Treat this as unassigned, i.e. do nothing.
5137 // TODO-CQ: Consider adjusting the heuristics (probably in the caller of this method)
5138 // to avoid reusing these registers.
5140 // If this interval isn't active, it's available if it isn't referenced
5141 // at this location (or the previous location, if the recent RefPosition
5142 // is a delayRegFree).
5143 else if (!assignedInterval->isActive &&
5144 (recentReference->refType == RefTypeExpUse || recentReference->getRefEndLocation() < currentLoc))
5146 // This interval must have a next reference (otherwise it wouldn't be assigned to this register)
5147 RefPosition* nextReference = recentReference->nextRefPosition;
5148 if (nextReference != nullptr)
5150 if (nextReference->nodeLocation < nextRefLocation)
5152 nextRefLocation = nextReference->nodeLocation;
5157 assert(recentReference->copyReg && recentReference->registerAssignment != regMask);
5165 if (nextRefLocation < *nextRefLocationPtr)
5167 *nextRefLocationPtr = nextRefLocation;
5171 if (regType == TYP_DOUBLE)
5173 // Recurse, but check the other half this time (TYP_FLOAT)
5174 if (!registerIsAvailable(getRegisterRecord(REG_NEXT(physRegRecord->regNum)), currentLoc, nextRefLocationPtr,
5177 nextRefLocation = *nextRefLocationPtr;
5179 #endif // _TARGET_ARM_
5181 return (nextRefLocation >= currentLoc);
5184 //------------------------------------------------------------------------
5185 // getRegisterType: Get the RegisterType to use for the given RefPosition
5188 // currentInterval: The interval for the current allocation
5189 // refPosition: The RefPosition of the current Interval for which a register is being allocated
5192 // The RegisterType that should be allocated for this RefPosition
5195 // This will nearly always be identical to the registerType of the interval, except in the case
5196 // of SIMD types of 8 bytes (currently only Vector2) when they are passed and returned in integer
5197 // registers, or copied to a return temp.
5198 // This method need only be called in situations where we may be dealing with the register requirements
5199 // of a RefTypeUse RefPosition (i.e. not when we are only looking at the type of an interval, nor when
5200 // we are interested in the "defining" type of the interval). This is because the situation of interest
5201 // only happens at the use (where it must be copied to an integer register).
5203 RegisterType LinearScan::getRegisterType(Interval* currentInterval, RefPosition* refPosition)
5205 assert(refPosition->getInterval() == currentInterval);
5206 RegisterType regType = currentInterval->registerType;
5207 regMaskTP candidates = refPosition->registerAssignment;
5209 assert((candidates & allRegs(regType)) != RBM_NONE);
5213 //------------------------------------------------------------------------
5214 // tryAllocateFreeReg: Find a free register that satisfies the requirements for refPosition,
5215 // and takes into account the preferences for the given Interval
5218 // currentInterval: The interval for the current allocation
5219 // refPosition: The RefPosition of the current Interval for which a register is being allocated
5222 // The regNumber, if any, allocated to the RefPositon. Returns REG_NA if no free register is found.
5225 // TODO-CQ: Consider whether we need to use a different order for tree temps than for vars, as
5228 static const regNumber lsraRegOrder[] = {REG_VAR_ORDER};
5229 const unsigned lsraRegOrderSize = ArrLen(lsraRegOrder);
5230 static const regNumber lsraRegOrderFlt[] = {REG_VAR_ORDER_FLT};
5231 const unsigned lsraRegOrderFltSize = ArrLen(lsraRegOrderFlt);
5233 regNumber LinearScan::tryAllocateFreeReg(Interval* currentInterval, RefPosition* refPosition)
5235 regNumber foundReg = REG_NA;
5237 RegisterType regType = getRegisterType(currentInterval, refPosition);
5238 const regNumber* regOrder;
5239 unsigned regOrderSize;
5240 if (useFloatReg(regType))
5242 regOrder = lsraRegOrderFlt;
5243 regOrderSize = lsraRegOrderFltSize;
5247 regOrder = lsraRegOrder;
5248 regOrderSize = lsraRegOrderSize;
5251 LsraLocation currentLocation = refPosition->nodeLocation;
5252 RefPosition* nextRefPos = refPosition->nextRefPosition;
5253 LsraLocation nextLocation = (nextRefPos == nullptr) ? currentLocation : nextRefPos->nodeLocation;
5254 regMaskTP candidates = refPosition->registerAssignment;
5255 regMaskTP preferences = currentInterval->registerPreferences;
5257 if (RefTypeIsDef(refPosition->refType))
5259 if (currentInterval->hasConflictingDefUse)
5261 resolveConflictingDefAndUse(currentInterval, refPosition);
5262 candidates = refPosition->registerAssignment;
5264 // Otherwise, check for the case of a fixed-reg def of a reg that will be killed before the
5265 // use, or interferes at the point of use (which shouldn't happen, but Lower doesn't mark
5266 // the contained nodes as interfering).
5267 // Note that we may have a ParamDef RefPosition that is marked isFixedRegRef, but which
5268 // has had its registerAssignment changed to no longer be a single register.
5269 else if (refPosition->isFixedRegRef && nextRefPos != nullptr && RefTypeIsUse(nextRefPos->refType) &&
5270 !nextRefPos->isFixedRegRef && genMaxOneBit(refPosition->registerAssignment))
5272 regNumber defReg = refPosition->assignedReg();
5273 RegRecord* defRegRecord = getRegisterRecord(defReg);
5275 RefPosition* currFixedRegRefPosition = defRegRecord->recentRefPosition;
5276 assert(currFixedRegRefPosition != nullptr &&
5277 currFixedRegRefPosition->nodeLocation == refPosition->nodeLocation);
5279 // If there is another fixed reference to this register before the use, change the candidates
5280 // on this RefPosition to include that of nextRefPos.
5281 if (currFixedRegRefPosition->nextRefPosition != nullptr &&
5282 currFixedRegRefPosition->nextRefPosition->nodeLocation <= nextRefPos->getRefEndLocation())
5284 candidates |= nextRefPos->registerAssignment;
5285 if (preferences == refPosition->registerAssignment)
5287 preferences = candidates;
5293 preferences &= candidates;
5294 if (preferences == RBM_NONE)
5296 preferences = candidates;
5298 regMaskTP relatedPreferences = RBM_NONE;
5301 candidates = stressLimitRegs(refPosition, candidates);
5303 bool mustAssignARegister = true;
5304 assert(candidates != RBM_NONE);
5306 // If the related interval has no further references, it is possible that it is a source of the
5307 // node that produces this interval. However, we don't want to use the relatedInterval for preferencing
5308 // if its next reference is not a new definition (as it either is or will become live).
5309 Interval* relatedInterval = currentInterval->relatedInterval;
5310 if (relatedInterval != nullptr)
5312 RefPosition* nextRelatedRefPosition = relatedInterval->getNextRefPosition();
5313 if (nextRelatedRefPosition != nullptr)
5315 // Don't use the relatedInterval for preferencing if its next reference is not a new definition.
5316 if (!RefTypeIsDef(nextRelatedRefPosition->refType))
5318 relatedInterval = nullptr;
5320 // Is the relatedInterval simply a copy to another relatedInterval?
5321 else if ((relatedInterval->relatedInterval != nullptr) &&
5322 (nextRelatedRefPosition->nextRefPosition != nullptr) &&
5323 (nextRelatedRefPosition->nextRefPosition->nextRefPosition == nullptr) &&
5324 (nextRelatedRefPosition->nextRefPosition->nodeLocation <
5325 relatedInterval->relatedInterval->getNextRefLocation()))
5327 // The current relatedInterval has only two remaining RefPositions, both of which
5328 // occur prior to the next RefPosition for its relatedInterval.
5329 // It is likely a copy.
5330 relatedInterval = relatedInterval->relatedInterval;
5335 if (relatedInterval != nullptr)
5337 // If the related interval already has an assigned register, then use that
5338 // as the related preference. We'll take the related
5339 // interval preferences into account in the loop over all the registers.
5341 if (relatedInterval->assignedReg != nullptr)
5343 relatedPreferences = genRegMask(relatedInterval->assignedReg->regNum);
5347 relatedPreferences = relatedInterval->registerPreferences;
5351 bool preferCalleeSave = currentInterval->preferCalleeSave;
5353 // For floating point, we want to be less aggressive about using callee-save registers.
5354 // So in that case, we just need to ensure that the current RefPosition is covered.
5355 RefPosition* rangeEndRefPosition;
5356 RefPosition* lastRefPosition = currentInterval->lastRefPosition;
5357 if (useFloatReg(currentInterval->registerType))
5359 rangeEndRefPosition = refPosition;
5363 rangeEndRefPosition = currentInterval->lastRefPosition;
5364 // If we have a relatedInterval that is not currently occupying a register,
5365 // and whose lifetime begins after this one ends,
5366 // we want to try to select a register that will cover its lifetime.
5367 if ((relatedInterval != nullptr) && (relatedInterval->assignedReg == nullptr) &&
5368 (relatedInterval->getNextRefLocation() >= rangeEndRefPosition->nodeLocation))
5370 lastRefPosition = relatedInterval->lastRefPosition;
5371 preferCalleeSave = relatedInterval->preferCalleeSave;
5375 // If this has a delayed use (due to being used in a rmw position of a
5376 // non-commutative operator), its endLocation is delayed until the "def"
5377 // position, which is one location past the use (getRefEndLocation() takes care of this).
5378 LsraLocation rangeEndLocation = rangeEndRefPosition->getRefEndLocation();
5379 LsraLocation lastLocation = lastRefPosition->getRefEndLocation();
5380 regNumber prevReg = REG_NA;
5382 if (currentInterval->assignedReg)
5384 bool useAssignedReg = false;
5385 // This was an interval that was previously allocated to the given
5386 // physical register, and we should try to allocate it to that register
5387 // again, if possible and reasonable.
5388 // Use it preemptively (i.e. before checking other available regs)
5389 // only if it is preferred and available.
5391 RegRecord* regRec = currentInterval->assignedReg;
5392 prevReg = regRec->regNum;
5393 regMaskTP prevRegBit = genRegMask(prevReg);
5395 // Is it in the preferred set of regs?
5396 if ((prevRegBit & preferences) != RBM_NONE)
5398 // Is it currently available?
5399 LsraLocation nextPhysRefLoc;
5400 if (registerIsAvailable(regRec, currentLocation, &nextPhysRefLoc, currentInterval->registerType))
5402 // If the register is next referenced at this location, only use it if
5403 // this has a fixed reg requirement (i.e. this is the reference that caused
5404 // the FixedReg ref to be created)
5406 if (!regRec->conflictingFixedRegReference(refPosition))
5408 useAssignedReg = true;
5414 regNumber foundReg = prevReg;
5415 assignPhysReg(regRec, currentInterval);
5416 refPosition->registerAssignment = genRegMask(foundReg);
5421 // Don't keep trying to allocate to this register
5422 currentInterval->assignedReg = nullptr;
5426 RegRecord* availablePhysRegInterval = nullptr;
5427 Interval* intervalToUnassign = nullptr;
5429 // Each register will receive a score which is the sum of the scoring criteria below.
5430 // These were selected on the assumption that they will have an impact on the "goodness"
5431 // of a register selection, and have been tuned to a certain extent by observing the impact
5432 // of the ordering on asmDiffs. However, there is probably much more room for tuning,
5433 // and perhaps additional criteria.
5435 // These are FLAGS (bits) so that we can easily order them and add them together.
5436 // If the scores are equal, but one covers more of the current interval's range,
5437 // then it wins. Otherwise, the one encountered earlier in the regOrder wins.
5441 VALUE_AVAILABLE = 0x40, // It is a constant value that is already in an acceptable register.
5442 COVERS = 0x20, // It is in the interval's preference set and it covers the entire lifetime.
5443 OWN_PREFERENCE = 0x10, // It is in the preference set of this interval.
5444 COVERS_RELATED = 0x08, // It is in the preference set of the related interval and covers the entire lifetime.
5445 RELATED_PREFERENCE = 0x04, // It is in the preference set of the related interval.
5446 CALLER_CALLEE = 0x02, // It is in the right "set" for the interval (caller or callee-save).
5447 UNASSIGNED = 0x01, // It is not currently assigned to an inactive interval.
5452 // Compute the best possible score so we can stop looping early if we find it.
5453 // TODO-Throughput: At some point we may want to short-circuit the computation of each score, but
5454 // probably not until we've tuned the order of these criteria. At that point,
5455 // we'll need to avoid the short-circuit if we've got a stress option to reverse
5457 int bestPossibleScore = COVERS + UNASSIGNED + OWN_PREFERENCE + CALLER_CALLEE;
5458 if (relatedPreferences != RBM_NONE)
5460 bestPossibleScore |= RELATED_PREFERENCE + COVERS_RELATED;
5463 LsraLocation bestLocation = MinLocation;
5465 // In non-debug builds, this will simply get optimized away
5466 bool reverseSelect = false;
5468 reverseSelect = doReverseSelect();
5471 // An optimization for the common case where there is only one candidate -
5472 // avoid looping over all the other registers
5474 regNumber singleReg = REG_NA;
5476 if (genMaxOneBit(candidates))
5479 singleReg = genRegNumFromMask(candidates);
5480 regOrder = &singleReg;
5483 for (unsigned i = 0; i < regOrderSize && (candidates != RBM_NONE); i++)
5485 regNumber regNum = regOrder[i];
5486 regMaskTP candidateBit = genRegMask(regNum);
5488 if (!(candidates & candidateBit))
5493 candidates &= ~candidateBit;
5495 RegRecord* physRegRecord = getRegisterRecord(regNum);
5498 LsraLocation nextPhysRefLocation = MaxLocation;
5500 // By chance, is this register already holding this interval, as a copyReg or having
5501 // been restored as inactive after a kill?
5502 if (physRegRecord->assignedInterval == currentInterval)
5504 availablePhysRegInterval = physRegRecord;
5505 intervalToUnassign = nullptr;
5509 // Find the next RefPosition of the physical register
5510 if (!registerIsAvailable(physRegRecord, currentLocation, &nextPhysRefLocation, regType))
5515 // If the register is next referenced at this location, only use it if
5516 // this has a fixed reg requirement (i.e. this is the reference that caused
5517 // the FixedReg ref to be created)
5519 if (physRegRecord->conflictingFixedRegReference(refPosition))
5524 // If this is a definition of a constant interval, check to see if its value is already in this register.
5525 if (currentInterval->isConstant && RefTypeIsDef(refPosition->refType) &&
5526 (physRegRecord->assignedInterval != nullptr) && physRegRecord->assignedInterval->isConstant)
5528 noway_assert(refPosition->treeNode != nullptr);
5529 GenTree* otherTreeNode = physRegRecord->assignedInterval->firstRefPosition->treeNode;
5530 noway_assert(otherTreeNode != nullptr);
5532 if (refPosition->treeNode->OperGet() == otherTreeNode->OperGet())
5534 switch (otherTreeNode->OperGet())
5537 if ((refPosition->treeNode->AsIntCon()->IconValue() ==
5538 otherTreeNode->AsIntCon()->IconValue()) &&
5539 (varTypeGCtype(refPosition->treeNode) == varTypeGCtype(otherTreeNode)))
5541 #ifdef _TARGET_64BIT_
5542 // If the constant is negative, only reuse registers of the same type.
5543 // This is because, on a 64-bit system, we do not sign-extend immediates in registers to
5544 // 64-bits unless they are actually longs, as this requires a longer instruction.
5545 // This doesn't apply to a 32-bit system, on which long values occupy multiple registers.
5546 // (We could sign-extend, but we would have to always sign-extend, because if we reuse more
5547 // than once, we won't have access to the instruction that originally defines the constant).
5548 if ((refPosition->treeNode->TypeGet() == otherTreeNode->TypeGet()) ||
5549 (refPosition->treeNode->AsIntCon()->IconValue() >= 0))
5550 #endif // _TARGET_64BIT_
5552 score |= VALUE_AVAILABLE;
5558 // For floating point constants, the values must be identical, not simply compare
5559 // equal. So we compare the bits.
5560 if (refPosition->treeNode->AsDblCon()->isBitwiseEqual(otherTreeNode->AsDblCon()) &&
5561 (refPosition->treeNode->TypeGet() == otherTreeNode->TypeGet()))
5563 score |= VALUE_AVAILABLE;
5568 // for all other 'otherTreeNode->OperGet()' kinds, we leave 'score' unchanged
5574 // If the nextPhysRefLocation is a fixedRef for the rangeEndRefPosition, increment it so that
5575 // we don't think it isn't covering the live range.
5576 // This doesn't handle the case where earlier RefPositions for this Interval are also
5577 // FixedRefs of this regNum, but at least those are only interesting in the case where those
5578 // are "local last uses" of the Interval - otherwise the liveRange would interfere with the reg.
5579 if (nextPhysRefLocation == rangeEndLocation && rangeEndRefPosition->isFixedRefOfReg(regNum))
5581 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_INCREMENT_RANGE_END, currentInterval, regNum));
5582 nextPhysRefLocation++;
5585 if ((candidateBit & preferences) != RBM_NONE)
5587 score |= OWN_PREFERENCE;
5588 if (nextPhysRefLocation > rangeEndLocation)
5593 if (relatedInterval != nullptr && (candidateBit & relatedPreferences) != RBM_NONE)
5595 score |= RELATED_PREFERENCE;
5596 if (nextPhysRefLocation > relatedInterval->lastRefPosition->nodeLocation)
5598 score |= COVERS_RELATED;
5602 // If we had a fixed-reg def of a reg that will be killed before the use, prefer it to any other registers
5603 // with the same score. (Note that we haven't changed the original registerAssignment on the RefPosition).
5604 // Overload the RELATED_PREFERENCE value.
5605 else if (candidateBit == refPosition->registerAssignment)
5607 score |= RELATED_PREFERENCE;
5610 if ((preferCalleeSave && physRegRecord->isCalleeSave) || (!preferCalleeSave && !physRegRecord->isCalleeSave))
5612 score |= CALLER_CALLEE;
5615 // The register is considered unassigned if it has no assignedInterval, OR
5616 // if its next reference is beyond the range of this interval.
5617 if (physRegRecord->assignedInterval == nullptr ||
5618 physRegRecord->assignedInterval->getNextRefLocation() > lastLocation)
5620 score |= UNASSIGNED;
5623 bool foundBetterCandidate = false;
5625 if (score > bestScore)
5627 foundBetterCandidate = true;
5629 else if (score == bestScore)
5631 // Prefer a register that covers the range.
5632 if (bestLocation <= lastLocation)
5634 if (nextPhysRefLocation > bestLocation)
5636 foundBetterCandidate = true;
5639 // If both cover the range, prefer a register that is killed sooner (leaving the longer range register
5640 // available). If both cover the range and also getting killed at the same location, prefer the one which
5641 // is same as previous assignment.
5642 else if (nextPhysRefLocation > lastLocation)
5644 if (nextPhysRefLocation < bestLocation)
5646 foundBetterCandidate = true;
5648 else if (nextPhysRefLocation == bestLocation && prevReg == regNum)
5650 foundBetterCandidate = true;
5656 if (doReverseSelect() && bestScore != 0)
5658 foundBetterCandidate = !foundBetterCandidate;
5662 if (foundBetterCandidate)
5664 bestLocation = nextPhysRefLocation;
5665 availablePhysRegInterval = physRegRecord;
5666 intervalToUnassign = physRegRecord->assignedInterval;
5670 // there is no way we can get a better score so break out
5671 if (!reverseSelect && score == bestPossibleScore && bestLocation == rangeEndLocation + 1)
5677 if (availablePhysRegInterval != nullptr)
5679 if (intervalToUnassign != nullptr)
5681 RegRecord* physRegToUnassign = availablePhysRegInterval;
5683 // We should unassign a double register if availablePhysRegInterval is part of the double register
5684 if (availablePhysRegInterval->assignedInterval->registerType == TYP_DOUBLE &&
5685 !genIsValidDoubleReg(availablePhysRegInterval->regNum))
5686 physRegToUnassign = findAnotherHalfRegRec(availablePhysRegInterval);
5688 unassignPhysReg(physRegToUnassign, intervalToUnassign->recentRefPosition);
5689 if (bestScore & VALUE_AVAILABLE)
5691 assert(intervalToUnassign->isConstant);
5692 refPosition->treeNode->SetReuseRegVal();
5694 // If we considered this "unassigned" because this interval's lifetime ends before
5695 // the next ref, remember it.
5696 else if ((bestScore & UNASSIGNED) != 0 && intervalToUnassign != nullptr)
5698 updatePreviousInterval(physRegToUnassign, intervalToUnassign, intervalToUnassign->registerType);
5703 assert((bestScore & VALUE_AVAILABLE) == 0);
5705 assignPhysReg(availablePhysRegInterval, currentInterval);
5706 foundReg = availablePhysRegInterval->regNum;
5707 regMaskTP foundRegMask = genRegMask(foundReg);
5708 refPosition->registerAssignment = foundRegMask;
5709 if (relatedInterval != nullptr)
5711 relatedInterval->updateRegisterPreferences(foundRegMask);
5717 //------------------------------------------------------------------------
5718 // canSpillReg: Determine whether we can spill physRegRecord
5721 // physRegRecord - reg to spill
5722 // refLocation - Location of RefPosition where this register will be spilled
5723 // recentAssignedRefWeight - Weight of recent assigned RefPosition which will be determined in this function
5724 // farthestRefPosWeight - Current farthestRefPosWeight at allocateBusyReg()
5727 // True - if we can spill physRegRecord
5728 // False - otherwise
5730 // Note: This helper is designed to be used only from allocateBusyReg() and canSpillDoubleReg()
5732 bool LinearScan::canSpillReg(RegRecord* physRegRecord,
5733 LsraLocation refLocation,
5734 unsigned* recentAssignedRefWeight,
5735 unsigned farthestRefPosWeight)
5737 assert(physRegRecord->assignedInterval != nullptr);
5738 RefPosition* recentAssignedRef = physRegRecord->assignedInterval->recentRefPosition;
5740 if (recentAssignedRef != nullptr)
5742 if (recentAssignedRef->nodeLocation == refLocation)
5744 // We can't spill a register that's being used at the current location
5748 // If the current position has the candidate register marked to be delayed,
5749 // check if the previous location is using this register, if that's the case we have to skip
5750 // since we can't spill this register.
5751 if (recentAssignedRef->delayRegFree && (refLocation == recentAssignedRef->nodeLocation + 1))
5756 // We don't prefer to spill a register if the weight of recentAssignedRef > weight
5757 // of the spill candidate found so far. We would consider spilling a greater weight
5758 // ref position only if the refPosition being allocated must need a reg.
5759 *recentAssignedRefWeight = getWeight(recentAssignedRef);
5760 if (*recentAssignedRefWeight > farthestRefPosWeight)
5769 bool LinearScan::canSpillDoubleReg(RegRecord* physRegRecord,
5770 LsraLocation refLocation,
5771 unsigned* recentAssignedRefWeight,
5772 unsigned farthestRefPosWeight)
5775 unsigned weight = BB_ZERO_WEIGHT;
5776 unsigned weight2 = BB_ZERO_WEIGHT;
5778 RegRecord* physRegRecord2 = findAnotherHalfRegRec(physRegRecord);
5780 if (physRegRecord->assignedInterval != nullptr)
5781 retVal &= canSpillReg(physRegRecord, refLocation, &weight, farthestRefPosWeight);
5783 if (physRegRecord2->assignedInterval != nullptr)
5784 retVal &= canSpillReg(physRegRecord2, refLocation, &weight2, farthestRefPosWeight);
5786 if (!(weight == BB_ZERO_WEIGHT && weight2 == BB_ZERO_WEIGHT))
5788 // weight and/or weight2 have been updated.
5789 *recentAssignedRefWeight = (weight > weight2) ? weight : weight2;
5796 //----------------------------------------------------------------------------
5797 // checkActiveInterval: Test activness of an interval
5798 // and check assertions if the interval is not active
5801 // interval - An interval to be tested
5802 // refLocation - Location where the interval is being tested
5805 // True - iff the interval is active
5806 // False - otherwise
5808 // Note: This helper is designed to be used only from checkActiveIntervals()
5810 bool LinearScan::checkActiveInterval(Interval* interval, LsraLocation refLocation)
5812 if (!interval->isActive)
5814 RefPosition* recentAssignedRef = interval->recentRefPosition;
5815 // Note that we may or may not have actually handled the reference yet, so it could either
5816 // be recentAssignedRef, or the next reference.
5817 assert(recentAssignedRef != nullptr);
5818 if (recentAssignedRef->nodeLocation != refLocation)
5820 if (recentAssignedRef->nodeLocation + 1 == refLocation)
5822 assert(recentAssignedRef->delayRegFree);
5826 RefPosition* nextAssignedRef = recentAssignedRef->nextRefPosition;
5827 assert(nextAssignedRef != nullptr);
5828 assert(nextAssignedRef->nodeLocation == refLocation ||
5829 (nextAssignedRef->nodeLocation + 1 == refLocation && nextAssignedRef->delayRegFree));
5837 //----------------------------------------------------------------------------------------
5838 // checkActiveIntervals: Test activness of a interval assinged to a register
5839 // and check assertions if the interval is not active.
5840 // We look into all intervals of two float registers consisting
5841 // a double regsiter for ARM32.
5844 // physRegRecord - A register
5845 // refLocation - Location where intervsl is being tested
5846 // registerType - Type of regsiter
5849 // True - iff the all intervals are active
5850 // False - otherwise
5852 // Note: This helper is designed to be used only from allocateBusyReg()
5854 bool LinearScan::checkActiveIntervals(RegRecord* physRegRecord, LsraLocation refLocation, RegisterType registerType)
5856 Interval* assignedInterval = physRegRecord->assignedInterval;
5859 // Check two intervals for a double register in ARM32
5860 Interval* assignedInterval2 = nullptr;
5861 if (registerType == TYP_DOUBLE)
5862 assignedInterval2 = findAnotherHalfRegRec(physRegRecord)->assignedInterval;
5864 // Both intervals should not be nullptr at the same time, becasue we already handle this case before.
5865 assert(!(assignedInterval == nullptr && assignedInterval2 == nullptr));
5867 if (assignedInterval != nullptr && !checkActiveInterval(assignedInterval, refLocation))
5870 if (assignedInterval2 != nullptr && !checkActiveInterval(assignedInterval2, refLocation))
5875 return checkActiveInterval(assignedInterval, refLocation);
5880 void LinearScan::unassignDoublePhysReg(RegRecord* doubleRegRecord)
5882 RegRecord* doubleRegRecordLo = doubleRegRecord;
5883 RegRecord* doubleRegRecordHi = findAnotherHalfRegRec(doubleRegRecordLo);
5884 // For a double register, we has following four cases.
5885 // Case 1: doubleRegRecLo is assigned to TYP_DOUBLE interval
5886 // Case 2: doubleRegRecLo and doubleRegRecHi are assigned to different TYP_FLOAT intervals
5887 // Case 3: doubelRegRecLo is assgined to TYP_FLOAT interval and doubleRegRecHi is nullptr
5888 // Case 4: doubleRegRecordLo is nullptr, and doubleRegRecordHi is assigned to a TYP_FLOAT interval
5889 if (doubleRegRecordLo->assignedInterval != nullptr)
5891 if (doubleRegRecordLo->assignedInterval->registerType == TYP_DOUBLE)
5893 // Case 1: doubleRegRecLo is assigned to TYP_DOUBLE interval
5894 unassignPhysReg(doubleRegRecordLo, doubleRegRecordLo->assignedInterval->recentRefPosition);
5898 // Case 2: doubleRegRecLo and doubleRegRecHi are assigned to different TYP_FLOAT intervals
5899 // Case 3: doubelRegRecLo is assgined to TYP_FLOAT interval and doubleRegRecHi is nullptr
5900 assert(doubleRegRecordLo->assignedInterval->registerType == TYP_FLOAT);
5901 unassignPhysReg(doubleRegRecordLo, doubleRegRecordLo->assignedInterval->recentRefPosition);
5903 if (doubleRegRecordHi != nullptr)
5905 assert(doubleRegRecordHi->assignedInterval->registerType == TYP_FLOAT);
5906 unassignPhysReg(doubleRegRecordHi, doubleRegRecordHi->assignedInterval->recentRefPosition);
5912 // Case 4: doubleRegRecordLo is nullptr, and doubleRegRecordHi is assigned to a TYP_FLOAT interval
5913 assert(doubleRegRecordHi->assignedInterval != nullptr);
5914 assert(doubleRegRecordHi->assignedInterval->registerType == TYP_FLOAT);
5915 unassignPhysReg(doubleRegRecordHi, doubleRegRecordHi->assignedInterval->recentRefPosition);
5919 #endif // _TARGET_ARM_
5921 //----------------------------------------------------------------------------------------
5922 // isRegInUse: Test whether regRec is being used at the refPosition
5925 // regRec - A register to be tested
5926 // refPosition - RefPosition where regRec is tested
5927 // nextLocation - next RefPosition of interval assigned to regRec
5930 // True - if regRec is beding used
5931 // False - otherwise
5933 // Note: This helper is designed to be used only from allocateBusyReg()
5935 bool LinearScan::isRegInUse(RegRecord* regRec, RefPosition* refPosition, LsraLocation* nextLocation)
5937 Interval* assignedInterval = regRec->assignedInterval;
5938 if (assignedInterval != nullptr)
5940 LsraLocation refLocation = refPosition->nodeLocation;
5941 RefPosition* nextRefPosition = assignedInterval->getNextRefPosition();
5942 *nextLocation = assignedInterval->getNextRefLocation();
5944 // We should never spill a register that's occupied by an Interval with its next use at the current
5946 // Normally this won't occur (unless we actually had more uses in a single node than there are registers),
5947 // because we'll always find something with a later nextLocation, but it can happen in stress when
5948 // we have LSRA_SELECT_NEAREST.
5949 if ((*nextLocation == refLocation) && !refPosition->isFixedRegRef && nextRefPosition->RequiresRegister())
5957 //------------------------------------------------------------------------
5958 // allocateBusyReg: Find a busy register that satisfies the requirements for refPosition,
5959 // and that can be spilled.
5962 // current The interval for the current allocation
5963 // refPosition The RefPosition of the current Interval for which a register is being allocated
5964 // allocateIfProfitable If true, a reg may not be allocated if all other ref positions currently
5965 // occupying registers are more important than the 'refPosition'.
5968 // The regNumber allocated to the RefPositon. Returns REG_NA if no free register is found.
5970 // Note: Currently this routine uses weight and farthest distance of next reference
5971 // to select a ref position for spilling.
5972 // a) if allocateIfProfitable = false
5973 // The ref position chosen for spilling will be the lowest weight
5974 // of all and if there is is more than one ref position with the
5975 // same lowest weight, among them choses the one with farthest
5976 // distance to its next reference.
5978 // b) if allocateIfProfitable = true
5979 // The ref position chosen for spilling will not only be lowest weight
5980 // of all but also has a weight lower than 'refPosition'. If there is
5981 // no such ref position, reg will not be allocated.
5982 regNumber LinearScan::allocateBusyReg(Interval* current, RefPosition* refPosition, bool allocateIfProfitable)
5984 regNumber foundReg = REG_NA;
5986 RegisterType regType = getRegisterType(current, refPosition);
5987 regMaskTP candidates = refPosition->registerAssignment;
5988 regMaskTP preferences = (current->registerPreferences & candidates);
5989 if (preferences == RBM_NONE)
5991 preferences = candidates;
5993 if (candidates == RBM_NONE)
5995 // This assumes only integer and floating point register types
5996 // if we target a processor with additional register types,
5997 // this would have to change
5998 candidates = allRegs(regType);
6002 candidates = stressLimitRegs(refPosition, candidates);
6005 // TODO-CQ: Determine whether/how to take preferences into account in addition to
6006 // prefering the one with the furthest ref position when considering
6007 // a candidate to spill
6008 RegRecord* farthestRefPhysRegRecord = nullptr;
6010 RegRecord* farthestRefPhysRegRecord2 = nullptr;
6012 LsraLocation farthestLocation = MinLocation;
6013 LsraLocation refLocation = refPosition->nodeLocation;
6014 unsigned farthestRefPosWeight;
6015 if (allocateIfProfitable)
6017 // If allocating a reg is optional, we will consider those ref positions
6018 // whose weight is less than 'refPosition' for spilling.
6019 farthestRefPosWeight = getWeight(refPosition);
6023 // If allocating a reg is a must, we start off with max weight so
6024 // that the first spill candidate will be selected based on
6025 // farthest distance alone. Since we start off with farthestLocation
6026 // initialized to MinLocation, the first available ref position
6027 // will be selected as spill candidate and its weight as the
6028 // fathestRefPosWeight.
6029 farthestRefPosWeight = BB_MAX_WEIGHT;
6032 for (regNumber regNum : Registers(regType))
6034 regMaskTP candidateBit = genRegMask(regNum);
6035 if (!(candidates & candidateBit))
6039 RegRecord* physRegRecord = getRegisterRecord(regNum);
6041 RegRecord* physRegRecord2 = nullptr;
6042 // For ARM32, let's consider two float registers consisting a double reg together,
6043 // when allocaing a double register.
6044 if (current->registerType == TYP_DOUBLE)
6046 assert(genIsValidDoubleReg(regNum));
6047 physRegRecord2 = findAnotherHalfRegRec(physRegRecord);
6051 if (physRegRecord->isBusyUntilNextKill)
6055 Interval* assignedInterval = physRegRecord->assignedInterval;
6057 Interval* assignedInterval2 = (physRegRecord2 == nullptr) ? nullptr : physRegRecord2->assignedInterval;
6060 // If there is a fixed reference at the same location (and it's not due to this reference),
6063 if (physRegRecord->conflictingFixedRegReference(refPosition))
6065 assert(candidates != candidateBit);
6069 LsraLocation physRegNextLocation = MaxLocation;
6070 if (refPosition->isFixedRefOfRegMask(candidateBit))
6072 // Either there is a fixed reference due to this node, or one associated with a
6073 // fixed use fed by a def at this node.
6074 // In either case, we must use this register as it's the only candidate
6075 // TODO-CQ: At the time we allocate a register to a fixed-reg def, if it's not going
6076 // to remain live until the use, we should set the candidates to allRegs(regType)
6077 // to avoid a spill - codegen can then insert the copy.
6078 assert(candidates == candidateBit);
6080 // If a refPosition has a fixed reg as its candidate and is also marked
6081 // as allocateIfProfitable, we should allocate fixed reg only if the
6082 // weight of this ref position is greater than the weight of the ref
6083 // position to which fixed reg is assigned. Such a case would arise
6084 // on x86 under LSRA stress.
6085 if (!allocateIfProfitable)
6087 physRegNextLocation = MaxLocation;
6088 farthestRefPosWeight = BB_MAX_WEIGHT;
6093 physRegNextLocation = physRegRecord->getNextRefLocation();
6095 // If refPosition requires a fixed register, we should reject all others.
6096 // Otherwise, we will still evaluate all phyRegs though their next location is
6097 // not better than farthestLocation found so far.
6099 // TODO: this method should be using an approach similar to tryAllocateFreeReg()
6100 // where it uses a regOrder array to avoid iterating over any but the single
6102 if (refPosition->isFixedRegRef && physRegNextLocation < farthestLocation)
6108 // If this register is not assigned to an interval, either
6109 // - it has a FixedReg reference at the current location that is not this reference, OR
6110 // - this is the special case of a fixed loReg, where this interval has a use at the same location
6111 // In either case, we cannot use it
6112 CLANG_FORMAT_COMMENT_ANCHOR;
6115 if (assignedInterval == nullptr && assignedInterval2 == nullptr)
6117 if (assignedInterval == nullptr)
6120 RefPosition* nextPhysRegPosition = physRegRecord->getNextRefPosition();
6122 #ifndef _TARGET_ARM64_
6123 // TODO-Cleanup: Revisit this after Issue #3524 is complete
6124 // On ARM64 the nodeLocation is not always == refLocation, Disabling this assert for now.
6125 assert(nextPhysRegPosition->nodeLocation == refLocation && candidateBit != candidates);
6131 RefPosition* recentAssignedRef = (assignedInterval == nullptr) ? nullptr : assignedInterval->recentRefPosition;
6132 RefPosition* recentAssignedRef2 =
6133 (assignedInterval2 == nullptr) ? nullptr : assignedInterval2->recentRefPosition;
6135 RefPosition* recentAssignedRef = assignedInterval->recentRefPosition;
6138 if (!checkActiveIntervals(physRegRecord, refLocation, current->registerType))
6143 // If we have a recentAssignedRef, check that it is going to be OK to spill it
6145 // TODO-Review: Under what conditions recentAssginedRef would be null?
6146 unsigned recentAssignedRefWeight = BB_ZERO_WEIGHT;
6149 if (current->registerType == TYP_DOUBLE)
6151 if (!canSpillDoubleReg(physRegRecord, refLocation, &recentAssignedRefWeight, farthestRefPosWeight))
6156 // This if-stmt is associated with the above else
6157 if (!canSpillReg(physRegRecord, refLocation, &recentAssignedRefWeight, farthestRefPosWeight))
6162 LsraLocation nextLocation = MinLocation;
6164 if (isRegInUse(physRegRecord, refPosition, &nextLocation))
6170 if (current->registerType == TYP_DOUBLE)
6172 LsraLocation nextLocation2 = MinLocation;
6173 if (isRegInUse(physRegRecord2, refPosition, &nextLocation2))
6177 nextLocation = (nextLocation > nextLocation2) ? nextLocation : nextLocation2;
6181 if (nextLocation > physRegNextLocation)
6183 nextLocation = physRegNextLocation;
6186 bool isBetterLocation;
6189 if (doSelectNearest() && farthestRefPhysRegRecord != nullptr)
6191 isBetterLocation = (nextLocation <= farthestLocation);
6195 // This if-stmt is associated with the above else
6196 if (recentAssignedRefWeight < farthestRefPosWeight)
6198 isBetterLocation = true;
6202 // This would mean the weight of spill ref position we found so far is equal
6203 // to the weight of the ref position that is being evaluated. In this case
6204 // we prefer to spill ref position whose distance to its next reference is
6206 assert(recentAssignedRefWeight == farthestRefPosWeight);
6208 // If allocateIfProfitable=true, the first spill candidate selected
6209 // will be based on weight alone. After we have found a spill
6210 // candidate whose weight is less than the 'refPosition', we will
6211 // consider farthest distance when there is a tie in weights.
6212 // This is to ensure that we don't spill a ref position whose
6213 // weight is equal to weight of 'refPosition'.
6214 if (allocateIfProfitable && farthestRefPhysRegRecord == nullptr)
6216 isBetterLocation = false;
6220 isBetterLocation = (nextLocation > farthestLocation);
6222 if (nextLocation > farthestLocation)
6224 isBetterLocation = true;
6226 else if (nextLocation == farthestLocation)
6228 // Both weight and distance are equal.
6229 // Prefer that ref position which is marked both reload and
6230 // allocate if profitable. These ref positions don't need
6231 // need to be spilled as they are already in memory and
6232 // codegen considers them as contained memory operands.
6233 CLANG_FORMAT_COMMENT_ANCHOR;
6235 // TODO-CQ-ARM: Just conservatively "and" two condition. We may implement better condision later.
6236 isBetterLocation = true;
6237 if (recentAssignedRef != nullptr)
6238 isBetterLocation &= (recentAssignedRef->reload && recentAssignedRef->AllocateIfProfitable());
6240 if (recentAssignedRef2 != nullptr)
6241 isBetterLocation &= (recentAssignedRef2->reload && recentAssignedRef2->AllocateIfProfitable());
6243 isBetterLocation = (recentAssignedRef != nullptr) && recentAssignedRef->reload &&
6244 recentAssignedRef->AllocateIfProfitable();
6249 isBetterLocation = false;
6254 if (isBetterLocation)
6256 farthestLocation = nextLocation;
6257 farthestRefPhysRegRecord = physRegRecord;
6259 farthestRefPhysRegRecord2 = physRegRecord2;
6261 farthestRefPosWeight = recentAssignedRefWeight;
6266 if (allocateIfProfitable)
6268 // There may not be a spill candidate or if one is found
6269 // its weight must be less than the weight of 'refPosition'
6270 assert((farthestRefPhysRegRecord == nullptr) || (farthestRefPosWeight < getWeight(refPosition)));
6274 // Must have found a spill candidate.
6275 assert(farthestRefPhysRegRecord != nullptr);
6276 if ((farthestLocation == refLocation) && !refPosition->isFixedRegRef)
6279 Interval* assignedInterval =
6280 (farthestRefPhysRegRecord == nullptr) ? nullptr : farthestRefPhysRegRecord->assignedInterval;
6281 Interval* assignedInterval2 =
6282 (farthestRefPhysRegRecord2 == nullptr) ? nullptr : farthestRefPhysRegRecord2->assignedInterval;
6283 RefPosition* nextRefPosition =
6284 (assignedInterval == nullptr) ? nullptr : assignedInterval->getNextRefPosition();
6285 RefPosition* nextRefPosition2 =
6286 (assignedInterval2 == nullptr) ? nullptr : assignedInterval2->getNextRefPosition();
6287 if (nextRefPosition != nullptr)
6289 if (nextRefPosition2 != nullptr)
6291 assert(!nextRefPosition->RequiresRegister() || !nextRefPosition2->RequiresRegister());
6295 assert(!nextRefPosition->RequiresRegister());
6300 assert(nextRefPosition2 != nullptr && !nextRefPosition2->RequiresRegister());
6302 #else // !_TARGET_ARM_
6303 Interval* assignedInterval = farthestRefPhysRegRecord->assignedInterval;
6304 RefPosition* nextRefPosition = assignedInterval->getNextRefPosition();
6305 assert(!nextRefPosition->RequiresRegister());
6306 #endif // !_TARGET_ARM_
6310 assert(farthestLocation > refLocation || refPosition->isFixedRegRef);
6315 if (farthestRefPhysRegRecord != nullptr)
6317 foundReg = farthestRefPhysRegRecord->regNum;
6320 if (current->registerType == TYP_DOUBLE)
6322 assert(genIsValidDoubleReg(foundReg));
6323 unassignDoublePhysReg(farthestRefPhysRegRecord);
6328 unassignPhysReg(farthestRefPhysRegRecord, farthestRefPhysRegRecord->assignedInterval->recentRefPosition);
6331 assignPhysReg(farthestRefPhysRegRecord, current);
6332 refPosition->registerAssignment = genRegMask(foundReg);
6337 refPosition->registerAssignment = RBM_NONE;
6343 // Grab a register to use to copy and then immediately use.
6344 // This is called only for localVar intervals that already have a register
6345 // assignment that is not compatible with the current RefPosition.
6346 // This is not like regular assignment, because we don't want to change
6347 // any preferences or existing register assignments.
6348 // Prefer a free register that's got the earliest next use.
6349 // Otherwise, spill something with the farthest next use
6351 regNumber LinearScan::assignCopyReg(RefPosition* refPosition)
6353 Interval* currentInterval = refPosition->getInterval();
6354 assert(currentInterval != nullptr);
6355 assert(currentInterval->isActive);
6357 bool foundFreeReg = false;
6358 RegRecord* bestPhysReg = nullptr;
6359 LsraLocation bestLocation = MinLocation;
6360 regMaskTP candidates = refPosition->registerAssignment;
6362 // Save the relatedInterval, if any, so that it doesn't get modified during allocation.
6363 Interval* savedRelatedInterval = currentInterval->relatedInterval;
6364 currentInterval->relatedInterval = nullptr;
6366 // We don't want really want to change the default assignment,
6367 // so 1) pretend this isn't active, and 2) remember the old reg
6368 regNumber oldPhysReg = currentInterval->physReg;
6369 RegRecord* oldRegRecord = currentInterval->assignedReg;
6370 assert(oldRegRecord->regNum == oldPhysReg);
6371 currentInterval->isActive = false;
6373 regNumber allocatedReg = tryAllocateFreeReg(currentInterval, refPosition);
6374 if (allocatedReg == REG_NA)
6376 allocatedReg = allocateBusyReg(currentInterval, refPosition, false);
6379 // Now restore the old info
6380 currentInterval->relatedInterval = savedRelatedInterval;
6381 currentInterval->physReg = oldPhysReg;
6382 currentInterval->assignedReg = oldRegRecord;
6383 currentInterval->isActive = true;
6385 refPosition->copyReg = true;
6386 return allocatedReg;
6389 // Check if the interval is already assigned and if it is then unassign the physical record
6390 // then set the assignedInterval to 'interval'
6392 void LinearScan::checkAndAssignInterval(RegRecord* regRec, Interval* interval)
6394 if (regRec->assignedInterval != nullptr && regRec->assignedInterval != interval)
6396 // This is allocated to another interval. Either it is inactive, or it was allocated as a
6397 // copyReg and is therefore not the "assignedReg" of the other interval. In the latter case,
6398 // we simply unassign it - in the former case we need to set the physReg on the interval to
6399 // REG_NA to indicate that it is no longer in that register.
6400 // The lack of checking for this case resulted in an assert in the retail version of System.dll,
6401 // in method SerialStream.GetDcbFlag.
6402 // Note that we can't check for the copyReg case, because we may have seen a more recent
6403 // RefPosition for the Interval that was NOT a copyReg.
6404 if (regRec->assignedInterval->assignedReg == regRec)
6406 assert(regRec->assignedInterval->isActive == false);
6407 regRec->assignedInterval->physReg = REG_NA;
6409 unassignPhysReg(regRec->regNum);
6412 updateAssignedInterval(regRec, interval, interval->registerType);
6415 // Assign the given physical register interval to the given interval
6416 void LinearScan::assignPhysReg(RegRecord* regRec, Interval* interval)
6418 regMaskTP assignedRegMask = genRegMask(regRec->regNum);
6419 compiler->codeGen->regSet.rsSetRegsModified(assignedRegMask DEBUGARG(dumpTerse));
6421 checkAndAssignInterval(regRec, interval);
6422 interval->assignedReg = regRec;
6424 interval->physReg = regRec->regNum;
6425 interval->isActive = true;
6426 if (interval->isLocalVar)
6428 // Prefer this register for future references
6429 interval->updateRegisterPreferences(assignedRegMask);
6433 //------------------------------------------------------------------------
6434 // setIntervalAsSplit: Set this Interval as being split
6437 // interval - The Interval which is being split
6443 // The given Interval will be marked as split, and it will be added to the
6444 // set of splitOrSpilledVars.
6447 // "interval" must be a lclVar interval, as tree temps are never split.
6448 // This is asserted in the call to getVarIndex().
6450 void LinearScan::setIntervalAsSplit(Interval* interval)
6452 if (interval->isLocalVar)
6454 unsigned varIndex = interval->getVarIndex(compiler);
6455 if (!interval->isSplit)
6457 VarSetOps::AddElemD(compiler, splitOrSpilledVars, varIndex);
6461 assert(VarSetOps::IsMember(compiler, splitOrSpilledVars, varIndex));
6464 interval->isSplit = true;
6467 //------------------------------------------------------------------------
6468 // setIntervalAsSpilled: Set this Interval as being spilled
6471 // interval - The Interval which is being spilled
6477 // The given Interval will be marked as spilled, and it will be added
6478 // to the set of splitOrSpilledVars.
6480 void LinearScan::setIntervalAsSpilled(Interval* interval)
6482 if (interval->isLocalVar)
6484 unsigned varIndex = interval->getVarIndex(compiler);
6485 if (!interval->isSpilled)
6487 VarSetOps::AddElemD(compiler, splitOrSpilledVars, varIndex);
6491 assert(VarSetOps::IsMember(compiler, splitOrSpilledVars, varIndex));
6494 interval->isSpilled = true;
6497 //------------------------------------------------------------------------
6498 // spill: Spill this Interval between "fromRefPosition" and "toRefPosition"
6501 // fromRefPosition - The RefPosition at which the Interval is to be spilled
6502 // toRefPosition - The RefPosition at which it must be reloaded
6508 // fromRefPosition and toRefPosition must not be null
6510 void LinearScan::spillInterval(Interval* interval, RefPosition* fromRefPosition, RefPosition* toRefPosition)
6512 assert(fromRefPosition != nullptr && toRefPosition != nullptr);
6513 assert(fromRefPosition->getInterval() == interval && toRefPosition->getInterval() == interval);
6514 assert(fromRefPosition->nextRefPosition == toRefPosition);
6516 if (!fromRefPosition->lastUse)
6518 // If not allocated a register, Lcl var def/use ref positions even if reg optional
6519 // should be marked as spillAfter.
6520 if (!fromRefPosition->RequiresRegister() && !(interval->isLocalVar && fromRefPosition->IsActualRef()))
6522 fromRefPosition->registerAssignment = RBM_NONE;
6526 fromRefPosition->spillAfter = true;
6529 assert(toRefPosition != nullptr);
6534 dumpLsraAllocationEvent(LSRA_EVENT_SPILL, interval);
6538 INTRACK_STATS(updateLsraStat(LSRA_STAT_SPILL, fromRefPosition->bbNum));
6540 interval->isActive = false;
6541 setIntervalAsSpilled(interval);
6543 // If fromRefPosition occurs before the beginning of this block, mark this as living in the stack
6544 // on entry to this block.
6545 if (fromRefPosition->nodeLocation <= curBBStartLocation)
6547 // This must be a lclVar interval
6548 assert(interval->isLocalVar);
6549 setInVarRegForBB(curBBNum, interval->varNum, REG_STK);
6553 //------------------------------------------------------------------------
6554 // unassignPhysRegNoSpill: Unassign the given physical register record from
6555 // an active interval, without spilling.
6558 // regRec - the RegRecord to be unasssigned
6564 // The assignedInterval must not be null, and must be active.
6567 // This method is used to unassign a register when an interval needs to be moved to a
6568 // different register, but not (yet) spilled.
6570 void LinearScan::unassignPhysRegNoSpill(RegRecord* regRec)
6572 Interval* assignedInterval = regRec->assignedInterval;
6573 assert(assignedInterval != nullptr && assignedInterval->isActive);
6574 assignedInterval->isActive = false;
6575 unassignPhysReg(regRec, nullptr);
6576 assignedInterval->isActive = true;
6579 //------------------------------------------------------------------------
6580 // checkAndClearInterval: Clear the assignedInterval for the given
6581 // physical register record
6584 // regRec - the physical RegRecord to be unasssigned
6585 // spillRefPosition - The RefPosition at which the assignedInterval is to be spilled
6586 // or nullptr if we aren't spilling
6592 // see unassignPhysReg
6594 void LinearScan::checkAndClearInterval(RegRecord* regRec, RefPosition* spillRefPosition)
6596 Interval* assignedInterval = regRec->assignedInterval;
6597 assert(assignedInterval != nullptr);
6598 regNumber thisRegNum = regRec->regNum;
6600 if (spillRefPosition == nullptr)
6602 // Note that we can't assert for the copyReg case
6604 if (assignedInterval->physReg == thisRegNum)
6606 assert(assignedInterval->isActive == false);
6611 assert(spillRefPosition->getInterval() == assignedInterval);
6614 updateAssignedInterval(regRec, nullptr, assignedInterval->registerType);
6617 //------------------------------------------------------------------------
6618 // unassignPhysReg: Unassign the given physical register record, and spill the
6619 // assignedInterval at the given spillRefPosition, if any.
6622 // regRec - the RegRecord to be unasssigned
6623 // spillRefPosition - The RefPosition at which the assignedInterval is to be spilled
6629 // The assignedInterval must not be null.
6630 // If spillRefPosition is null, the assignedInterval must be inactive, or not currently
6631 // assigned to this register (e.g. this is a copyReg for that Interval).
6632 // Otherwise, spillRefPosition must be associated with the assignedInterval.
6634 void LinearScan::unassignPhysReg(RegRecord* regRec, RefPosition* spillRefPosition)
6636 Interval* assignedInterval = regRec->assignedInterval;
6637 assert(assignedInterval != nullptr);
6639 regNumber thisRegNum = regRec->regNum;
6642 regNumber nextRegNum = REG_NA;
6643 RegRecord* nextRegRec = nullptr;
6645 // Prepare second half RegRecord of a double register for TYP_DOUBLE
6646 if (assignedInterval->registerType == TYP_DOUBLE)
6648 assert(isFloatRegType(regRec->registerType));
6649 assert(genIsValidDoubleReg(regRec->regNum));
6651 nextRegNum = REG_NEXT(regRec->regNum);
6652 nextRegRec = getRegisterRecord(nextRegNum);
6654 // Both two RegRecords should have been assigned to the same interval.
6655 assert(assignedInterval == nextRegRec->assignedInterval);
6657 #endif // _TARGET_ARM_
6659 checkAndClearInterval(regRec, spillRefPosition);
6662 if (assignedInterval->registerType == TYP_DOUBLE)
6664 // Both two RegRecords should have been unassigned together.
6665 assert(regRec->assignedInterval == nullptr);
6666 assert(nextRegRec->assignedInterval == nullptr);
6668 #endif // _TARGET_ARM_
6671 if (VERBOSE && !dumpTerse)
6673 printf("unassigning %s: ", getRegName(regRec->regNum));
6674 assignedInterval->dump();
6679 RefPosition* nextRefPosition = nullptr;
6680 if (spillRefPosition != nullptr)
6682 nextRefPosition = spillRefPosition->nextRefPosition;
6685 if (assignedInterval->physReg != REG_NA && assignedInterval->physReg != thisRegNum)
6687 // This must have been a temporary copy reg, but we can't assert that because there
6688 // may have been intervening RefPositions that were not copyRegs.
6690 // reg->assignedInterval has already been set to nullptr by checkAndClearInterval()
6691 assert(regRec->assignedInterval == nullptr);
6695 regNumber victimAssignedReg = assignedInterval->physReg;
6696 assignedInterval->physReg = REG_NA;
6698 bool spill = assignedInterval->isActive && nextRefPosition != nullptr;
6701 // If this is an active interval, it must have a recentRefPosition,
6702 // otherwise it would not be active
6703 assert(spillRefPosition != nullptr);
6706 // TODO-CQ: Enable this and insert an explicit GT_COPY (otherwise there's no way to communicate
6707 // to codegen that we want the copyReg to be the new home location).
6708 // If the last reference was a copyReg, and we're spilling the register
6709 // it was copied from, then make the copyReg the new primary location
6711 if (spillRefPosition->copyReg)
6713 regNumber copyFromRegNum = victimAssignedReg;
6714 regNumber copyRegNum = genRegNumFromMask(spillRefPosition->registerAssignment);
6715 if (copyFromRegNum == thisRegNum &&
6716 getRegisterRecord(copyRegNum)->assignedInterval == assignedInterval)
6718 assert(copyRegNum != thisRegNum);
6719 assignedInterval->physReg = copyRegNum;
6720 assignedInterval->assignedReg = this->getRegisterRecord(copyRegNum);
6726 // With JitStressRegs == 0x80 (LSRA_EXTEND_LIFETIMES), we may have a RefPosition
6727 // that is not marked lastUse even though the treeNode is a lastUse. In that case
6728 // we must not mark it for spill because the register will have been immediately freed
6729 // after use. While we could conceivably add special handling for this case in codegen,
6730 // it would be messy and undesirably cause the "bleeding" of LSRA stress modes outside
6732 if (extendLifetimes() && assignedInterval->isLocalVar && RefTypeIsUse(spillRefPosition->refType) &&
6733 spillRefPosition->treeNode != nullptr && (spillRefPosition->treeNode->gtFlags & GTF_VAR_DEATH) != 0)
6735 dumpLsraAllocationEvent(LSRA_EVENT_SPILL_EXTENDED_LIFETIME, assignedInterval);
6736 assignedInterval->isActive = false;
6738 // If the spillRefPosition occurs before the beginning of this block, it will have
6739 // been marked as living in this register on entry to this block, but we now need
6740 // to mark this as living on the stack.
6741 if (spillRefPosition->nodeLocation <= curBBStartLocation)
6743 setInVarRegForBB(curBBNum, assignedInterval->varNum, REG_STK);
6744 if (spillRefPosition->nextRefPosition != nullptr)
6746 setIntervalAsSpilled(assignedInterval);
6751 // Otherwise, we need to mark spillRefPosition as lastUse, or the interval
6752 // will remain active beyond its allocated range during the resolution phase.
6753 spillRefPosition->lastUse = true;
6759 spillInterval(assignedInterval, spillRefPosition, nextRefPosition);
6762 // Maintain the association with the interval, if it has more references.
6763 // Or, if we "remembered" an interval assigned to this register, restore it.
6764 if (nextRefPosition != nullptr)
6766 assignedInterval->assignedReg = regRec;
6768 else if (canRestorePreviousInterval(regRec, assignedInterval))
6770 regRec->assignedInterval = regRec->previousInterval;
6771 regRec->previousInterval = nullptr;
6775 // We can not use updateAssignedInterval() and updatePreviousInterval() here,
6776 // because regRec may not be a even-numbered float register.
6778 // Update second half RegRecord of a double register for TYP_DOUBLE
6779 if (regRec->assignedInterval->registerType == TYP_DOUBLE)
6781 RegRecord* anotherHalfRegRec = findAnotherHalfRegRec(regRec);
6783 anotherHalfRegRec->assignedInterval = regRec->assignedInterval;
6784 anotherHalfRegRec->previousInterval = nullptr;
6786 #endif // _TARGET_ARM_
6791 dumpLsraAllocationEvent(LSRA_EVENT_RESTORE_PREVIOUS_INTERVAL_AFTER_SPILL, regRec->assignedInterval,
6796 dumpLsraAllocationEvent(LSRA_EVENT_RESTORE_PREVIOUS_INTERVAL, regRec->assignedInterval, thisRegNum);
6802 updateAssignedInterval(regRec, nullptr, assignedInterval->registerType);
6803 updatePreviousInterval(regRec, nullptr, assignedInterval->registerType);
6807 //------------------------------------------------------------------------
6808 // spillGCRefs: Spill any GC-type intervals that are currently in registers.a
6811 // killRefPosition - The RefPosition for the kill
6816 void LinearScan::spillGCRefs(RefPosition* killRefPosition)
6818 // For each physical register that can hold a GC type,
6819 // if it is occupied by an interval of a GC type, spill that interval.
6820 regMaskTP candidateRegs = killRefPosition->registerAssignment;
6821 while (candidateRegs != RBM_NONE)
6823 regMaskTP nextRegBit = genFindLowestBit(candidateRegs);
6824 candidateRegs &= ~nextRegBit;
6825 regNumber nextReg = genRegNumFromMask(nextRegBit);
6826 RegRecord* regRecord = getRegisterRecord(nextReg);
6827 Interval* assignedInterval = regRecord->assignedInterval;
6828 if (assignedInterval == nullptr || (assignedInterval->isActive == false) ||
6829 !varTypeIsGC(assignedInterval->registerType))
6833 unassignPhysReg(regRecord, assignedInterval->recentRefPosition);
6835 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_DONE_KILL_GC_REFS, nullptr, REG_NA, nullptr));
6838 //------------------------------------------------------------------------
6839 // processBlockEndAllocation: Update var locations after 'currentBlock' has been allocated
6842 // currentBlock - the BasicBlock we have just finished allocating registers for
6848 // Calls processBlockEndLocations() to set the outVarToRegMap, then gets the next block,
6849 // and sets the inVarToRegMap appropriately.
6851 void LinearScan::processBlockEndAllocation(BasicBlock* currentBlock)
6853 assert(currentBlock != nullptr);
6854 if (enregisterLocalVars)
6856 processBlockEndLocations(currentBlock);
6858 markBlockVisited(currentBlock);
6860 // Get the next block to allocate.
6861 // When the last block in the method has successors, there will be a final "RefTypeBB" to
6862 // ensure that we get the varToRegMap set appropriately, but in that case we don't need
6863 // to worry about "nextBlock".
6864 BasicBlock* nextBlock = getNextBlock();
6865 if (nextBlock != nullptr)
6867 processBlockStartLocations(nextBlock, true);
6871 //------------------------------------------------------------------------
6872 // rotateBlockStartLocation: When in the LSRA_BLOCK_BOUNDARY_ROTATE stress mode, attempt to
6873 // "rotate" the register assignment for a localVar to the next higher
6874 // register that is available.
6877 // interval - the Interval for the variable whose register is getting rotated
6878 // targetReg - its register assignment from the predecessor block being used for live-in
6879 // availableRegs - registers available for use
6882 // The new register to use.
6885 regNumber LinearScan::rotateBlockStartLocation(Interval* interval, regNumber targetReg, regMaskTP availableRegs)
6887 if (targetReg != REG_STK && getLsraBlockBoundaryLocations() == LSRA_BLOCK_BOUNDARY_ROTATE)
6889 // If we're rotating the register locations at block boundaries, try to use
6890 // the next higher register number of the appropriate register type.
6891 regMaskTP candidateRegs = allRegs(interval->registerType) & availableRegs;
6892 regNumber firstReg = REG_NA;
6893 regNumber newReg = REG_NA;
6894 while (candidateRegs != RBM_NONE)
6896 regMaskTP nextRegBit = genFindLowestBit(candidateRegs);
6897 candidateRegs &= ~nextRegBit;
6898 regNumber nextReg = genRegNumFromMask(nextRegBit);
6899 if (nextReg > targetReg)
6904 else if (firstReg == REG_NA)
6909 if (newReg == REG_NA)
6911 assert(firstReg != REG_NA);
6921 //--------------------------------------------------------------------------------------
6922 // isSecondHalfReg: Test if recRec is second half of double register
6923 // which is assigned to an interval.
6926 // regRec - a register to be tested
6927 // interval - an interval which is assigned to some register
6933 // True only if regRec is second half of assignedReg in interval
6935 bool LinearScan::isSecondHalfReg(RegRecord* regRec, Interval* interval)
6937 RegRecord* assignedReg = interval->assignedReg;
6939 if (assignedReg != nullptr && interval->registerType == TYP_DOUBLE)
6941 // interval should have been allocated to a valid double register
6942 assert(genIsValidDoubleReg(assignedReg->regNum));
6944 // Find a second half RegRecord of double register
6945 regNumber firstRegNum = assignedReg->regNum;
6946 regNumber secondRegNum = REG_NEXT(firstRegNum);
6948 assert(genIsValidFloatReg(secondRegNum) && !genIsValidDoubleReg(secondRegNum));
6950 RegRecord* secondRegRec = getRegisterRecord(secondRegNum);
6952 return secondRegRec == regRec;
6958 //------------------------------------------------------------------------------------------
6959 // findAnotherHalfRegRec: Find another half RegRecord which forms same ARM32 double register
6962 // regRec - A float RegRecord
6968 // A RegRecord which forms same double register with regRec
6970 RegRecord* LinearScan::findAnotherHalfRegRec(RegRecord* regRec)
6972 regNumber anotherHalfRegNum;
6973 RegRecord* anotherHalfRegRec;
6975 assert(genIsValidFloatReg(regRec->regNum));
6977 // Find another half register for TYP_DOUBLE interval,
6978 // following same logic in canRestorePreviousInterval().
6979 if (genIsValidDoubleReg(regRec->regNum))
6981 anotherHalfRegNum = REG_NEXT(regRec->regNum);
6982 assert(!genIsValidDoubleReg(anotherHalfRegNum));
6986 anotherHalfRegNum = REG_PREV(regRec->regNum);
6987 assert(genIsValidDoubleReg(anotherHalfRegNum));
6989 anotherHalfRegRec = getRegisterRecord(anotherHalfRegNum);
6991 return anotherHalfRegRec;
6995 //--------------------------------------------------------------------------------------
6996 // canRestorePreviousInterval: Test if we can restore previous interval
6999 // regRec - a register which contains previous interval to be restored
7000 // assignedInterval - an interval just unassigned
7006 // True only if previous interval of regRec can be restored
7008 bool LinearScan::canRestorePreviousInterval(RegRecord* regRec, Interval* assignedInterval)
7011 (regRec->previousInterval != nullptr && regRec->previousInterval != assignedInterval &&
7012 regRec->previousInterval->assignedReg == regRec && regRec->previousInterval->getNextRefPosition() != nullptr);
7015 if (retVal && regRec->previousInterval->registerType == TYP_DOUBLE)
7017 RegRecord* anotherHalfRegRec = findAnotherHalfRegRec(regRec);
7019 retVal = retVal && anotherHalfRegRec->assignedInterval == nullptr;
7026 bool LinearScan::isAssignedToInterval(Interval* interval, RegRecord* regRec)
7028 bool isAssigned = (interval->assignedReg == regRec);
7030 isAssigned |= isSecondHalfReg(regRec, interval);
7035 //------------------------------------------------------------------------
7036 // processBlockStartLocations: Update var locations on entry to 'currentBlock' and clear constant
7040 // currentBlock - the BasicBlock we are about to allocate registers for
7041 // allocationPass - true if we are currently allocating registers (versus writing them back)
7047 // During the allocation pass, we use the outVarToRegMap of the selected predecessor to
7048 // determine the lclVar locations for the inVarToRegMap.
7049 // During the resolution (write-back) pass, we only modify the inVarToRegMap in cases where
7050 // a lclVar was spilled after the block had been completed.
7051 void LinearScan::processBlockStartLocations(BasicBlock* currentBlock, bool allocationPass)
7053 // If we have no register candidates we should only call this method during allocation.
7055 assert(enregisterLocalVars || allocationPass);
7057 if (!enregisterLocalVars)
7059 // Just clear any constant registers and return.
7060 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
7062 RegRecord* physRegRecord = getRegisterRecord(reg);
7063 Interval* assignedInterval = physRegRecord->assignedInterval;
7065 if (assignedInterval != nullptr)
7067 assert(assignedInterval->isConstant);
7068 physRegRecord->assignedInterval = nullptr;
7071 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_START_BB, nullptr, REG_NA, currentBlock));
7075 unsigned predBBNum = blockInfo[currentBlock->bbNum].predBBNum;
7076 VarToRegMap predVarToRegMap = getOutVarToRegMap(predBBNum);
7077 VarToRegMap inVarToRegMap = getInVarToRegMap(currentBlock->bbNum);
7078 bool hasCriticalInEdge = blockInfo[currentBlock->bbNum].hasCriticalInEdge;
7080 VarSetOps::AssignNoCopy(compiler, currentLiveVars,
7081 VarSetOps::Intersection(compiler, registerCandidateVars, currentBlock->bbLiveIn));
7083 if (getLsraExtendLifeTimes())
7085 VarSetOps::AssignNoCopy(compiler, currentLiveVars, registerCandidateVars);
7087 // If we are rotating register assignments at block boundaries, we want to make the
7088 // inactive registers available for the rotation.
7089 regMaskTP inactiveRegs = RBM_NONE;
7091 regMaskTP liveRegs = RBM_NONE;
7092 VarSetOps::Iter iter(compiler, currentLiveVars);
7093 unsigned varIndex = 0;
7094 while (iter.NextElem(&varIndex))
7096 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
7097 if (!compiler->lvaTable[varNum].lvLRACandidate)
7101 regNumber targetReg;
7102 Interval* interval = getIntervalForLocalVar(varIndex);
7103 RefPosition* nextRefPosition = interval->getNextRefPosition();
7104 assert(nextRefPosition != nullptr);
7108 targetReg = getVarReg(predVarToRegMap, varIndex);
7110 regNumber newTargetReg = rotateBlockStartLocation(interval, targetReg, (~liveRegs | inactiveRegs));
7111 if (newTargetReg != targetReg)
7113 targetReg = newTargetReg;
7114 setIntervalAsSplit(interval);
7117 setVarReg(inVarToRegMap, varIndex, targetReg);
7119 else // !allocationPass (i.e. resolution/write-back pass)
7121 targetReg = getVarReg(inVarToRegMap, varIndex);
7122 // There are four cases that we need to consider during the resolution pass:
7123 // 1. This variable had a register allocated initially, and it was not spilled in the RefPosition
7124 // that feeds this block. In this case, both targetReg and predVarToRegMap[varIndex] will be targetReg.
7125 // 2. This variable had not been spilled prior to the end of predBB, but was later spilled, so
7126 // predVarToRegMap[varIndex] will be REG_STK, but targetReg is its former allocated value.
7127 // In this case, we will normally change it to REG_STK. We will update its "spilled" status when we
7128 // encounter it in resolveLocalRef().
7129 // 2a. If the next RefPosition is marked as a copyReg, we need to retain the allocated register. This is
7130 // because the copyReg RefPosition will not have recorded the "home" register, yet downstream
7131 // RefPositions rely on the correct "home" register.
7132 // 3. This variable was spilled before we reached the end of predBB. In this case, both targetReg and
7133 // predVarToRegMap[varIndex] will be REG_STK, and the next RefPosition will have been marked
7134 // as reload during allocation time if necessary (note that by the time we actually reach the next
7135 // RefPosition, we may be using a different predecessor, at which it is still in a register).
7136 // 4. This variable was spilled during the allocation of this block, so targetReg is REG_STK
7137 // (because we set inVarToRegMap at the time we spilled it), but predVarToRegMap[varIndex]
7138 // is not REG_STK. We retain the REG_STK value in the inVarToRegMap.
7139 if (targetReg != REG_STK)
7141 if (getVarReg(predVarToRegMap, varIndex) != REG_STK)
7144 assert(getVarReg(predVarToRegMap, varIndex) == targetReg ||
7145 getLsraBlockBoundaryLocations() == LSRA_BLOCK_BOUNDARY_ROTATE);
7147 else if (!nextRefPosition->copyReg)
7150 setVarReg(inVarToRegMap, varIndex, REG_STK);
7151 targetReg = REG_STK;
7153 // Else case 2a. - retain targetReg.
7155 // Else case #3 or #4, we retain targetReg and nothing further to do or assert.
7157 if (interval->physReg == targetReg)
7159 if (interval->isActive)
7161 assert(targetReg != REG_STK);
7162 assert(interval->assignedReg != nullptr && interval->assignedReg->regNum == targetReg &&
7163 interval->assignedReg->assignedInterval == interval);
7164 liveRegs |= genRegMask(targetReg);
7168 else if (interval->physReg != REG_NA)
7170 // This can happen if we are using the locations from a basic block other than the
7171 // immediately preceding one - where the variable was in a different location.
7172 if (targetReg != REG_STK)
7174 // Unassign it from the register (it will get a new register below).
7175 if (interval->assignedReg != nullptr && interval->assignedReg->assignedInterval == interval)
7177 interval->isActive = false;
7178 unassignPhysReg(getRegisterRecord(interval->physReg), nullptr);
7182 // This interval was live in this register the last time we saw a reference to it,
7183 // but has since been displaced.
7184 interval->physReg = REG_NA;
7187 else if (allocationPass)
7189 // Keep the register assignment - if another var has it, it will get unassigned.
7190 // Otherwise, resolution will fix it up later, and it will be more
7191 // likely to match other assignments this way.
7192 interval->isActive = true;
7193 liveRegs |= genRegMask(interval->physReg);
7194 INDEBUG(inactiveRegs |= genRegMask(interval->physReg));
7195 setVarReg(inVarToRegMap, varIndex, interval->physReg);
7199 interval->physReg = REG_NA;
7202 if (targetReg != REG_STK)
7204 RegRecord* targetRegRecord = getRegisterRecord(targetReg);
7205 liveRegs |= genRegMask(targetReg);
7206 if (!interval->isActive)
7208 interval->isActive = true;
7209 interval->physReg = targetReg;
7210 interval->assignedReg = targetRegRecord;
7212 Interval* assignedInterval = targetRegRecord->assignedInterval;
7213 if (assignedInterval != interval)
7215 // Is there another interval currently assigned to this register? If so unassign it.
7216 if (assignedInterval != nullptr)
7218 if (isAssignedToInterval(assignedInterval, targetRegRecord))
7220 regNumber assignedRegNum = assignedInterval->assignedReg->regNum;
7222 // If the interval is active, it will be set to active when we reach its new
7223 // register assignment (which we must not yet have done, or it wouldn't still be
7224 // assigned to this register).
7225 assignedInterval->isActive = false;
7226 unassignPhysReg(assignedInterval->assignedReg, nullptr);
7227 if (allocationPass && assignedInterval->isLocalVar &&
7228 inVarToRegMap[assignedInterval->getVarIndex(compiler)] == assignedRegNum)
7230 inVarToRegMap[assignedInterval->getVarIndex(compiler)] = REG_STK;
7235 // This interval is no longer assigned to this register.
7236 updateAssignedInterval(targetRegRecord, nullptr, assignedInterval->registerType);
7239 assignPhysReg(targetRegRecord, interval);
7241 if (interval->recentRefPosition != nullptr && !interval->recentRefPosition->copyReg &&
7242 interval->recentRefPosition->registerAssignment != genRegMask(targetReg))
7244 interval->getNextRefPosition()->outOfOrder = true;
7249 // Unassign any registers that are no longer live.
7250 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
7252 if ((liveRegs & genRegMask(reg)) == 0)
7254 RegRecord* physRegRecord = getRegisterRecord(reg);
7255 Interval* assignedInterval = physRegRecord->assignedInterval;
7257 if (assignedInterval != nullptr)
7259 assert(assignedInterval->isLocalVar || assignedInterval->isConstant);
7261 if (!assignedInterval->isConstant && assignedInterval->assignedReg == physRegRecord)
7263 assignedInterval->isActive = false;
7264 if (assignedInterval->getNextRefPosition() == nullptr)
7266 unassignPhysReg(physRegRecord, nullptr);
7268 inVarToRegMap[assignedInterval->getVarIndex(compiler)] = REG_STK;
7272 // This interval may still be active, but was in another register in an
7273 // intervening block.
7274 updateAssignedInterval(physRegRecord, nullptr, assignedInterval->registerType);
7278 if (assignedInterval->registerType == TYP_DOUBLE)
7280 // Skip next float register, because we already addressed a double register
7281 assert(genIsValidDoubleReg(reg));
7282 reg = REG_NEXT(reg);
7284 #endif // _TARGET_ARM_
7290 RegRecord* physRegRecord = getRegisterRecord(reg);
7291 Interval* assignedInterval = physRegRecord->assignedInterval;
7293 if (assignedInterval != nullptr && assignedInterval->registerType == TYP_DOUBLE)
7295 // Skip next float register, because we already addressed a double register
7296 assert(genIsValidDoubleReg(reg));
7297 reg = REG_NEXT(reg);
7300 #endif // _TARGET_ARM_
7302 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_START_BB, nullptr, REG_NA, currentBlock));
7305 //------------------------------------------------------------------------
7306 // processBlockEndLocations: Record the variables occupying registers after completing the current block.
7309 // currentBlock - the block we have just completed.
7315 // This must be called both during the allocation and resolution (write-back) phases.
7316 // This is because we need to have the outVarToRegMap locations in order to set the locations
7317 // at successor blocks during allocation time, but if lclVars are spilled after a block has been
7318 // completed, we need to record the REG_STK location for those variables at resolution time.
7320 void LinearScan::processBlockEndLocations(BasicBlock* currentBlock)
7322 assert(currentBlock != nullptr && currentBlock->bbNum == curBBNum);
7323 VarToRegMap outVarToRegMap = getOutVarToRegMap(curBBNum);
7325 VarSetOps::AssignNoCopy(compiler, currentLiveVars,
7326 VarSetOps::Intersection(compiler, registerCandidateVars, currentBlock->bbLiveOut));
7328 if (getLsraExtendLifeTimes())
7330 VarSetOps::Assign(compiler, currentLiveVars, registerCandidateVars);
7333 regMaskTP liveRegs = RBM_NONE;
7334 VarSetOps::Iter iter(compiler, currentLiveVars);
7335 unsigned varIndex = 0;
7336 while (iter.NextElem(&varIndex))
7338 Interval* interval = getIntervalForLocalVar(varIndex);
7339 if (interval->isActive)
7341 assert(interval->physReg != REG_NA && interval->physReg != REG_STK);
7342 setVarReg(outVarToRegMap, varIndex, interval->physReg);
7346 outVarToRegMap[varIndex] = REG_STK;
7349 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_END_BB));
7353 void LinearScan::dumpRefPositions(const char* str)
7355 printf("------------\n");
7356 printf("REFPOSITIONS %s: \n", str);
7357 printf("------------\n");
7358 for (auto& refPos : refPositions)
7365 bool LinearScan::registerIsFree(regNumber regNum, RegisterType regType)
7367 RegRecord* physRegRecord = getRegisterRecord(regNum);
7369 bool isFree = physRegRecord->isFree();
7372 if (isFree && regType == TYP_DOUBLE)
7374 isFree = getRegisterRecord(REG_NEXT(regNum))->isFree();
7376 #endif // _TARGET_ARM_
7381 //------------------------------------------------------------------------
7382 // LinearScan::freeRegister: Make a register available for use
7385 // physRegRecord - the RegRecord for the register to be freed.
7392 // It may be that the RegRecord has already been freed, e.g. due to a kill,
7393 // in which case this method has no effect.
7396 // If there is currently an Interval assigned to this register, and it has
7397 // more references (i.e. this is a local last-use, but more uses and/or
7398 // defs remain), it will remain assigned to the physRegRecord. However, since
7399 // it is marked inactive, the register will be available, albeit less desirable
7401 void LinearScan::freeRegister(RegRecord* physRegRecord)
7403 Interval* assignedInterval = physRegRecord->assignedInterval;
7404 // It may have already been freed by a "Kill"
7405 if (assignedInterval != nullptr)
7407 assignedInterval->isActive = false;
7408 // If this is a constant node, that we may encounter again (e.g. constant),
7409 // don't unassign it until we need the register.
7410 if (!assignedInterval->isConstant)
7412 RefPosition* nextRefPosition = assignedInterval->getNextRefPosition();
7413 // Unassign the register only if there are no more RefPositions, or the next
7414 // one is a def. Note that the latter condition doesn't actually ensure that
7415 // there aren't subsequent uses that could be reached by a def in the assigned
7416 // register, but is merely a heuristic to avoid tying up the register (or using
7417 // it when it's non-optimal). A better alternative would be to use SSA, so that
7418 // we wouldn't unnecessarily link separate live ranges to the same register.
7419 if (nextRefPosition == nullptr || RefTypeIsDef(nextRefPosition->refType))
7422 assert((assignedInterval->registerType != TYP_DOUBLE) || genIsValidDoubleReg(physRegRecord->regNum));
7423 #endif // _TARGET_ARM_
7424 unassignPhysReg(physRegRecord, nullptr);
7430 void LinearScan::freeRegisters(regMaskTP regsToFree)
7432 if (regsToFree == RBM_NONE)
7437 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_FREE_REGS));
7438 while (regsToFree != RBM_NONE)
7440 regMaskTP nextRegBit = genFindLowestBit(regsToFree);
7441 regsToFree &= ~nextRegBit;
7442 regNumber nextReg = genRegNumFromMask(nextRegBit);
7443 freeRegister(getRegisterRecord(nextReg));
7447 // Actual register allocation, accomplished by iterating over all of the previously
7448 // constructed Intervals
7449 // Loosely based on raAssignVars()
7451 void LinearScan::allocateRegisters()
7453 JITDUMP("*************** In LinearScan::allocateRegisters()\n");
7454 DBEXEC(VERBOSE, lsraDumpIntervals("before allocateRegisters"));
7456 // at start, nothing is active except for register args
7457 for (auto& interval : intervals)
7459 Interval* currentInterval = &interval;
7460 currentInterval->recentRefPosition = nullptr;
7461 currentInterval->isActive = false;
7462 if (currentInterval->isLocalVar)
7464 LclVarDsc* varDsc = currentInterval->getLocalVar(compiler);
7465 if (varDsc->lvIsRegArg && currentInterval->firstRefPosition != nullptr)
7467 currentInterval->isActive = true;
7472 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
7474 getRegisterRecord(reg)->recentRefPosition = nullptr;
7475 getRegisterRecord(reg)->isActive = false;
7479 regNumber lastAllocatedReg = REG_NA;
7482 dumpRefPositions("BEFORE ALLOCATION");
7483 dumpVarRefPositions("BEFORE ALLOCATION");
7485 printf("\n\nAllocating Registers\n"
7486 "--------------------\n");
7489 dumpRegRecordHeader();
7490 // Now print an empty indent
7491 printf(indentFormat, "");
7496 BasicBlock* currentBlock = nullptr;
7498 LsraLocation prevLocation = MinLocation;
7499 regMaskTP regsToFree = RBM_NONE;
7500 regMaskTP delayRegsToFree = RBM_NONE;
7502 // This is the most recent RefPosition for which a register was allocated
7503 // - currently only used for DEBUG but maintained in non-debug, for clarity of code
7504 // (and will be optimized away because in non-debug spillAlways() unconditionally returns false)
7505 RefPosition* lastAllocatedRefPosition = nullptr;
7507 bool handledBlockEnd = false;
7509 for (auto& refPosition : refPositions)
7511 RefPosition* currentRefPosition = &refPosition;
7514 // Set the activeRefPosition to null until we're done with any boundary handling.
7515 activeRefPosition = nullptr;
7520 // We're really dumping the RegRecords "after" the previous RefPosition, but it's more convenient
7521 // to do this here, since there are a number of "continue"s in this loop.
7531 // This is the previousRefPosition of the current Referent, if any
7532 RefPosition* previousRefPosition = nullptr;
7534 Interval* currentInterval = nullptr;
7535 Referenceable* currentReferent = nullptr;
7536 bool isInternalRef = false;
7537 RefType refType = currentRefPosition->refType;
7539 currentReferent = currentRefPosition->referent;
7541 if (spillAlways() && lastAllocatedRefPosition != nullptr && !lastAllocatedRefPosition->isPhysRegRef &&
7542 !lastAllocatedRefPosition->getInterval()->isInternal &&
7543 (RefTypeIsDef(lastAllocatedRefPosition->refType) || lastAllocatedRefPosition->getInterval()->isLocalVar))
7545 assert(lastAllocatedRefPosition->registerAssignment != RBM_NONE);
7546 RegRecord* regRecord = lastAllocatedRefPosition->getInterval()->assignedReg;
7547 unassignPhysReg(regRecord, lastAllocatedRefPosition);
7548 // Now set lastAllocatedRefPosition to null, so that we don't try to spill it again
7549 lastAllocatedRefPosition = nullptr;
7552 // We wait to free any registers until we've completed all the
7553 // uses for the current node.
7554 // This avoids reusing registers too soon.
7555 // We free before the last true def (after all the uses & internal
7556 // registers), and then again at the beginning of the next node.
7557 // This is made easier by assigning two LsraLocations per node - one
7558 // for all the uses, internal registers & all but the last def, and
7559 // another for the final def (if any).
7561 LsraLocation currentLocation = currentRefPosition->nodeLocation;
7563 if ((regsToFree | delayRegsToFree) != RBM_NONE)
7565 bool doFreeRegs = false;
7566 // Free at a new location, or at a basic block boundary
7567 if (currentLocation > prevLocation || refType == RefTypeBB)
7574 freeRegisters(regsToFree);
7575 regsToFree = delayRegsToFree;
7576 delayRegsToFree = RBM_NONE;
7579 prevLocation = currentLocation;
7581 // get previous refposition, then current refpos is the new previous
7582 if (currentReferent != nullptr)
7584 previousRefPosition = currentReferent->recentRefPosition;
7585 currentReferent->recentRefPosition = currentRefPosition;
7589 assert((refType == RefTypeBB) || (refType == RefTypeKillGCRefs));
7592 // For the purposes of register resolution, we handle the DummyDefs before
7593 // the block boundary - so the RefTypeBB is after all the DummyDefs.
7594 // However, for the purposes of allocation, we want to handle the block
7595 // boundary first, so that we can free any registers occupied by lclVars
7596 // that aren't live in the next block and make them available for the
7599 if (!handledBlockEnd && (refType == RefTypeBB || refType == RefTypeDummyDef))
7601 // Free any delayed regs (now in regsToFree) before processing the block boundary
7602 freeRegisters(regsToFree);
7603 regsToFree = RBM_NONE;
7604 handledBlockEnd = true;
7605 curBBStartLocation = currentRefPosition->nodeLocation;
7606 if (currentBlock == nullptr)
7608 currentBlock = startBlockSequence();
7612 processBlockEndAllocation(currentBlock);
7613 currentBlock = moveToNextBlock();
7616 if (VERBOSE && currentBlock != nullptr && !dumpTerse)
7618 currentBlock->dspBlockHeader(compiler);
7625 activeRefPosition = currentRefPosition;
7630 dumpRefPositionShort(currentRefPosition, currentBlock);
7634 currentRefPosition->dump();
7639 if (refType == RefTypeBB)
7641 handledBlockEnd = false;
7645 if (refType == RefTypeKillGCRefs)
7647 spillGCRefs(currentRefPosition);
7651 // If this is a FixedReg, disassociate any inactive constant interval from this register.
7652 // Otherwise, do nothing.
7653 if (refType == RefTypeFixedReg)
7655 RegRecord* regRecord = currentRefPosition->getReg();
7656 Interval* assignedInterval = regRecord->assignedInterval;
7658 if (assignedInterval != nullptr && !assignedInterval->isActive && assignedInterval->isConstant)
7660 regRecord->assignedInterval = nullptr;
7663 // Update overlapping floating point register for TYP_DOUBLE
7664 if (assignedInterval->registerType == TYP_DOUBLE)
7666 regRecord = getRegisterRecord(REG_NEXT(regRecord->regNum));
7667 assignedInterval = regRecord->assignedInterval;
7669 assert(assignedInterval != nullptr && !assignedInterval->isActive && assignedInterval->isConstant);
7670 regRecord->assignedInterval = nullptr;
7674 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_FIXED_REG, nullptr, currentRefPosition->assignedReg()));
7678 // If this is an exposed use, do nothing - this is merely a placeholder to attempt to
7679 // ensure that a register is allocated for the full lifetime. The resolution logic
7680 // will take care of moving to the appropriate register if needed.
7682 if (refType == RefTypeExpUse)
7684 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_EXP_USE));
7688 regNumber assignedRegister = REG_NA;
7690 if (currentRefPosition->isIntervalRef())
7692 currentInterval = currentRefPosition->getInterval();
7693 assignedRegister = currentInterval->physReg;
7695 if (VERBOSE && !dumpTerse)
7697 currentInterval->dump();
7701 // Identify the special cases where we decide up-front not to allocate
7702 bool allocate = true;
7703 bool didDump = false;
7705 if (refType == RefTypeParamDef || refType == RefTypeZeroInit)
7707 // For a ParamDef with a weighted refCount less than unity, don't enregister it at entry.
7708 // TODO-CQ: Consider doing this only for stack parameters, since otherwise we may be needlessly
7709 // inserting a store.
7710 LclVarDsc* varDsc = currentInterval->getLocalVar(compiler);
7711 assert(varDsc != nullptr);
7712 if (refType == RefTypeParamDef && varDsc->lvRefCntWtd <= BB_UNITY_WEIGHT)
7714 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_NO_ENTRY_REG_ALLOCATED, currentInterval));
7717 setIntervalAsSpilled(currentInterval);
7719 // If it has no actual references, mark it as "lastUse"; since they're not actually part
7720 // of any flow they won't have been marked during dataflow. Otherwise, if we allocate a
7721 // register we won't unassign it.
7722 else if (currentRefPosition->nextRefPosition == nullptr)
7724 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_ZERO_REF, currentInterval));
7725 currentRefPosition->lastUse = true;
7729 else if (refType == RefTypeUpperVectorSaveDef || refType == RefTypeUpperVectorSaveUse)
7731 Interval* lclVarInterval = currentInterval->relatedInterval;
7732 if (lclVarInterval->physReg == REG_NA)
7737 #endif // FEATURE_SIMD
7739 if (allocate == false)
7741 if (assignedRegister != REG_NA)
7743 unassignPhysReg(getRegisterRecord(assignedRegister), currentRefPosition);
7747 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_NO_REG_ALLOCATED, currentInterval));
7750 currentRefPosition->registerAssignment = RBM_NONE;
7754 if (currentInterval->isSpecialPutArg)
7756 assert(!currentInterval->isLocalVar);
7757 Interval* srcInterval = currentInterval->relatedInterval;
7758 assert(srcInterval->isLocalVar);
7759 if (refType == RefTypeDef)
7761 assert(srcInterval->recentRefPosition->nodeLocation == currentLocation - 1);
7762 RegRecord* physRegRecord = srcInterval->assignedReg;
7764 // For a putarg_reg to be special, its next use location has to be the same
7765 // as fixed reg's next kill location. Otherwise, if source lcl var's next use
7766 // is after the kill of fixed reg but before putarg_reg's next use, fixed reg's
7767 // kill would lead to spill of source but not the putarg_reg if it were treated
7769 if (srcInterval->isActive &&
7770 genRegMask(srcInterval->physReg) == currentRefPosition->registerAssignment &&
7771 currentInterval->getNextRefLocation() == physRegRecord->getNextRefLocation())
7773 assert(physRegRecord->regNum == srcInterval->physReg);
7775 // Special putarg_reg acts as a pass-thru since both source lcl var
7776 // and putarg_reg have the same register allocated. Physical reg
7777 // record of reg continue to point to source lcl var's interval
7778 // instead of to putarg_reg's interval. So if a spill of reg
7779 // allocated to source lcl var happens, to reallocate to another
7780 // tree node, before its use at call node it will lead to spill of
7781 // lcl var instead of putarg_reg since physical reg record is pointing
7782 // to lcl var's interval. As a result, arg reg would get trashed leading
7783 // to bad codegen. The assumption here is that source lcl var of a
7784 // special putarg_reg doesn't get spilled and re-allocated prior to
7785 // its use at the call node. This is ensured by marking physical reg
7786 // record as busy until next kill.
7787 physRegRecord->isBusyUntilNextKill = true;
7791 currentInterval->isSpecialPutArg = false;
7794 // If this is still a SpecialPutArg, continue;
7795 if (currentInterval->isSpecialPutArg)
7797 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_SPECIAL_PUTARG, currentInterval,
7798 currentRefPosition->assignedReg()));
7803 if (assignedRegister == REG_NA && RefTypeIsUse(refType))
7805 currentRefPosition->reload = true;
7806 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_RELOAD, currentInterval, assignedRegister));
7810 regMaskTP assignedRegBit = RBM_NONE;
7811 bool isInRegister = false;
7812 if (assignedRegister != REG_NA)
7814 isInRegister = true;
7815 assignedRegBit = genRegMask(assignedRegister);
7816 if (!currentInterval->isActive)
7818 // If this is a use, it must have started the block on the stack, but the register
7819 // was available for use so we kept the association.
7820 if (RefTypeIsUse(refType))
7822 assert(enregisterLocalVars);
7823 assert(inVarToRegMaps[curBBNum][currentInterval->getVarIndex(compiler)] == REG_STK &&
7824 previousRefPosition->nodeLocation <= curBBStartLocation);
7825 isInRegister = false;
7829 currentInterval->isActive = true;
7832 assert(currentInterval->assignedReg != nullptr &&
7833 currentInterval->assignedReg->regNum == assignedRegister &&
7834 currentInterval->assignedReg->assignedInterval == currentInterval);
7837 // If this is a physical register, we unconditionally assign it to itself!
7838 if (currentRefPosition->isPhysRegRef)
7840 RegRecord* currentReg = currentRefPosition->getReg();
7841 Interval* assignedInterval = currentReg->assignedInterval;
7843 if (assignedInterval != nullptr)
7845 unassignPhysReg(currentReg, assignedInterval->recentRefPosition);
7847 currentReg->isActive = true;
7848 assignedRegister = currentReg->regNum;
7849 assignedRegBit = genRegMask(assignedRegister);
7850 if (refType == RefTypeKill)
7852 currentReg->isBusyUntilNextKill = false;
7855 else if (previousRefPosition != nullptr)
7857 assert(previousRefPosition->nextRefPosition == currentRefPosition);
7858 assert(assignedRegister == REG_NA || assignedRegBit == previousRefPosition->registerAssignment ||
7859 currentRefPosition->outOfOrder || previousRefPosition->copyReg ||
7860 previousRefPosition->refType == RefTypeExpUse || currentRefPosition->refType == RefTypeDummyDef);
7862 else if (assignedRegister != REG_NA)
7864 // Handle the case where this is a preassigned register (i.e. parameter).
7865 // We don't want to actually use the preassigned register if it's not
7866 // going to cover the lifetime - but we had to preallocate it to ensure
7867 // that it remained live.
7868 // TODO-CQ: At some point we may want to refine the analysis here, in case
7869 // it might be beneficial to keep it in this reg for PART of the lifetime
7870 if (currentInterval->isLocalVar)
7872 regMaskTP preferences = currentInterval->registerPreferences;
7873 bool keepAssignment = true;
7874 bool matchesPreferences = (preferences & genRegMask(assignedRegister)) != RBM_NONE;
7876 // Will the assigned register cover the lifetime? If not, does it at least
7877 // meet the preferences for the next RefPosition?
7878 RegRecord* physRegRecord = getRegisterRecord(currentInterval->physReg);
7879 RefPosition* nextPhysRegRefPos = physRegRecord->getNextRefPosition();
7880 if (nextPhysRegRefPos != nullptr &&
7881 nextPhysRegRefPos->nodeLocation <= currentInterval->lastRefPosition->nodeLocation)
7883 // Check to see if the existing assignment matches the preferences (e.g. callee save registers)
7884 // and ensure that the next use of this localVar does not occur after the nextPhysRegRefPos
7885 // There must be a next RefPosition, because we know that the Interval extends beyond the
7886 // nextPhysRegRefPos.
7887 RefPosition* nextLclVarRefPos = currentRefPosition->nextRefPosition;
7888 assert(nextLclVarRefPos != nullptr);
7889 if (!matchesPreferences || nextPhysRegRefPos->nodeLocation < nextLclVarRefPos->nodeLocation ||
7890 physRegRecord->conflictingFixedRegReference(nextLclVarRefPos))
7892 keepAssignment = false;
7895 else if (refType == RefTypeParamDef && !matchesPreferences)
7897 // Don't use the register, even if available, if it doesn't match the preferences.
7898 // Note that this case is only for ParamDefs, for which we haven't yet taken preferences
7899 // into account (we've just automatically got the initial location). In other cases,
7900 // we would already have put it in a preferenced register, if it was available.
7901 // TODO-CQ: Consider expanding this to check availability - that would duplicate
7902 // code here, but otherwise we may wind up in this register anyway.
7903 keepAssignment = false;
7906 if (keepAssignment == false)
7908 currentRefPosition->registerAssignment = allRegs(currentInterval->registerType);
7909 unassignPhysRegNoSpill(physRegRecord);
7911 // If the preferences are currently set to just this register, reset them to allRegs
7912 // of the appropriate type (just as we just reset the registerAssignment for this
7914 // Otherwise, simply remove this register from the preferences, if it's there.
7916 if (currentInterval->registerPreferences == assignedRegBit)
7918 currentInterval->registerPreferences = currentRefPosition->registerAssignment;
7922 currentInterval->registerPreferences &= ~assignedRegBit;
7925 assignedRegister = REG_NA;
7926 assignedRegBit = RBM_NONE;
7931 if (assignedRegister != REG_NA)
7933 // If there is a conflicting fixed reference, insert a copy.
7934 RegRecord* physRegRecord = getRegisterRecord(assignedRegister);
7935 if (physRegRecord->conflictingFixedRegReference(currentRefPosition))
7937 // We may have already reassigned the register to the conflicting reference.
7938 // If not, we need to unassign this interval.
7939 if (physRegRecord->assignedInterval == currentInterval)
7941 unassignPhysRegNoSpill(physRegRecord);
7943 currentRefPosition->moveReg = true;
7944 assignedRegister = REG_NA;
7945 setIntervalAsSplit(currentInterval);
7946 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_MOVE_REG, currentInterval, assignedRegister));
7948 else if ((genRegMask(assignedRegister) & currentRefPosition->registerAssignment) != 0)
7950 currentRefPosition->registerAssignment = assignedRegBit;
7951 if (!currentReferent->isActive)
7953 // If we've got an exposed use at the top of a block, the
7954 // interval might not have been active. Otherwise if it's a use,
7955 // the interval must be active.
7956 if (refType == RefTypeDummyDef)
7958 currentReferent->isActive = true;
7959 assert(getRegisterRecord(assignedRegister)->assignedInterval == currentInterval);
7963 currentRefPosition->reload = true;
7966 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_KEPT_ALLOCATION, currentInterval, assignedRegister));
7970 assert(currentInterval != nullptr);
7972 // It's already in a register, but not one we need.
7973 if (!RefTypeIsDef(currentRefPosition->refType))
7975 regNumber copyReg = assignCopyReg(currentRefPosition);
7976 assert(copyReg != REG_NA);
7977 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_COPY_REG, currentInterval, copyReg));
7978 lastAllocatedRefPosition = currentRefPosition;
7979 if (currentRefPosition->lastUse)
7981 if (currentRefPosition->delayRegFree)
7983 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_LAST_USE_DELAYED, currentInterval,
7985 delayRegsToFree |= (genRegMask(assignedRegister) | currentRefPosition->registerAssignment);
7989 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_LAST_USE, currentInterval, assignedRegister));
7990 regsToFree |= (genRegMask(assignedRegister) | currentRefPosition->registerAssignment);
7993 // If this is a tree temp (non-localVar) interval, we will need an explicit move.
7994 if (!currentInterval->isLocalVar)
7996 currentRefPosition->moveReg = true;
7997 currentRefPosition->copyReg = false;
8003 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_NEEDS_NEW_REG, nullptr, assignedRegister));
8004 regsToFree |= genRegMask(assignedRegister);
8005 // We want a new register, but we don't want this to be considered a spill.
8006 assignedRegister = REG_NA;
8007 if (physRegRecord->assignedInterval == currentInterval)
8009 unassignPhysRegNoSpill(physRegRecord);
8015 if (assignedRegister == REG_NA)
8017 bool allocateReg = true;
8019 if (currentRefPosition->AllocateIfProfitable())
8021 // We can avoid allocating a register if it is a the last use requiring a reload.
8022 if (currentRefPosition->lastUse && currentRefPosition->reload)
8024 allocateReg = false;
8028 // Under stress mode, don't attempt to allocate a reg to
8029 // reg optional ref position.
8030 if (allocateReg && regOptionalNoAlloc())
8032 allocateReg = false;
8039 // Try to allocate a register
8040 assignedRegister = tryAllocateFreeReg(currentInterval, currentRefPosition);
8043 // If no register was found, and if the currentRefPosition must have a register,
8044 // then find a register to spill
8045 if (assignedRegister == REG_NA)
8047 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8048 if (refType == RefTypeUpperVectorSaveDef)
8050 // TODO-CQ: Determine whether copying to two integer callee-save registers would be profitable.
8052 // SaveDef position occurs after the Use of args and at the same location as Kill/Def
8053 // positions of a call node. But SaveDef position cannot use any of the arg regs as
8054 // they are needed for call node.
8055 currentRefPosition->registerAssignment =
8056 (allRegs(TYP_FLOAT) & RBM_FLT_CALLEE_TRASH & ~RBM_FLTARG_REGS);
8057 assignedRegister = tryAllocateFreeReg(currentInterval, currentRefPosition);
8059 // There MUST be caller-save registers available, because they have all just been killed.
8060 // Amd64 Windows: xmm4-xmm5 are guaranteed to be available as xmm0-xmm3 are used for passing args.
8061 // Amd64 Unix: xmm8-xmm15 are guaranteed to be avilable as xmm0-xmm7 are used for passing args.
8062 // X86 RyuJIT Windows: xmm4-xmm7 are guanrateed to be available.
8063 assert(assignedRegister != REG_NA);
8067 // i) The reason we have to spill is that SaveDef position is allocated after the Kill positions
8068 // of the call node are processed. Since callee-trash registers are killed by call node
8069 // we explicity spill and unassign the register.
8070 // ii) These will look a bit backward in the dump, but it's a pain to dump the alloc before the
8072 unassignPhysReg(getRegisterRecord(assignedRegister), currentRefPosition);
8073 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_ALLOC_REG, currentInterval, assignedRegister));
8075 // Now set assignedRegister to REG_NA again so that we don't re-activate it.
8076 assignedRegister = REG_NA;
8079 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8080 if (currentRefPosition->RequiresRegister() || currentRefPosition->AllocateIfProfitable())
8084 assignedRegister = allocateBusyReg(currentInterval, currentRefPosition,
8085 currentRefPosition->AllocateIfProfitable());
8088 if (assignedRegister != REG_NA)
8091 dumpLsraAllocationEvent(LSRA_EVENT_ALLOC_SPILLED_REG, currentInterval, assignedRegister));
8095 // This can happen only for those ref positions that are to be allocated
8096 // only if profitable.
8097 noway_assert(currentRefPosition->AllocateIfProfitable());
8099 currentRefPosition->registerAssignment = RBM_NONE;
8100 currentRefPosition->reload = false;
8101 setIntervalAsSpilled(currentInterval);
8103 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_NO_REG_ALLOCATED, currentInterval));
8108 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_NO_REG_ALLOCATED, currentInterval));
8109 currentRefPosition->registerAssignment = RBM_NONE;
8110 currentInterval->isActive = false;
8111 setIntervalAsSpilled(currentInterval);
8119 if (currentInterval->isConstant && (currentRefPosition->treeNode != nullptr) &&
8120 currentRefPosition->treeNode->IsReuseRegVal())
8122 dumpLsraAllocationEvent(LSRA_EVENT_REUSE_REG, nullptr, assignedRegister, currentBlock);
8126 dumpLsraAllocationEvent(LSRA_EVENT_ALLOC_REG, nullptr, assignedRegister, currentBlock);
8132 if (refType == RefTypeDummyDef && assignedRegister != REG_NA)
8134 setInVarRegForBB(curBBNum, currentInterval->varNum, assignedRegister);
8137 // If we allocated a register, and this is a use of a spilled value,
8138 // it should have been marked for reload above.
8139 if (assignedRegister != REG_NA && RefTypeIsUse(refType) && !isInRegister)
8141 assert(currentRefPosition->reload);
8145 // If we allocated a register, record it
8146 if (currentInterval != nullptr && assignedRegister != REG_NA)
8148 assignedRegBit = genRegMask(assignedRegister);
8149 currentRefPosition->registerAssignment = assignedRegBit;
8150 currentInterval->physReg = assignedRegister;
8151 regsToFree &= ~assignedRegBit; // we'll set it again later if it's dead
8153 // If this interval is dead, free the register.
8154 // The interval could be dead if this is a user variable, or if the
8155 // node is being evaluated for side effects, or a call whose result
8156 // is not used, etc.
8157 if (currentRefPosition->lastUse || currentRefPosition->nextRefPosition == nullptr)
8159 assert(currentRefPosition->isIntervalRef());
8161 if (refType != RefTypeExpUse && currentRefPosition->nextRefPosition == nullptr)
8163 if (currentRefPosition->delayRegFree)
8165 delayRegsToFree |= assignedRegBit;
8167 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_LAST_USE_DELAYED));
8171 regsToFree |= assignedRegBit;
8173 INDEBUG(dumpLsraAllocationEvent(LSRA_EVENT_LAST_USE));
8178 currentInterval->isActive = false;
8182 lastAllocatedRefPosition = currentRefPosition;
8186 // Free registers to clear associated intervals for resolution phase
8187 CLANG_FORMAT_COMMENT_ANCHOR;
8190 if (getLsraExtendLifeTimes())
8192 // If we have extended lifetimes, we need to make sure all the registers are freed.
8193 for (int regNumIndex = 0; regNumIndex <= REG_FP_LAST; regNumIndex++)
8195 RegRecord& regRecord = physRegs[regNumIndex];
8196 Interval* interval = regRecord.assignedInterval;
8197 if (interval != nullptr)
8199 interval->isActive = false;
8200 unassignPhysReg(®Record, nullptr);
8207 freeRegisters(regsToFree | delayRegsToFree);
8215 // Dump the RegRecords after the last RefPosition is handled.
8220 dumpRefPositions("AFTER ALLOCATION");
8221 dumpVarRefPositions("AFTER ALLOCATION");
8223 // Dump the intervals that remain active
8224 printf("Active intervals at end of allocation:\n");
8226 // We COULD just reuse the intervalIter from above, but ArrayListIterator doesn't
8227 // provide a Reset function (!) - we'll probably replace this so don't bother
8230 for (auto& interval : intervals)
8232 if (interval.isActive)
8244 //-----------------------------------------------------------------------------
8245 // updateAssignedInterval: Update assigned interval of register.
8248 // reg - register to be updated
8249 // interval - interval to be assigned
8250 // regType - regsiter type
8256 // For ARM32, when "regType" is TYP_DOUBLE, "reg" should be a even-numbered
8257 // float register, i.e. lower half of double register.
8260 // For ARM32, two float registers consisting a double register are updated
8261 // together when "regType" is TYP_DOUBLE.
8263 void LinearScan::updateAssignedInterval(RegRecord* reg, Interval* interval, RegisterType regType)
8265 reg->assignedInterval = interval;
8268 // Update overlapping floating point register for TYP_DOUBLE
8269 if (regType == TYP_DOUBLE)
8271 assert(genIsValidDoubleReg(reg->regNum));
8273 RegRecord* anotherHalfReg = findAnotherHalfRegRec(reg);
8275 anotherHalfReg->assignedInterval = interval;
8280 //-----------------------------------------------------------------------------
8281 // updatePreviousInterval: Update previous interval of register.
8284 // reg - register to be updated
8285 // interval - interval to be assigned
8286 // regType - regsiter type
8292 // For ARM32, when "regType" is TYP_DOUBLE, "reg" should be a even-numbered
8293 // float register, i.e. lower half of double register.
8296 // For ARM32, two float registers consisting a double register are updated
8297 // together when "regType" is TYP_DOUBLE.
8299 void LinearScan::updatePreviousInterval(RegRecord* reg, Interval* interval, RegisterType regType)
8301 reg->previousInterval = interval;
8304 // Update overlapping floating point register for TYP_DOUBLE
8305 if (regType == TYP_DOUBLE)
8307 assert(genIsValidDoubleReg(reg->regNum));
8309 RegRecord* anotherHalfReg = findAnotherHalfRegRec(reg);
8311 anotherHalfReg->previousInterval = interval;
8316 // LinearScan::resolveLocalRef
8318 // Update the graph for a local reference.
8319 // Also, track the register (if any) that is currently occupied.
8321 // treeNode: The lclVar that's being resolved
8322 // currentRefPosition: the RefPosition associated with the treeNode
8325 // This method is called for each local reference, during the resolveRegisters
8326 // phase of LSRA. It is responsible for keeping the following in sync:
8327 // - varDsc->lvRegNum (and lvOtherReg) contain the unique register location.
8328 // If it is not in the same register through its lifetime, it is set to REG_STK.
8329 // - interval->physReg is set to the assigned register
8330 // (i.e. at the code location which is currently being handled by resolveRegisters())
8331 // - interval->isActive is true iff the interval is live and occupying a register
8332 // - interval->isSpilled should have already been set to true if the interval is EVER spilled
8333 // - interval->isSplit is set to true if the interval does not occupy the same
8334 // register throughout the method
8335 // - RegRecord->assignedInterval points to the interval which currently occupies
8337 // - For each lclVar node:
8338 // - gtRegNum/gtRegPair is set to the currently allocated register(s).
8339 // - GTF_SPILLED is set on a use if it must be reloaded prior to use.
8340 // - GTF_SPILL is set if it must be spilled after use.
8342 // A copyReg is an ugly case where the variable must be in a specific (fixed) register,
8343 // but it currently resides elsewhere. The register allocator must track the use of the
8344 // fixed register, but it marks the lclVar node with the register it currently lives in
8345 // and the code generator does the necessary move.
8347 // Before beginning, the varDsc for each parameter must be set to its initial location.
8349 // NICE: Consider tracking whether an Interval is always in the same location (register/stack)
8350 // in which case it will require no resolution.
8352 void LinearScan::resolveLocalRef(BasicBlock* block, GenTreePtr treeNode, RefPosition* currentRefPosition)
8354 assert((block == nullptr) == (treeNode == nullptr));
8355 assert(enregisterLocalVars);
8357 // Is this a tracked local? Or just a register allocated for loading
8358 // a non-tracked one?
8359 Interval* interval = currentRefPosition->getInterval();
8360 if (!interval->isLocalVar)
8364 interval->recentRefPosition = currentRefPosition;
8365 LclVarDsc* varDsc = interval->getLocalVar(compiler);
8367 // NOTE: we set the GTF_VAR_DEATH flag here unless we are extending lifetimes, in which case we write
8368 // this bit in checkLastUses. This is a bit of a hack, but is necessary because codegen requires
8369 // accurate last use info that is not reflected in the lastUse bit on ref positions when we are extending
8370 // lifetimes. See also the comments in checkLastUses.
8371 if ((treeNode != nullptr) && !extendLifetimes())
8373 if (currentRefPosition->lastUse)
8375 treeNode->gtFlags |= GTF_VAR_DEATH;
8379 treeNode->gtFlags &= ~GTF_VAR_DEATH;
8383 if (currentRefPosition->registerAssignment == RBM_NONE)
8385 assert(!currentRefPosition->RequiresRegister());
8386 assert(interval->isSpilled);
8388 varDsc->lvRegNum = REG_STK;
8389 if (interval->assignedReg != nullptr && interval->assignedReg->assignedInterval == interval)
8391 updateAssignedInterval(interval->assignedReg, nullptr, interval->registerType);
8393 interval->assignedReg = nullptr;
8394 interval->physReg = REG_NA;
8395 if (treeNode != nullptr)
8397 treeNode->SetContained();
8403 // In most cases, assigned and home registers will be the same
8404 // The exception is the copyReg case, where we've assigned a register
8405 // for a specific purpose, but will be keeping the register assignment
8406 regNumber assignedReg = currentRefPosition->assignedReg();
8407 regNumber homeReg = assignedReg;
8409 // Undo any previous association with a physical register, UNLESS this
8411 if (!currentRefPosition->copyReg)
8413 regNumber oldAssignedReg = interval->physReg;
8414 if (oldAssignedReg != REG_NA && assignedReg != oldAssignedReg)
8416 RegRecord* oldRegRecord = getRegisterRecord(oldAssignedReg);
8417 if (oldRegRecord->assignedInterval == interval)
8419 updateAssignedInterval(oldRegRecord, nullptr, interval->registerType);
8424 if (currentRefPosition->refType == RefTypeUse && !currentRefPosition->reload)
8426 // Was this spilled after our predecessor was scheduled?
8427 if (interval->physReg == REG_NA)
8429 assert(inVarToRegMaps[curBBNum][varDsc->lvVarIndex] == REG_STK);
8430 currentRefPosition->reload = true;
8434 bool reload = currentRefPosition->reload;
8435 bool spillAfter = currentRefPosition->spillAfter;
8437 // In the reload case we either:
8438 // - Set the register to REG_STK if it will be referenced only from the home location, or
8439 // - Set the register to the assigned register and set GTF_SPILLED if it must be loaded into a register.
8442 assert(currentRefPosition->refType != RefTypeDef);
8443 assert(interval->isSpilled);
8444 varDsc->lvRegNum = REG_STK;
8447 interval->physReg = assignedReg;
8450 // If there is no treeNode, this must be a RefTypeExpUse, in
8451 // which case we did the reload already
8452 if (treeNode != nullptr)
8454 treeNode->gtFlags |= GTF_SPILLED;
8457 if (currentRefPosition->AllocateIfProfitable())
8459 // This is a use of lclVar that is flagged as reg-optional
8460 // by lower/codegen and marked for both reload and spillAfter.
8461 // In this case we can avoid unnecessary reload and spill
8462 // by setting reg on lclVar to REG_STK and reg on tree node
8463 // to REG_NA. Codegen will generate the code by considering
8464 // it as a contained memory operand.
8466 // Note that varDsc->lvRegNum is already to REG_STK above.
8467 interval->physReg = REG_NA;
8468 treeNode->gtRegNum = REG_NA;
8469 treeNode->gtFlags &= ~GTF_SPILLED;
8470 treeNode->SetContained();
8474 treeNode->gtFlags |= GTF_SPILL;
8480 assert(currentRefPosition->refType == RefTypeExpUse);
8483 else if (spillAfter && !RefTypeIsUse(currentRefPosition->refType))
8485 // In the case of a pure def, don't bother spilling - just assign it to the
8486 // stack. However, we need to remember that it was spilled.
8488 assert(interval->isSpilled);
8489 varDsc->lvRegNum = REG_STK;
8490 interval->physReg = REG_NA;
8491 if (treeNode != nullptr)
8493 treeNode->gtRegNum = REG_NA;
8498 // Not reload and Not pure-def that's spillAfter
8500 if (currentRefPosition->copyReg || currentRefPosition->moveReg)
8502 // For a copyReg or moveReg, we have two cases:
8503 // - In the first case, we have a fixedReg - i.e. a register which the code
8504 // generator is constrained to use.
8505 // The code generator will generate the appropriate move to meet the requirement.
8506 // - In the second case, we were forced to use a different register because of
8507 // interference (or JitStressRegs).
8508 // In this case, we generate a GT_COPY.
8509 // In either case, we annotate the treeNode with the register in which the value
8510 // currently lives. For moveReg, the homeReg is the new register (as assigned above).
8511 // But for copyReg, the homeReg remains unchanged.
8513 assert(treeNode != nullptr);
8514 treeNode->gtRegNum = interval->physReg;
8516 if (currentRefPosition->copyReg)
8518 homeReg = interval->physReg;
8522 assert(interval->isSplit);
8523 interval->physReg = assignedReg;
8526 if (!currentRefPosition->isFixedRegRef || currentRefPosition->moveReg)
8528 // This is the second case, where we need to generate a copy
8529 insertCopyOrReload(block, treeNode, currentRefPosition->getMultiRegIdx(), currentRefPosition);
8534 interval->physReg = assignedReg;
8536 if (!interval->isSpilled && !interval->isSplit)
8538 if (varDsc->lvRegNum != REG_STK)
8540 // If the register assignments don't match, then this interval is split.
8541 if (varDsc->lvRegNum != assignedReg)
8543 setIntervalAsSplit(interval);
8544 varDsc->lvRegNum = REG_STK;
8549 varDsc->lvRegNum = assignedReg;
8555 if (treeNode != nullptr)
8557 treeNode->gtFlags |= GTF_SPILL;
8559 assert(interval->isSpilled);
8560 interval->physReg = REG_NA;
8561 varDsc->lvRegNum = REG_STK;
8565 // Update the physRegRecord for the register, so that we know what vars are in
8566 // regs at the block boundaries
8567 RegRecord* physRegRecord = getRegisterRecord(homeReg);
8568 if (spillAfter || currentRefPosition->lastUse)
8570 interval->isActive = false;
8571 interval->assignedReg = nullptr;
8572 interval->physReg = REG_NA;
8574 updateAssignedInterval(physRegRecord, nullptr, interval->registerType);
8578 interval->isActive = true;
8579 interval->assignedReg = physRegRecord;
8581 updateAssignedInterval(physRegRecord, interval, interval->registerType);
8585 void LinearScan::writeRegisters(RefPosition* currentRefPosition, GenTree* tree)
8587 lsraAssignRegToTree(tree, currentRefPosition->assignedReg(), currentRefPosition->getMultiRegIdx());
8590 //------------------------------------------------------------------------
8591 // insertCopyOrReload: Insert a copy in the case where a tree node value must be moved
8592 // to a different register at the point of use (GT_COPY), or it is reloaded to a different register
8593 // than the one it was spilled from (GT_RELOAD).
8596 // block - basic block in which GT_COPY/GT_RELOAD is inserted.
8597 // tree - This is the node to copy or reload.
8598 // Insert copy or reload node between this node and its parent.
8599 // multiRegIdx - register position of tree node for which copy or reload is needed.
8600 // refPosition - The RefPosition at which copy or reload will take place.
8603 // The GT_COPY or GT_RELOAD will be inserted in the proper spot in execution order where the reload is to occur.
8605 // For example, for this tree (numbers are execution order, lower is earlier and higher is later):
8607 // +---------+----------+
8609 // +---------+----------+
8614 // +-------------------+ +----------------------+
8615 // | x (1) | "tree" | y (2) |
8616 // +-------------------+ +----------------------+
8618 // generate this tree:
8620 // +---------+----------+
8622 // +---------+----------+
8627 // +-------------------+ +----------------------+
8628 // | GT_RELOAD (3) | | y (2) |
8629 // +-------------------+ +----------------------+
8631 // +-------------------+
8633 // +-------------------+
8635 // Note in particular that the GT_RELOAD node gets inserted in execution order immediately before the parent of "tree",
8636 // which seems a bit weird since normally a node's parent (in this case, the parent of "x", GT_RELOAD in the "after"
8637 // picture) immediately follows all of its children (that is, normally the execution ordering is postorder).
8638 // The ordering must be this weird "out of normal order" way because the "x" node is being spilled, probably
8639 // because the expression in the tree represented above by "y" has high register requirements. We don't want
8640 // to reload immediately, of course. So we put GT_RELOAD where the reload should actually happen.
8642 // Note that GT_RELOAD is required when we reload to a different register than the one we spilled to. It can also be
8643 // used if we reload to the same register. Normally, though, in that case we just mark the node with GTF_SPILLED,
8644 // and the unspilling code automatically reuses the same register, and does the reload when it notices that flag
8645 // when considering a node's operands.
8647 void LinearScan::insertCopyOrReload(BasicBlock* block, GenTreePtr tree, unsigned multiRegIdx, RefPosition* refPosition)
8649 LIR::Range& blockRange = LIR::AsRange(block);
8652 bool foundUse = blockRange.TryGetUse(tree, &treeUse);
8655 GenTree* parent = treeUse.User();
8658 if (refPosition->reload)
8666 #if TRACK_LSRA_STATS
8667 updateLsraStat(LSRA_STAT_COPY_REG, block->bbNum);
8671 // If the parent is a reload/copy node, then tree must be a multi-reg call node
8672 // that has already had one of its registers spilled. This is Because multi-reg
8673 // call node is the only node whose RefTypeDef positions get independently
8674 // spilled or reloaded. It is possible that one of its RefTypeDef position got
8675 // spilled and the next use of it requires it to be in a different register.
8677 // In this case set the ith position reg of reload/copy node to the reg allocated
8678 // for copy/reload refPosition. Essentially a copy/reload node will have a reg
8679 // for each multi-reg position of its child. If there is a valid reg in ith
8680 // position of GT_COPY or GT_RELOAD node then the corresponding result of its
8681 // child needs to be copied or reloaded to that reg.
8682 if (parent->IsCopyOrReload())
8684 noway_assert(parent->OperGet() == oper);
8685 noway_assert(tree->IsMultiRegCall());
8686 GenTreeCall* call = tree->AsCall();
8687 GenTreeCopyOrReload* copyOrReload = parent->AsCopyOrReload();
8688 noway_assert(copyOrReload->GetRegNumByIdx(multiRegIdx) == REG_NA);
8689 copyOrReload->SetRegNumByIdx(refPosition->assignedReg(), multiRegIdx);
8693 // Create the new node, with "tree" as its only child.
8694 var_types treeType = tree->TypeGet();
8696 GenTreeCopyOrReload* newNode = new (compiler, oper) GenTreeCopyOrReload(oper, treeType, tree);
8697 assert(refPosition->registerAssignment != RBM_NONE);
8698 newNode->SetRegNumByIdx(refPosition->assignedReg(), multiRegIdx);
8699 newNode->gtLsraInfo.isLsraAdded = true;
8700 newNode->gtLsraInfo.isLocalDefUse = false;
8701 if (refPosition->copyReg)
8703 // This is a TEMPORARY copy
8704 assert(isCandidateLocalRef(tree));
8705 newNode->gtFlags |= GTF_VAR_DEATH;
8708 // Insert the copy/reload after the spilled node and replace the use of the original node with a use
8709 // of the copy/reload.
8710 blockRange.InsertAfter(tree, newNode);
8711 treeUse.ReplaceWith(compiler, newNode);
8715 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8716 //------------------------------------------------------------------------
8717 // insertUpperVectorSaveAndReload: Insert code to save and restore the upper half of a vector that lives
8718 // in a callee-save register at the point of a kill (the upper half is
8722 // tree - This is the node around which we will insert the Save & Reload.
8723 // It will be a call or some node that turns into a call.
8724 // refPosition - The RefTypeUpperVectorSaveDef RefPosition.
8726 void LinearScan::insertUpperVectorSaveAndReload(GenTreePtr tree, RefPosition* refPosition, BasicBlock* block)
8728 Interval* lclVarInterval = refPosition->getInterval()->relatedInterval;
8729 assert(lclVarInterval->isLocalVar == true);
8730 LclVarDsc* varDsc = compiler->lvaTable + lclVarInterval->varNum;
8731 assert(varDsc->lvType == LargeVectorType);
8732 regNumber lclVarReg = lclVarInterval->physReg;
8733 if (lclVarReg == REG_NA)
8738 assert((genRegMask(lclVarReg) & RBM_FLT_CALLEE_SAVED) != RBM_NONE);
8740 regNumber spillReg = refPosition->assignedReg();
8741 bool spillToMem = refPosition->spillAfter;
8743 LIR::Range& blockRange = LIR::AsRange(block);
8745 // First, insert the save before the call.
8747 GenTreePtr saveLcl = compiler->gtNewLclvNode(lclVarInterval->varNum, LargeVectorType);
8748 saveLcl->gtLsraInfo.isLsraAdded = true;
8749 saveLcl->gtRegNum = lclVarReg;
8750 saveLcl->gtLsraInfo.isLocalDefUse = false;
8752 GenTreeSIMD* simdNode =
8753 new (compiler, GT_SIMD) GenTreeSIMD(LargeVectorSaveType, saveLcl, nullptr, SIMDIntrinsicUpperSave,
8754 varDsc->lvBaseType, genTypeSize(LargeVectorType));
8755 simdNode->gtLsraInfo.isLsraAdded = true;
8756 simdNode->gtRegNum = spillReg;
8759 simdNode->gtFlags |= GTF_SPILL;
8762 blockRange.InsertBefore(tree, LIR::SeqTree(compiler, simdNode));
8764 // Now insert the restore after the call.
8766 GenTreePtr restoreLcl = compiler->gtNewLclvNode(lclVarInterval->varNum, LargeVectorType);
8767 restoreLcl->gtLsraInfo.isLsraAdded = true;
8768 restoreLcl->gtRegNum = lclVarReg;
8769 restoreLcl->gtLsraInfo.isLocalDefUse = false;
8771 simdNode = new (compiler, GT_SIMD)
8772 GenTreeSIMD(LargeVectorType, restoreLcl, nullptr, SIMDIntrinsicUpperRestore, varDsc->lvBaseType, 32);
8773 simdNode->gtLsraInfo.isLsraAdded = true;
8774 simdNode->gtRegNum = spillReg;
8777 simdNode->gtFlags |= GTF_SPILLED;
8780 blockRange.InsertAfter(tree, LIR::SeqTree(compiler, simdNode));
8782 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8784 //------------------------------------------------------------------------
8785 // initMaxSpill: Initializes the LinearScan members used to track the max number
8786 // of concurrent spills. This is needed so that we can set the
8787 // fields in Compiler, so that the code generator, in turn can
8788 // allocate the right number of spill locations.
8797 // This is called before any calls to updateMaxSpill().
8799 void LinearScan::initMaxSpill()
8801 needDoubleTmpForFPCall = false;
8802 needFloatTmpForFPCall = false;
8803 for (int i = 0; i < TYP_COUNT; i++)
8806 currentSpill[i] = 0;
8810 //------------------------------------------------------------------------
8811 // recordMaxSpill: Sets the fields in Compiler for the max number of concurrent spills.
8812 // (See the comment on initMaxSpill.)
8821 // This is called after updateMaxSpill() has been called for all "real"
8824 void LinearScan::recordMaxSpill()
8826 // Note: due to the temp normalization process (see tmpNormalizeType)
8827 // only a few types should actually be seen here.
8828 JITDUMP("Recording the maximum number of concurrent spills:\n");
8830 var_types returnType = compiler->tmpNormalizeType(compiler->info.compRetType);
8831 if (needDoubleTmpForFPCall || (returnType == TYP_DOUBLE))
8833 JITDUMP("Adding a spill temp for moving a double call/return value between xmm reg and x87 stack.\n");
8834 maxSpill[TYP_DOUBLE] += 1;
8836 if (needFloatTmpForFPCall || (returnType == TYP_FLOAT))
8838 JITDUMP("Adding a spill temp for moving a float call/return value between xmm reg and x87 stack.\n");
8839 maxSpill[TYP_FLOAT] += 1;
8841 #endif // _TARGET_X86_
8842 for (int i = 0; i < TYP_COUNT; i++)
8844 if (var_types(i) != compiler->tmpNormalizeType(var_types(i)))
8846 // Only normalized types should have anything in the maxSpill array.
8847 // We assume here that if type 'i' does not normalize to itself, then
8848 // nothing else normalizes to 'i', either.
8849 assert(maxSpill[i] == 0);
8851 if (maxSpill[i] != 0)
8853 JITDUMP(" %s: %d\n", varTypeName(var_types(i)), maxSpill[i]);
8854 compiler->tmpPreAllocateTemps(var_types(i), maxSpill[i]);
8860 //------------------------------------------------------------------------
8861 // updateMaxSpill: Update the maximum number of concurrent spills
8864 // refPosition - the current RefPosition being handled
8870 // The RefPosition has an associated interval (getInterval() will
8871 // otherwise assert).
8874 // This is called for each "real" RefPosition during the writeback
8875 // phase of LSRA. It keeps track of how many concurrently-live
8876 // spills there are, and the largest number seen so far.
8878 void LinearScan::updateMaxSpill(RefPosition* refPosition)
8880 RefType refType = refPosition->refType;
8882 if (refPosition->spillAfter || refPosition->reload ||
8883 (refPosition->AllocateIfProfitable() && refPosition->assignedReg() == REG_NA))
8885 Interval* interval = refPosition->getInterval();
8886 if (!interval->isLocalVar)
8888 // The tmp allocation logic 'normalizes' types to a small number of
8889 // types that need distinct stack locations from each other.
8890 // Those types are currently gc refs, byrefs, <= 4 byte non-GC items,
8891 // 8-byte non-GC items, and 16-byte or 32-byte SIMD vectors.
8892 // LSRA is agnostic to those choices but needs
8893 // to know what they are here.
8896 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8897 if ((refType == RefTypeUpperVectorSaveDef) || (refType == RefTypeUpperVectorSaveUse))
8899 typ = LargeVectorSaveType;
8902 #endif // !FEATURE_PARTIAL_SIMD_CALLEE_SAVE
8904 GenTreePtr treeNode = refPosition->treeNode;
8905 if (treeNode == nullptr)
8907 assert(RefTypeIsUse(refType));
8908 treeNode = interval->firstRefPosition->treeNode;
8910 assert(treeNode != nullptr);
8912 // In case of multi-reg call nodes, we need to use the type
8913 // of the return register given by multiRegIdx of the refposition.
8914 if (treeNode->IsMultiRegCall())
8916 ReturnTypeDesc* retTypeDesc = treeNode->AsCall()->GetReturnTypeDesc();
8917 typ = retTypeDesc->GetReturnRegType(refPosition->getMultiRegIdx());
8921 typ = treeNode->TypeGet();
8923 typ = compiler->tmpNormalizeType(typ);
8926 if (refPosition->spillAfter && !refPosition->reload)
8928 currentSpill[typ]++;
8929 if (currentSpill[typ] > maxSpill[typ])
8931 maxSpill[typ] = currentSpill[typ];
8934 else if (refPosition->reload)
8936 assert(currentSpill[typ] > 0);
8937 currentSpill[typ]--;
8939 else if (refPosition->AllocateIfProfitable() && refPosition->assignedReg() == REG_NA)
8941 // A spill temp not getting reloaded into a reg because it is
8942 // marked as allocate if profitable and getting used from its
8943 // memory location. To properly account max spill for typ we
8944 // decrement spill count.
8945 assert(RefTypeIsUse(refType));
8946 assert(currentSpill[typ] > 0);
8947 currentSpill[typ]--;
8949 JITDUMP(" Max spill for %s is %d\n", varTypeName(typ), maxSpill[typ]);
8954 // This is the final phase of register allocation. It writes the register assignments to
8955 // the tree, and performs resolution across joins and backedges.
8957 void LinearScan::resolveRegisters()
8959 // Iterate over the tree and the RefPositions in lockstep
8960 // - annotate the tree with register assignments by setting gtRegNum or gtRegPair (for longs)
8962 // - track globally-live var locations
8963 // - add resolution points at split/merge/critical points as needed
8965 // Need to use the same traversal order as the one that assigns the location numbers.
8967 // Dummy RefPositions have been added at any split, join or critical edge, at the
8968 // point where resolution may be required. These are located:
8969 // - for a split, at the top of the non-adjacent block
8970 // - for a join, at the bottom of the non-adjacent joining block
8971 // - for a critical edge, at the top of the target block of each critical
8973 // Note that a target block may have multiple incoming critical or split edges
8975 // These RefPositions record the expected location of the Interval at that point.
8976 // At each branch, we identify the location of each liveOut interval, and check
8977 // against the RefPositions at the target.
8980 LsraLocation currentLocation = MinLocation;
8982 // Clear register assignments - these will be reestablished as lclVar defs (including RefTypeParamDefs)
8984 if (enregisterLocalVars)
8986 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
8988 RegRecord* physRegRecord = getRegisterRecord(reg);
8989 Interval* assignedInterval = physRegRecord->assignedInterval;
8990 if (assignedInterval != nullptr)
8992 assignedInterval->assignedReg = nullptr;
8993 assignedInterval->physReg = REG_NA;
8995 physRegRecord->assignedInterval = nullptr;
8996 physRegRecord->recentRefPosition = nullptr;
8999 // Clear "recentRefPosition" for lclVar intervals
9000 for (unsigned varIndex = 0; varIndex < compiler->lvaTrackedCount; varIndex++)
9002 if (localVarIntervals[varIndex] != nullptr)
9004 localVarIntervals[varIndex]->recentRefPosition = nullptr;
9005 localVarIntervals[varIndex]->isActive = false;
9009 assert(compiler->lvaTable[compiler->lvaTrackedToVarNum[varIndex]].lvLRACandidate == false);
9014 // handle incoming arguments and special temps
9015 auto currentRefPosition = refPositions.begin();
9017 if (enregisterLocalVars)
9019 VarToRegMap entryVarToRegMap = inVarToRegMaps[compiler->fgFirstBB->bbNum];
9020 while (currentRefPosition != refPositions.end() &&
9021 (currentRefPosition->refType == RefTypeParamDef || currentRefPosition->refType == RefTypeZeroInit))
9023 Interval* interval = currentRefPosition->getInterval();
9024 assert(interval != nullptr && interval->isLocalVar);
9025 resolveLocalRef(nullptr, nullptr, currentRefPosition);
9026 regNumber reg = REG_STK;
9027 int varIndex = interval->getVarIndex(compiler);
9029 if (!currentRefPosition->spillAfter && currentRefPosition->registerAssignment != RBM_NONE)
9031 reg = currentRefPosition->assignedReg();
9036 interval->isActive = false;
9038 setVarReg(entryVarToRegMap, varIndex, reg);
9039 ++currentRefPosition;
9044 assert(currentRefPosition == refPositions.end() ||
9045 (currentRefPosition->refType != RefTypeParamDef && currentRefPosition->refType != RefTypeZeroInit));
9048 BasicBlock* insertionBlock = compiler->fgFirstBB;
9049 GenTreePtr insertionPoint = LIR::AsRange(insertionBlock).FirstNonPhiNode();
9051 // write back assignments
9052 for (block = startBlockSequence(); block != nullptr; block = moveToNextBlock())
9054 assert(curBBNum == block->bbNum);
9056 if (enregisterLocalVars)
9058 // Record the var locations at the start of this block.
9059 // (If it's fgFirstBB, we've already done that above, see entryVarToRegMap)
9061 curBBStartLocation = currentRefPosition->nodeLocation;
9062 if (block != compiler->fgFirstBB)
9064 processBlockStartLocations(block, false);
9067 // Handle the DummyDefs, updating the incoming var location.
9068 for (; currentRefPosition != refPositions.end() && currentRefPosition->refType == RefTypeDummyDef;
9069 ++currentRefPosition)
9071 assert(currentRefPosition->isIntervalRef());
9072 // Don't mark dummy defs as reload
9073 currentRefPosition->reload = false;
9074 resolveLocalRef(nullptr, nullptr, currentRefPosition);
9076 if (currentRefPosition->registerAssignment != RBM_NONE)
9078 reg = currentRefPosition->assignedReg();
9083 currentRefPosition->getInterval()->isActive = false;
9085 setInVarRegForBB(curBBNum, currentRefPosition->getInterval()->varNum, reg);
9089 // The next RefPosition should be for the block. Move past it.
9090 assert(currentRefPosition != refPositions.end());
9091 assert(currentRefPosition->refType == RefTypeBB);
9092 ++currentRefPosition;
9094 // Handle the RefPositions for the block
9095 for (; currentRefPosition != refPositions.end() && currentRefPosition->refType != RefTypeBB &&
9096 currentRefPosition->refType != RefTypeDummyDef;
9097 ++currentRefPosition)
9099 currentLocation = currentRefPosition->nodeLocation;
9101 // Ensure that the spill & copy info is valid.
9102 // First, if it's reload, it must not be copyReg or moveReg
9103 assert(!currentRefPosition->reload || (!currentRefPosition->copyReg && !currentRefPosition->moveReg));
9104 // If it's copyReg it must not be moveReg, and vice-versa
9105 assert(!currentRefPosition->copyReg || !currentRefPosition->moveReg);
9107 switch (currentRefPosition->refType)
9110 case RefTypeUpperVectorSaveUse:
9111 case RefTypeUpperVectorSaveDef:
9112 #endif // FEATURE_SIMD
9115 // These are the ones we're interested in
9118 case RefTypeFixedReg:
9119 // These require no handling at resolution time
9120 assert(currentRefPosition->referent != nullptr);
9121 currentRefPosition->referent->recentRefPosition = currentRefPosition;
9124 // Ignore the ExpUse cases - a RefTypeExpUse would only exist if the
9125 // variable is dead at the entry to the next block. So we'll mark
9126 // it as in its current location and resolution will take care of any
9128 assert(getNextBlock() == nullptr ||
9129 !VarSetOps::IsMember(compiler, getNextBlock()->bbLiveIn,
9130 currentRefPosition->getInterval()->getVarIndex(compiler)));
9131 currentRefPosition->referent->recentRefPosition = currentRefPosition;
9133 case RefTypeKillGCRefs:
9134 // No action to take at resolution time, and no interval to update recentRefPosition for.
9136 case RefTypeDummyDef:
9137 case RefTypeParamDef:
9138 case RefTypeZeroInit:
9139 // Should have handled all of these already
9144 updateMaxSpill(currentRefPosition);
9145 GenTree* treeNode = currentRefPosition->treeNode;
9147 #if FEATURE_PARTIAL_SIMD_CALLEE_SAVE
9148 if (currentRefPosition->refType == RefTypeUpperVectorSaveDef)
9150 // The treeNode must be a call, and this must be a RefPosition for a LargeVectorType LocalVar.
9151 // If the LocalVar is in a callee-save register, we are going to spill its upper half around the call.
9152 // If we have allocated a register to spill it to, we will use that; otherwise, we will spill it
9153 // to the stack. We can use as a temp register any non-arg caller-save register.
9154 noway_assert(treeNode != nullptr);
9155 currentRefPosition->referent->recentRefPosition = currentRefPosition;
9156 insertUpperVectorSaveAndReload(treeNode, currentRefPosition, block);
9158 else if (currentRefPosition->refType == RefTypeUpperVectorSaveUse)
9162 #endif // FEATURE_PARTIAL_SIMD_CALLEE_SAVE
9164 // Most uses won't actually need to be recorded (they're on the def).
9165 // In those cases, treeNode will be nullptr.
9166 if (treeNode == nullptr)
9168 // This is either a use, a dead def, or a field of a struct
9169 Interval* interval = currentRefPosition->getInterval();
9170 assert(currentRefPosition->refType == RefTypeUse ||
9171 currentRefPosition->registerAssignment == RBM_NONE || interval->isStructField);
9173 // TODO-Review: Need to handle the case where any of the struct fields
9174 // are reloaded/spilled at this use
9175 assert(!interval->isStructField ||
9176 (currentRefPosition->reload == false && currentRefPosition->spillAfter == false));
9178 if (interval->isLocalVar && !interval->isStructField)
9180 LclVarDsc* varDsc = interval->getLocalVar(compiler);
9182 // This must be a dead definition. We need to mark the lclVar
9183 // so that it's not considered a candidate for lvRegister, as
9184 // this dead def will have to go to the stack.
9185 assert(currentRefPosition->refType == RefTypeDef);
9186 varDsc->lvRegNum = REG_STK;
9191 LsraLocation loc = treeNode->gtLsraInfo.loc;
9192 assert(treeNode->IsLocal() || currentLocation == loc || currentLocation == loc + 1);
9194 if (currentRefPosition->isIntervalRef() && currentRefPosition->getInterval()->isInternal)
9196 treeNode->gtRsvdRegs |= currentRefPosition->registerAssignment;
9200 writeRegisters(currentRefPosition, treeNode);
9202 if (treeNode->IsLocal() && currentRefPosition->getInterval()->isLocalVar)
9204 resolveLocalRef(block, treeNode, currentRefPosition);
9207 // Mark spill locations on temps
9208 // (local vars are handled in resolveLocalRef, above)
9209 // Note that the tree node will be changed from GTF_SPILL to GTF_SPILLED
9210 // in codegen, taking care of the "reload" case for temps
9211 else if (currentRefPosition->spillAfter || (currentRefPosition->nextRefPosition != nullptr &&
9212 currentRefPosition->nextRefPosition->moveReg))
9214 if (treeNode != nullptr && currentRefPosition->isIntervalRef())
9216 if (currentRefPosition->spillAfter)
9218 treeNode->gtFlags |= GTF_SPILL;
9220 // If this is a constant interval that is reusing a pre-existing value, we actually need
9221 // to generate the value at this point in order to spill it.
9222 if (treeNode->IsReuseRegVal())
9224 treeNode->ResetReuseRegVal();
9227 // In case of multi-reg call node, also set spill flag on the
9228 // register specified by multi-reg index of current RefPosition.
9229 // Note that the spill flag on treeNode indicates that one or
9230 // more its allocated registers are in that state.
9231 if (treeNode->IsMultiRegCall())
9233 GenTreeCall* call = treeNode->AsCall();
9234 call->SetRegSpillFlagByIdx(GTF_SPILL, currentRefPosition->getMultiRegIdx());
9237 else if (treeNode->OperIsPutArgSplit())
9239 GenTreePutArgSplit* splitArg = treeNode->AsPutArgSplit();
9240 splitArg->SetRegSpillFlagByIdx(GTF_SPILL, currentRefPosition->getMultiRegIdx());
9245 // If the value is reloaded or moved to a different register, we need to insert
9246 // a node to hold the register to which it should be reloaded
9247 RefPosition* nextRefPosition = currentRefPosition->nextRefPosition;
9248 assert(nextRefPosition != nullptr);
9249 if (INDEBUG(alwaysInsertReload() ||)
9250 nextRefPosition->assignedReg() != currentRefPosition->assignedReg())
9252 if (nextRefPosition->assignedReg() != REG_NA)
9254 insertCopyOrReload(block, treeNode, currentRefPosition->getMultiRegIdx(),
9259 assert(nextRefPosition->AllocateIfProfitable());
9261 // In case of tree temps, if def is spilled and use didn't
9262 // get a register, set a flag on tree node to be treated as
9263 // contained at the point of its use.
9264 if (currentRefPosition->spillAfter && currentRefPosition->refType == RefTypeDef &&
9265 nextRefPosition->refType == RefTypeUse)
9267 assert(nextRefPosition->treeNode == nullptr);
9268 treeNode->gtFlags |= GTF_NOREG_AT_USE;
9274 // We should never have to "spill after" a temp use, since
9275 // they're single use
9284 if (enregisterLocalVars)
9286 processBlockEndLocations(block);
9290 if (enregisterLocalVars)
9295 printf("-----------------------\n");
9296 printf("RESOLVING BB BOUNDARIES\n");
9297 printf("-----------------------\n");
9299 printf("Resolution Candidates: ");
9300 dumpConvertedVarSet(compiler, resolutionCandidateVars);
9302 printf("Has %sCritical Edges\n\n", hasCriticalEdges ? "" : "No");
9304 printf("Prior to Resolution\n");
9305 foreach_block(compiler, block)
9307 printf("\nBB%02u use def in out\n", block->bbNum);
9308 dumpConvertedVarSet(compiler, block->bbVarUse);
9310 dumpConvertedVarSet(compiler, block->bbVarDef);
9312 dumpConvertedVarSet(compiler, block->bbLiveIn);
9314 dumpConvertedVarSet(compiler, block->bbLiveOut);
9317 dumpInVarToRegMap(block);
9318 dumpOutVarToRegMap(block);
9327 // Verify register assignments on variables
9330 for (lclNum = 0, varDsc = compiler->lvaTable; lclNum < compiler->lvaCount; lclNum++, varDsc++)
9332 if (!isCandidateVar(varDsc))
9334 varDsc->lvRegNum = REG_STK;
9338 Interval* interval = getIntervalForLocalVar(varDsc->lvVarIndex);
9340 // Determine initial position for parameters
9342 if (varDsc->lvIsParam)
9344 regMaskTP initialRegMask = interval->firstRefPosition->registerAssignment;
9345 regNumber initialReg = (initialRegMask == RBM_NONE || interval->firstRefPosition->spillAfter)
9347 : genRegNumFromMask(initialRegMask);
9348 regNumber sourceReg = (varDsc->lvIsRegArg) ? varDsc->lvArgReg : REG_STK;
9351 if (varTypeIsMultiReg(varDsc))
9353 // TODO-ARM-NYI: Map the hi/lo intervals back to lvRegNum and lvOtherReg (these should NYI
9355 assert(!"Multi-reg types not yet supported");
9358 #endif // _TARGET_ARM_
9360 varDsc->lvArgInitReg = initialReg;
9361 JITDUMP(" Set V%02u argument initial register to %s\n", lclNum, getRegName(initialReg));
9364 // Stack args that are part of dependently-promoted structs should never be register candidates (see
9365 // LinearScan::isRegCandidate).
9366 assert(varDsc->lvIsRegArg || !compiler->lvaIsFieldOfDependentlyPromotedStruct(varDsc));
9369 // If lvRegNum is REG_STK, that means that either no register
9370 // was assigned, or (more likely) that the same register was not
9371 // used for all references. In that case, codegen gets the register
9372 // from the tree node.
9373 if (varDsc->lvRegNum == REG_STK || interval->isSpilled || interval->isSplit)
9375 // For codegen purposes, we'll set lvRegNum to whatever register
9376 // it's currently in as we go.
9377 // However, we never mark an interval as lvRegister if it has either been spilled
9379 varDsc->lvRegister = false;
9381 // Skip any dead defs or exposed uses
9382 // (first use exposed will only occur when there is no explicit initialization)
9383 RefPosition* firstRefPosition = interval->firstRefPosition;
9384 while ((firstRefPosition != nullptr) && (firstRefPosition->refType == RefTypeExpUse))
9386 firstRefPosition = firstRefPosition->nextRefPosition;
9388 if (firstRefPosition == nullptr)
9391 varDsc->lvLRACandidate = false;
9392 if (varDsc->lvRefCnt == 0)
9394 varDsc->lvOnFrame = false;
9398 // We may encounter cases where a lclVar actually has no references, but
9399 // a non-zero refCnt. For safety (in case this is some "hidden" lclVar that we're
9400 // not correctly recognizing), we'll mark those as needing a stack location.
9401 // TODO-Cleanup: Make this an assert if/when we correct the refCnt
9403 varDsc->lvOnFrame = true;
9408 // If the interval was not spilled, it doesn't need a stack location.
9409 if (!interval->isSpilled)
9411 varDsc->lvOnFrame = false;
9413 if (firstRefPosition->registerAssignment == RBM_NONE || firstRefPosition->spillAfter)
9415 // Either this RefPosition is spilled, or regOptional or it is not a "real" def or use
9417 firstRefPosition->spillAfter || firstRefPosition->AllocateIfProfitable() ||
9418 (firstRefPosition->refType != RefTypeDef && firstRefPosition->refType != RefTypeUse));
9419 varDsc->lvRegNum = REG_STK;
9423 varDsc->lvRegNum = firstRefPosition->assignedReg();
9430 varDsc->lvRegister = true;
9431 varDsc->lvOnFrame = false;
9434 regMaskTP registerAssignment = genRegMask(varDsc->lvRegNum);
9435 assert(!interval->isSpilled && !interval->isSplit);
9436 RefPosition* refPosition = interval->firstRefPosition;
9437 assert(refPosition != nullptr);
9439 while (refPosition != nullptr)
9441 // All RefPositions must match, except for dead definitions,
9442 // copyReg/moveReg and RefTypeExpUse positions
9443 if (refPosition->registerAssignment != RBM_NONE && !refPosition->copyReg &&
9444 !refPosition->moveReg && refPosition->refType != RefTypeExpUse)
9446 assert(refPosition->registerAssignment == registerAssignment);
9448 refPosition = refPosition->nextRefPosition;
9459 printf("Trees after linear scan register allocator (LSRA)\n");
9460 compiler->fgDispBasicBlocks(true);
9463 verifyFinalAllocation();
9466 compiler->raMarkStkVars();
9469 // TODO-CQ: Review this comment and address as needed.
9470 // Change all unused promoted non-argument struct locals to a non-GC type (in this case TYP_INT)
9471 // so that the gc tracking logic and lvMustInit logic will ignore them.
9472 // Extract the code that does this from raAssignVars, and call it here.
9473 // PRECONDITIONS: Ensure that lvPromoted is set on promoted structs, if and
9474 // only if it is promoted on all paths.
9475 // Call might be something like:
9476 // compiler->BashUnusedStructLocals();
9480 //------------------------------------------------------------------------
9481 // insertMove: Insert a move of a lclVar with the given lclNum into the given block.
9484 // block - the BasicBlock into which the move will be inserted.
9485 // insertionPoint - the instruction before which to insert the move
9486 // lclNum - the lclNum of the var to be moved
9487 // fromReg - the register from which the var is moving
9488 // toReg - the register to which the var is moving
9494 // If insertionPoint is non-NULL, insert before that instruction;
9495 // otherwise, insert "near" the end (prior to the branch, if any).
9496 // If fromReg or toReg is REG_STK, then move from/to memory, respectively.
9498 void LinearScan::insertMove(
9499 BasicBlock* block, GenTreePtr insertionPoint, unsigned lclNum, regNumber fromReg, regNumber toReg)
9501 LclVarDsc* varDsc = compiler->lvaTable + lclNum;
9502 // the lclVar must be a register candidate
9503 assert(isRegCandidate(varDsc));
9504 // One or both MUST be a register
9505 assert(fromReg != REG_STK || toReg != REG_STK);
9506 // They must not be the same register.
9507 assert(fromReg != toReg);
9509 // This var can't be marked lvRegister now
9510 varDsc->lvRegNum = REG_STK;
9512 GenTreePtr src = compiler->gtNewLclvNode(lclNum, varDsc->TypeGet());
9513 src->gtLsraInfo.isLsraAdded = true;
9515 // There are three cases we need to handle:
9516 // - We are loading a lclVar from the stack.
9517 // - We are storing a lclVar to the stack.
9518 // - We are copying a lclVar between registers.
9520 // In the first and second cases, the lclVar node will be marked with GTF_SPILLED and GTF_SPILL, respectively.
9521 // It is up to the code generator to ensure that any necessary normalization is done when loading or storing the
9524 // In the third case, we generate GT_COPY(GT_LCL_VAR) and type each node with the normalized type of the lclVar.
9525 // This is safe because a lclVar is always normalized once it is in a register.
9528 if (fromReg == REG_STK)
9530 src->gtFlags |= GTF_SPILLED;
9531 src->gtRegNum = toReg;
9533 else if (toReg == REG_STK)
9535 src->gtFlags |= GTF_SPILL;
9536 src->gtRegNum = fromReg;
9540 var_types movType = genActualType(varDsc->TypeGet());
9541 src->gtType = movType;
9543 dst = new (compiler, GT_COPY) GenTreeCopyOrReload(GT_COPY, movType, src);
9544 // This is the new home of the lclVar - indicate that by clearing the GTF_VAR_DEATH flag.
9545 // Note that if src is itself a lastUse, this will have no effect.
9546 dst->gtFlags &= ~(GTF_VAR_DEATH);
9547 src->gtRegNum = fromReg;
9548 dst->gtRegNum = toReg;
9549 src->gtLsraInfo.isLocalDefUse = false;
9550 dst->gtLsraInfo.isLsraAdded = true;
9552 dst->gtLsraInfo.isLocalDefUse = true;
9554 LIR::Range treeRange = LIR::SeqTree(compiler, dst);
9555 LIR::Range& blockRange = LIR::AsRange(block);
9557 if (insertionPoint != nullptr)
9559 blockRange.InsertBefore(insertionPoint, std::move(treeRange));
9563 // Put the copy at the bottom
9564 // If there's a branch, make an embedded statement that executes just prior to the branch
9565 if (block->bbJumpKind == BBJ_COND || block->bbJumpKind == BBJ_SWITCH)
9567 noway_assert(!blockRange.IsEmpty());
9569 GenTree* branch = blockRange.LastNode();
9570 assert(branch->OperIsConditionalJump() || branch->OperGet() == GT_SWITCH_TABLE ||
9571 branch->OperGet() == GT_SWITCH);
9573 blockRange.InsertBefore(branch, std::move(treeRange));
9577 assert(block->bbJumpKind == BBJ_NONE || block->bbJumpKind == BBJ_ALWAYS);
9578 blockRange.InsertAtEnd(std::move(treeRange));
9583 void LinearScan::insertSwap(
9584 BasicBlock* block, GenTreePtr insertionPoint, unsigned lclNum1, regNumber reg1, unsigned lclNum2, regNumber reg2)
9589 const char* insertionPointString = "top";
9590 if (insertionPoint == nullptr)
9592 insertionPointString = "bottom";
9594 printf(" BB%02u %s: swap V%02u in %s with V%02u in %s\n", block->bbNum, insertionPointString, lclNum1,
9595 getRegName(reg1), lclNum2, getRegName(reg2));
9599 LclVarDsc* varDsc1 = compiler->lvaTable + lclNum1;
9600 LclVarDsc* varDsc2 = compiler->lvaTable + lclNum2;
9601 assert(reg1 != REG_STK && reg1 != REG_NA && reg2 != REG_STK && reg2 != REG_NA);
9603 GenTreePtr lcl1 = compiler->gtNewLclvNode(lclNum1, varDsc1->TypeGet());
9604 lcl1->gtLsraInfo.isLsraAdded = true;
9605 lcl1->gtLsraInfo.isLocalDefUse = false;
9606 lcl1->gtRegNum = reg1;
9608 GenTreePtr lcl2 = compiler->gtNewLclvNode(lclNum2, varDsc2->TypeGet());
9609 lcl2->gtLsraInfo.isLsraAdded = true;
9610 lcl2->gtLsraInfo.isLocalDefUse = false;
9611 lcl2->gtRegNum = reg2;
9613 GenTreePtr swap = compiler->gtNewOperNode(GT_SWAP, TYP_VOID, lcl1, lcl2);
9614 swap->gtLsraInfo.isLsraAdded = true;
9615 swap->gtLsraInfo.isLocalDefUse = false;
9616 swap->gtRegNum = REG_NA;
9618 lcl1->gtNext = lcl2;
9619 lcl2->gtPrev = lcl1;
9620 lcl2->gtNext = swap;
9621 swap->gtPrev = lcl2;
9623 LIR::Range swapRange = LIR::SeqTree(compiler, swap);
9624 LIR::Range& blockRange = LIR::AsRange(block);
9626 if (insertionPoint != nullptr)
9628 blockRange.InsertBefore(insertionPoint, std::move(swapRange));
9632 // Put the copy at the bottom
9633 // If there's a branch, make an embedded statement that executes just prior to the branch
9634 if (block->bbJumpKind == BBJ_COND || block->bbJumpKind == BBJ_SWITCH)
9636 noway_assert(!blockRange.IsEmpty());
9638 GenTree* branch = blockRange.LastNode();
9639 assert(branch->OperIsConditionalJump() || branch->OperGet() == GT_SWITCH_TABLE ||
9640 branch->OperGet() == GT_SWITCH);
9642 blockRange.InsertBefore(branch, std::move(swapRange));
9646 assert(block->bbJumpKind == BBJ_NONE || block->bbJumpKind == BBJ_ALWAYS);
9647 blockRange.InsertAtEnd(std::move(swapRange));
9652 //------------------------------------------------------------------------
9653 // getTempRegForResolution: Get a free register to use for resolution code.
9656 // fromBlock - The "from" block on the edge being resolved.
9657 // toBlock - The "to"block on the edge
9658 // type - the type of register required
9661 // Returns a register that is free on the given edge, or REG_NA if none is available.
9664 // It is up to the caller to check the return value, and to determine whether a register is
9665 // available, and to handle that case appropriately.
9666 // It is also up to the caller to cache the return value, as this is not cheap to compute.
9668 regNumber LinearScan::getTempRegForResolution(BasicBlock* fromBlock, BasicBlock* toBlock, var_types type)
9670 // TODO-Throughput: This would be much more efficient if we add RegToVarMaps instead of VarToRegMaps
9671 // and they would be more space-efficient as well.
9672 VarToRegMap fromVarToRegMap = getOutVarToRegMap(fromBlock->bbNum);
9673 VarToRegMap toVarToRegMap = getInVarToRegMap(toBlock->bbNum);
9675 regMaskTP freeRegs = allRegs(type);
9677 if (getStressLimitRegs() == LSRA_LIMIT_SMALL_SET)
9682 INDEBUG(freeRegs = stressLimitRegs(nullptr, freeRegs));
9684 // We are only interested in the variables that are live-in to the "to" block.
9685 VarSetOps::Iter iter(compiler, toBlock->bbLiveIn);
9686 unsigned varIndex = 0;
9687 while (iter.NextElem(&varIndex) && freeRegs != RBM_NONE)
9689 regNumber fromReg = getVarReg(fromVarToRegMap, varIndex);
9690 regNumber toReg = getVarReg(toVarToRegMap, varIndex);
9691 assert(fromReg != REG_NA && toReg != REG_NA);
9692 if (fromReg != REG_STK)
9694 freeRegs &= ~genRegMask(fromReg);
9696 if (toReg != REG_STK)
9698 freeRegs &= ~genRegMask(toReg);
9701 if (freeRegs == RBM_NONE)
9707 regNumber tempReg = genRegNumFromMask(genFindLowestBit(freeRegs));
9712 //------------------------------------------------------------------------
9713 // addResolution: Add a resolution move of the given interval
9716 // block - the BasicBlock into which the move will be inserted.
9717 // insertionPoint - the instruction before which to insert the move
9718 // interval - the interval of the var to be moved
9719 // toReg - the register to which the var is moving
9720 // fromReg - the register from which the var is moving
9726 // For joins, we insert at the bottom (indicated by an insertionPoint
9727 // of nullptr), while for splits we insert at the top.
9728 // This is because for joins 'block' is a pred of the join, while for splits it is a succ.
9729 // For critical edges, this function may be called twice - once to move from
9730 // the source (fromReg), if any, to the stack, in which case toReg will be
9731 // REG_STK, and we insert at the bottom (leave insertionPoint as nullptr).
9732 // The next time, we want to move from the stack to the destination (toReg),
9733 // in which case fromReg will be REG_STK, and we insert at the top.
9735 void LinearScan::addResolution(
9736 BasicBlock* block, GenTreePtr insertionPoint, Interval* interval, regNumber toReg, regNumber fromReg)
9739 const char* insertionPointString = "top";
9741 if (insertionPoint == nullptr)
9744 insertionPointString = "bottom";
9748 JITDUMP(" BB%02u %s: move V%02u from ", block->bbNum, insertionPointString, interval->varNum);
9749 JITDUMP("%s to %s", getRegName(fromReg), getRegName(toReg));
9751 insertMove(block, insertionPoint, interval->varNum, fromReg, toReg);
9752 if (fromReg == REG_STK || toReg == REG_STK)
9754 assert(interval->isSpilled);
9758 // We should have already marked this as spilled or split.
9759 assert((interval->isSpilled) || (interval->isSplit));
9762 INTRACK_STATS(updateLsraStat(LSRA_STAT_RESOLUTION_MOV, block->bbNum));
9765 //------------------------------------------------------------------------
9766 // handleOutgoingCriticalEdges: Performs the necessary resolution on all critical edges that feed out of 'block'
9769 // block - the block with outgoing critical edges.
9775 // For all outgoing critical edges (i.e. any successor of this block which is
9776 // a join edge), if there are any conflicts, split the edge by adding a new block,
9777 // and generate the resolution code into that block.
9779 void LinearScan::handleOutgoingCriticalEdges(BasicBlock* block)
9781 VARSET_TP outResolutionSet(VarSetOps::Intersection(compiler, block->bbLiveOut, resolutionCandidateVars));
9782 if (VarSetOps::IsEmpty(compiler, outResolutionSet))
9786 VARSET_TP sameResolutionSet(VarSetOps::MakeEmpty(compiler));
9787 VARSET_TP sameLivePathsSet(VarSetOps::MakeEmpty(compiler));
9788 VARSET_TP singleTargetSet(VarSetOps::MakeEmpty(compiler));
9789 VARSET_TP diffResolutionSet(VarSetOps::MakeEmpty(compiler));
9791 // Get the outVarToRegMap for this block
9792 VarToRegMap outVarToRegMap = getOutVarToRegMap(block->bbNum);
9793 unsigned succCount = block->NumSucc(compiler);
9794 assert(succCount > 1);
9795 VarToRegMap firstSuccInVarToRegMap = nullptr;
9796 BasicBlock* firstSucc = nullptr;
9798 // First, determine the live regs at the end of this block so that we know what regs are
9799 // available to copy into.
9800 // Note that for this purpose we use the full live-out set, because we must ensure that
9801 // even the registers that remain the same across the edge are preserved correctly.
9802 regMaskTP liveOutRegs = RBM_NONE;
9803 VarSetOps::Iter liveOutIter(compiler, block->bbLiveOut);
9804 unsigned liveOutVarIndex = 0;
9805 while (liveOutIter.NextElem(&liveOutVarIndex))
9807 regNumber fromReg = getVarReg(outVarToRegMap, liveOutVarIndex);
9808 if (fromReg != REG_STK)
9810 liveOutRegs |= genRegMask(fromReg);
9814 // Next, if this blocks ends with a switch table, we have to make sure not to copy
9815 // into the registers that it uses.
9816 regMaskTP switchRegs = RBM_NONE;
9817 if (block->bbJumpKind == BBJ_SWITCH)
9819 // At this point, Lowering has transformed any non-switch-table blocks into
9821 GenTree* switchTable = LIR::AsRange(block).LastNode();
9822 assert(switchTable != nullptr && switchTable->OperGet() == GT_SWITCH_TABLE);
9824 switchRegs = switchTable->gtRsvdRegs;
9825 GenTree* op1 = switchTable->gtGetOp1();
9826 GenTree* op2 = switchTable->gtGetOp2();
9827 noway_assert(op1 != nullptr && op2 != nullptr);
9828 assert(op1->gtRegNum != REG_NA && op2->gtRegNum != REG_NA);
9829 switchRegs |= genRegMask(op1->gtRegNum);
9830 switchRegs |= genRegMask(op2->gtRegNum);
9833 VarToRegMap sameVarToRegMap = sharedCriticalVarToRegMap;
9834 regMaskTP sameWriteRegs = RBM_NONE;
9835 regMaskTP diffReadRegs = RBM_NONE;
9837 // For each var that may require resolution, classify them as:
9838 // - in the same register at the end of this block and at each target (no resolution needed)
9839 // - in different registers at different targets (resolve separately):
9840 // diffResolutionSet
9841 // - in the same register at each target at which it's live, but different from the end of
9842 // this block. We may be able to resolve these as if it is "join", but only if they do not
9843 // write to any registers that are read by those in the diffResolutionSet:
9844 // sameResolutionSet
9846 VarSetOps::Iter outResolutionSetIter(compiler, outResolutionSet);
9847 unsigned outResolutionSetVarIndex = 0;
9848 while (outResolutionSetIter.NextElem(&outResolutionSetVarIndex))
9850 regNumber fromReg = getVarReg(outVarToRegMap, outResolutionSetVarIndex);
9851 bool isMatch = true;
9852 bool isSame = false;
9853 bool maybeSingleTarget = false;
9854 bool maybeSameLivePaths = false;
9855 bool liveOnlyAtSplitEdge = true;
9856 regNumber sameToReg = REG_NA;
9857 for (unsigned succIndex = 0; succIndex < succCount; succIndex++)
9859 BasicBlock* succBlock = block->GetSucc(succIndex, compiler);
9860 if (!VarSetOps::IsMember(compiler, succBlock->bbLiveIn, outResolutionSetVarIndex))
9862 maybeSameLivePaths = true;
9865 else if (liveOnlyAtSplitEdge)
9867 // Is the var live only at those target blocks which are connected by a split edge to this block
9868 liveOnlyAtSplitEdge = ((succBlock->bbPreds->flNext == nullptr) && (succBlock != compiler->fgFirstBB));
9871 regNumber toReg = getVarReg(getInVarToRegMap(succBlock->bbNum), outResolutionSetVarIndex);
9872 if (sameToReg == REG_NA)
9877 if (toReg == sameToReg)
9885 // Check for the cases where we can't write to a register.
9886 // We only need to check for these cases if sameToReg is an actual register (not REG_STK).
9887 if (sameToReg != REG_NA && sameToReg != REG_STK)
9889 // If there's a path on which this var isn't live, it may use the original value in sameToReg.
9890 // In this case, sameToReg will be in the liveOutRegs of this block.
9891 // Similarly, if sameToReg is in sameWriteRegs, it has already been used (i.e. for a lclVar that's
9892 // live only at another target), and we can't copy another lclVar into that reg in this block.
9893 regMaskTP sameToRegMask = genRegMask(sameToReg);
9894 if (maybeSameLivePaths &&
9895 (((sameToRegMask & liveOutRegs) != RBM_NONE) || ((sameToRegMask & sameWriteRegs) != RBM_NONE)))
9899 // If this register is used by a switch table at the end of the block, we can't do the copy
9900 // in this block (since we can't insert it after the switch).
9901 if ((sameToRegMask & switchRegs) != RBM_NONE)
9906 // If the var is live only at those blocks connected by a split edge and not live-in at some of the
9907 // target blocks, we will resolve it the same way as if it were in diffResolutionSet and resolution
9908 // will be deferred to the handling of split edges, which means copy will only be at those target(s).
9910 // Another way to achieve similar resolution for vars live only at split edges is by removing them
9911 // from consideration up-front but it requires that we traverse those edges anyway to account for
9912 // the registers that must note be overwritten.
9913 if (liveOnlyAtSplitEdge && maybeSameLivePaths)
9919 if (sameToReg == REG_NA)
9921 VarSetOps::AddElemD(compiler, diffResolutionSet, outResolutionSetVarIndex);
9922 if (fromReg != REG_STK)
9924 diffReadRegs |= genRegMask(fromReg);
9927 else if (sameToReg != fromReg)
9929 VarSetOps::AddElemD(compiler, sameResolutionSet, outResolutionSetVarIndex);
9930 setVarReg(sameVarToRegMap, outResolutionSetVarIndex, sameToReg);
9931 if (sameToReg != REG_STK)
9933 sameWriteRegs |= genRegMask(sameToReg);
9938 if (!VarSetOps::IsEmpty(compiler, sameResolutionSet))
9940 if ((sameWriteRegs & diffReadRegs) != RBM_NONE)
9942 // We cannot split the "same" and "diff" regs if the "same" set writes registers
9943 // that must be read by the "diff" set. (Note that when these are done as a "batch"
9944 // we carefully order them to ensure all the input regs are read before they are
9946 VarSetOps::UnionD(compiler, diffResolutionSet, sameResolutionSet);
9947 VarSetOps::ClearD(compiler, sameResolutionSet);
9951 // For any vars in the sameResolutionSet, we can simply add the move at the end of "block".
9952 resolveEdge(block, nullptr, ResolveSharedCritical, sameResolutionSet);
9955 if (!VarSetOps::IsEmpty(compiler, diffResolutionSet))
9957 for (unsigned succIndex = 0; succIndex < succCount; succIndex++)
9959 BasicBlock* succBlock = block->GetSucc(succIndex, compiler);
9961 // Any "diffResolutionSet" resolution for a block with no other predecessors will be handled later
9962 // as split resolution.
9963 if ((succBlock->bbPreds->flNext == nullptr) && (succBlock != compiler->fgFirstBB))
9968 // Now collect the resolution set for just this edge, if any.
9969 // Check only the vars in diffResolutionSet that are live-in to this successor.
9970 bool needsResolution = false;
9971 VarToRegMap succInVarToRegMap = getInVarToRegMap(succBlock->bbNum);
9972 VARSET_TP edgeResolutionSet(VarSetOps::Intersection(compiler, diffResolutionSet, succBlock->bbLiveIn));
9973 VarSetOps::Iter iter(compiler, edgeResolutionSet);
9974 unsigned varIndex = 0;
9975 while (iter.NextElem(&varIndex))
9977 regNumber fromReg = getVarReg(outVarToRegMap, varIndex);
9978 regNumber toReg = getVarReg(succInVarToRegMap, varIndex);
9980 if (fromReg == toReg)
9982 VarSetOps::RemoveElemD(compiler, edgeResolutionSet, varIndex);
9985 if (!VarSetOps::IsEmpty(compiler, edgeResolutionSet))
9987 resolveEdge(block, succBlock, ResolveCritical, edgeResolutionSet);
9993 //------------------------------------------------------------------------
9994 // resolveEdges: Perform resolution across basic block edges
10003 // Traverse the basic blocks.
10004 // - If this block has a single predecessor that is not the immediately
10005 // preceding block, perform any needed 'split' resolution at the beginning of this block
10006 // - Otherwise if this block has critical incoming edges, handle them.
10007 // - If this block has a single successor that has multiple predecesors, perform any needed
10008 // 'join' resolution at the end of this block.
10009 // Note that a block may have both 'split' or 'critical' incoming edge(s) and 'join' outgoing
10012 void LinearScan::resolveEdges()
10014 JITDUMP("RESOLVING EDGES\n");
10016 // The resolutionCandidateVars set was initialized with all the lclVars that are live-in to
10017 // any block. We now intersect that set with any lclVars that ever spilled or split.
10018 // If there are no candidates for resoultion, simply return.
10020 VarSetOps::IntersectionD(compiler, resolutionCandidateVars, splitOrSpilledVars);
10021 if (VarSetOps::IsEmpty(compiler, resolutionCandidateVars))
10026 BasicBlock *block, *prevBlock = nullptr;
10028 // Handle all the critical edges first.
10029 // We will try to avoid resolution across critical edges in cases where all the critical-edge
10030 // targets of a block have the same home. We will then split the edges only for the
10031 // remaining mismatches. We visit the out-edges, as that allows us to share the moves that are
10032 // common among allt he targets.
10034 if (hasCriticalEdges)
10036 foreach_block(compiler, block)
10038 if (block->bbNum > bbNumMaxBeforeResolution)
10040 // This is a new block added during resolution - we don't need to visit these now.
10043 if (blockInfo[block->bbNum].hasCriticalOutEdge)
10045 handleOutgoingCriticalEdges(block);
10051 prevBlock = nullptr;
10052 foreach_block(compiler, block)
10054 if (block->bbNum > bbNumMaxBeforeResolution)
10056 // This is a new block added during resolution - we don't need to visit these now.
10060 unsigned succCount = block->NumSucc(compiler);
10061 flowList* preds = block->bbPreds;
10062 BasicBlock* uniquePredBlock = block->GetUniquePred(compiler);
10064 // First, if this block has a single predecessor,
10065 // we may need resolution at the beginning of this block.
10066 // This may be true even if it's the block we used for starting locations,
10067 // if a variable was spilled.
10068 VARSET_TP inResolutionSet(VarSetOps::Intersection(compiler, block->bbLiveIn, resolutionCandidateVars));
10069 if (!VarSetOps::IsEmpty(compiler, inResolutionSet))
10071 if (uniquePredBlock != nullptr)
10073 // We may have split edges during critical edge resolution, and in the process split
10074 // a non-critical edge as well.
10075 // It is unlikely that we would ever have more than one of these in sequence (indeed,
10076 // I don't think it's possible), but there's no need to assume that it can't.
10077 while (uniquePredBlock->bbNum > bbNumMaxBeforeResolution)
10079 uniquePredBlock = uniquePredBlock->GetUniquePred(compiler);
10080 noway_assert(uniquePredBlock != nullptr);
10082 resolveEdge(uniquePredBlock, block, ResolveSplit, inResolutionSet);
10086 // Finally, if this block has a single successor:
10087 // - and that has at least one other predecessor (otherwise we will do the resolution at the
10088 // top of the successor),
10089 // - and that is not the target of a critical edge (otherwise we've already handled it)
10090 // we may need resolution at the end of this block.
10092 if (succCount == 1)
10094 BasicBlock* succBlock = block->GetSucc(0, compiler);
10095 if (succBlock->GetUniquePred(compiler) == nullptr)
10097 VARSET_TP outResolutionSet(
10098 VarSetOps::Intersection(compiler, succBlock->bbLiveIn, resolutionCandidateVars));
10099 if (!VarSetOps::IsEmpty(compiler, outResolutionSet))
10101 resolveEdge(block, succBlock, ResolveJoin, outResolutionSet);
10107 // Now, fixup the mapping for any blocks that were adding for edge splitting.
10108 // See the comment prior to the call to fgSplitEdge() in resolveEdge().
10109 // Note that we could fold this loop in with the checking code below, but that
10110 // would only improve the debug case, and would clutter up the code somewhat.
10111 if (compiler->fgBBNumMax > bbNumMaxBeforeResolution)
10113 foreach_block(compiler, block)
10115 if (block->bbNum > bbNumMaxBeforeResolution)
10117 // There may be multiple blocks inserted when we split. But we must always have exactly
10118 // one path (i.e. all blocks must be single-successor and single-predecessor),
10119 // and only one block along the path may be non-empty.
10120 // Note that we may have a newly-inserted block that is empty, but which connects
10121 // two non-resolution blocks. This happens when an edge is split that requires it.
10123 BasicBlock* succBlock = block;
10126 succBlock = succBlock->GetUniqueSucc();
10127 noway_assert(succBlock != nullptr);
10128 } while ((succBlock->bbNum > bbNumMaxBeforeResolution) && succBlock->isEmpty());
10130 BasicBlock* predBlock = block;
10133 predBlock = predBlock->GetUniquePred(compiler);
10134 noway_assert(predBlock != nullptr);
10135 } while ((predBlock->bbNum > bbNumMaxBeforeResolution) && predBlock->isEmpty());
10137 unsigned succBBNum = succBlock->bbNum;
10138 unsigned predBBNum = predBlock->bbNum;
10139 if (block->isEmpty())
10141 // For the case of the empty block, find the non-resolution block (succ or pred).
10142 if (predBBNum > bbNumMaxBeforeResolution)
10144 assert(succBBNum <= bbNumMaxBeforeResolution);
10154 assert((succBBNum <= bbNumMaxBeforeResolution) && (predBBNum <= bbNumMaxBeforeResolution));
10156 SplitEdgeInfo info = {predBBNum, succBBNum};
10157 getSplitBBNumToTargetBBNumMap()->Set(block->bbNum, info);
10163 // Make sure the varToRegMaps match up on all edges.
10164 bool foundMismatch = false;
10165 foreach_block(compiler, block)
10167 if (block->isEmpty() && block->bbNum > bbNumMaxBeforeResolution)
10171 VarToRegMap toVarToRegMap = getInVarToRegMap(block->bbNum);
10172 for (flowList* pred = block->bbPreds; pred != nullptr; pred = pred->flNext)
10174 BasicBlock* predBlock = pred->flBlock;
10175 VarToRegMap fromVarToRegMap = getOutVarToRegMap(predBlock->bbNum);
10176 VarSetOps::Iter iter(compiler, block->bbLiveIn);
10177 unsigned varIndex = 0;
10178 while (iter.NextElem(&varIndex))
10180 regNumber fromReg = getVarReg(fromVarToRegMap, varIndex);
10181 regNumber toReg = getVarReg(toVarToRegMap, varIndex);
10182 if (fromReg != toReg)
10184 if (!foundMismatch)
10186 foundMismatch = true;
10187 printf("Found mismatched var locations after resolution!\n");
10189 unsigned varNum = compiler->lvaTrackedToVarNum[varIndex];
10190 printf(" V%02u: BB%02u to BB%02u: %s to %s\n", varNum, predBlock->bbNum, block->bbNum,
10191 getRegName(fromReg), getRegName(toReg));
10196 assert(!foundMismatch);
10201 //------------------------------------------------------------------------
10202 // resolveEdge: Perform the specified type of resolution between two blocks.
10205 // fromBlock - the block from which the edge originates
10206 // toBlock - the block at which the edge terminates
10207 // resolveType - the type of resolution to be performed
10208 // liveSet - the set of tracked lclVar indices which may require resolution
10214 // The caller must have performed the analysis to determine the type of the edge.
10217 // This method emits the correctly ordered moves necessary to place variables in the
10218 // correct registers across a Split, Join or Critical edge.
10219 // In order to avoid overwriting register values before they have been moved to their
10220 // new home (register/stack), it first does the register-to-stack moves (to free those
10221 // registers), then the register to register moves, ensuring that the target register
10222 // is free before the move, and then finally the stack to register moves.
10224 void LinearScan::resolveEdge(BasicBlock* fromBlock,
10225 BasicBlock* toBlock,
10226 ResolveType resolveType,
10227 VARSET_VALARG_TP liveSet)
10229 VarToRegMap fromVarToRegMap = getOutVarToRegMap(fromBlock->bbNum);
10230 VarToRegMap toVarToRegMap;
10231 if (resolveType == ResolveSharedCritical)
10233 toVarToRegMap = sharedCriticalVarToRegMap;
10237 toVarToRegMap = getInVarToRegMap(toBlock->bbNum);
10240 // The block to which we add the resolution moves depends on the resolveType
10242 switch (resolveType)
10245 case ResolveSharedCritical:
10251 case ResolveCritical:
10252 // fgSplitEdge may add one or two BasicBlocks. It returns the block that splits
10253 // the edge from 'fromBlock' and 'toBlock', but if it inserts that block right after
10254 // a block with a fall-through it will have to create another block to handle that edge.
10255 // These new blocks can be mapped to existing blocks in order to correctly handle
10256 // the calls to recordVarLocationsAtStartOfBB() from codegen. That mapping is handled
10257 // in resolveEdges(), after all the edge resolution has been done (by calling this
10258 // method for each edge).
10259 block = compiler->fgSplitEdge(fromBlock, toBlock);
10261 // Split edges are counted against fromBlock.
10262 INTRACK_STATS(updateLsraStat(LSRA_STAT_SPLIT_EDGE, fromBlock->bbNum));
10269 #ifndef _TARGET_XARCH_
10270 // We record tempregs for beginning and end of each block.
10271 // For amd64/x86 we only need a tempReg for float - we'll use xchg for int.
10272 // TODO-Throughput: It would be better to determine the tempRegs on demand, but the code below
10273 // modifies the varToRegMaps so we don't have all the correct registers at the time
10274 // we need to get the tempReg.
10275 regNumber tempRegInt =
10276 (resolveType == ResolveSharedCritical) ? REG_NA : getTempRegForResolution(fromBlock, toBlock, TYP_INT);
10277 #endif // !_TARGET_XARCH_
10278 regNumber tempRegFlt = REG_NA;
10279 if ((compiler->compFloatingPointUsed) && (resolveType != ResolveSharedCritical))
10282 #ifdef _TARGET_ARM_
10283 // Let's try to reserve a double register for TYP_FLOAT and TYP_DOUBLE
10284 tempRegFlt = getTempRegForResolution(fromBlock, toBlock, TYP_DOUBLE);
10285 if (tempRegFlt == REG_NA)
10287 // If fails, try to reserve a float register for TYP_FLOAT
10288 tempRegFlt = getTempRegForResolution(fromBlock, toBlock, TYP_FLOAT);
10291 tempRegFlt = getTempRegForResolution(fromBlock, toBlock, TYP_FLOAT);
10295 regMaskTP targetRegsToDo = RBM_NONE;
10296 regMaskTP targetRegsReady = RBM_NONE;
10297 regMaskTP targetRegsFromStack = RBM_NONE;
10299 // The following arrays capture the location of the registers as they are moved:
10300 // - location[reg] gives the current location of the var that was originally in 'reg'.
10301 // (Note that a var may be moved more than once.)
10302 // - source[reg] gives the original location of the var that needs to be moved to 'reg'.
10303 // For example, if a var is in rax and needs to be moved to rsi, then we would start with:
10304 // location[rax] == rax
10305 // source[rsi] == rax -- this doesn't change
10306 // Then, if for some reason we need to move it temporary to rbx, we would have:
10307 // location[rax] == rbx
10308 // Once we have completed the move, we will have:
10309 // location[rax] == REG_NA
10310 // This indicates that the var originally in rax is now in its target register.
10312 regNumberSmall location[REG_COUNT];
10313 C_ASSERT(sizeof(char) == sizeof(regNumberSmall)); // for memset to work
10314 memset(location, REG_NA, REG_COUNT);
10315 regNumberSmall source[REG_COUNT];
10316 memset(source, REG_NA, REG_COUNT);
10318 // What interval is this register associated with?
10319 // (associated with incoming reg)
10320 Interval* sourceIntervals[REG_COUNT];
10321 memset(&sourceIntervals, 0, sizeof(sourceIntervals));
10323 // Intervals for vars that need to be loaded from the stack
10324 Interval* stackToRegIntervals[REG_COUNT];
10325 memset(&stackToRegIntervals, 0, sizeof(stackToRegIntervals));
10327 // Get the starting insertion point for the "to" resolution
10328 GenTreePtr insertionPoint = nullptr;
10329 if (resolveType == ResolveSplit || resolveType == ResolveCritical)
10331 insertionPoint = LIR::AsRange(block).FirstNonPhiNode();
10335 // - Perform all moves from reg to stack (no ordering needed on these)
10336 // - For reg to reg moves, record the current location, associating their
10337 // source location with the target register they need to go into
10338 // - For stack to reg moves (done last, no ordering needed between them)
10339 // record the interval associated with the target reg
10340 // TODO-Throughput: We should be looping over the liveIn and liveOut registers, since
10341 // that will scale better than the live variables
10343 VarSetOps::Iter iter(compiler, liveSet);
10344 unsigned varIndex = 0;
10345 while (iter.NextElem(&varIndex))
10347 regNumber fromReg = getVarReg(fromVarToRegMap, varIndex);
10348 regNumber toReg = getVarReg(toVarToRegMap, varIndex);
10349 if (fromReg == toReg)
10354 // For Critical edges, the location will not change on either side of the edge,
10355 // since we'll add a new block to do the move.
10356 if (resolveType == ResolveSplit)
10358 setVarReg(toVarToRegMap, varIndex, fromReg);
10360 else if (resolveType == ResolveJoin || resolveType == ResolveSharedCritical)
10362 setVarReg(fromVarToRegMap, varIndex, toReg);
10365 assert(fromReg < UCHAR_MAX && toReg < UCHAR_MAX);
10367 Interval* interval = getIntervalForLocalVar(varIndex);
10369 if (fromReg == REG_STK)
10371 stackToRegIntervals[toReg] = interval;
10372 targetRegsFromStack |= genRegMask(toReg);
10374 else if (toReg == REG_STK)
10376 // Do the reg to stack moves now
10377 addResolution(block, insertionPoint, interval, REG_STK, fromReg);
10378 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10382 location[fromReg] = (regNumberSmall)fromReg;
10383 source[toReg] = (regNumberSmall)fromReg;
10384 sourceIntervals[fromReg] = interval;
10385 targetRegsToDo |= genRegMask(toReg);
10389 // REGISTER to REGISTER MOVES
10391 // First, find all the ones that are ready to move now
10392 regMaskTP targetCandidates = targetRegsToDo;
10393 while (targetCandidates != RBM_NONE)
10395 regMaskTP targetRegMask = genFindLowestBit(targetCandidates);
10396 targetCandidates &= ~targetRegMask;
10397 regNumber targetReg = genRegNumFromMask(targetRegMask);
10398 if (location[targetReg] == REG_NA)
10400 targetRegsReady |= targetRegMask;
10404 // Perform reg to reg moves
10405 while (targetRegsToDo != RBM_NONE)
10407 while (targetRegsReady != RBM_NONE)
10409 regMaskTP targetRegMask = genFindLowestBit(targetRegsReady);
10410 targetRegsToDo &= ~targetRegMask;
10411 targetRegsReady &= ~targetRegMask;
10412 regNumber targetReg = genRegNumFromMask(targetRegMask);
10413 assert(location[targetReg] != targetReg);
10414 regNumber sourceReg = (regNumber)source[targetReg];
10415 regNumber fromReg = (regNumber)location[sourceReg];
10416 assert(fromReg < UCHAR_MAX && sourceReg < UCHAR_MAX);
10417 Interval* interval = sourceIntervals[sourceReg];
10418 assert(interval != nullptr);
10419 addResolution(block, insertionPoint, interval, targetReg, fromReg);
10420 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10421 sourceIntervals[sourceReg] = nullptr;
10422 location[sourceReg] = REG_NA;
10424 // Do we have a free targetReg?
10425 if (fromReg == sourceReg && source[fromReg] != REG_NA)
10427 regMaskTP fromRegMask = genRegMask(fromReg);
10428 targetRegsReady |= fromRegMask;
10431 if (targetRegsToDo != RBM_NONE)
10433 regMaskTP targetRegMask = genFindLowestBit(targetRegsToDo);
10434 regNumber targetReg = genRegNumFromMask(targetRegMask);
10436 // Is it already there due to other moves?
10437 // If not, move it to the temp reg, OR swap it with another register
10438 regNumber sourceReg = (regNumber)source[targetReg];
10439 regNumber fromReg = (regNumber)location[sourceReg];
10440 if (targetReg == fromReg)
10442 targetRegsToDo &= ~targetRegMask;
10446 regNumber tempReg = REG_NA;
10447 bool useSwap = false;
10448 if (emitter::isFloatReg(targetReg))
10450 #ifdef _TARGET_ARM_
10451 if (sourceIntervals[fromReg]->registerType == TYP_DOUBLE)
10453 // ARM32 requires a double temp register for TYP_DOUBLE.
10454 // We tried to reserve a double temp register first, but sometimes we can't.
10455 tempReg = genIsValidDoubleReg(tempRegFlt) ? tempRegFlt : REG_NA;
10458 #endif // _TARGET_ARM_
10459 tempReg = tempRegFlt;
10461 #ifdef _TARGET_XARCH_
10466 #else // !_TARGET_XARCH_
10470 tempReg = tempRegInt;
10473 #endif // !_TARGET_XARCH_
10474 if (useSwap || tempReg == REG_NA)
10476 // First, we have to figure out the destination register for what's currently in fromReg,
10477 // so that we can find its sourceInterval.
10478 regNumber otherTargetReg = REG_NA;
10480 // By chance, is fromReg going where it belongs?
10481 if (location[source[fromReg]] == targetReg)
10483 otherTargetReg = fromReg;
10484 // If we can swap, we will be done with otherTargetReg as well.
10485 // Otherwise, we'll spill it to the stack and reload it later.
10488 regMaskTP fromRegMask = genRegMask(fromReg);
10489 targetRegsToDo &= ~fromRegMask;
10494 // Look at the remaining registers from targetRegsToDo (which we expect to be relatively
10495 // small at this point) to find out what's currently in targetReg.
10496 regMaskTP mask = targetRegsToDo;
10497 while (mask != RBM_NONE && otherTargetReg == REG_NA)
10499 regMaskTP nextRegMask = genFindLowestBit(mask);
10500 regNumber nextReg = genRegNumFromMask(nextRegMask);
10501 mask &= ~nextRegMask;
10502 if (location[source[nextReg]] == targetReg)
10504 otherTargetReg = nextReg;
10508 assert(otherTargetReg != REG_NA);
10512 // Generate a "swap" of fromReg and targetReg
10513 insertSwap(block, insertionPoint, sourceIntervals[source[otherTargetReg]]->varNum, targetReg,
10514 sourceIntervals[sourceReg]->varNum, fromReg);
10515 location[sourceReg] = REG_NA;
10516 location[source[otherTargetReg]] = (regNumberSmall)fromReg;
10518 INTRACK_STATS(updateLsraStat(LSRA_STAT_RESOLUTION_MOV, block->bbNum));
10522 // Spill "targetReg" to the stack and add its eventual target (otherTargetReg)
10523 // to "targetRegsFromStack", which will be handled below.
10524 // NOTE: This condition is very rare. Setting COMPlus_JitStressRegs=0x203
10525 // has been known to trigger it in JIT SH.
10527 // First, spill "otherInterval" from targetReg to the stack.
10528 Interval* otherInterval = sourceIntervals[source[otherTargetReg]];
10529 setIntervalAsSpilled(otherInterval);
10530 addResolution(block, insertionPoint, otherInterval, REG_STK, targetReg);
10531 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10532 location[source[otherTargetReg]] = REG_STK;
10534 // Now, move the interval that is going to targetReg, and add its "fromReg" to
10535 // "targetRegsReady".
10536 addResolution(block, insertionPoint, sourceIntervals[sourceReg], targetReg, fromReg);
10537 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10538 location[sourceReg] = REG_NA;
10539 targetRegsReady |= genRegMask(fromReg);
10541 targetRegsToDo &= ~targetRegMask;
10545 compiler->codeGen->regSet.rsSetRegsModified(genRegMask(tempReg) DEBUGARG(dumpTerse));
10546 assert(sourceIntervals[targetReg] != nullptr);
10547 addResolution(block, insertionPoint, sourceIntervals[targetReg], tempReg, targetReg);
10548 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10549 location[targetReg] = (regNumberSmall)tempReg;
10550 targetRegsReady |= targetRegMask;
10556 // Finally, perform stack to reg moves
10557 // All the target regs will be empty at this point
10558 while (targetRegsFromStack != RBM_NONE)
10560 regMaskTP targetRegMask = genFindLowestBit(targetRegsFromStack);
10561 targetRegsFromStack &= ~targetRegMask;
10562 regNumber targetReg = genRegNumFromMask(targetRegMask);
10564 Interval* interval = stackToRegIntervals[targetReg];
10565 assert(interval != nullptr);
10567 addResolution(block, insertionPoint, interval, targetReg, REG_STK);
10568 JITDUMP(" (%s)\n", resolveTypeName[resolveType]);
10572 void TreeNodeInfo::Initialize(LinearScan* lsra, GenTree* node, LsraLocation location)
10574 regMaskTP dstCandidates;
10576 // if there is a reg indicated on the tree node, use that for dstCandidates
10577 // the exception is the NOP, which sometimes show up around late args.
10578 // TODO-Cleanup: get rid of those NOPs.
10579 if (node->gtRegNum == REG_NA || node->gtOper == GT_NOP)
10582 if (node->OperGet() == GT_PUTARG_REG)
10584 dstCandidates = lsra->allRegs(TYP_INT);
10589 dstCandidates = lsra->allRegs(node->TypeGet());
10594 dstCandidates = genRegMask(node->gtRegNum);
10597 internalIntCount = 0;
10598 internalFloatCount = 0;
10599 isLocalDefUse = false;
10600 isLsraAdded = false;
10601 definesAnyRegisters = false;
10603 setDstCandidates(lsra, dstCandidates);
10604 srcCandsIndex = dstCandsIndex;
10606 setInternalCandidates(lsra, lsra->allRegs(TYP_INT));
10610 isInitialized = true;
10613 assert(IsValid(lsra));
10616 regMaskTP TreeNodeInfo::getSrcCandidates(LinearScan* lsra)
10618 return lsra->GetRegMaskForIndex(srcCandsIndex);
10621 void TreeNodeInfo::setSrcCandidates(LinearScan* lsra, regMaskTP mask)
10623 LinearScan::RegMaskIndex i = lsra->GetIndexForRegMask(mask);
10624 assert(FitsIn<unsigned char>(i));
10625 srcCandsIndex = (unsigned char)i;
10628 regMaskTP TreeNodeInfo::getDstCandidates(LinearScan* lsra)
10630 return lsra->GetRegMaskForIndex(dstCandsIndex);
10633 void TreeNodeInfo::setDstCandidates(LinearScan* lsra, regMaskTP mask)
10635 LinearScan::RegMaskIndex i = lsra->GetIndexForRegMask(mask);
10636 assert(FitsIn<unsigned char>(i));
10637 dstCandsIndex = (unsigned char)i;
10640 regMaskTP TreeNodeInfo::getInternalCandidates(LinearScan* lsra)
10642 return lsra->GetRegMaskForIndex(internalCandsIndex);
10645 void TreeNodeInfo::setInternalCandidates(LinearScan* lsra, regMaskTP mask)
10647 LinearScan::RegMaskIndex i = lsra->GetIndexForRegMask(mask);
10648 assert(FitsIn<unsigned char>(i));
10649 internalCandsIndex = (unsigned char)i;
10652 void TreeNodeInfo::addInternalCandidates(LinearScan* lsra, regMaskTP mask)
10654 LinearScan::RegMaskIndex i = lsra->GetIndexForRegMask(lsra->GetRegMaskForIndex(internalCandsIndex) | mask);
10655 assert(FitsIn<unsigned char>(i));
10656 internalCandsIndex = (unsigned char)i;
10659 #if TRACK_LSRA_STATS
10660 // ----------------------------------------------------------
10661 // updateLsraStat: Increment LSRA stat counter.
10664 // stat - LSRA stat enum
10665 // bbNum - Basic block to which LSRA stat needs to be
10666 // associated with.
10668 void LinearScan::updateLsraStat(LsraStat stat, unsigned bbNum)
10670 if (bbNum > bbNumMaxBeforeResolution)
10672 // This is a newly created basic block as part of resolution.
10673 // These blocks contain resolution moves that are already accounted.
10679 case LSRA_STAT_SPILL:
10680 ++(blockInfo[bbNum].spillCount);
10683 case LSRA_STAT_COPY_REG:
10684 ++(blockInfo[bbNum].copyRegCount);
10687 case LSRA_STAT_RESOLUTION_MOV:
10688 ++(blockInfo[bbNum].resolutionMovCount);
10691 case LSRA_STAT_SPLIT_EDGE:
10692 ++(blockInfo[bbNum].splitEdgeCount);
10700 // -----------------------------------------------------------
10701 // dumpLsraStats - dumps Lsra stats to given file.
10704 // file - file to which stats are to be written.
10706 void LinearScan::dumpLsraStats(FILE* file)
10708 unsigned sumSpillCount = 0;
10709 unsigned sumCopyRegCount = 0;
10710 unsigned sumResolutionMovCount = 0;
10711 unsigned sumSplitEdgeCount = 0;
10712 UINT64 wtdSpillCount = 0;
10713 UINT64 wtdCopyRegCount = 0;
10714 UINT64 wtdResolutionMovCount = 0;
10716 fprintf(file, "----------\n");
10717 fprintf(file, "LSRA Stats");
10721 fprintf(file, " : %s\n", compiler->info.compFullName);
10725 // In verbose mode no need to print full name
10726 // while printing lsra stats.
10727 fprintf(file, "\n");
10730 fprintf(file, " : %s\n", compiler->eeGetMethodFullName(compiler->info.compCompHnd));
10733 fprintf(file, "----------\n");
10735 for (BasicBlock* block = compiler->fgFirstBB; block != nullptr; block = block->bbNext)
10737 if (block->bbNum > bbNumMaxBeforeResolution)
10742 unsigned spillCount = blockInfo[block->bbNum].spillCount;
10743 unsigned copyRegCount = blockInfo[block->bbNum].copyRegCount;
10744 unsigned resolutionMovCount = blockInfo[block->bbNum].resolutionMovCount;
10745 unsigned splitEdgeCount = blockInfo[block->bbNum].splitEdgeCount;
10747 if (spillCount != 0 || copyRegCount != 0 || resolutionMovCount != 0 || splitEdgeCount != 0)
10749 fprintf(file, "BB%02u [%8d]: ", block->bbNum, block->bbWeight);
10750 fprintf(file, "SpillCount = %d, ResolutionMovs = %d, SplitEdges = %d, CopyReg = %d\n", spillCount,
10751 resolutionMovCount, splitEdgeCount, copyRegCount);
10754 sumSpillCount += spillCount;
10755 sumCopyRegCount += copyRegCount;
10756 sumResolutionMovCount += resolutionMovCount;
10757 sumSplitEdgeCount += splitEdgeCount;
10759 wtdSpillCount += (UINT64)spillCount * block->bbWeight;
10760 wtdCopyRegCount += (UINT64)copyRegCount * block->bbWeight;
10761 wtdResolutionMovCount += (UINT64)resolutionMovCount * block->bbWeight;
10764 fprintf(file, "Total Tracked Vars: %d\n", compiler->lvaTrackedCount);
10765 fprintf(file, "Total Reg Cand Vars: %d\n", regCandidateVarCount);
10766 fprintf(file, "Total number of Intervals: %d\n", static_cast<unsigned>(intervals.size() - 1));
10767 fprintf(file, "Total number of RefPositions: %d\n", static_cast<unsigned>(refPositions.size() - 1));
10768 fprintf(file, "Total Spill Count: %d Weighted: %I64u\n", sumSpillCount, wtdSpillCount);
10769 fprintf(file, "Total CopyReg Count: %d Weighted: %I64u\n", sumCopyRegCount, wtdCopyRegCount);
10770 fprintf(file, "Total ResolutionMov Count: %d Weighted: %I64u\n", sumResolutionMovCount, wtdResolutionMovCount);
10771 fprintf(file, "Total number of split edges: %d\n", sumSplitEdgeCount);
10773 // compute total number of spill temps created
10774 unsigned numSpillTemps = 0;
10775 for (int i = 0; i < TYP_COUNT; i++)
10777 numSpillTemps += maxSpill[i];
10779 fprintf(file, "Total Number of spill temps created: %d\n\n", numSpillTemps);
10781 #endif // TRACK_LSRA_STATS
10784 void dumpRegMask(regMaskTP regs)
10786 if (regs == RBM_ALLINT)
10788 printf("[allInt]");
10790 else if (regs == (RBM_ALLINT & ~RBM_FPBASE))
10792 printf("[allIntButFP]");
10794 else if (regs == RBM_ALLFLOAT)
10796 printf("[allFloat]");
10798 else if (regs == RBM_ALLDOUBLE)
10800 printf("[allDouble]");
10808 static const char* getRefTypeName(RefType refType)
10812 #define DEF_REFTYPE(memberName, memberValue, shortName) \
10814 return #memberName;
10815 #include "lsra_reftypes.h"
10822 static const char* getRefTypeShortName(RefType refType)
10826 #define DEF_REFTYPE(memberName, memberValue, shortName) \
10829 #include "lsra_reftypes.h"
10836 void RefPosition::dump()
10838 printf("<RefPosition #%-3u @%-3u", rpNum, nodeLocation);
10840 if (nextRefPosition)
10842 printf(" ->#%-3u", nextRefPosition->rpNum);
10845 printf(" %s ", getRefTypeName(refType));
10847 if (this->isPhysRegRef)
10849 this->getReg()->tinyDump();
10851 else if (getInterval())
10853 this->getInterval()->tinyDump();
10856 if (this->treeNode)
10858 printf("%s ", treeNode->OpName(treeNode->OperGet()));
10860 printf("BB%02u ", this->bbNum);
10862 printf("regmask=");
10863 dumpRegMask(registerAssignment);
10873 if (this->spillAfter)
10875 printf(" spillAfter");
10885 if (this->isFixedRegRef)
10889 if (this->isLocalDefUse)
10893 if (this->delayRegFree)
10897 if (this->outOfOrder)
10899 printf(" outOfOrder");
10902 if (this->AllocateIfProfitable())
10904 printf(" regOptional");
10909 void RegRecord::dump()
10914 void Interval::dump()
10916 printf("Interval %2u:", intervalIndex);
10920 printf(" (V%02u)", varNum);
10924 printf(" (INTERNAL)");
10928 printf(" (SPILLED)");
10932 printf(" (SPLIT)");
10936 printf(" (struct)");
10938 if (isSpecialPutArg)
10940 printf(" (specialPutArg)");
10944 printf(" (constant)");
10947 printf(" RefPositions {");
10948 for (RefPosition* refPosition = this->firstRefPosition; refPosition != nullptr;
10949 refPosition = refPosition->nextRefPosition)
10951 printf("#%u@%u", refPosition->rpNum, refPosition->nodeLocation);
10952 if (refPosition->nextRefPosition)
10959 // this is not used (yet?)
10960 // printf(" SpillOffset %d", this->spillOffset);
10962 printf(" physReg:%s", getRegName(physReg));
10964 printf(" Preferences=");
10965 dumpRegMask(this->registerPreferences);
10967 if (relatedInterval)
10969 printf(" RelatedInterval ");
10970 relatedInterval->microDump();
10971 printf("[%p]", dspPtr(relatedInterval));
10977 // print out very concise representation
10978 void Interval::tinyDump()
10980 printf("<Ivl:%u", intervalIndex);
10983 printf(" V%02u", varNum);
10987 printf(" internal");
10992 // print out extremely concise representation
10993 void Interval::microDump()
10995 char intervalTypeChar = 'I';
10998 intervalTypeChar = 'T';
11000 else if (isLocalVar)
11002 intervalTypeChar = 'L';
11005 printf("<%c%u>", intervalTypeChar, intervalIndex);
11008 void RegRecord::tinyDump()
11010 printf("<Reg:%-3s> ", getRegName(regNum));
11013 void TreeNodeInfo::dump(LinearScan* lsra)
11015 printf("<TreeNodeInfo @ %2u %d=%d %di %df", loc, dstCount, srcCount, internalIntCount, internalFloatCount);
11017 dumpRegMask(getSrcCandidates(lsra));
11019 dumpRegMask(getInternalCandidates(lsra));
11021 dumpRegMask(getDstCandidates(lsra));
11046 if (isInternalRegDelayFree)
11053 void LinearScan::lsraDumpIntervals(const char* msg)
11055 Interval* interval;
11057 printf("\nLinear scan intervals %s:\n", msg);
11058 for (auto& interval : intervals)
11060 // only dump something if it has references
11061 // if (interval->firstRefPosition)
11068 // Dumps a tree node as a destination or source operand, with the style
11069 // of dump dependent on the mode
11070 void LinearScan::lsraGetOperandString(GenTreePtr tree,
11071 LsraTupleDumpMode mode,
11072 char* operandString,
11073 unsigned operandStringLength)
11075 const char* lastUseChar = "";
11076 if ((tree->gtFlags & GTF_VAR_DEATH) != 0)
11082 case LinearScan::LSRA_DUMP_PRE:
11083 _snprintf_s(operandString, operandStringLength, operandStringLength, "t%d%s", tree->gtTreeID, lastUseChar);
11085 case LinearScan::LSRA_DUMP_REFPOS:
11086 _snprintf_s(operandString, operandStringLength, operandStringLength, "t%d%s", tree->gtTreeID, lastUseChar);
11088 case LinearScan::LSRA_DUMP_POST:
11090 Compiler* compiler = JitTls::GetCompiler();
11092 if (!tree->gtHasReg())
11094 _snprintf_s(operandString, operandStringLength, operandStringLength, "STK%s", lastUseChar);
11098 _snprintf_s(operandString, operandStringLength, operandStringLength, "%s%s",
11099 getRegName(tree->gtRegNum, useFloatReg(tree->TypeGet())), lastUseChar);
11104 printf("ERROR: INVALID TUPLE DUMP MODE\n");
11108 void LinearScan::lsraDispNode(GenTreePtr tree, LsraTupleDumpMode mode, bool hasDest)
11110 Compiler* compiler = JitTls::GetCompiler();
11111 const unsigned operandStringLength = 16;
11112 char operandString[operandStringLength];
11113 const char* emptyDestOperand = " ";
11114 char spillChar = ' ';
11116 if (mode == LinearScan::LSRA_DUMP_POST)
11118 if ((tree->gtFlags & GTF_SPILL) != 0)
11122 if (!hasDest && tree->gtHasReg())
11124 // A node can define a register, but not produce a value for a parent to consume,
11125 // i.e. in the "localDefUse" case.
11126 // There used to be an assert here that we wouldn't spill such a node.
11127 // However, we can have unused lclVars that wind up being the node at which
11128 // it is spilled. This probably indicates a bug, but we don't realy want to
11129 // assert during a dump.
11130 if (spillChar == 'S')
11141 printf("%c N%03u. ", spillChar, tree->gtSeqNum);
11143 LclVarDsc* varDsc = nullptr;
11144 unsigned varNum = UINT_MAX;
11145 if (tree->IsLocal())
11147 varNum = tree->gtLclVarCommon.gtLclNum;
11148 varDsc = &(compiler->lvaTable[varNum]);
11149 if (varDsc->lvLRACandidate)
11156 if (mode == LinearScan::LSRA_DUMP_POST && tree->gtFlags & GTF_SPILLED)
11158 assert(tree->gtHasReg());
11160 lsraGetOperandString(tree, mode, operandString, operandStringLength);
11161 printf("%-15s =", operandString);
11165 printf("%-15s ", emptyDestOperand);
11167 if (varDsc != nullptr)
11169 if (varDsc->lvLRACandidate)
11171 if (mode == LSRA_DUMP_REFPOS)
11173 printf(" V%02u(L%d)", varNum, getIntervalForLocalVar(varDsc->lvVarIndex)->intervalIndex);
11177 lsraGetOperandString(tree, mode, operandString, operandStringLength);
11178 printf(" V%02u(%s)", varNum, operandString);
11179 if (mode == LinearScan::LSRA_DUMP_POST && tree->gtFlags & GTF_SPILLED)
11187 printf(" V%02u MEM", varNum);
11190 else if (tree->OperIsAssignment())
11192 assert(!tree->gtHasReg());
11193 printf(" asg%s ", GenTree::OpName(tree->OperGet()));
11197 compiler->gtDispNodeName(tree);
11198 if (tree->OperKind() & GTK_LEAF)
11200 compiler->gtDispLeaf(tree, nullptr);
11205 //------------------------------------------------------------------------
11206 // ComputeOperandDstCount: computes the number of registers defined by a
11209 // For most nodes, this is simple:
11210 // - Nodes that do not produce values (e.g. stores and other void-typed
11211 // nodes) and nodes that immediately use the registers they define
11212 // produce no registers
11213 // - Nodes that are marked as defining N registers define N registers.
11215 // For contained nodes, however, things are more complicated: for purposes
11216 // of bookkeeping, a contained node is treated as producing the transitive
11217 // closure of the registers produced by its sources.
11220 // operand - The operand for which to compute a register count.
11223 // The number of registers defined by `operand`.
11225 void LinearScan::DumpOperandDefs(
11226 GenTree* operand, bool& first, LsraTupleDumpMode mode, char* operandString, const unsigned operandStringLength)
11228 assert(operand != nullptr);
11229 assert(operandString != nullptr);
11231 if (ComputeOperandDstCount(operand) == 0)
11236 if (operand->gtLsraInfo.dstCount != 0)
11238 // This operand directly produces registers; print it.
11239 for (int i = 0; i < operand->gtLsraInfo.dstCount; i++)
11246 lsraGetOperandString(operand, mode, operandString, operandStringLength);
11247 printf("%s", operandString);
11254 // This is a contained node. Dump the defs produced by its operands.
11255 for (GenTree* op : operand->Operands())
11257 DumpOperandDefs(op, first, mode, operandString, operandStringLength);
11262 void LinearScan::TupleStyleDump(LsraTupleDumpMode mode)
11265 LsraLocation currentLoc = 1; // 0 is the entry
11266 const unsigned operandStringLength = 16;
11267 char operandString[operandStringLength];
11269 // currentRefPosition is not used for LSRA_DUMP_PRE
11270 // We keep separate iterators for defs, so that we can print them
11271 // on the lhs of the dump
11272 auto currentRefPosition = refPositions.begin();
11276 case LSRA_DUMP_PRE:
11277 printf("TUPLE STYLE DUMP BEFORE LSRA\n");
11279 case LSRA_DUMP_REFPOS:
11280 printf("TUPLE STYLE DUMP WITH REF POSITIONS\n");
11282 case LSRA_DUMP_POST:
11283 printf("TUPLE STYLE DUMP WITH REGISTER ASSIGNMENTS\n");
11286 printf("ERROR: INVALID TUPLE DUMP MODE\n");
11290 if (mode != LSRA_DUMP_PRE)
11292 printf("Incoming Parameters: ");
11293 for (; currentRefPosition != refPositions.end() && currentRefPosition->refType != RefTypeBB;
11294 ++currentRefPosition)
11296 Interval* interval = currentRefPosition->getInterval();
11297 assert(interval != nullptr && interval->isLocalVar);
11298 printf(" V%02d", interval->varNum);
11299 if (mode == LSRA_DUMP_POST)
11302 if (currentRefPosition->registerAssignment == RBM_NONE)
11308 reg = currentRefPosition->assignedReg();
11310 LclVarDsc* varDsc = &(compiler->lvaTable[interval->varNum]);
11312 regNumber assignedReg = varDsc->lvRegNum;
11313 regNumber argReg = (varDsc->lvIsRegArg) ? varDsc->lvArgReg : REG_STK;
11315 assert(reg == assignedReg || varDsc->lvRegister == false);
11318 printf(getRegName(argReg, isFloatRegType(interval->registerType)));
11321 printf("%s)", getRegName(reg, isFloatRegType(interval->registerType)));
11327 for (block = startBlockSequence(); block != nullptr; block = moveToNextBlock())
11331 if (mode == LSRA_DUMP_REFPOS)
11333 bool printedBlockHeader = false;
11334 // We should find the boundary RefPositions in the order of exposed uses, dummy defs, and the blocks
11335 for (; currentRefPosition != refPositions.end() &&
11336 (currentRefPosition->refType == RefTypeExpUse || currentRefPosition->refType == RefTypeDummyDef ||
11337 (currentRefPosition->refType == RefTypeBB && !printedBlockHeader));
11338 ++currentRefPosition)
11340 Interval* interval = nullptr;
11341 if (currentRefPosition->isIntervalRef())
11343 interval = currentRefPosition->getInterval();
11345 switch (currentRefPosition->refType)
11347 case RefTypeExpUse:
11348 assert(interval != nullptr);
11349 assert(interval->isLocalVar);
11350 printf(" Exposed use of V%02u at #%d\n", interval->varNum, currentRefPosition->rpNum);
11352 case RefTypeDummyDef:
11353 assert(interval != nullptr);
11354 assert(interval->isLocalVar);
11355 printf(" Dummy def of V%02u at #%d\n", interval->varNum, currentRefPosition->rpNum);
11358 block->dspBlockHeader(compiler);
11359 printedBlockHeader = true;
11363 printf("Unexpected RefPosition type at #%d\n", currentRefPosition->rpNum);
11370 block->dspBlockHeader(compiler);
11373 if (enregisterLocalVars && mode == LSRA_DUMP_POST && block != compiler->fgFirstBB &&
11374 block->bbNum <= bbNumMaxBeforeResolution)
11376 printf("Predecessor for variable locations: BB%02u\n", blockInfo[block->bbNum].predBBNum);
11377 dumpInVarToRegMap(block);
11379 if (block->bbNum > bbNumMaxBeforeResolution)
11381 SplitEdgeInfo splitEdgeInfo;
11382 splitBBNumToTargetBBNumMap->Lookup(block->bbNum, &splitEdgeInfo);
11383 assert(splitEdgeInfo.toBBNum <= bbNumMaxBeforeResolution);
11384 assert(splitEdgeInfo.fromBBNum <= bbNumMaxBeforeResolution);
11385 printf("New block introduced for resolution from BB%02u to BB%02u\n", splitEdgeInfo.fromBBNum,
11386 splitEdgeInfo.toBBNum);
11389 for (GenTree* node : LIR::AsRange(block).NonPhiNodes())
11391 GenTree* tree = node;
11393 genTreeOps oper = tree->OperGet();
11394 TreeNodeInfo& info = tree->gtLsraInfo;
11395 if (tree->gtLsraInfo.isLsraAdded)
11397 // This must be one of the nodes that we add during LSRA
11399 if (oper == GT_LCL_VAR)
11404 else if (oper == GT_RELOAD || oper == GT_COPY)
11409 #ifdef FEATURE_SIMD
11410 else if (oper == GT_SIMD)
11412 if (tree->gtSIMD.gtSIMDIntrinsicID == SIMDIntrinsicUpperSave)
11419 assert(tree->gtSIMD.gtSIMDIntrinsicID == SIMDIntrinsicUpperRestore);
11424 #endif // FEATURE_SIMD
11427 assert(oper == GT_SWAP);
11431 info.internalIntCount = 0;
11432 info.internalFloatCount = 0;
11435 int consume = info.srcCount;
11436 int produce = info.dstCount;
11437 regMaskTP killMask = RBM_NONE;
11438 regMaskTP fixedMask = RBM_NONE;
11440 lsraDispNode(tree, mode, produce != 0 && mode != LSRA_DUMP_REFPOS);
11442 if (mode != LSRA_DUMP_REFPOS)
11449 for (GenTree* operand : tree->Operands())
11451 DumpOperandDefs(operand, first, mode, operandString, operandStringLength);
11457 // Print each RefPosition on a new line, but
11458 // printing all the kills for each node on a single line
11459 // and combining the fixed regs with their associated def or use
11460 bool killPrinted = false;
11461 RefPosition* lastFixedRegRefPos = nullptr;
11462 for (; currentRefPosition != refPositions.end() &&
11463 (currentRefPosition->refType == RefTypeUse || currentRefPosition->refType == RefTypeFixedReg ||
11464 currentRefPosition->refType == RefTypeKill || currentRefPosition->refType == RefTypeDef) &&
11465 (currentRefPosition->nodeLocation == tree->gtSeqNum ||
11466 currentRefPosition->nodeLocation == tree->gtSeqNum + 1);
11467 ++currentRefPosition)
11469 Interval* interval = nullptr;
11470 if (currentRefPosition->isIntervalRef())
11472 interval = currentRefPosition->getInterval();
11474 switch (currentRefPosition->refType)
11477 if (currentRefPosition->isPhysRegRef)
11479 printf("\n Use:R%d(#%d)",
11480 currentRefPosition->getReg()->regNum, currentRefPosition->rpNum);
11484 assert(interval != nullptr);
11486 interval->microDump();
11487 printf("(#%d)", currentRefPosition->rpNum);
11488 if (currentRefPosition->isFixedRegRef)
11490 assert(genMaxOneBit(currentRefPosition->registerAssignment));
11491 assert(lastFixedRegRefPos != nullptr);
11492 printf(" Fixed:%s(#%d)", getRegName(currentRefPosition->assignedReg(),
11493 isFloatRegType(interval->registerType)),
11494 lastFixedRegRefPos->rpNum);
11495 lastFixedRegRefPos = nullptr;
11497 if (currentRefPosition->isLocalDefUse)
11499 printf(" LocalDefUse");
11501 if (currentRefPosition->lastUse)
11509 // Print each def on a new line
11510 assert(interval != nullptr);
11512 interval->microDump();
11513 printf("(#%d)", currentRefPosition->rpNum);
11514 if (currentRefPosition->isFixedRegRef)
11516 assert(genMaxOneBit(currentRefPosition->registerAssignment));
11517 printf(" %s", getRegName(currentRefPosition->assignedReg(),
11518 isFloatRegType(interval->registerType)));
11520 if (currentRefPosition->isLocalDefUse)
11522 printf(" LocalDefUse");
11524 if (currentRefPosition->lastUse)
11528 if (interval->relatedInterval != nullptr)
11531 interval->relatedInterval->microDump();
11538 printf("\n Kill: ");
11539 killPrinted = true;
11541 printf(getRegName(currentRefPosition->assignedReg(),
11542 isFloatRegType(currentRefPosition->getReg()->registerType)));
11545 case RefTypeFixedReg:
11546 lastFixedRegRefPos = currentRefPosition;
11549 printf("Unexpected RefPosition type at #%d\n", currentRefPosition->rpNum);
11555 if (info.internalIntCount != 0 && mode != LSRA_DUMP_REFPOS)
11557 printf("\tinternal (%d):\t", info.internalIntCount);
11558 if (mode == LSRA_DUMP_POST)
11560 dumpRegMask(tree->gtRsvdRegs);
11562 else if ((info.getInternalCandidates(this) & allRegs(TYP_INT)) != allRegs(TYP_INT))
11564 dumpRegMask(info.getInternalCandidates(this) & allRegs(TYP_INT));
11568 if (info.internalFloatCount != 0 && mode != LSRA_DUMP_REFPOS)
11570 printf("\tinternal (%d):\t", info.internalFloatCount);
11571 if (mode == LSRA_DUMP_POST)
11573 dumpRegMask(tree->gtRsvdRegs);
11575 else if ((info.getInternalCandidates(this) & allRegs(TYP_INT)) != allRegs(TYP_INT))
11577 dumpRegMask(info.getInternalCandidates(this) & allRegs(TYP_INT));
11582 if (enregisterLocalVars && mode == LSRA_DUMP_POST)
11584 dumpOutVarToRegMap(block);
11591 void LinearScan::dumpLsraAllocationEvent(LsraDumpEvent event,
11592 Interval* interval,
11594 BasicBlock* currentBlock)
11602 // Conflicting def/use
11603 case LSRA_EVENT_DEFUSE_CONFLICT:
11606 printf(" Def and Use have conflicting register requirements:");
11610 printf("DUconflict ");
11614 case LSRA_EVENT_DEFUSE_FIXED_DELAY_USE:
11617 printf(" Can't change useAssignment ");
11620 case LSRA_EVENT_DEFUSE_CASE1:
11623 printf(" case #1, use the defRegAssignment\n");
11627 printf(indentFormat, " case #1 use defRegAssignment");
11629 dumpEmptyRefPosition();
11632 case LSRA_EVENT_DEFUSE_CASE2:
11635 printf(" case #2, use the useRegAssignment\n");
11639 printf(indentFormat, " case #2 use useRegAssignment");
11641 dumpEmptyRefPosition();
11644 case LSRA_EVENT_DEFUSE_CASE3:
11647 printf(" case #3, change the defRegAssignment to the use regs\n");
11651 printf(indentFormat, " case #3 use useRegAssignment");
11653 dumpEmptyRefPosition();
11656 case LSRA_EVENT_DEFUSE_CASE4:
11659 printf(" case #4, change the useRegAssignment to the def regs\n");
11663 printf(indentFormat, " case #4 use defRegAssignment");
11665 dumpEmptyRefPosition();
11668 case LSRA_EVENT_DEFUSE_CASE5:
11671 printf(" case #5, Conflicting Def and Use single-register requirements require copies - set def to all "
11672 "regs of the appropriate type\n");
11676 printf(indentFormat, " case #5 set def to all regs");
11678 dumpEmptyRefPosition();
11681 case LSRA_EVENT_DEFUSE_CASE6:
11684 printf(" case #6, Conflicting Def and Use register requirements require a copy\n");
11688 printf(indentFormat, " case #6 need a copy");
11690 dumpEmptyRefPosition();
11694 case LSRA_EVENT_SPILL:
11697 printf("Spilled:\n");
11702 assert(interval != nullptr && interval->assignedReg != nullptr);
11703 printf("Spill %-4s ", getRegName(interval->assignedReg->regNum));
11705 dumpEmptyRefPosition();
11708 case LSRA_EVENT_SPILL_EXTENDED_LIFETIME:
11711 printf(" Spilled extended lifetime var V%02u at last use; not marked for actual spill.",
11712 interval->intervalIndex);
11716 // Restoring the previous register
11717 case LSRA_EVENT_RESTORE_PREVIOUS_INTERVAL_AFTER_SPILL:
11718 assert(interval != nullptr);
11721 printf(" Assign register %s to previous interval Ivl:%d after spill\n", getRegName(reg),
11722 interval->intervalIndex);
11726 // If we spilled, then the dump is already pre-indented, but we need to pre-indent for the subsequent
11728 // with a dumpEmptyRefPosition().
11729 printf("SRstr %-4s ", getRegName(reg));
11731 dumpEmptyRefPosition();
11734 case LSRA_EVENT_RESTORE_PREVIOUS_INTERVAL:
11735 assert(interval != nullptr);
11738 printf(" Assign register %s to previous interval Ivl:%d\n", getRegName(reg), interval->intervalIndex);
11742 if (activeRefPosition == nullptr)
11744 printf(emptyRefPositionFormat, "");
11746 printf("Restr %-4s ", getRegName(reg));
11748 if (activeRefPosition != nullptr)
11750 printf(emptyRefPositionFormat, "");
11755 // Done with GC Kills
11756 case LSRA_EVENT_DONE_KILL_GC_REFS:
11757 printf("DoneKillGC ");
11760 // Block boundaries
11761 case LSRA_EVENT_START_BB:
11762 assert(currentBlock != nullptr);
11765 printf("\n\n Live Vars(Regs) at start of BB%02u (from pred BB%02u):", currentBlock->bbNum,
11766 blockInfo[currentBlock->bbNum].predBBNum);
11767 dumpVarToRegMap(inVarToRegMaps[currentBlock->bbNum]);
11770 case LSRA_EVENT_END_BB:
11773 printf("\n\n Live Vars(Regs) after BB%02u:", currentBlock->bbNum);
11774 dumpVarToRegMap(outVarToRegMaps[currentBlock->bbNum]);
11778 case LSRA_EVENT_FREE_REGS:
11781 printf("Freeing registers:\n");
11785 // Characteristics of the current RefPosition
11786 case LSRA_EVENT_INCREMENT_RANGE_END:
11789 printf(" Incrementing nextPhysRegLocation for %s\n", getRegName(reg));
11793 case LSRA_EVENT_LAST_USE:
11796 printf(" Last use, marked to be freed\n");
11799 case LSRA_EVENT_LAST_USE_DELAYED:
11802 printf(" Last use, marked to be freed (delayed)\n");
11805 case LSRA_EVENT_NEEDS_NEW_REG:
11808 printf(" Needs new register; mark %s to be freed\n", getRegName(reg));
11812 printf("Free %-4s ", getRegName(reg));
11814 dumpEmptyRefPosition();
11818 // Allocation decisions
11819 case LSRA_EVENT_FIXED_REG:
11820 case LSRA_EVENT_EXP_USE:
11823 printf("No allocation\n");
11827 printf("Keep %-4s ", getRegName(reg));
11830 case LSRA_EVENT_ZERO_REF:
11831 assert(interval != nullptr && interval->isLocalVar);
11834 printf("Marking V%02u as last use there are no actual references\n", interval->varNum);
11840 dumpEmptyRefPosition();
11843 case LSRA_EVENT_KEPT_ALLOCATION:
11846 printf("already allocated %4s\n", getRegName(reg));
11850 printf("Keep %-4s ", getRegName(reg));
11853 case LSRA_EVENT_COPY_REG:
11854 assert(interval != nullptr && interval->recentRefPosition != nullptr);
11857 printf("allocated %s as copyReg\n\n", getRegName(reg));
11861 printf("Copy %-4s ", getRegName(reg));
11864 case LSRA_EVENT_MOVE_REG:
11865 assert(interval != nullptr && interval->recentRefPosition != nullptr);
11868 printf(" needs a new register; marked as moveReg\n");
11872 printf("Move %-4s ", getRegName(reg));
11874 dumpEmptyRefPosition();
11877 case LSRA_EVENT_ALLOC_REG:
11880 printf("allocated %s\n", getRegName(reg));
11884 printf("Alloc %-4s ", getRegName(reg));
11887 case LSRA_EVENT_REUSE_REG:
11890 printf("reused constant in %s\n", getRegName(reg));
11894 printf("Reuse %-4s ", getRegName(reg));
11897 case LSRA_EVENT_ALLOC_SPILLED_REG:
11900 printf("allocated spilled register %s\n", getRegName(reg));
11904 printf("Steal %-4s ", getRegName(reg));
11907 case LSRA_EVENT_NO_ENTRY_REG_ALLOCATED:
11908 assert(interval != nullptr && interval->isLocalVar);
11911 printf("Not allocating an entry register for V%02u due to low ref count\n", interval->varNum);
11918 case LSRA_EVENT_NO_REG_ALLOCATED:
11921 printf("no register allocated\n");
11928 case LSRA_EVENT_RELOAD:
11931 printf(" Marked for reload\n");
11935 printf("ReLod %-4s ", getRegName(reg));
11937 dumpEmptyRefPosition();
11940 case LSRA_EVENT_SPECIAL_PUTARG:
11943 printf(" Special case of putArg - using lclVar that's in the expected reg\n");
11947 printf("PtArg %-4s ", getRegName(reg));
11955 //------------------------------------------------------------------------
11956 // dumpRegRecordHeader: Dump the header for a column-based dump of the register state.
11965 // Reg names fit in 4 characters (minimum width of the columns)
11968 // In order to make the table as dense as possible (for ease of reading the dumps),
11969 // we determine the minimum regColumnWidth width required to represent:
11970 // regs, by name (e.g. eax or xmm0) - this is fixed at 4 characters.
11971 // intervals, as Vnn for lclVar intervals, or as I<num> for other intervals.
11972 // The table is indented by the amount needed for dumpRefPositionShort, which is
11973 // captured in shortRefPositionDumpWidth.
11975 void LinearScan::dumpRegRecordHeader()
11977 printf("The following table has one or more rows for each RefPosition that is handled during allocation.\n"
11978 "The first column provides the basic information about the RefPosition, with its type (e.g. Def,\n"
11979 "Use, Fixd) followed by a '*' if it is a last use, and a 'D' if it is delayRegFree, and then the\n"
11980 "action taken during allocation (e.g. Alloc a new register, or Keep an existing one).\n"
11981 "The subsequent columns show the Interval occupying each register, if any, followed by 'a' if it is\n"
11982 "active, and 'i'if it is inactive. Columns are only printed up to the last modifed register, which\n"
11983 "may increase during allocation, in which case additional columns will appear. Registers which are\n"
11984 "not marked modified have ---- in their column.\n\n");
11986 // First, determine the width of each register column (which holds a reg name in the
11987 // header, and an interval name in each subsequent row).
11988 int intervalNumberWidth = (int)log10((double)intervals.size()) + 1;
11989 // The regColumnWidth includes the identifying character (I or V) and an 'i' or 'a' (inactive or active)
11990 regColumnWidth = intervalNumberWidth + 2;
11991 if (regColumnWidth < 4)
11993 regColumnWidth = 4;
11995 sprintf_s(intervalNameFormat, MAX_FORMAT_CHARS, "%%c%%-%dd", regColumnWidth - 2);
11996 sprintf_s(regNameFormat, MAX_FORMAT_CHARS, "%%-%ds", regColumnWidth);
11998 // Next, determine the width of the short RefPosition (see dumpRefPositionShort()).
11999 // This is in the form:
12000 // nnn.#mmm NAME TYPEld
12002 // nnn is the Location, right-justified to the width needed for the highest location.
12003 // mmm is the RefPosition rpNum, left-justified to the width needed for the highest rpNum.
12004 // NAME is dumped by dumpReferentName(), and is "regColumnWidth".
12005 // TYPE is RefTypeNameShort, and is 4 characters
12006 // l is either '*' (if a last use) or ' ' (otherwise)
12007 // d is either 'D' (if a delayed use) or ' ' (otherwise)
12009 maxNodeLocation = (maxNodeLocation == 0)
12011 : maxNodeLocation; // corner case of a method with an infinite loop without any gentree nodes
12012 assert(maxNodeLocation >= 1);
12013 assert(refPositions.size() >= 1);
12014 int nodeLocationWidth = (int)log10((double)maxNodeLocation) + 1;
12015 int refPositionWidth = (int)log10((double)refPositions.size()) + 1;
12016 int refTypeInfoWidth = 4 /*TYPE*/ + 2 /* last-use and delayed */ + 1 /* space */;
12017 int locationAndRPNumWidth = nodeLocationWidth + 2 /* .# */ + refPositionWidth + 1 /* space */;
12018 int shortRefPositionDumpWidth = locationAndRPNumWidth + regColumnWidth + 1 /* space */ + refTypeInfoWidth;
12019 sprintf_s(shortRefPositionFormat, MAX_FORMAT_CHARS, "%%%dd.#%%-%dd ", nodeLocationWidth, refPositionWidth);
12020 sprintf_s(emptyRefPositionFormat, MAX_FORMAT_CHARS, "%%-%ds", shortRefPositionDumpWidth);
12022 // The width of the "allocation info"
12023 // - a 5-character allocation decision
12025 // - a 4-character register
12027 int allocationInfoWidth = 5 + 1 + 4 + 1;
12029 // Next, determine the width of the legend for each row. This includes:
12030 // - a short RefPosition dump (shortRefPositionDumpWidth), which includes a space
12031 // - the allocation info (allocationInfoWidth), which also includes a space
12033 regTableIndent = shortRefPositionDumpWidth + allocationInfoWidth;
12035 // BBnn printed left-justified in the NAME Typeld and allocationInfo space.
12036 int bbDumpWidth = regColumnWidth + 1 + refTypeInfoWidth + allocationInfoWidth;
12037 int bbNumWidth = (int)log10((double)compiler->fgBBNumMax) + 1;
12038 // In the unlikely event that BB numbers overflow the space, we'll simply omit the predBB
12039 int predBBNumDumpSpace = regTableIndent - locationAndRPNumWidth - bbNumWidth - 9; // 'BB' + ' PredBB'
12040 if (predBBNumDumpSpace < bbNumWidth)
12042 sprintf_s(bbRefPosFormat, MAX_LEGEND_FORMAT_CHARS, "BB%%-%dd", shortRefPositionDumpWidth - 2);
12046 sprintf_s(bbRefPosFormat, MAX_LEGEND_FORMAT_CHARS, "BB%%-%dd PredBB%%-%dd", bbNumWidth, predBBNumDumpSpace);
12049 if (compiler->shouldDumpASCIITrees())
12051 columnSeparator = "|";
12059 columnSeparator = "\xe2\x94\x82";
12060 line = "\xe2\x94\x80";
12061 leftBox = "\xe2\x94\x9c";
12062 middleBox = "\xe2\x94\xbc";
12063 rightBox = "\xe2\x94\xa4";
12065 sprintf_s(indentFormat, MAX_FORMAT_CHARS, "%%-%ds", regTableIndent);
12067 // Now, set up the legend format for the RefPosition info
12068 sprintf_s(legendFormat, MAX_LEGEND_FORMAT_CHARS, "%%-%d.%ds%%-%d.%ds%%-%ds%%s", nodeLocationWidth + 1,
12069 nodeLocationWidth + 1, refPositionWidth + 2, refPositionWidth + 2, regColumnWidth + 1);
12071 // Finally, print a "title row" including the legend and the reg names
12072 dumpRegRecordTitle();
12075 int LinearScan::getLastUsedRegNumIndex()
12077 int lastUsedRegNumIndex = 0;
12078 regMaskTP usedRegsMask = compiler->codeGen->regSet.rsGetModifiedRegsMask();
12079 int lastRegNumIndex = compiler->compFloatingPointUsed ? REG_FP_LAST : REG_INT_LAST;
12080 for (int regNumIndex = 0; regNumIndex <= lastRegNumIndex; regNumIndex++)
12082 if ((usedRegsMask & genRegMask((regNumber)regNumIndex)) != 0)
12084 lastUsedRegNumIndex = regNumIndex;
12087 return lastUsedRegNumIndex;
12090 void LinearScan::dumpRegRecordTitleLines()
12092 for (int i = 0; i < regTableIndent; i++)
12094 printf("%s", line);
12096 int lastUsedRegNumIndex = getLastUsedRegNumIndex();
12097 for (int regNumIndex = 0; regNumIndex <= lastUsedRegNumIndex; regNumIndex++)
12099 printf("%s", middleBox);
12100 for (int i = 0; i < regColumnWidth; i++)
12102 printf("%s", line);
12105 printf("%s\n", rightBox);
12107 void LinearScan::dumpRegRecordTitle()
12109 dumpRegRecordTitleLines();
12111 // Print out the legend for the RefPosition info
12112 printf(legendFormat, "Loc ", "RP# ", "Name ", "Type Action Reg ");
12114 // Print out the register name column headers
12115 char columnFormatArray[MAX_FORMAT_CHARS];
12116 sprintf_s(columnFormatArray, MAX_FORMAT_CHARS, "%s%%-%d.%ds", columnSeparator, regColumnWidth, regColumnWidth);
12117 int lastUsedRegNumIndex = getLastUsedRegNumIndex();
12118 for (int regNumIndex = 0; regNumIndex <= lastUsedRegNumIndex; regNumIndex++)
12120 regNumber regNum = (regNumber)regNumIndex;
12121 const char* regName = getRegName(regNum);
12122 printf(columnFormatArray, regName);
12124 printf("%s\n", columnSeparator);
12126 rowCountSinceLastTitle = 0;
12128 dumpRegRecordTitleLines();
12131 void LinearScan::dumpRegRecords()
12133 static char columnFormatArray[18];
12134 int lastUsedRegNumIndex = getLastUsedRegNumIndex();
12135 regMaskTP usedRegsMask = compiler->codeGen->regSet.rsGetModifiedRegsMask();
12137 for (int regNumIndex = 0; regNumIndex <= lastUsedRegNumIndex; regNumIndex++)
12139 printf("%s", columnSeparator);
12140 RegRecord& regRecord = physRegs[regNumIndex];
12141 Interval* interval = regRecord.assignedInterval;
12142 if (interval != nullptr)
12144 dumpIntervalName(interval);
12145 char activeChar = interval->isActive ? 'a' : 'i';
12146 printf("%c", activeChar);
12148 else if (regRecord.isBusyUntilNextKill)
12150 printf(columnFormatArray, "Busy");
12152 else if ((usedRegsMask & genRegMask((regNumber)regNumIndex)) == 0)
12154 sprintf_s(columnFormatArray, MAX_FORMAT_CHARS, "%%-%ds", regColumnWidth);
12155 printf(columnFormatArray, "----");
12159 sprintf_s(columnFormatArray, MAX_FORMAT_CHARS, "%%-%ds", regColumnWidth);
12160 printf(columnFormatArray, "");
12163 printf("%s\n", columnSeparator);
12165 if (rowCountSinceLastTitle > MAX_ROWS_BETWEEN_TITLES)
12167 dumpRegRecordTitle();
12169 rowCountSinceLastTitle++;
12172 void LinearScan::dumpIntervalName(Interval* interval)
12174 if (interval->isLocalVar)
12176 printf(intervalNameFormat, 'V', interval->varNum);
12178 else if (interval->isConstant)
12180 printf(intervalNameFormat, 'C', interval->intervalIndex);
12184 printf(intervalNameFormat, 'I', interval->intervalIndex);
12188 void LinearScan::dumpEmptyRefPosition()
12190 printf(emptyRefPositionFormat, "");
12193 // Note that the size of this dump is computed in dumpRegRecordHeader().
12195 void LinearScan::dumpRefPositionShort(RefPosition* refPosition, BasicBlock* currentBlock)
12197 BasicBlock* block = currentBlock;
12198 if (refPosition->refType == RefTypeBB)
12200 // Always print a title row before a RefTypeBB (except for the first, because we
12201 // will already have printed it before the parameters)
12202 if (refPosition->refType == RefTypeBB && block != compiler->fgFirstBB && block != nullptr)
12204 dumpRegRecordTitle();
12207 printf(shortRefPositionFormat, refPosition->nodeLocation, refPosition->rpNum);
12208 if (refPosition->refType == RefTypeBB)
12210 if (block == nullptr)
12212 printf(regNameFormat, "END");
12214 printf(regNameFormat, "");
12218 printf(bbRefPosFormat, block->bbNum, block == compiler->fgFirstBB ? 0 : blockInfo[block->bbNum].predBBNum);
12221 else if (refPosition->isIntervalRef())
12223 Interval* interval = refPosition->getInterval();
12224 dumpIntervalName(interval);
12225 char lastUseChar = ' ';
12226 char delayChar = ' ';
12227 if (refPosition->lastUse)
12230 if (refPosition->delayRegFree)
12235 printf(" %s%c%c ", getRefTypeShortName(refPosition->refType), lastUseChar, delayChar);
12237 else if (refPosition->isPhysRegRef)
12239 RegRecord* regRecord = refPosition->getReg();
12240 printf(regNameFormat, getRegName(regRecord->regNum));
12241 printf(" %s ", getRefTypeShortName(refPosition->refType));
12245 assert(refPosition->refType == RefTypeKillGCRefs);
12246 // There's no interval or reg name associated with this.
12247 printf(regNameFormat, " ");
12248 printf(" %s ", getRefTypeShortName(refPosition->refType));
12252 //------------------------------------------------------------------------
12253 // LinearScan::IsResolutionMove:
12254 // Returns true if the given node is a move inserted by LSRA
12258 // node - the node to check.
12260 bool LinearScan::IsResolutionMove(GenTree* node)
12262 if (!node->gtLsraInfo.isLsraAdded)
12267 switch (node->OperGet())
12271 return node->gtLsraInfo.isLocalDefUse;
12281 //------------------------------------------------------------------------
12282 // LinearScan::IsResolutionNode:
12283 // Returns true if the given node is either a move inserted by LSRA
12284 // resolution or an operand to such a move.
12287 // containingRange - the range that contains the node to check.
12288 // node - the node to check.
12290 bool LinearScan::IsResolutionNode(LIR::Range& containingRange, GenTree* node)
12294 if (IsResolutionMove(node))
12299 if (!node->gtLsraInfo.isLsraAdded || (node->OperGet() != GT_LCL_VAR))
12305 bool foundUse = containingRange.TryGetUse(node, &use);
12312 //------------------------------------------------------------------------
12313 // verifyFinalAllocation: Traverse the RefPositions and verify various invariants.
12322 // If verbose is set, this will also dump a table of the final allocations.
12323 void LinearScan::verifyFinalAllocation()
12327 printf("\nFinal allocation\n");
12330 // Clear register assignments.
12331 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
12333 RegRecord* physRegRecord = getRegisterRecord(reg);
12334 physRegRecord->assignedInterval = nullptr;
12337 for (auto& interval : intervals)
12339 interval.assignedReg = nullptr;
12340 interval.physReg = REG_NA;
12343 DBEXEC(VERBOSE, dumpRegRecordTitle());
12345 BasicBlock* currentBlock = nullptr;
12346 GenTree* firstBlockEndResolutionNode = nullptr;
12347 regMaskTP regsToFree = RBM_NONE;
12348 regMaskTP delayRegsToFree = RBM_NONE;
12349 LsraLocation currentLocation = MinLocation;
12350 for (auto& refPosition : refPositions)
12352 RefPosition* currentRefPosition = &refPosition;
12353 Interval* interval = nullptr;
12354 RegRecord* regRecord = nullptr;
12355 regNumber regNum = REG_NA;
12356 if (currentRefPosition->refType == RefTypeBB)
12358 regsToFree |= delayRegsToFree;
12359 delayRegsToFree = RBM_NONE;
12360 // For BB RefPositions, wait until we dump the "end of block" info before dumping the basic RefPosition
12365 // For other RefPosition types, we can dump the basic RefPosition info now.
12366 DBEXEC(VERBOSE, dumpRefPositionShort(currentRefPosition, currentBlock));
12368 if (currentRefPosition->isPhysRegRef)
12370 regRecord = currentRefPosition->getReg();
12371 regRecord->recentRefPosition = currentRefPosition;
12372 regNum = regRecord->regNum;
12374 else if (currentRefPosition->isIntervalRef())
12376 interval = currentRefPosition->getInterval();
12377 interval->recentRefPosition = currentRefPosition;
12378 if (currentRefPosition->registerAssignment != RBM_NONE)
12380 if (!genMaxOneBit(currentRefPosition->registerAssignment))
12382 assert(currentRefPosition->refType == RefTypeExpUse ||
12383 currentRefPosition->refType == RefTypeDummyDef);
12387 regNum = currentRefPosition->assignedReg();
12388 regRecord = getRegisterRecord(regNum);
12394 LsraLocation newLocation = currentRefPosition->nodeLocation;
12396 if (newLocation > currentLocation)
12399 // We could use the freeRegisters() method, but we'd have to carefully manage the active intervals.
12400 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
12402 regMaskTP regMask = genRegMask(reg);
12403 if ((regsToFree & regMask) != RBM_NONE)
12405 RegRecord* physRegRecord = getRegisterRecord(reg);
12406 physRegRecord->assignedInterval = nullptr;
12409 regsToFree = delayRegsToFree;
12410 regsToFree = RBM_NONE;
12412 currentLocation = newLocation;
12414 switch (currentRefPosition->refType)
12418 if (currentBlock == nullptr)
12420 currentBlock = startBlockSequence();
12424 // Verify the resolution moves at the end of the previous block.
12425 for (GenTree* node = firstBlockEndResolutionNode; node != nullptr; node = node->gtNext)
12427 assert(enregisterLocalVars);
12428 // Only verify nodes that are actually moves; don't bother with the nodes that are
12429 // operands to moves.
12430 if (IsResolutionMove(node))
12432 verifyResolutionMove(node, currentLocation);
12436 // Validate the locations at the end of the previous block.
12437 if (enregisterLocalVars)
12439 VarToRegMap outVarToRegMap = outVarToRegMaps[currentBlock->bbNum];
12440 VarSetOps::Iter iter(compiler, currentBlock->bbLiveOut);
12441 unsigned varIndex = 0;
12442 while (iter.NextElem(&varIndex))
12444 if (localVarIntervals[varIndex] == nullptr)
12446 assert(!compiler->lvaTable[compiler->lvaTrackedToVarNum[varIndex]].lvLRACandidate);
12449 regNumber regNum = getVarReg(outVarToRegMap, varIndex);
12450 interval = getIntervalForLocalVar(varIndex);
12451 assert(interval->physReg == regNum || (interval->physReg == REG_NA && regNum == REG_STK));
12452 interval->physReg = REG_NA;
12453 interval->assignedReg = nullptr;
12454 interval->isActive = false;
12458 // Clear register assignments.
12459 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
12461 RegRecord* physRegRecord = getRegisterRecord(reg);
12462 physRegRecord->assignedInterval = nullptr;
12465 // Now, record the locations at the beginning of this block.
12466 currentBlock = moveToNextBlock();
12469 if (currentBlock != nullptr)
12471 if (enregisterLocalVars)
12473 VarToRegMap inVarToRegMap = inVarToRegMaps[currentBlock->bbNum];
12474 VarSetOps::Iter iter(compiler, currentBlock->bbLiveIn);
12475 unsigned varIndex = 0;
12476 while (iter.NextElem(&varIndex))
12478 if (localVarIntervals[varIndex] == nullptr)
12480 assert(!compiler->lvaTable[compiler->lvaTrackedToVarNum[varIndex]].lvLRACandidate);
12483 regNumber regNum = getVarReg(inVarToRegMap, varIndex);
12484 interval = getIntervalForLocalVar(varIndex);
12485 interval->physReg = regNum;
12486 interval->assignedReg = &(physRegs[regNum]);
12487 interval->isActive = true;
12488 physRegs[regNum].assignedInterval = interval;
12494 dumpRefPositionShort(currentRefPosition, currentBlock);
12498 // Finally, handle the resolution moves, if any, at the beginning of the next block.
12499 firstBlockEndResolutionNode = nullptr;
12500 bool foundNonResolutionNode = false;
12502 LIR::Range& currentBlockRange = LIR::AsRange(currentBlock);
12503 for (GenTree* node : currentBlockRange.NonPhiNodes())
12505 if (IsResolutionNode(currentBlockRange, node))
12507 assert(enregisterLocalVars);
12508 if (foundNonResolutionNode)
12510 firstBlockEndResolutionNode = node;
12513 else if (IsResolutionMove(node))
12515 // Only verify nodes that are actually moves; don't bother with the nodes that are
12516 // operands to moves.
12517 verifyResolutionMove(node, currentLocation);
12522 foundNonResolutionNode = true;
12531 assert(regRecord != nullptr);
12532 assert(regRecord->assignedInterval == nullptr);
12533 dumpLsraAllocationEvent(LSRA_EVENT_KEPT_ALLOCATION, nullptr, regRecord->regNum, currentBlock);
12535 case RefTypeFixedReg:
12536 assert(regRecord != nullptr);
12537 dumpLsraAllocationEvent(LSRA_EVENT_KEPT_ALLOCATION, nullptr, regRecord->regNum, currentBlock);
12540 case RefTypeUpperVectorSaveDef:
12541 case RefTypeUpperVectorSaveUse:
12544 case RefTypeParamDef:
12545 case RefTypeZeroInit:
12546 assert(interval != nullptr);
12548 if (interval->isSpecialPutArg)
12550 dumpLsraAllocationEvent(LSRA_EVENT_SPECIAL_PUTARG, interval, regNum);
12553 if (currentRefPosition->reload)
12555 interval->isActive = true;
12556 assert(regNum != REG_NA);
12557 interval->physReg = regNum;
12558 interval->assignedReg = regRecord;
12559 regRecord->assignedInterval = interval;
12560 dumpLsraAllocationEvent(LSRA_EVENT_RELOAD, nullptr, regRecord->regNum, currentBlock);
12562 if (regNum == REG_NA)
12564 dumpLsraAllocationEvent(LSRA_EVENT_NO_REG_ALLOCATED, interval);
12566 else if (RefTypeIsDef(currentRefPosition->refType))
12568 interval->isActive = true;
12571 if (interval->isConstant && (currentRefPosition->treeNode != nullptr) &&
12572 currentRefPosition->treeNode->IsReuseRegVal())
12574 dumpLsraAllocationEvent(LSRA_EVENT_REUSE_REG, nullptr, regRecord->regNum, currentBlock);
12578 dumpLsraAllocationEvent(LSRA_EVENT_ALLOC_REG, nullptr, regRecord->regNum, currentBlock);
12582 else if (currentRefPosition->copyReg)
12584 dumpLsraAllocationEvent(LSRA_EVENT_COPY_REG, interval, regRecord->regNum, currentBlock);
12586 else if (currentRefPosition->moveReg)
12588 assert(interval->assignedReg != nullptr);
12589 interval->assignedReg->assignedInterval = nullptr;
12590 interval->physReg = regNum;
12591 interval->assignedReg = regRecord;
12592 regRecord->assignedInterval = interval;
12595 printf("Move %-4s ", getRegName(regRecord->regNum));
12600 dumpLsraAllocationEvent(LSRA_EVENT_KEPT_ALLOCATION, nullptr, regRecord->regNum, currentBlock);
12602 if (currentRefPosition->lastUse || currentRefPosition->spillAfter)
12604 interval->isActive = false;
12606 if (regNum != REG_NA)
12608 if (currentRefPosition->spillAfter)
12612 // If refPos is marked as copyReg, then the reg that is spilled
12613 // is the homeReg of the interval not the reg currently assigned
12615 regNumber spillReg = regNum;
12616 if (currentRefPosition->copyReg)
12618 assert(interval != nullptr);
12619 spillReg = interval->physReg;
12622 dumpEmptyRefPosition();
12623 printf("Spill %-4s ", getRegName(spillReg));
12626 else if (currentRefPosition->copyReg)
12628 regRecord->assignedInterval = interval;
12632 interval->physReg = regNum;
12633 interval->assignedReg = regRecord;
12634 regRecord->assignedInterval = interval;
12638 case RefTypeKillGCRefs:
12639 // No action to take.
12640 // However, we will assert that, at resolution time, no registers contain GC refs.
12642 DBEXEC(VERBOSE, printf(" "));
12643 regMaskTP candidateRegs = currentRefPosition->registerAssignment;
12644 while (candidateRegs != RBM_NONE)
12646 regMaskTP nextRegBit = genFindLowestBit(candidateRegs);
12647 candidateRegs &= ~nextRegBit;
12648 regNumber nextReg = genRegNumFromMask(nextRegBit);
12649 RegRecord* regRecord = getRegisterRecord(nextReg);
12650 Interval* assignedInterval = regRecord->assignedInterval;
12651 assert(assignedInterval == nullptr || !varTypeIsGC(assignedInterval->registerType));
12656 case RefTypeExpUse:
12657 case RefTypeDummyDef:
12658 // Do nothing; these will be handled by the RefTypeBB.
12659 DBEXEC(VERBOSE, printf(" "));
12662 case RefTypeInvalid:
12663 // for these 'currentRefPosition->refType' values, No action to take
12667 if (currentRefPosition->refType != RefTypeBB)
12669 DBEXEC(VERBOSE, dumpRegRecords());
12670 if (interval != nullptr)
12672 if (currentRefPosition->copyReg)
12674 assert(interval->physReg != regNum);
12675 regRecord->assignedInterval = nullptr;
12676 assert(interval->assignedReg != nullptr);
12677 regRecord = interval->assignedReg;
12679 if (currentRefPosition->spillAfter || currentRefPosition->lastUse)
12681 interval->physReg = REG_NA;
12682 interval->assignedReg = nullptr;
12684 // regRegcord could be null if the RefPosition does not require a register.
12685 if (regRecord != nullptr)
12687 regRecord->assignedInterval = nullptr;
12691 assert(!currentRefPosition->RequiresRegister());
12698 // Now, verify the resolution blocks.
12699 // Currently these are nearly always at the end of the method, but that may not alwyas be the case.
12700 // So, we'll go through all the BBs looking for blocks whose bbNum is greater than bbNumMaxBeforeResolution.
12701 for (BasicBlock* currentBlock = compiler->fgFirstBB; currentBlock != nullptr; currentBlock = currentBlock->bbNext)
12703 if (currentBlock->bbNum > bbNumMaxBeforeResolution)
12705 // If we haven't enregistered an lclVars, we have no resolution blocks.
12706 assert(enregisterLocalVars);
12710 dumpRegRecordTitle();
12711 printf(shortRefPositionFormat, 0, 0);
12712 assert(currentBlock->bbPreds != nullptr && currentBlock->bbPreds->flBlock != nullptr);
12713 printf(bbRefPosFormat, currentBlock->bbNum, currentBlock->bbPreds->flBlock->bbNum);
12717 // Clear register assignments.
12718 for (regNumber reg = REG_FIRST; reg < ACTUAL_REG_COUNT; reg = REG_NEXT(reg))
12720 RegRecord* physRegRecord = getRegisterRecord(reg);
12721 physRegRecord->assignedInterval = nullptr;
12724 // Set the incoming register assignments
12725 VarToRegMap inVarToRegMap = getInVarToRegMap(currentBlock->bbNum);
12726 VarSetOps::Iter iter(compiler, currentBlock->bbLiveIn);
12727 unsigned varIndex = 0;
12728 while (iter.NextElem(&varIndex))
12730 if (localVarIntervals[varIndex] == nullptr)
12732 assert(!compiler->lvaTable[compiler->lvaTrackedToVarNum[varIndex]].lvLRACandidate);
12735 regNumber regNum = getVarReg(inVarToRegMap, varIndex);
12736 Interval* interval = getIntervalForLocalVar(varIndex);
12737 interval->physReg = regNum;
12738 interval->assignedReg = &(physRegs[regNum]);
12739 interval->isActive = true;
12740 physRegs[regNum].assignedInterval = interval;
12743 // Verify the moves in this block
12744 LIR::Range& currentBlockRange = LIR::AsRange(currentBlock);
12745 for (GenTree* node : currentBlockRange.NonPhiNodes())
12747 assert(IsResolutionNode(currentBlockRange, node));
12748 if (IsResolutionMove(node))
12750 // Only verify nodes that are actually moves; don't bother with the nodes that are
12751 // operands to moves.
12752 verifyResolutionMove(node, currentLocation);
12756 // Verify the outgoing register assignments
12758 VarToRegMap outVarToRegMap = getOutVarToRegMap(currentBlock->bbNum);
12759 VarSetOps::Iter iter(compiler, currentBlock->bbLiveOut);
12760 unsigned varIndex = 0;
12761 while (iter.NextElem(&varIndex))
12763 if (localVarIntervals[varIndex] == nullptr)
12765 assert(!compiler->lvaTable[compiler->lvaTrackedToVarNum[varIndex]].lvLRACandidate);
12768 regNumber regNum = getVarReg(outVarToRegMap, varIndex);
12769 Interval* interval = getIntervalForLocalVar(varIndex);
12770 assert(interval->physReg == regNum || (interval->physReg == REG_NA && regNum == REG_STK));
12771 interval->physReg = REG_NA;
12772 interval->assignedReg = nullptr;
12773 interval->isActive = false;
12779 DBEXEC(VERBOSE, printf("\n"));
12782 //------------------------------------------------------------------------
12783 // verifyResolutionMove: Verify a resolution statement. Called by verifyFinalAllocation()
12786 // resolutionMove - A GenTree* that must be a resolution move.
12787 // currentLocation - The LsraLocation of the most recent RefPosition that has been verified.
12793 // If verbose is set, this will also dump the moves into the table of final allocations.
12794 void LinearScan::verifyResolutionMove(GenTree* resolutionMove, LsraLocation currentLocation)
12796 GenTree* dst = resolutionMove;
12797 assert(IsResolutionMove(dst));
12799 if (dst->OperGet() == GT_SWAP)
12801 GenTreeLclVarCommon* left = dst->gtGetOp1()->AsLclVarCommon();
12802 GenTreeLclVarCommon* right = dst->gtGetOp2()->AsLclVarCommon();
12803 regNumber leftRegNum = left->gtRegNum;
12804 regNumber rightRegNum = right->gtRegNum;
12805 LclVarDsc* leftVarDsc = compiler->lvaTable + left->gtLclNum;
12806 LclVarDsc* rightVarDsc = compiler->lvaTable + right->gtLclNum;
12807 Interval* leftInterval = getIntervalForLocalVar(leftVarDsc->lvVarIndex);
12808 Interval* rightInterval = getIntervalForLocalVar(rightVarDsc->lvVarIndex);
12809 assert(leftInterval->physReg == leftRegNum && rightInterval->physReg == rightRegNum);
12810 leftInterval->physReg = rightRegNum;
12811 rightInterval->physReg = leftRegNum;
12812 leftInterval->assignedReg = &physRegs[rightRegNum];
12813 rightInterval->assignedReg = &physRegs[leftRegNum];
12814 physRegs[rightRegNum].assignedInterval = leftInterval;
12815 physRegs[leftRegNum].assignedInterval = rightInterval;
12818 printf(shortRefPositionFormat, currentLocation, 0);
12819 dumpIntervalName(leftInterval);
12821 printf(" %-4s ", getRegName(rightRegNum));
12823 printf(shortRefPositionFormat, currentLocation, 0);
12824 dumpIntervalName(rightInterval);
12826 printf(" %-4s ", getRegName(leftRegNum));
12831 regNumber dstRegNum = dst->gtRegNum;
12832 regNumber srcRegNum;
12833 GenTreeLclVarCommon* lcl;
12834 if (dst->OperGet() == GT_COPY)
12836 lcl = dst->gtGetOp1()->AsLclVarCommon();
12837 srcRegNum = lcl->gtRegNum;
12841 lcl = dst->AsLclVarCommon();
12842 if ((lcl->gtFlags & GTF_SPILLED) != 0)
12844 srcRegNum = REG_STK;
12848 assert((lcl->gtFlags & GTF_SPILL) != 0);
12849 srcRegNum = dstRegNum;
12850 dstRegNum = REG_STK;
12854 Interval* interval = getIntervalForLocalVarNode(lcl);
12855 assert(interval->physReg == srcRegNum || (srcRegNum == REG_STK && interval->physReg == REG_NA));
12856 if (srcRegNum != REG_STK)
12858 physRegs[srcRegNum].assignedInterval = nullptr;
12860 if (dstRegNum != REG_STK)
12862 interval->physReg = dstRegNum;
12863 interval->assignedReg = &(physRegs[dstRegNum]);
12864 physRegs[dstRegNum].assignedInterval = interval;
12865 interval->isActive = true;
12869 interval->physReg = REG_NA;
12870 interval->assignedReg = nullptr;
12871 interval->isActive = false;
12875 printf(shortRefPositionFormat, currentLocation, 0);
12876 dumpIntervalName(interval);
12878 printf(" %-4s ", getRegName(dstRegNum));
12884 #endif // !LEGACY_BACKEND