summaryrefslogtreecommitdiffstats
path: root/src/cmd/compile/internal/ssa/layout.go
blob: e4a8c6ffbf0dde5d4958c177642547216c1f2cdf (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
// Copyright 2015 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package ssa

// layout orders basic blocks in f with the goal of minimizing control flow instructions.
// After this phase returns, the order of f.Blocks matters and is the order
// in which those blocks will appear in the assembly output.
func layout(f *Func) {
	f.Blocks = layoutOrder(f)
}

// Register allocation may use a different order which has constraints
// imposed by the linear-scan algorithm.
func layoutRegallocOrder(f *Func) []*Block {
	// remnant of an experiment; perhaps there will be another.
	return layoutOrder(f)
}

func layoutOrder(f *Func) []*Block {
	order := make([]*Block, 0, f.NumBlocks())
	scheduled := f.Cache.allocBoolSlice(f.NumBlocks())
	defer f.Cache.freeBoolSlice(scheduled)
	idToBlock := f.Cache.allocBlockSlice(f.NumBlocks())
	defer f.Cache.freeBlockSlice(idToBlock)
	indegree := f.Cache.allocIntSlice(f.NumBlocks())
	defer f.Cache.freeIntSlice(indegree)
	posdegree := f.newSparseSet(f.NumBlocks()) // blocks with positive remaining degree
	defer f.retSparseSet(posdegree)
	// blocks with zero remaining degree. Use slice to simulate a LIFO queue to implement
	// the depth-first topology sorting algorithm.
	var zerodegree []ID
	// LIFO queue. Track the successor blocks of the scheduled block so that when we
	// encounter loops, we choose to schedule the successor block of the most recently
	// scheduled block.
	var succs []ID
	exit := f.newSparseSet(f.NumBlocks()) // exit blocks
	defer f.retSparseSet(exit)

	// Populate idToBlock and find exit blocks.
	for _, b := range f.Blocks {
		idToBlock[b.ID] = b
		if b.Kind == BlockExit {
			exit.add(b.ID)
		}
	}

	// Expand exit to include blocks post-dominated by exit blocks.
	for {
		changed := false
		for _, id := range exit.contents() {
			b := idToBlock[id]
		NextPred:
			for _, pe := range b.Preds {
				p := pe.b
				if exit.contains(p.ID) {
					continue
				}
				for _, s := range p.Succs {
					if !exit.contains(s.b.ID) {
						continue NextPred
					}
				}
				// All Succs are in exit; add p.
				exit.add(p.ID)
				changed = true
			}
		}
		if !changed {
			break
		}
	}

	// Initialize indegree of each block
	for _, b := range f.Blocks {
		if exit.contains(b.ID) {
			// exit blocks are always scheduled last
			continue
		}
		indegree[b.ID] = len(b.Preds)
		if len(b.Preds) == 0 {
			// Push an element to the tail of the queue.
			zerodegree = append(zerodegree, b.ID)
		} else {
			posdegree.add(b.ID)
		}
	}

	bid := f.Entry.ID
blockloop:
	for {
		// add block to schedule
		b := idToBlock[bid]
		order = append(order, b)
		scheduled[bid] = true
		if len(order) == len(f.Blocks) {
			break
		}

		// Here, the order of traversing the b.Succs affects the direction in which the topological
		// sort advances in depth. Take the following cfg as an example, regardless of other factors.
		//           b1
		//         0/ \1
		//        b2   b3
		// Traverse b.Succs in order, the right child node b3 will be scheduled immediately after
		// b1, traverse b.Succs in reverse order, the left child node b2 will be scheduled
		// immediately after b1. The test results show that reverse traversal performs a little
		// better.
		// Note: You need to consider both layout and register allocation when testing performance.
		for i := len(b.Succs) - 1; i >= 0; i-- {
			c := b.Succs[i].b
			indegree[c.ID]--
			if indegree[c.ID] == 0 {
				posdegree.remove(c.ID)
				zerodegree = append(zerodegree, c.ID)
			} else {
				succs = append(succs, c.ID)
			}
		}

		// Pick the next block to schedule
		// Pick among the successor blocks that have not been scheduled yet.

		// Use likely direction if we have it.
		var likely *Block
		switch b.Likely {
		case BranchLikely:
			likely = b.Succs[0].b
		case BranchUnlikely:
			likely = b.Succs[1].b
		}
		if likely != nil && !scheduled[likely.ID] {
			bid = likely.ID
			continue
		}

		// Use degree for now.
		bid = 0
		// TODO: improve this part
		// No successor of the previously scheduled block works.
		// Pick a zero-degree block if we can.
		for len(zerodegree) > 0 {
			// Pop an element from the tail of the queue.
			cid := zerodegree[len(zerodegree)-1]
			zerodegree = zerodegree[:len(zerodegree)-1]
			if !scheduled[cid] {
				bid = cid
				continue blockloop
			}
		}

		// Still nothing, pick the unscheduled successor block encountered most recently.
		for len(succs) > 0 {
			// Pop an element from the tail of the queue.
			cid := succs[len(succs)-1]
			succs = succs[:len(succs)-1]
			if !scheduled[cid] {
				bid = cid
				continue blockloop
			}
		}

		// Still nothing, pick any non-exit block.
		for posdegree.size() > 0 {
			cid := posdegree.pop()
			if !scheduled[cid] {
				bid = cid
				continue blockloop
			}
		}
		// Pick any exit block.
		// TODO: Order these to minimize jump distances?
		for {
			cid := exit.pop()
			if !scheduled[cid] {
				bid = cid
				continue blockloop
			}
		}
	}
	f.laidout = true
	return order
	//f.Blocks = order
}