clang 20.0.0git
ThreadSafetyTIL.cpp
Go to the documentation of this file.
1//===- ThreadSafetyTIL.cpp ------------------------------------------------===//
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
10#include "clang/Basic/LLVM.h"
11#include "llvm/Support/Casting.h"
12#include <cassert>
13#include <cstddef>
14
15using namespace clang;
16using namespace threadSafety;
17using namespace til;
18
20 switch (Op) {
21 case UOP_Minus: return "-";
22 case UOP_BitNot: return "~";
23 case UOP_LogicNot: return "!";
24 }
25 return {};
26}
27
29 switch (Op) {
30 case BOP_Mul: return "*";
31 case BOP_Div: return "/";
32 case BOP_Rem: return "%";
33 case BOP_Add: return "+";
34 case BOP_Sub: return "-";
35 case BOP_Shl: return "<<";
36 case BOP_Shr: return ">>";
37 case BOP_BitAnd: return "&";
38 case BOP_BitXor: return "^";
39 case BOP_BitOr: return "|";
40 case BOP_Eq: return "==";
41 case BOP_Neq: return "!=";
42 case BOP_Lt: return "<";
43 case BOP_Leq: return "<=";
44 case BOP_Cmp: return "<=>";
45 case BOP_LogicAnd: return "&&";
46 case BOP_LogicOr: return "||";
47 }
48 return {};
49}
50
51SExpr* Future::force() {
52 Status = FS_evaluating;
53 Result = compute();
54 Status = FS_done;
55 return Result;
56}
57
59 unsigned Idx = Predecessors.size();
60 Predecessors.reserveCheck(1, Arena);
61 Predecessors.push_back(Pred);
62 for (auto *E : Args) {
63 if (auto *Ph = dyn_cast<Phi>(E)) {
64 Ph->values().reserveCheck(1, Arena);
65 Ph->values().push_back(nullptr);
66 }
67 }
68 return Idx;
69}
70
71void BasicBlock::reservePredecessors(unsigned NumPreds) {
72 Predecessors.reserve(NumPreds, Arena);
73 for (auto *E : Args) {
74 if (auto *Ph = dyn_cast<Phi>(E)) {
75 Ph->values().reserve(NumPreds, Arena);
76 }
77 }
78}
79
80// If E is a variable, then trace back through any aliases or redundant
81// Phi nodes to find the canonical definition.
83 while (true) {
84 if (const auto *V = dyn_cast<Variable>(E)) {
85 if (V->kind() == Variable::VK_Let) {
86 E = V->definition();
87 continue;
88 }
89 }
90 if (const auto *Ph = dyn_cast<Phi>(E)) {
91 if (Ph->status() == Phi::PH_SingleVal) {
92 E = Ph->values()[0];
93 continue;
94 }
95 }
96 break;
97 }
98 return E;
99}
100
101// If E is a variable, then trace back through any aliases or redundant
102// Phi nodes to find the canonical definition.
103// The non-const version will simplify incomplete Phi nodes.
105 while (true) {
106 if (auto *V = dyn_cast<Variable>(E)) {
107 if (V->kind() != Variable::VK_Let)
108 return V;
109 // Eliminate redundant variables, e.g. x = y, or x = 5,
110 // but keep anything more complicated.
111 if (til::ThreadSafetyTIL::isTrivial(V->definition())) {
112 E = V->definition();
113 continue;
114 }
115 return V;
116 }
117 if (auto *Ph = dyn_cast<Phi>(E)) {
118 if (Ph->status() == Phi::PH_Incomplete)
120 // Eliminate redundant Phi nodes.
121 if (Ph->status() == Phi::PH_SingleVal) {
122 E = Ph->values()[0];
123 continue;
124 }
125 }
126 return E;
127 }
128}
129
130// Trace the arguments of an incomplete Phi node to see if they have the same
131// canonical definition. If so, mark the Phi node as redundant.
132// getCanonicalVal() will recursively call simplifyIncompletePhi().
134 assert(Ph && Ph->status() == Phi::PH_Incomplete);
135
136 // eliminate infinite recursion -- assume that this node is not redundant.
138
139 SExpr *E0 = simplifyToCanonicalVal(Ph->values()[0]);
140 for (unsigned i = 1, n = Ph->values().size(); i < n; ++i) {
141 SExpr *Ei = simplifyToCanonicalVal(Ph->values()[i]);
142 if (Ei == Ph)
143 continue; // Recursive reference to itself. Don't count.
144 if (Ei != E0) {
145 return; // Status is already set to MultiVal.
146 }
147 }
149}
150
151// Renumbers the arguments and instructions to have unique, sequential IDs.
152unsigned BasicBlock::renumberInstrs(unsigned ID) {
153 for (auto *Arg : Args)
154 Arg->setID(this, ID++);
155 for (auto *Instr : Instrs)
156 Instr->setID(this, ID++);
157 TermInstr->setID(this, ID++);
158 return ID;
159}
160
161// Sorts the CFGs blocks using a reverse post-order depth-first traversal.
162// Each block will be written into the Blocks array in order, and its BlockID
163// will be set to the index in the array. Sorting should start from the entry
164// block, and ID should be the total number of blocks.
165unsigned BasicBlock::topologicalSort(SimpleArray<BasicBlock *> &Blocks,
166 unsigned ID) {
167 if (Visited) return ID;
168 Visited = true;
169 for (auto *Block : successors())
170 ID = Block->topologicalSort(Blocks, ID);
171 // set ID and update block array in place.
172 // We may lose pointers to unreachable blocks.
173 assert(ID > 0);
174 BlockID = --ID;
175 Blocks[BlockID] = this;
176 return ID;
177}
178
179// Performs a reverse topological traversal, starting from the exit block and
180// following back-edges. The dominator is serialized before any predecessors,
181// which guarantees that all blocks are serialized after their dominator and
182// before their post-dominator (because it's a reverse topological traversal).
183// ID should be initially set to 0.
184//
185// This sort assumes that (1) dominators have been computed, (2) there are no
186// critical edges, and (3) the entry block is reachable from the exit block
187// and no blocks are accessible via traversal of back-edges from the exit that
188// weren't accessible via forward edges from the entry.
189unsigned BasicBlock::topologicalFinalSort(SimpleArray<BasicBlock *> &Blocks,
190 unsigned ID) {
191 // Visited is assumed to have been set by the topologicalSort. This pass
192 // assumes !Visited means that we've visited this node before.
193 if (!Visited) return ID;
194 Visited = false;
195 if (DominatorNode.Parent)
196 ID = DominatorNode.Parent->topologicalFinalSort(Blocks, ID);
197 for (auto *Pred : Predecessors)
198 ID = Pred->topologicalFinalSort(Blocks, ID);
199 assert(static_cast<size_t>(ID) < Blocks.size());
200 BlockID = ID++;
201 Blocks[BlockID] = this;
202 return ID;
203}
204
205// Computes the immediate dominator of the current block. Assumes that all of
206// its predecessors have already computed their dominators. This is achieved
207// by visiting the nodes in topological order.
208void BasicBlock::computeDominator() {
209 BasicBlock *Candidate = nullptr;
210 // Walk backwards from each predecessor to find the common dominator node.
211 for (auto *Pred : Predecessors) {
212 // Skip back-edges
213 if (Pred->BlockID >= BlockID) continue;
214 // If we don't yet have a candidate for dominator yet, take this one.
215 if (Candidate == nullptr) {
216 Candidate = Pred;
217 continue;
218 }
219 // Walk the alternate and current candidate back to find a common ancestor.
220 auto *Alternate = Pred;
221 while (Alternate != Candidate) {
222 if (Candidate->BlockID > Alternate->BlockID)
223 Candidate = Candidate->DominatorNode.Parent;
224 else
225 Alternate = Alternate->DominatorNode.Parent;
226 }
227 }
228 DominatorNode.Parent = Candidate;
229 DominatorNode.SizeOfSubTree = 1;
230}
231
232// Computes the immediate post-dominator of the current block. Assumes that all
233// of its successors have already computed their post-dominators. This is
234// achieved visiting the nodes in reverse topological order.
235void BasicBlock::computePostDominator() {
236 BasicBlock *Candidate = nullptr;
237 // Walk back from each predecessor to find the common post-dominator node.
238 for (auto *Succ : successors()) {
239 // Skip back-edges
240 if (Succ->BlockID <= BlockID) continue;
241 // If we don't yet have a candidate for post-dominator yet, take this one.
242 if (Candidate == nullptr) {
243 Candidate = Succ;
244 continue;
245 }
246 // Walk the alternate and current candidate back to find a common ancestor.
247 auto *Alternate = Succ;
248 while (Alternate != Candidate) {
249 if (Candidate->BlockID < Alternate->BlockID)
250 Candidate = Candidate->PostDominatorNode.Parent;
251 else
252 Alternate = Alternate->PostDominatorNode.Parent;
253 }
254 }
255 PostDominatorNode.Parent = Candidate;
256 PostDominatorNode.SizeOfSubTree = 1;
257}
258
259// Renumber instructions in all blocks
260void SCFG::renumberInstrs() {
261 unsigned InstrID = 0;
262 for (auto *Block : Blocks)
263 InstrID = Block->renumberInstrs(InstrID);
264}
265
266static inline void computeNodeSize(BasicBlock *B,
268 BasicBlock::TopologyNode *N = &(B->*TN);
269 if (N->Parent) {
270 BasicBlock::TopologyNode *P = &(N->Parent->*TN);
271 // Initially set ID relative to the (as yet uncomputed) parent ID
272 N->NodeID = P->SizeOfSubTree;
273 P->SizeOfSubTree += N->SizeOfSubTree;
274 }
275}
276
277static inline void computeNodeID(BasicBlock *B,
279 BasicBlock::TopologyNode *N = &(B->*TN);
280 if (N->Parent) {
281 BasicBlock::TopologyNode *P = &(N->Parent->*TN);
282 N->NodeID += P->NodeID; // Fix NodeIDs relative to starting node.
283 }
284}
285
286// Normalizes a CFG. Normalization has a few major components:
287// 1) Removing unreachable blocks.
288// 2) Computing dominators and post-dominators
289// 3) Topologically sorting the blocks into the "Blocks" array.
291 // Topologically sort the blocks starting from the entry block.
292 unsigned NumUnreachableBlocks = Entry->topologicalSort(Blocks, Blocks.size());
293 if (NumUnreachableBlocks > 0) {
294 // If there were unreachable blocks shift everything down, and delete them.
295 for (unsigned I = NumUnreachableBlocks, E = Blocks.size(); I < E; ++I) {
296 unsigned NI = I - NumUnreachableBlocks;
297 Blocks[NI] = Blocks[I];
298 Blocks[NI]->BlockID = NI;
299 // FIXME: clean up predecessor pointers to unreachable blocks?
300 }
301 Blocks.drop(NumUnreachableBlocks);
302 }
303
304 // Compute dominators.
305 for (auto *Block : Blocks)
306 Block->computeDominator();
307
308 // Once dominators have been computed, the final sort may be performed.
309 unsigned NumBlocks = Exit->topologicalFinalSort(Blocks, 0);
310 assert(static_cast<size_t>(NumBlocks) == Blocks.size());
311 (void) NumBlocks;
312
313 // Renumber the instructions now that we have a final sort.
314 renumberInstrs();
315
316 // Compute post-dominators and compute the sizes of each node in the
317 // dominator tree.
318 for (auto *Block : Blocks.reverse()) {
319 Block->computePostDominator();
320 computeNodeSize(Block, &BasicBlock::DominatorNode);
321 }
322 // Compute the sizes of each node in the post-dominator tree and assign IDs in
323 // the dominator tree.
324 for (auto *Block : Blocks) {
325 computeNodeID(Block, &BasicBlock::DominatorNode);
326 computeNodeSize(Block, &BasicBlock::PostDominatorNode);
327 }
328 // Assign IDs in the post-dominator tree.
329 for (auto *Block : Blocks.reverse()) {
330 computeNodeID(Block, &BasicBlock::PostDominatorNode);
331 }
332}
#define V(N, I)
Definition: ASTContext.h:3341
StringRef P
Expr * E
Forward-declares and imports various common LLVM datatypes that clang wants to use unqualified.
static void computeNodeSize(BasicBlock *B, BasicBlock::TopologyNode BasicBlock::*TN)
static void computeNodeID(BasicBlock *B, BasicBlock::TopologyNode BasicBlock::*TN)
A basic block is part of an SCFG.
unsigned addPredecessor(BasicBlock *Pred)
ArrayRef< BasicBlock * > successors()
void reservePredecessors(unsigned NumPreds)
Phi Node, for code in SSA form.
const ValArray & values() const
Base class for AST nodes in the typed intermediate language.
void setID(BasicBlock *B, unsigned id)
Set the basic block and instruction ID for this expression.
void reserve(size_t Ncp, MemRegionRef A)
void reserveCheck(size_t N, MemRegionRef A)
TIL_UnaryOpcode
Opcode for unary arithmetic operations.
void simplifyIncompleteArg(til::Phi *Ph)
StringRef getBinaryOpcodeString(TIL_BinaryOpcode Op)
Return the name of a binary opcode.
StringRef getUnaryOpcodeString(TIL_UnaryOpcode Op)
Return the name of a unary opcode.
TIL_BinaryOpcode
Opcode for binary arithmetic operations.
SExpr * simplifyToCanonicalVal(SExpr *E)
const SExpr * getCanonicalVal(const SExpr *E)
The JSON file list parser is used to communicate input to InstallAPI.