xref: /llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.cpp (revision e6b24955455d74ee748d1e9986d67de6d40ed22e)
1 //===-- LegalizeTypes.cpp - Common code for DAG type legalizer ------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the SelectionDAG::LegalizeTypes method.  It transforms
10 // an arbitrary well-formed SelectionDAG to only consist of legal types.  This
11 // is common code shared among the LegalizeTypes*.cpp files.
12 //
13 //===----------------------------------------------------------------------===//
14 
15 #include "LegalizeTypes.h"
16 #include "llvm/ADT/SetVector.h"
17 #include "llvm/IR/DataLayout.h"
18 #include "llvm/Support/CommandLine.h"
19 #include "llvm/Support/ErrorHandling.h"
20 #include "llvm/Support/raw_ostream.h"
21 using namespace llvm;
22 
23 #define DEBUG_TYPE "legalize-types"
24 
25 static cl::opt<bool>
26 EnableExpensiveChecks("enable-legalize-types-checking", cl::Hidden);
27 
28 /// Do extensive, expensive, basic correctness checking.
29 void DAGTypeLegalizer::PerformExpensiveChecks() {
30   // If a node is not processed, then none of its values should be mapped by any
31   // of PromotedIntegers, ExpandedIntegers, ..., ReplacedValues.
32 
33   // If a node is processed, then each value with an illegal type must be mapped
34   // by exactly one of PromotedIntegers, ExpandedIntegers, ..., ReplacedValues.
35   // Values with a legal type may be mapped by ReplacedValues, but not by any of
36   // the other maps.
37 
38   // Note that these invariants may not hold momentarily when processing a node:
39   // the node being processed may be put in a map before being marked Processed.
40 
41   // Note that it is possible to have nodes marked NewNode in the DAG.  This can
42   // occur in two ways.  Firstly, a node may be created during legalization but
43   // never passed to the legalization core.  This is usually due to the implicit
44   // folding that occurs when using the DAG.getNode operators.  Secondly, a new
45   // node may be passed to the legalization core, but when analyzed may morph
46   // into a different node, leaving the original node as a NewNode in the DAG.
47   // A node may morph if one of its operands changes during analysis.  Whether
48   // it actually morphs or not depends on whether, after updating its operands,
49   // it is equivalent to an existing node: if so, it morphs into that existing
50   // node (CSE).  An operand can change during analysis if the operand is a new
51   // node that morphs, or it is a processed value that was mapped to some other
52   // value (as recorded in ReplacedValues) in which case the operand is turned
53   // into that other value.  If a node morphs then the node it morphed into will
54   // be used instead of it for legalization, however the original node continues
55   // to live on in the DAG.
56   // The conclusion is that though there may be nodes marked NewNode in the DAG,
57   // all uses of such nodes are also marked NewNode: the result is a fungus of
58   // NewNodes growing on top of the useful nodes, and perhaps using them, but
59   // not used by them.
60 
61   // If a value is mapped by ReplacedValues, then it must have no uses, except
62   // by nodes marked NewNode (see above).
63 
64   // The final node obtained by mapping by ReplacedValues is not marked NewNode.
65   // Note that ReplacedValues should be applied iteratively.
66 
67   // Note that the ReplacedValues map may also map deleted nodes (by iterating
68   // over the DAG we never dereference deleted nodes).  This means that it may
69   // also map nodes marked NewNode if the deallocated memory was reallocated as
70   // another node, and that new node was not seen by the LegalizeTypes machinery
71   // (for example because it was created but not used).  In general, we cannot
72   // distinguish between new nodes and deleted nodes.
73   SmallVector<SDNode*, 16> NewNodes;
74   for (SDNode &Node : DAG.allnodes()) {
75     // Remember nodes marked NewNode - they are subject to extra checking below.
76     if (Node.getNodeId() == NewNode)
77       NewNodes.push_back(&Node);
78 
79     for (unsigned i = 0, e = Node.getNumValues(); i != e; ++i) {
80       SDValue Res(&Node, i);
81       bool Failed = false;
82       // Don't create a value in map.
83       auto ResId = ValueToIdMap.lookup(Res);
84 
85       unsigned Mapped = 0;
86       if (ResId) {
87         auto I = ReplacedValues.find(ResId);
88         if (I != ReplacedValues.end()) {
89           Mapped |= 1;
90           // Check that remapped values are only used by nodes marked NewNode.
91           for (SDUse &U : Node.uses())
92             if (U.getResNo() == i)
93               assert(U.getUser()->getNodeId() == NewNode &&
94                      "Remapped value has non-trivial use!");
95 
96           // Check that the final result of applying ReplacedValues is not
97           // marked NewNode.
98           auto NewValId = I->second;
99           I = ReplacedValues.find(NewValId);
100           while (I != ReplacedValues.end()) {
101             NewValId = I->second;
102             I = ReplacedValues.find(NewValId);
103           }
104           SDValue NewVal = getSDValue(NewValId);
105           (void)NewVal;
106           assert(NewVal.getNode()->getNodeId() != NewNode &&
107                  "ReplacedValues maps to a new node!");
108         }
109         if (PromotedIntegers.count(ResId))
110           Mapped |= 2;
111         if (SoftenedFloats.count(ResId))
112           Mapped |= 4;
113         if (ScalarizedVectors.count(ResId))
114           Mapped |= 8;
115         if (ExpandedIntegers.count(ResId))
116           Mapped |= 16;
117         if (ExpandedFloats.count(ResId))
118           Mapped |= 32;
119         if (SplitVectors.count(ResId))
120           Mapped |= 64;
121         if (WidenedVectors.count(ResId))
122           Mapped |= 128;
123         if (PromotedFloats.count(ResId))
124           Mapped |= 256;
125         if (SoftPromotedHalfs.count(ResId))
126           Mapped |= 512;
127       }
128 
129       if (Node.getNodeId() != Processed) {
130         // Since we allow ReplacedValues to map deleted nodes, it may map nodes
131         // marked NewNode too, since a deleted node may have been reallocated as
132         // another node that has not been seen by the LegalizeTypes machinery.
133         if ((Node.getNodeId() == NewNode && Mapped > 1) ||
134             (Node.getNodeId() != NewNode && Mapped != 0)) {
135           dbgs() << "Unprocessed value in a map!";
136           Failed = true;
137         }
138       } else if (isTypeLegal(Res.getValueType()) || IgnoreNodeResults(&Node)) {
139         if (Mapped > 1) {
140           dbgs() << "Value with legal type was transformed!";
141           Failed = true;
142         }
143       } else {
144         if (Mapped == 0) {
145           SDValue NodeById = IdToValueMap.lookup(ResId);
146           // It is possible the node has been remapped to another node and had
147           // its Id updated in the Value to Id table. The node it remapped to
148           // may not have been processed yet. Look up the Id in the Id to Value
149           // table and re-check the Processed state. If the node hasn't been
150           // remapped we'll get the same state as we got earlier.
151           if (NodeById->getNodeId() == Processed) {
152             dbgs() << "Processed value not in any map!";
153             Failed = true;
154           }
155         } else if (Mapped & (Mapped - 1)) {
156           dbgs() << "Value in multiple maps!";
157           Failed = true;
158         }
159       }
160 
161       if (Failed) {
162         if (Mapped & 1)
163           dbgs() << " ReplacedValues";
164         if (Mapped & 2)
165           dbgs() << " PromotedIntegers";
166         if (Mapped & 4)
167           dbgs() << " SoftenedFloats";
168         if (Mapped & 8)
169           dbgs() << " ScalarizedVectors";
170         if (Mapped & 16)
171           dbgs() << " ExpandedIntegers";
172         if (Mapped & 32)
173           dbgs() << " ExpandedFloats";
174         if (Mapped & 64)
175           dbgs() << " SplitVectors";
176         if (Mapped & 128)
177           dbgs() << " WidenedVectors";
178         if (Mapped & 256)
179           dbgs() << " PromotedFloats";
180         if (Mapped & 512)
181           dbgs() << " SoftPromoteHalfs";
182         dbgs() << "\n";
183         llvm_unreachable(nullptr);
184       }
185     }
186   }
187 
188 #ifndef NDEBUG
189   // Checked that NewNodes are only used by other NewNodes.
190   for (SDNode *N : NewNodes) {
191     for (SDNode *U : N->users())
192       assert(U->getNodeId() == NewNode && "NewNode used by non-NewNode!");
193   }
194 #endif
195 }
196 
197 /// This is the main entry point for the type legalizer. This does a top-down
198 /// traversal of the dag, legalizing types as it goes. Returns "true" if it made
199 /// any changes.
200 bool DAGTypeLegalizer::run() {
201   bool Changed = false;
202 
203   // Create a dummy node (which is not added to allnodes), that adds a reference
204   // to the root node, preventing it from being deleted, and tracking any
205   // changes of the root.
206   HandleSDNode Dummy(DAG.getRoot());
207   Dummy.setNodeId(Unanalyzed);
208 
209   // The root of the dag may dangle to deleted nodes until the type legalizer is
210   // done.  Set it to null to avoid confusion.
211   DAG.setRoot(SDValue());
212 
213   // Walk all nodes in the graph, assigning them a NodeId of 'ReadyToProcess'
214   // (and remembering them) if they are leaves and assigning 'Unanalyzed' if
215   // non-leaves.
216   for (SDNode &Node : DAG.allnodes()) {
217     if (Node.getNumOperands() == 0) {
218       Node.setNodeId(ReadyToProcess);
219       Worklist.push_back(&Node);
220     } else {
221       Node.setNodeId(Unanalyzed);
222     }
223   }
224 
225   // Now that we have a set of nodes to process, handle them all.
226   while (!Worklist.empty()) {
227 #ifndef EXPENSIVE_CHECKS
228     if (EnableExpensiveChecks)
229 #endif
230       PerformExpensiveChecks();
231 
232     SDNode *N = Worklist.pop_back_val();
233     assert(N->getNodeId() == ReadyToProcess &&
234            "Node should be ready if on worklist!");
235 
236     LLVM_DEBUG(dbgs() << "\nLegalizing node: "; N->dump(&DAG));
237     if (IgnoreNodeResults(N)) {
238       LLVM_DEBUG(dbgs() << "Ignoring node results\n");
239       goto ScanOperands;
240     }
241 
242     // Scan the values produced by the node, checking to see if any result
243     // types are illegal.
244     for (unsigned i = 0, NumResults = N->getNumValues(); i < NumResults; ++i) {
245       EVT ResultVT = N->getValueType(i);
246       LLVM_DEBUG(dbgs() << "Analyzing result type: " << ResultVT << "\n");
247       switch (getTypeAction(ResultVT)) {
248       case TargetLowering::TypeLegal:
249         LLVM_DEBUG(dbgs() << "Legal result type\n");
250         break;
251       case TargetLowering::TypeScalarizeScalableVector:
252         report_fatal_error(
253             "Scalarization of scalable vectors is not supported.");
254       // The following calls must take care of *all* of the node's results,
255       // not just the illegal result they were passed (this includes results
256       // with a legal type).  Results can be remapped using ReplaceValueWith,
257       // or their promoted/expanded/etc values registered in PromotedIntegers,
258       // ExpandedIntegers etc.
259       case TargetLowering::TypePromoteInteger:
260         PromoteIntegerResult(N, i);
261         Changed = true;
262         goto NodeDone;
263       case TargetLowering::TypeExpandInteger:
264         ExpandIntegerResult(N, i);
265         Changed = true;
266         goto NodeDone;
267       case TargetLowering::TypeSoftenFloat:
268         SoftenFloatResult(N, i);
269         Changed = true;
270         goto NodeDone;
271       case TargetLowering::TypeExpandFloat:
272         ExpandFloatResult(N, i);
273         Changed = true;
274         goto NodeDone;
275       case TargetLowering::TypeScalarizeVector:
276         ScalarizeVectorResult(N, i);
277         Changed = true;
278         goto NodeDone;
279       case TargetLowering::TypeSplitVector:
280         SplitVectorResult(N, i);
281         Changed = true;
282         goto NodeDone;
283       case TargetLowering::TypeWidenVector:
284         WidenVectorResult(N, i);
285         Changed = true;
286         goto NodeDone;
287       case TargetLowering::TypePromoteFloat:
288         PromoteFloatResult(N, i);
289         Changed = true;
290         goto NodeDone;
291       case TargetLowering::TypeSoftPromoteHalf:
292         SoftPromoteHalfResult(N, i);
293         Changed = true;
294         goto NodeDone;
295       }
296     }
297 
298 ScanOperands:
299     // Scan the operand list for the node, handling any nodes with operands that
300     // are illegal.
301     {
302     unsigned NumOperands = N->getNumOperands();
303     bool NeedsReanalyzing = false;
304     unsigned i;
305     for (i = 0; i != NumOperands; ++i) {
306       if (IgnoreNodeResults(N->getOperand(i).getNode()))
307         continue;
308 
309       const auto &Op = N->getOperand(i);
310       LLVM_DEBUG(dbgs() << "Analyzing operand: "; Op.dump(&DAG));
311       EVT OpVT = Op.getValueType();
312       switch (getTypeAction(OpVT)) {
313       case TargetLowering::TypeLegal:
314         LLVM_DEBUG(dbgs() << "Legal operand\n");
315         continue;
316       case TargetLowering::TypeScalarizeScalableVector:
317         report_fatal_error(
318             "Scalarization of scalable vectors is not supported.");
319       // The following calls must either replace all of the node's results
320       // using ReplaceValueWith, and return "false"; or update the node's
321       // operands in place, and return "true".
322       case TargetLowering::TypePromoteInteger:
323         NeedsReanalyzing = PromoteIntegerOperand(N, i);
324         Changed = true;
325         break;
326       case TargetLowering::TypeExpandInteger:
327         NeedsReanalyzing = ExpandIntegerOperand(N, i);
328         Changed = true;
329         break;
330       case TargetLowering::TypeSoftenFloat:
331         NeedsReanalyzing = SoftenFloatOperand(N, i);
332         Changed = true;
333         break;
334       case TargetLowering::TypeExpandFloat:
335         NeedsReanalyzing = ExpandFloatOperand(N, i);
336         Changed = true;
337         break;
338       case TargetLowering::TypeScalarizeVector:
339         NeedsReanalyzing = ScalarizeVectorOperand(N, i);
340         Changed = true;
341         break;
342       case TargetLowering::TypeSplitVector:
343         NeedsReanalyzing = SplitVectorOperand(N, i);
344         Changed = true;
345         break;
346       case TargetLowering::TypeWidenVector:
347         NeedsReanalyzing = WidenVectorOperand(N, i);
348         Changed = true;
349         break;
350       case TargetLowering::TypePromoteFloat:
351         NeedsReanalyzing = PromoteFloatOperand(N, i);
352         Changed = true;
353         break;
354       case TargetLowering::TypeSoftPromoteHalf:
355         NeedsReanalyzing = SoftPromoteHalfOperand(N, i);
356         Changed = true;
357         break;
358       }
359       break;
360     }
361 
362     // The sub-method updated N in place.  Check to see if any operands are new,
363     // and if so, mark them.  If the node needs revisiting, don't add all users
364     // to the worklist etc.
365     if (NeedsReanalyzing) {
366       assert(N->getNodeId() == ReadyToProcess && "Node ID recalculated?");
367 
368       N->setNodeId(NewNode);
369       // Recompute the NodeId and correct processed operands, adding the node to
370       // the worklist if ready.
371       SDNode *M = AnalyzeNewNode(N);
372       if (M == N)
373         // The node didn't morph - nothing special to do, it will be revisited.
374         continue;
375 
376       // The node morphed - this is equivalent to legalizing by replacing every
377       // value of N with the corresponding value of M.  So do that now.
378       assert(N->getNumValues() == M->getNumValues() &&
379              "Node morphing changed the number of results!");
380       for (unsigned i = 0, e = N->getNumValues(); i != e; ++i)
381         // Replacing the value takes care of remapping the new value.
382         ReplaceValueWith(SDValue(N, i), SDValue(M, i));
383       assert(N->getNodeId() == NewNode && "Unexpected node state!");
384       // The node continues to live on as part of the NewNode fungus that
385       // grows on top of the useful nodes.  Nothing more needs to be done
386       // with it - move on to the next node.
387       continue;
388     }
389 
390     if (i == NumOperands) {
391       LLVM_DEBUG(dbgs() << "Legally typed node: "; N->dump(&DAG));
392     }
393     }
394 NodeDone:
395 
396     // If we reach here, the node was processed, potentially creating new nodes.
397     // Mark it as processed and add its users to the worklist as appropriate.
398     assert(N->getNodeId() == ReadyToProcess && "Node ID recalculated?");
399     N->setNodeId(Processed);
400 
401     for (SDNode *User : N->users()) {
402       int NodeId = User->getNodeId();
403 
404       // This node has two options: it can either be a new node or its Node ID
405       // may be a count of the number of operands it has that are not ready.
406       if (NodeId > 0) {
407         User->setNodeId(NodeId-1);
408 
409         // If this was the last use it was waiting on, add it to the ready list.
410         if (NodeId-1 == ReadyToProcess)
411           Worklist.push_back(User);
412         continue;
413       }
414 
415       // If this is an unreachable new node, then ignore it.  If it ever becomes
416       // reachable by being used by a newly created node then it will be handled
417       // by AnalyzeNewNode.
418       if (NodeId == NewNode)
419         continue;
420 
421       // Otherwise, this node is new: this is the first operand of it that
422       // became ready.  Its new NodeId is the number of operands it has minus 1
423       // (as this node is now processed).
424       assert(NodeId == Unanalyzed && "Unknown node ID!");
425       User->setNodeId(User->getNumOperands() - 1);
426 
427       // If the node only has a single operand, it is now ready.
428       if (User->getNumOperands() == 1)
429         Worklist.push_back(User);
430     }
431   }
432 
433 #ifndef EXPENSIVE_CHECKS
434   if (EnableExpensiveChecks)
435 #endif
436     PerformExpensiveChecks();
437 
438   // If the root changed (e.g. it was a dead load) update the root.
439   DAG.setRoot(Dummy.getValue());
440 
441   // Remove dead nodes.  This is important to do for cleanliness but also before
442   // the checking loop below.  Implicit folding by the DAG.getNode operators and
443   // node morphing can cause unreachable nodes to be around with their flags set
444   // to new.
445   DAG.RemoveDeadNodes();
446 
447   // In a debug build, scan all the nodes to make sure we found them all.  This
448   // ensures that there are no cycles and that everything got processed.
449 #ifndef NDEBUG
450   for (SDNode &Node : DAG.allnodes()) {
451     bool Failed = false;
452 
453     // Check that all result types are legal.
454     if (!IgnoreNodeResults(&Node))
455       for (unsigned i = 0, NumVals = Node.getNumValues(); i < NumVals; ++i)
456         if (!isTypeLegal(Node.getValueType(i))) {
457           dbgs() << "Result type " << i << " illegal: ";
458           Node.dump(&DAG);
459           Failed = true;
460         }
461 
462     // Check that all operand types are legal.
463     for (unsigned i = 0, NumOps = Node.getNumOperands(); i < NumOps; ++i)
464       if (!IgnoreNodeResults(Node.getOperand(i).getNode()) &&
465           !isTypeLegal(Node.getOperand(i).getValueType())) {
466         dbgs() << "Operand type " << i << " illegal: ";
467         Node.getOperand(i).dump(&DAG);
468         Failed = true;
469       }
470 
471     if (Node.getNodeId() != Processed) {
472        if (Node.getNodeId() == NewNode)
473          dbgs() << "New node not analyzed?\n";
474        else if (Node.getNodeId() == Unanalyzed)
475          dbgs() << "Unanalyzed node not noticed?\n";
476        else if (Node.getNodeId() > 0)
477          dbgs() << "Operand not processed?\n";
478        else if (Node.getNodeId() == ReadyToProcess)
479          dbgs() << "Not added to worklist?\n";
480        Failed = true;
481     }
482 
483     if (Failed) {
484       Node.dump(&DAG); dbgs() << "\n";
485       llvm_unreachable(nullptr);
486     }
487   }
488 #endif
489 
490   return Changed;
491 }
492 
493 /// The specified node is the root of a subtree of potentially new nodes.
494 /// Correct any processed operands (this may change the node) and calculate the
495 /// NodeId. If the node itself changes to a processed node, it is not remapped -
496 /// the caller needs to take care of this. Returns the potentially changed node.
497 SDNode *DAGTypeLegalizer::AnalyzeNewNode(SDNode *N) {
498   // If this was an existing node that is already done, we're done.
499   if (N->getNodeId() != NewNode && N->getNodeId() != Unanalyzed)
500     return N;
501 
502   // Okay, we know that this node is new.  Recursively walk all of its operands
503   // to see if they are new also.  The depth of this walk is bounded by the size
504   // of the new tree that was constructed (usually 2-3 nodes), so we don't worry
505   // about revisiting of nodes.
506   //
507   // As we walk the operands, keep track of the number of nodes that are
508   // processed.  If non-zero, this will become the new nodeid of this node.
509   // Operands may morph when they are analyzed.  If so, the node will be
510   // updated after all operands have been analyzed.  Since this is rare,
511   // the code tries to minimize overhead in the non-morphing case.
512 
513   std::vector<SDValue> NewOps;
514   unsigned NumProcessed = 0;
515   for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) {
516     SDValue OrigOp = N->getOperand(i);
517     SDValue Op = OrigOp;
518 
519     AnalyzeNewValue(Op); // Op may morph.
520 
521     if (Op.getNode()->getNodeId() == Processed)
522       ++NumProcessed;
523 
524     if (!NewOps.empty()) {
525       // Some previous operand changed.  Add this one to the list.
526       NewOps.push_back(Op);
527     } else if (Op != OrigOp) {
528       // This is the first operand to change - add all operands so far.
529       NewOps.insert(NewOps.end(), N->op_begin(), N->op_begin() + i);
530       NewOps.push_back(Op);
531     }
532   }
533 
534   // Some operands changed - update the node.
535   if (!NewOps.empty()) {
536     SDNode *M = DAG.UpdateNodeOperands(N, NewOps);
537     if (M != N) {
538       // The node morphed into a different node.  Normally for this to happen
539       // the original node would have to be marked NewNode.  However this can
540       // in theory momentarily not be the case while ReplaceValueWith is doing
541       // its stuff.  Mark the original node NewNode to help basic correctness
542       // checking.
543       N->setNodeId(NewNode);
544       if (M->getNodeId() != NewNode && M->getNodeId() != Unanalyzed)
545         // It morphed into a previously analyzed node - nothing more to do.
546         return M;
547 
548       // It morphed into a different new node.  Do the equivalent of passing
549       // it to AnalyzeNewNode: expunge it and calculate the NodeId.  No need
550       // to remap the operands, since they are the same as the operands we
551       // remapped above.
552       N = M;
553     }
554   }
555 
556   // Calculate the NodeId.
557   N->setNodeId(N->getNumOperands() - NumProcessed);
558   if (N->getNodeId() == ReadyToProcess)
559     Worklist.push_back(N);
560 
561   return N;
562 }
563 
564 /// Call AnalyzeNewNode, updating the node in Val if needed.
565 /// If the node changes to a processed node, then remap it.
566 void DAGTypeLegalizer::AnalyzeNewValue(SDValue &Val) {
567   Val.setNode(AnalyzeNewNode(Val.getNode()));
568   if (Val.getNode()->getNodeId() == Processed)
569     // We were passed a processed node, or it morphed into one - remap it.
570     RemapValue(Val);
571 }
572 
573 /// If the specified value was already legalized to another value,
574 /// replace it by that value.
575 void DAGTypeLegalizer::RemapValue(SDValue &V) {
576   auto Id = getTableId(V);
577   V = getSDValue(Id);
578 }
579 
580 void DAGTypeLegalizer::RemapId(TableId &Id) {
581   auto I = ReplacedValues.find(Id);
582   if (I != ReplacedValues.end()) {
583     assert(Id != I->second && "Id is mapped to itself.");
584     // Use path compression to speed up future lookups if values get multiply
585     // replaced with other values.
586     RemapId(I->second);
587     Id = I->second;
588 
589     // Note that N = IdToValueMap[Id] it is possible to have
590     // N.getNode()->getNodeId() == NewNode at this point because it is possible
591     // for a node to be put in the map before being processed.
592   }
593 }
594 
595 namespace {
596   /// This class is a DAGUpdateListener that listens for updates to nodes and
597   /// recomputes their ready state.
598   class NodeUpdateListener : public SelectionDAG::DAGUpdateListener {
599     DAGTypeLegalizer &DTL;
600     SmallSetVector<SDNode*, 16> &NodesToAnalyze;
601   public:
602     explicit NodeUpdateListener(DAGTypeLegalizer &dtl,
603                                 SmallSetVector<SDNode*, 16> &nta)
604       : SelectionDAG::DAGUpdateListener(dtl.getDAG()),
605         DTL(dtl), NodesToAnalyze(nta) {}
606 
607     void NodeDeleted(SDNode *N, SDNode *E) override {
608       assert(N->getNodeId() != DAGTypeLegalizer::ReadyToProcess &&
609              N->getNodeId() != DAGTypeLegalizer::Processed &&
610              "Invalid node ID for RAUW deletion!");
611       // It is possible, though rare, for the deleted node N to occur as a
612       // target in a map, so note the replacement N -> E in ReplacedValues.
613       assert(E && "Node not replaced?");
614       DTL.NoteDeletion(N, E);
615 
616       // In theory the deleted node could also have been scheduled for analysis.
617       // So remove it from the set of nodes which will be analyzed.
618       NodesToAnalyze.remove(N);
619 
620       // In general nothing needs to be done for E, since it didn't change but
621       // only gained new uses.  However N -> E was just added to ReplacedValues,
622       // and the result of a ReplacedValues mapping is not allowed to be marked
623       // NewNode.  So if E is marked NewNode, then it needs to be analyzed.
624       if (E->getNodeId() == DAGTypeLegalizer::NewNode)
625         NodesToAnalyze.insert(E);
626     }
627 
628     void NodeUpdated(SDNode *N) override {
629       // Node updates can mean pretty much anything.  It is possible that an
630       // operand was set to something already processed (f.e.) in which case
631       // this node could become ready.  Recompute its flags.
632       assert(N->getNodeId() != DAGTypeLegalizer::ReadyToProcess &&
633              N->getNodeId() != DAGTypeLegalizer::Processed &&
634              "Invalid node ID for RAUW deletion!");
635       N->setNodeId(DAGTypeLegalizer::NewNode);
636       NodesToAnalyze.insert(N);
637     }
638   };
639 }
640 
641 
642 /// The specified value was legalized to the specified other value.
643 /// Update the DAG and NodeIds replacing any uses of From to use To instead.
644 void DAGTypeLegalizer::ReplaceValueWith(SDValue From, SDValue To) {
645   assert(From.getNode() != To.getNode() && "Potential legalization loop!");
646 
647   // If expansion produced new nodes, make sure they are properly marked.
648   AnalyzeNewValue(To);
649 
650   // Anything that used the old node should now use the new one.  Note that this
651   // can potentially cause recursive merging.
652   SmallSetVector<SDNode*, 16> NodesToAnalyze;
653   NodeUpdateListener NUL(*this, NodesToAnalyze);
654   do {
655 
656     // The old node may be present in a map like ExpandedIntegers or
657     // PromotedIntegers. Inform maps about the replacement.
658     auto FromId = getTableId(From);
659     auto ToId = getTableId(To);
660 
661     if (FromId != ToId)
662       ReplacedValues[FromId] = ToId;
663     DAG.ReplaceAllUsesOfValueWith(From, To);
664 
665     // Process the list of nodes that need to be reanalyzed.
666     while (!NodesToAnalyze.empty()) {
667       SDNode *N = NodesToAnalyze.pop_back_val();
668       if (N->getNodeId() != DAGTypeLegalizer::NewNode)
669         // The node was analyzed while reanalyzing an earlier node - it is safe
670         // to skip.  Note that this is not a morphing node - otherwise it would
671         // still be marked NewNode.
672         continue;
673 
674       // Analyze the node's operands and recalculate the node ID.
675       SDNode *M = AnalyzeNewNode(N);
676       if (M != N) {
677         // The node morphed into a different node.  Make everyone use the new
678         // node instead.
679         assert(M->getNodeId() != NewNode && "Analysis resulted in NewNode!");
680         assert(N->getNumValues() == M->getNumValues() &&
681                "Node morphing changed the number of results!");
682         for (unsigned i = 0, e = N->getNumValues(); i != e; ++i) {
683           SDValue OldVal(N, i);
684           SDValue NewVal(M, i);
685           if (M->getNodeId() == Processed)
686             RemapValue(NewVal);
687           // OldVal may be a target of the ReplacedValues map which was marked
688           // NewNode to force reanalysis because it was updated.  Ensure that
689           // anything that ReplacedValues mapped to OldVal will now be mapped
690           // all the way to NewVal.
691           auto OldValId = getTableId(OldVal);
692           auto NewValId = getTableId(NewVal);
693           DAG.ReplaceAllUsesOfValueWith(OldVal, NewVal);
694           if (OldValId != NewValId)
695             ReplacedValues[OldValId] = NewValId;
696         }
697         // The original node continues to exist in the DAG, marked NewNode.
698       }
699     }
700     // When recursively update nodes with new nodes, it is possible to have
701     // new uses of From due to CSE. If this happens, replace the new uses of
702     // From with To.
703   } while (!From.use_empty());
704 }
705 
706 void DAGTypeLegalizer::SetPromotedInteger(SDValue Op, SDValue Result) {
707   assert(Result.getValueType() ==
708          TLI.getTypeToTransformTo(*DAG.getContext(), Op.getValueType()) &&
709          "Invalid type for promoted integer");
710   AnalyzeNewValue(Result);
711 
712   auto &OpIdEntry = PromotedIntegers[getTableId(Op)];
713   assert((OpIdEntry == 0) && "Node is already promoted!");
714   OpIdEntry = getTableId(Result);
715 
716   DAG.transferDbgValues(Op, Result);
717 }
718 
719 void DAGTypeLegalizer::SetSoftenedFloat(SDValue Op, SDValue Result) {
720 #ifndef NDEBUG
721   EVT VT = Result.getValueType();
722   LLVMContext &Ctx = *DAG.getContext();
723   assert((VT == EVT::getIntegerVT(Ctx, 80) ||
724           VT == TLI.getTypeToTransformTo(Ctx, Op.getValueType())) &&
725          "Invalid type for softened float");
726 #endif
727   AnalyzeNewValue(Result);
728 
729   auto &OpIdEntry = SoftenedFloats[getTableId(Op)];
730   assert((OpIdEntry == 0) && "Node is already converted to integer!");
731   OpIdEntry = getTableId(Result);
732 }
733 
734 void DAGTypeLegalizer::SetPromotedFloat(SDValue Op, SDValue Result) {
735   assert(Result.getValueType() ==
736          TLI.getTypeToTransformTo(*DAG.getContext(), Op.getValueType()) &&
737          "Invalid type for promoted float");
738   AnalyzeNewValue(Result);
739 
740   auto &OpIdEntry = PromotedFloats[getTableId(Op)];
741   assert((OpIdEntry == 0) && "Node is already promoted!");
742   OpIdEntry = getTableId(Result);
743 }
744 
745 void DAGTypeLegalizer::SetSoftPromotedHalf(SDValue Op, SDValue Result) {
746   assert(Result.getValueType() == MVT::i16 &&
747          "Invalid type for soft-promoted half");
748   AnalyzeNewValue(Result);
749 
750   auto &OpIdEntry = SoftPromotedHalfs[getTableId(Op)];
751   assert((OpIdEntry == 0) && "Node is already promoted!");
752   OpIdEntry = getTableId(Result);
753 }
754 
755 void DAGTypeLegalizer::SetScalarizedVector(SDValue Op, SDValue Result) {
756   // Note that in some cases vector operation operands may be greater than
757   // the vector element type. For example BUILD_VECTOR of type <1 x i1> with
758   // a constant i8 operand.
759 
760   // We don't currently support the scalarization of scalable vector types.
761   assert(Result.getValueSizeInBits().getFixedValue() >=
762              Op.getScalarValueSizeInBits() &&
763          "Invalid type for scalarized vector");
764   AnalyzeNewValue(Result);
765 
766   auto &OpIdEntry = ScalarizedVectors[getTableId(Op)];
767   assert((OpIdEntry == 0) && "Node is already scalarized!");
768   OpIdEntry = getTableId(Result);
769 }
770 
771 void DAGTypeLegalizer::GetExpandedInteger(SDValue Op, SDValue &Lo,
772                                           SDValue &Hi) {
773   std::pair<TableId, TableId> &Entry = ExpandedIntegers[getTableId(Op)];
774   assert((Entry.first != 0) && "Operand isn't expanded");
775   Lo = getSDValue(Entry.first);
776   Hi = getSDValue(Entry.second);
777 }
778 
779 void DAGTypeLegalizer::SetExpandedInteger(SDValue Op, SDValue Lo,
780                                           SDValue Hi) {
781   assert(Lo.getValueType() ==
782          TLI.getTypeToTransformTo(*DAG.getContext(), Op.getValueType()) &&
783          Hi.getValueType() == Lo.getValueType() &&
784          "Invalid type for expanded integer");
785   // Lo/Hi may have been newly allocated, if so, add nodeid's as relevant.
786   AnalyzeNewValue(Lo);
787   AnalyzeNewValue(Hi);
788 
789   // Transfer debug values. Don't invalidate the source debug value until it's
790   // been transferred to the high and low bits.
791   if (DAG.getDataLayout().isBigEndian()) {
792     DAG.transferDbgValues(Op, Hi, 0, Hi.getValueSizeInBits(), false);
793     DAG.transferDbgValues(Op, Lo, Hi.getValueSizeInBits(),
794                           Lo.getValueSizeInBits());
795   } else {
796     DAG.transferDbgValues(Op, Lo, 0, Lo.getValueSizeInBits(), false);
797     DAG.transferDbgValues(Op, Hi, Lo.getValueSizeInBits(),
798                           Hi.getValueSizeInBits());
799   }
800 
801   // Remember that this is the result of the node.
802   std::pair<TableId, TableId> &Entry = ExpandedIntegers[getTableId(Op)];
803   assert((Entry.first == 0) && "Node already expanded");
804   Entry.first = getTableId(Lo);
805   Entry.second = getTableId(Hi);
806 }
807 
808 void DAGTypeLegalizer::GetExpandedFloat(SDValue Op, SDValue &Lo,
809                                         SDValue &Hi) {
810   std::pair<TableId, TableId> &Entry = ExpandedFloats[getTableId(Op)];
811   assert((Entry.first != 0) && "Operand isn't expanded");
812   Lo = getSDValue(Entry.first);
813   Hi = getSDValue(Entry.second);
814 }
815 
816 void DAGTypeLegalizer::SetExpandedFloat(SDValue Op, SDValue Lo,
817                                         SDValue Hi) {
818   assert(Lo.getValueType() ==
819          TLI.getTypeToTransformTo(*DAG.getContext(), Op.getValueType()) &&
820          Hi.getValueType() == Lo.getValueType() &&
821          "Invalid type for expanded float");
822   // Lo/Hi may have been newly allocated, if so, add nodeid's as relevant.
823   AnalyzeNewValue(Lo);
824   AnalyzeNewValue(Hi);
825 
826   std::pair<TableId, TableId> &Entry = ExpandedFloats[getTableId(Op)];
827   assert((Entry.first == 0) && "Node already expanded");
828   Entry.first = getTableId(Lo);
829   Entry.second = getTableId(Hi);
830 }
831 
832 void DAGTypeLegalizer::GetSplitVector(SDValue Op, SDValue &Lo,
833                                       SDValue &Hi) {
834   std::pair<TableId, TableId> &Entry = SplitVectors[getTableId(Op)];
835   Lo = getSDValue(Entry.first);
836   Hi = getSDValue(Entry.second);
837   assert(Lo.getNode() && "Operand isn't split");
838   ;
839 }
840 
841 void DAGTypeLegalizer::SetSplitVector(SDValue Op, SDValue Lo,
842                                       SDValue Hi) {
843   assert(Lo.getValueType().getVectorElementType() ==
844              Op.getValueType().getVectorElementType() &&
845          Lo.getValueType().getVectorElementCount() * 2 ==
846              Op.getValueType().getVectorElementCount() &&
847          Hi.getValueType() == Lo.getValueType() &&
848          "Invalid type for split vector");
849   // Lo/Hi may have been newly allocated, if so, add nodeid's as relevant.
850   AnalyzeNewValue(Lo);
851   AnalyzeNewValue(Hi);
852 
853   // Remember that this is the result of the node.
854   std::pair<TableId, TableId> &Entry = SplitVectors[getTableId(Op)];
855   assert((Entry.first == 0) && "Node already split");
856   Entry.first = getTableId(Lo);
857   Entry.second = getTableId(Hi);
858 }
859 
860 void DAGTypeLegalizer::SetWidenedVector(SDValue Op, SDValue Result) {
861   assert(Result.getValueType() ==
862          TLI.getTypeToTransformTo(*DAG.getContext(), Op.getValueType()) &&
863          "Invalid type for widened vector");
864   AnalyzeNewValue(Result);
865 
866   auto &OpIdEntry = WidenedVectors[getTableId(Op)];
867   assert((OpIdEntry == 0) && "Node already widened!");
868   OpIdEntry = getTableId(Result);
869 }
870 
871 
872 //===----------------------------------------------------------------------===//
873 // Utilities.
874 //===----------------------------------------------------------------------===//
875 
876 /// Convert to an integer of the same size.
877 SDValue DAGTypeLegalizer::BitConvertToInteger(SDValue Op) {
878   unsigned BitWidth = Op.getValueSizeInBits();
879   return DAG.getNode(ISD::BITCAST, SDLoc(Op),
880                      EVT::getIntegerVT(*DAG.getContext(), BitWidth), Op);
881 }
882 
883 /// Convert to a vector of integers of the same size.
884 SDValue DAGTypeLegalizer::BitConvertVectorToIntegerVector(SDValue Op) {
885   assert(Op.getValueType().isVector() && "Only applies to vectors!");
886   unsigned EltWidth = Op.getScalarValueSizeInBits();
887   EVT EltNVT = EVT::getIntegerVT(*DAG.getContext(), EltWidth);
888   auto EltCnt = Op.getValueType().getVectorElementCount();
889   return DAG.getNode(ISD::BITCAST, SDLoc(Op),
890                      EVT::getVectorVT(*DAG.getContext(), EltNVT, EltCnt), Op);
891 }
892 
893 SDValue DAGTypeLegalizer::CreateStackStoreLoad(SDValue Op,
894                                                EVT DestVT) {
895   SDLoc dl(Op);
896   // Create the stack frame object.  Make sure it is aligned for both
897   // the source and destination types.
898 
899   // In cases where the vector is illegal it will be broken down into parts
900   // and stored in parts - we should use the alignment for the smallest part.
901   Align DestAlign = DAG.getReducedAlign(DestVT, /*UseABI=*/false);
902   Align OpAlign = DAG.getReducedAlign(Op.getValueType(), /*UseABI=*/false);
903   Align Align = std::max(DestAlign, OpAlign);
904   SDValue StackPtr =
905       DAG.CreateStackTemporary(Op.getValueType().getStoreSize(), Align);
906   // Emit a store to the stack slot.
907   SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Op, StackPtr,
908                                MachinePointerInfo(), Align);
909   // Result is a load from the stack slot.
910   return DAG.getLoad(DestVT, dl, Store, StackPtr, MachinePointerInfo(), Align);
911 }
912 
913 /// Replace the node's results with custom code provided by the target and
914 /// return "true", or do nothing and return "false".
915 /// The last parameter is FALSE if we are dealing with a node with legal
916 /// result types and illegal operand. The second parameter denotes the type of
917 /// illegal OperandNo in that case.
918 /// The last parameter being TRUE means we are dealing with a
919 /// node with illegal result types. The second parameter denotes the type of
920 /// illegal ResNo in that case.
921 bool DAGTypeLegalizer::CustomLowerNode(SDNode *N, EVT VT, bool LegalizeResult) {
922   // See if the target wants to custom lower this node.
923   if (TLI.getOperationAction(N->getOpcode(), VT) != TargetLowering::Custom)
924     return false;
925 
926   SmallVector<SDValue, 8> Results;
927   if (LegalizeResult)
928     TLI.ReplaceNodeResults(N, Results, DAG);
929   else
930     TLI.LowerOperationWrapper(N, Results, DAG);
931 
932   if (Results.empty())
933     // The target didn't want to custom lower it after all.
934     return false;
935 
936   // Make everything that once used N's values now use those in Results instead.
937   assert(Results.size() == N->getNumValues() &&
938          "Custom lowering returned the wrong number of results!");
939   for (unsigned i = 0, e = Results.size(); i != e; ++i) {
940     ReplaceValueWith(SDValue(N, i), Results[i]);
941   }
942   return true;
943 }
944 
945 
946 /// Widen the node's results with custom code provided by the target and return
947 /// "true", or do nothing and return "false".
948 bool DAGTypeLegalizer::CustomWidenLowerNode(SDNode *N, EVT VT) {
949   // See if the target wants to custom lower this node.
950   if (TLI.getOperationAction(N->getOpcode(), VT) != TargetLowering::Custom)
951     return false;
952 
953   SmallVector<SDValue, 8> Results;
954   TLI.ReplaceNodeResults(N, Results, DAG);
955 
956   if (Results.empty())
957     // The target didn't want to custom widen lower its result after all.
958     return false;
959 
960   // Update the widening map.
961   assert(Results.size() == N->getNumValues() &&
962          "Custom lowering returned the wrong number of results!");
963   for (unsigned i = 0, e = Results.size(); i != e; ++i) {
964     // If this is a chain output or already widened just replace it.
965     bool WasWidened = SDValue(N, i).getValueType() != Results[i].getValueType();
966     if (WasWidened)
967       SetWidenedVector(SDValue(N, i), Results[i]);
968     else
969       ReplaceValueWith(SDValue(N, i), Results[i]);
970   }
971   return true;
972 }
973 
974 SDValue DAGTypeLegalizer::DisintegrateMERGE_VALUES(SDNode *N, unsigned ResNo) {
975   for (unsigned i = 0, e = N->getNumValues(); i != e; ++i)
976     if (i != ResNo)
977       ReplaceValueWith(SDValue(N, i), SDValue(N->getOperand(i)));
978   return SDValue(N->getOperand(ResNo));
979 }
980 
981 /// Use ISD::EXTRACT_ELEMENT nodes to extract the low and high parts of the
982 /// given value.
983 void DAGTypeLegalizer::GetPairElements(SDValue Pair,
984                                        SDValue &Lo, SDValue &Hi) {
985   SDLoc dl(Pair);
986   EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), Pair.getValueType());
987   std::tie(Lo, Hi) = DAG.SplitScalar(Pair, dl, NVT, NVT);
988 }
989 
990 /// Build an integer with low bits Lo and high bits Hi.
991 SDValue DAGTypeLegalizer::JoinIntegers(SDValue Lo, SDValue Hi) {
992   // Arbitrarily use dlHi for result SDLoc
993   SDLoc dlHi(Hi);
994   SDLoc dlLo(Lo);
995   EVT LVT = Lo.getValueType();
996   EVT HVT = Hi.getValueType();
997   EVT NVT = EVT::getIntegerVT(*DAG.getContext(),
998                               LVT.getSizeInBits() + HVT.getSizeInBits());
999 
1000   EVT ShiftAmtVT = TLI.getShiftAmountTy(NVT, DAG.getDataLayout());
1001   Lo = DAG.getNode(ISD::ZERO_EXTEND, dlLo, NVT, Lo);
1002   Hi = DAG.getNode(ISD::ANY_EXTEND, dlHi, NVT, Hi);
1003   Hi = DAG.getNode(ISD::SHL, dlHi, NVT, Hi,
1004                    DAG.getConstant(LVT.getSizeInBits(), dlHi, ShiftAmtVT));
1005   return DAG.getNode(ISD::OR, dlHi, NVT, Lo, Hi);
1006 }
1007 
1008 /// Promote the given target boolean to a target boolean of the given type.
1009 /// A target boolean is an integer value, not necessarily of type i1, the bits
1010 /// of which conform to getBooleanContents.
1011 ///
1012 /// ValVT is the type of values that produced the boolean.
1013 SDValue DAGTypeLegalizer::PromoteTargetBoolean(SDValue Bool, EVT ValVT) {
1014   return TLI.promoteTargetBoolean(DAG, Bool, ValVT);
1015 }
1016 
1017 /// Return the lower LoVT bits of Op in Lo and the upper HiVT bits in Hi.
1018 void DAGTypeLegalizer::SplitInteger(SDValue Op,
1019                                     EVT LoVT, EVT HiVT,
1020                                     SDValue &Lo, SDValue &Hi) {
1021   SDLoc dl(Op);
1022   assert(LoVT.getSizeInBits() + HiVT.getSizeInBits() ==
1023          Op.getValueSizeInBits() && "Invalid integer splitting!");
1024   Lo = DAG.getNode(ISD::TRUNCATE, dl, LoVT, Op);
1025   unsigned ReqShiftAmountInBits =
1026       Log2_32_Ceil(Op.getValueType().getSizeInBits());
1027   MVT ShiftAmountTy =
1028       TLI.getScalarShiftAmountTy(DAG.getDataLayout(), Op.getValueType());
1029   if (ReqShiftAmountInBits > ShiftAmountTy.getSizeInBits())
1030     ShiftAmountTy = MVT::getIntegerVT(NextPowerOf2(ReqShiftAmountInBits));
1031   Hi = DAG.getNode(ISD::SRL, dl, Op.getValueType(), Op,
1032                    DAG.getConstant(LoVT.getSizeInBits(), dl, ShiftAmountTy));
1033   Hi = DAG.getNode(ISD::TRUNCATE, dl, HiVT, Hi);
1034 }
1035 
1036 /// Return the lower and upper halves of Op's bits in a value type half the
1037 /// size of Op's.
1038 void DAGTypeLegalizer::SplitInteger(SDValue Op,
1039                                     SDValue &Lo, SDValue &Hi) {
1040   EVT HalfVT =
1041       EVT::getIntegerVT(*DAG.getContext(), Op.getValueSizeInBits() / 2);
1042   SplitInteger(Op, HalfVT, HalfVT, Lo, Hi);
1043 }
1044 
1045 
1046 //===----------------------------------------------------------------------===//
1047 //  Entry Point
1048 //===----------------------------------------------------------------------===//
1049 
1050 /// This transforms the SelectionDAG into a SelectionDAG that only uses types
1051 /// natively supported by the target. Returns "true" if it made any changes.
1052 ///
1053 /// Note that this is an involved process that may invalidate pointers into
1054 /// the graph.
1055 bool SelectionDAG::LegalizeTypes() {
1056   return DAGTypeLegalizer(*this).run();
1057 }
1058