1 //===-- SpillPlacement.cpp - Optimal Spill Code Placement -----------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This file implements the spill code placement analysis. 11 // 12 // Each edge bundle corresponds to a node in a Hopfield network. Constraints on 13 // basic blocks are weighted by the block frequency and added to become the node 14 // bias. 15 // 16 // Transparent basic blocks have the variable live through, but don't care if it 17 // is spilled or in a register. These blocks become connections in the Hopfield 18 // network, again weighted by block frequency. 19 // 20 // The Hopfield network minimizes (possibly locally) its energy function: 21 // 22 // E = -sum_n V_n * ( B_n + sum_{n, m linked by b} V_m * F_b ) 23 // 24 // The energy function represents the expected spill code execution frequency, 25 // or the cost of spilling. This is a Lyapunov function which never increases 26 // when a node is updated. It is guaranteed to converge to a local minimum. 27 // 28 //===----------------------------------------------------------------------===// 29 30 #define DEBUG_TYPE "spillplacement" 31 #include "SpillPlacement.h" 32 #include "llvm/CodeGen/EdgeBundles.h" 33 #include "llvm/CodeGen/LiveIntervalAnalysis.h" 34 #include "llvm/CodeGen/MachineBasicBlock.h" 35 #include "llvm/CodeGen/MachineFunction.h" 36 #include "llvm/CodeGen/MachineLoopInfo.h" 37 #include "llvm/CodeGen/Passes.h" 38 #include "llvm/Support/Debug.h" 39 #include "llvm/Support/Format.h" 40 41 using namespace llvm; 42 43 char SpillPlacement::ID = 0; 44 INITIALIZE_PASS_BEGIN(SpillPlacement, "spill-code-placement", 45 "Spill Code Placement Analysis", true, true) 46 INITIALIZE_PASS_DEPENDENCY(EdgeBundles) 47 INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo) 48 INITIALIZE_PASS_END(SpillPlacement, "spill-code-placement", 49 "Spill Code Placement Analysis", true, true) 50 51 char &llvm::SpillPlacementID = SpillPlacement::ID; 52 53 void SpillPlacement::getAnalysisUsage(AnalysisUsage &AU) const { 54 AU.setPreservesAll(); 55 AU.addRequiredTransitive<EdgeBundles>(); 56 AU.addRequiredTransitive<MachineLoopInfo>(); 57 MachineFunctionPass::getAnalysisUsage(AU); 58 } 59 60 /// Node - Each edge bundle corresponds to a Hopfield node. 61 /// 62 /// The node contains precomputed frequency data that only depends on the CFG, 63 /// but Bias and Links are computed each time placeSpills is called. 64 /// 65 /// The node Value is positive when the variable should be in a register. The 66 /// value can change when linked nodes change, but convergence is very fast 67 /// because all weights are positive. 68 /// 69 struct SpillPlacement::Node { 70 /// Scale - Inverse block frequency feeding into[0] or out of[1] the bundle. 71 /// Ideally, these two numbers should be identical, but inaccuracies in the 72 /// block frequency estimates means that we need to normalize ingoing and 73 /// outgoing frequencies separately so they are commensurate. 74 float Scale[2]; 75 76 /// Bias - Normalized contributions from non-transparent blocks. 77 /// A bundle connected to a MustSpill block has a huge negative bias, 78 /// otherwise it is a number in the range [-2;2]. 79 float Bias; 80 81 /// Value - Output value of this node computed from the Bias and links. 82 /// This is always in the range [-1;1]. A positive number means the variable 83 /// should go in a register through this bundle. 84 float Value; 85 86 typedef SmallVector<std::pair<float, unsigned>, 4> LinkVector; 87 88 /// Links - (Weight, BundleNo) for all transparent blocks connecting to other 89 /// bundles. The weights are all positive and add up to at most 2, weights 90 /// from ingoing and outgoing nodes separately add up to a most 1. The weight 91 /// sum can be less than 2 when the variable is not live into / out of some 92 /// connected basic blocks. 93 LinkVector Links; 94 95 /// preferReg - Return true when this node prefers to be in a register. 96 bool preferReg() const { 97 // Undecided nodes (Value==0) go on the stack. 98 return Value > 0; 99 } 100 101 /// mustSpill - Return True if this node is so biased that it must spill. 102 bool mustSpill() const { 103 // Actually, we must spill if Bias < sum(weights). 104 // It may be worth it to compute the weight sum here? 105 return Bias < -2.0f; 106 } 107 108 /// Node - Create a blank Node. 109 Node() { 110 Scale[0] = Scale[1] = 0; 111 } 112 113 /// clear - Reset per-query data, but preserve frequencies that only depend on 114 // the CFG. 115 void clear() { 116 Bias = Value = 0; 117 Links.clear(); 118 } 119 120 /// addLink - Add a link to bundle b with weight w. 121 /// out=0 for an ingoing link, and 1 for an outgoing link. 122 void addLink(unsigned b, float w, bool out) { 123 // Normalize w relative to all connected blocks from that direction. 124 w *= Scale[out]; 125 126 // There can be multiple links to the same bundle, add them up. 127 for (LinkVector::iterator I = Links.begin(), E = Links.end(); I != E; ++I) 128 if (I->second == b) { 129 I->first += w; 130 return; 131 } 132 // This must be the first link to b. 133 Links.push_back(std::make_pair(w, b)); 134 } 135 136 /// addBias - Bias this node from an ingoing[0] or outgoing[1] link. 137 /// Return the change to the total number of positive biases. 138 void addBias(float w, bool out) { 139 // Normalize w relative to all connected blocks from that direction. 140 w *= Scale[out]; 141 Bias += w; 142 } 143 144 /// update - Recompute Value from Bias and Links. Return true when node 145 /// preference changes. 146 bool update(const Node nodes[]) { 147 // Compute the weighted sum of inputs. 148 float Sum = Bias; 149 for (LinkVector::iterator I = Links.begin(), E = Links.end(); I != E; ++I) 150 Sum += I->first * nodes[I->second].Value; 151 152 // The weighted sum is going to be in the range [-2;2]. Ideally, we should 153 // simply set Value = sign(Sum), but we will add a dead zone around 0 for 154 // two reasons: 155 // 1. It avoids arbitrary bias when all links are 0 as is possible during 156 // initial iterations. 157 // 2. It helps tame rounding errors when the links nominally sum to 0. 158 const float Thres = 1e-4f; 159 bool Before = preferReg(); 160 if (Sum < -Thres) 161 Value = -1; 162 else if (Sum > Thres) 163 Value = 1; 164 else 165 Value = 0; 166 return Before != preferReg(); 167 } 168 }; 169 170 bool SpillPlacement::runOnMachineFunction(MachineFunction &mf) { 171 MF = &mf; 172 bundles = &getAnalysis<EdgeBundles>(); 173 loops = &getAnalysis<MachineLoopInfo>(); 174 175 assert(!nodes && "Leaking node array"); 176 nodes = new Node[bundles->getNumBundles()]; 177 178 // Compute total ingoing and outgoing block frequencies for all bundles. 179 BlockFrequency.resize(mf.getNumBlockIDs()); 180 for (MachineFunction::iterator I = mf.begin(), E = mf.end(); I != E; ++I) { 181 float Freq = LiveIntervals::getSpillWeight(true, false, 182 loops->getLoopDepth(I)); 183 unsigned Num = I->getNumber(); 184 BlockFrequency[Num] = Freq; 185 nodes[bundles->getBundle(Num, 1)].Scale[0] += Freq; 186 nodes[bundles->getBundle(Num, 0)].Scale[1] += Freq; 187 } 188 189 // Scales are reciprocal frequencies. 190 for (unsigned i = 0, e = bundles->getNumBundles(); i != e; ++i) 191 for (unsigned d = 0; d != 2; ++d) 192 if (nodes[i].Scale[d] > 0) 193 nodes[i].Scale[d] = 1 / nodes[i].Scale[d]; 194 195 // We never change the function. 196 return false; 197 } 198 199 void SpillPlacement::releaseMemory() { 200 delete[] nodes; 201 nodes = 0; 202 } 203 204 /// activate - mark node n as active if it wasn't already. 205 void SpillPlacement::activate(unsigned n) { 206 if (ActiveNodes->test(n)) 207 return; 208 ActiveNodes->set(n); 209 nodes[n].clear(); 210 } 211 212 213 /// addConstraints - Compute node biases and weights from a set of constraints. 214 /// Set a bit in NodeMask for each active node. 215 void SpillPlacement::addConstraints(ArrayRef<BlockConstraint> LiveBlocks) { 216 for (ArrayRef<BlockConstraint>::iterator I = LiveBlocks.begin(), 217 E = LiveBlocks.end(); I != E; ++I) { 218 float Freq = getBlockFrequency(I->Number); 219 const float Bias[] = { 220 0, // DontCare, 221 1, // PrefReg, 222 -1, // PrefSpill 223 -HUGE_VALF // MustSpill 224 }; 225 226 // Live-in to block? 227 if (I->Entry != DontCare) { 228 unsigned ib = bundles->getBundle(I->Number, 0); 229 activate(ib); 230 nodes[ib].addBias(Freq * Bias[I->Entry], 1); 231 } 232 233 // Live-out from block? 234 if (I->Exit != DontCare) { 235 unsigned ob = bundles->getBundle(I->Number, 1); 236 activate(ob); 237 nodes[ob].addBias(Freq * Bias[I->Exit], 0); 238 } 239 } 240 } 241 242 void SpillPlacement::addLinks(ArrayRef<unsigned> Links) { 243 for (ArrayRef<unsigned>::iterator I = Links.begin(), E = Links.end(); I != E; 244 ++I) { 245 unsigned Number = *I; 246 unsigned ib = bundles->getBundle(Number, 0); 247 unsigned ob = bundles->getBundle(Number, 1); 248 249 // Ignore self-loops. 250 if (ib == ob) 251 continue; 252 activate(ib); 253 activate(ob); 254 if (nodes[ib].Links.empty() && !nodes[ib].mustSpill()) 255 Linked.push_back(ib); 256 if (nodes[ob].Links.empty() && !nodes[ob].mustSpill()) 257 Linked.push_back(ob); 258 float Freq = getBlockFrequency(Number); 259 nodes[ib].addLink(ob, Freq, 1); 260 nodes[ob].addLink(ib, Freq, 0); 261 } 262 } 263 264 bool SpillPlacement::scanActiveBundles() { 265 Linked.clear(); 266 RecentPositive.clear(); 267 for (int n = ActiveNodes->find_first(); n>=0; n = ActiveNodes->find_next(n)) { 268 nodes[n].update(nodes); 269 // A node that must spill, or a node without any links is not going to 270 // change its value ever again, so exclude it from iterations. 271 if (nodes[n].mustSpill()) 272 continue; 273 if (!nodes[n].Links.empty()) 274 Linked.push_back(n); 275 if (nodes[n].preferReg()) 276 RecentPositive.push_back(n); 277 } 278 return !RecentPositive.empty(); 279 } 280 281 /// iterate - Repeatedly update the Hopfield nodes until stability or the 282 /// maximum number of iterations is reached. 283 /// @param Linked - Numbers of linked nodes that need updating. 284 void SpillPlacement::iterate() { 285 // First update the recently positive nodes. They have likely received new 286 // negative bias that will turn them off. 287 while (!RecentPositive.empty()) 288 nodes[RecentPositive.pop_back_val()].update(nodes); 289 290 if (Linked.empty()) 291 return; 292 293 // Run up to 10 iterations. The edge bundle numbering is closely related to 294 // basic block numbering, so there is a strong tendency towards chains of 295 // linked nodes with sequential numbers. By scanning the linked nodes 296 // backwards and forwards, we make it very likely that a single node can 297 // affect the entire network in a single iteration. That means very fast 298 // convergence, usually in a single iteration. 299 for (unsigned iteration = 0; iteration != 10; ++iteration) { 300 // Scan backwards, skipping the last node which was just updated. 301 bool Changed = false; 302 for (SmallVectorImpl<unsigned>::const_reverse_iterator I = 303 llvm::next(Linked.rbegin()), E = Linked.rend(); I != E; ++I) { 304 unsigned n = *I; 305 if (nodes[n].update(nodes)) { 306 Changed = true; 307 if (nodes[n].preferReg()) 308 RecentPositive.push_back(n); 309 } 310 } 311 if (!Changed || !RecentPositive.empty()) 312 return; 313 314 // Scan forwards, skipping the first node which was just updated. 315 Changed = false; 316 for (SmallVectorImpl<unsigned>::const_iterator I = 317 llvm::next(Linked.begin()), E = Linked.end(); I != E; ++I) { 318 unsigned n = *I; 319 if (nodes[n].update(nodes)) { 320 Changed = true; 321 if (nodes[n].preferReg()) 322 RecentPositive.push_back(n); 323 } 324 } 325 if (!Changed || !RecentPositive.empty()) 326 return; 327 } 328 } 329 330 void SpillPlacement::prepare(BitVector &RegBundles) { 331 Linked.clear(); 332 RecentPositive.clear(); 333 // Reuse RegBundles as our ActiveNodes vector. 334 ActiveNodes = &RegBundles; 335 ActiveNodes->clear(); 336 ActiveNodes->resize(bundles->getNumBundles()); 337 } 338 339 bool 340 SpillPlacement::finish() { 341 assert(ActiveNodes && "Call prepare() first"); 342 343 // Write preferences back to ActiveNodes. 344 bool Perfect = true; 345 for (int n = ActiveNodes->find_first(); n>=0; n = ActiveNodes->find_next(n)) 346 if (!nodes[n].preferReg()) { 347 ActiveNodes->reset(n); 348 Perfect = false; 349 } 350 ActiveNodes = 0; 351 return Perfect; 352 } 353