github.com/bir3/gocompiler@v0.9.2202/src/cmd/compile/internal/ssa/dom.go (about) 1 // Copyright 2015 The Go Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 package ssa 6 7 // This file contains code to compute the dominator tree 8 // of a control-flow graph. 9 10 // postorder computes a postorder traversal ordering for the 11 // basic blocks in f. Unreachable blocks will not appear. 12 func postorder(f *Func) []*Block { 13 return postorderWithNumbering(f, nil) 14 } 15 16 type blockAndIndex struct { 17 b *Block 18 index int // index is the number of successor edges of b that have already been explored. 19 } 20 21 // postorderWithNumbering provides a DFS postordering. 22 // This seems to make loop-finding more robust. 23 func postorderWithNumbering(f *Func, ponums []int32) []*Block { 24 seen := f.Cache.allocBoolSlice(f.NumBlocks()) 25 defer f.Cache.freeBoolSlice(seen) 26 27 // result ordering 28 order := make([]*Block, 0, len(f.Blocks)) 29 30 // stack of blocks and next child to visit 31 // A constant bound allows this to be stack-allocated. 32 is 32 // enough to cover almost every postorderWithNumbering call. 33 s := make([]blockAndIndex, 0, 32) 34 s = append(s, blockAndIndex{b: f.Entry}) 35 seen[f.Entry.ID] = true 36 for len(s) > 0 { 37 tos := len(s) - 1 38 x := s[tos] 39 b := x.b 40 if i := x.index; i < len(b.Succs) { 41 s[tos].index++ 42 bb := b.Succs[i].Block() 43 if !seen[bb.ID] { 44 seen[bb.ID] = true 45 s = append(s, blockAndIndex{b: bb}) 46 } 47 continue 48 } 49 s = s[:tos] 50 if ponums != nil { 51 ponums[b.ID] = int32(len(order)) 52 } 53 order = append(order, b) 54 } 55 return order 56 } 57 58 type linkedBlocks func(*Block) []Edge 59 60 func dominators(f *Func) []*Block { 61 preds := func(b *Block) []Edge { return b.Preds } 62 succs := func(b *Block) []Edge { return b.Succs } 63 64 //TODO: benchmark and try to find criteria for swapping between 65 // dominatorsSimple and dominatorsLT 66 return f.dominatorsLTOrig(f.Entry, preds, succs) 67 } 68 69 // dominatorsLTOrig runs Lengauer-Tarjan to compute a dominator tree starting at 70 // entry and using predFn/succFn to find predecessors/successors to allow 71 // computing both dominator and post-dominator trees. 72 func (f *Func) dominatorsLTOrig(entry *Block, predFn linkedBlocks, succFn linkedBlocks) []*Block { 73 // Adapted directly from the original TOPLAS article's "simple" algorithm 74 75 maxBlockID := entry.Func.NumBlocks() 76 scratch := f.Cache.allocIDSlice(7 * maxBlockID) 77 defer f.Cache.freeIDSlice(scratch) 78 semi := scratch[0*maxBlockID : 1*maxBlockID] 79 vertex := scratch[1*maxBlockID : 2*maxBlockID] 80 label := scratch[2*maxBlockID : 3*maxBlockID] 81 parent := scratch[3*maxBlockID : 4*maxBlockID] 82 ancestor := scratch[4*maxBlockID : 5*maxBlockID] 83 bucketHead := scratch[5*maxBlockID : 6*maxBlockID] 84 bucketLink := scratch[6*maxBlockID : 7*maxBlockID] 85 86 // This version uses integers for most of the computation, 87 // to make the work arrays smaller and pointer-free. 88 // fromID translates from ID to *Block where that is needed. 89 fromID := f.Cache.allocBlockSlice(maxBlockID) 90 defer f.Cache.freeBlockSlice(fromID) 91 for _, v := range f.Blocks { 92 fromID[v.ID] = v 93 } 94 idom := make([]*Block, maxBlockID) 95 96 // Step 1. Carry out a depth first search of the problem graph. Number 97 // the vertices from 1 to n as they are reached during the search. 98 n := f.dfsOrig(entry, succFn, semi, vertex, label, parent) 99 100 for i := n; i >= 2; i-- { 101 w := vertex[i] 102 103 // step2 in TOPLAS paper 104 for _, e := range predFn(fromID[w]) { 105 v := e.b 106 if semi[v.ID] == 0 { 107 // skip unreachable predecessor 108 // not in original, but we're using existing pred instead of building one. 109 continue 110 } 111 u := evalOrig(v.ID, ancestor, semi, label) 112 if semi[u] < semi[w] { 113 semi[w] = semi[u] 114 } 115 } 116 117 // add w to bucket[vertex[semi[w]]] 118 // implement bucket as a linked list implemented 119 // in a pair of arrays. 120 vsw := vertex[semi[w]] 121 bucketLink[w] = bucketHead[vsw] 122 bucketHead[vsw] = w 123 124 linkOrig(parent[w], w, ancestor) 125 126 // step3 in TOPLAS paper 127 for v := bucketHead[parent[w]]; v != 0; v = bucketLink[v] { 128 u := evalOrig(v, ancestor, semi, label) 129 if semi[u] < semi[v] { 130 idom[v] = fromID[u] 131 } else { 132 idom[v] = fromID[parent[w]] 133 } 134 } 135 } 136 // step 4 in toplas paper 137 for i := ID(2); i <= n; i++ { 138 w := vertex[i] 139 if idom[w].ID != vertex[semi[w]] { 140 idom[w] = idom[idom[w].ID] 141 } 142 } 143 144 return idom 145 } 146 147 // dfsOrig performs a depth first search over the blocks starting at entry block 148 // (in arbitrary order). This is a de-recursed version of dfs from the 149 // original Tarjan-Lengauer TOPLAS article. It's important to return the 150 // same values for parent as the original algorithm. 151 func (f *Func) dfsOrig(entry *Block, succFn linkedBlocks, semi, vertex, label, parent []ID) ID { 152 n := ID(0) 153 s := make([]*Block, 0, 256) 154 s = append(s, entry) 155 156 for len(s) > 0 { 157 v := s[len(s)-1] 158 s = s[:len(s)-1] 159 // recursing on v 160 161 if semi[v.ID] != 0 { 162 continue // already visited 163 } 164 n++ 165 semi[v.ID] = n 166 vertex[n] = v.ID 167 label[v.ID] = v.ID 168 // ancestor[v] already zero 169 for _, e := range succFn(v) { 170 w := e.b 171 // if it has a dfnum, we've already visited it 172 if semi[w.ID] == 0 { 173 // yes, w can be pushed multiple times. 174 s = append(s, w) 175 parent[w.ID] = v.ID // keep overwriting this till it is visited. 176 } 177 } 178 } 179 return n 180 } 181 182 // compressOrig is the "simple" compress function from LT paper. 183 func compressOrig(v ID, ancestor, semi, label []ID) { 184 if ancestor[ancestor[v]] != 0 { 185 compressOrig(ancestor[v], ancestor, semi, label) 186 if semi[label[ancestor[v]]] < semi[label[v]] { 187 label[v] = label[ancestor[v]] 188 } 189 ancestor[v] = ancestor[ancestor[v]] 190 } 191 } 192 193 // evalOrig is the "simple" eval function from LT paper. 194 func evalOrig(v ID, ancestor, semi, label []ID) ID { 195 if ancestor[v] == 0 { 196 return v 197 } 198 compressOrig(v, ancestor, semi, label) 199 return label[v] 200 } 201 202 func linkOrig(v, w ID, ancestor []ID) { 203 ancestor[w] = v 204 } 205 206 // dominatorsSimple computes the dominator tree for f. It returns a slice 207 // which maps block ID to the immediate dominator of that block. 208 // Unreachable blocks map to nil. The entry block maps to nil. 209 func dominatorsSimple(f *Func) []*Block { 210 // A simple algorithm for now 211 // Cooper, Harvey, Kennedy 212 idom := make([]*Block, f.NumBlocks()) 213 214 // Compute postorder walk 215 post := f.postorder() 216 217 // Make map from block id to order index (for intersect call) 218 postnum := f.Cache.allocIntSlice(f.NumBlocks()) 219 defer f.Cache.freeIntSlice(postnum) 220 for i, b := range post { 221 postnum[b.ID] = i 222 } 223 224 // Make the entry block a self-loop 225 idom[f.Entry.ID] = f.Entry 226 if postnum[f.Entry.ID] != len(post)-1 { 227 f.Fatalf("entry block %v not last in postorder", f.Entry) 228 } 229 230 // Compute relaxation of idom entries 231 for { 232 changed := false 233 234 for i := len(post) - 2; i >= 0; i-- { 235 b := post[i] 236 var d *Block 237 for _, e := range b.Preds { 238 p := e.b 239 if idom[p.ID] == nil { 240 continue 241 } 242 if d == nil { 243 d = p 244 continue 245 } 246 d = intersect(d, p, postnum, idom) 247 } 248 if d != idom[b.ID] { 249 idom[b.ID] = d 250 changed = true 251 } 252 } 253 if !changed { 254 break 255 } 256 } 257 // Set idom of entry block to nil instead of itself. 258 idom[f.Entry.ID] = nil 259 return idom 260 } 261 262 // intersect finds the closest dominator of both b and c. 263 // It requires a postorder numbering of all the blocks. 264 func intersect(b, c *Block, postnum []int, idom []*Block) *Block { 265 // TODO: This loop is O(n^2). It used to be used in nilcheck, 266 // see BenchmarkNilCheckDeep*. 267 for b != c { 268 if postnum[b.ID] < postnum[c.ID] { 269 b = idom[b.ID] 270 } else { 271 c = idom[c.ID] 272 } 273 } 274 return b 275 }