Static Single Assignment Form
静态单一分配形式,每个变量只赋值一次
intermediate representation(IR): 中间代码表示
作用
- 常数传播（constant propagation）
- 值域传播（value range propagation）
- 稀疏有条件的常数传播（sparse conditional constant propagation）
- 消除无用的程式码（dead code elimination）
- 全域数值编号（global value numbering）
- 消除部分的冗余（partial redundancy elimination）
- 强度折减（strength reduction）
- 寄存器分配（register allocation）
demo
x = 1
y = x
变成了
y = 1
转化
{
"title": {
"text": "Graph 简单示例"
},
"tooltip": {},
"animationDurationUpdate": 1500,
"animationEasingUpdate": "quinticInOut",
"series": [{
"type": "graph",
"layout": "none",
"symbolSize": 50,
"roam": true,
"label": {
"show": true
},
"edgeSymbol": ["circle", "arrow"],
"edgeSymbolSize": [4, 10],
"edgeLabel": {
"fontSize": 20
},
"data": [{
"name": "x=5;x=x-3;x<3?",
"x": 400,
"y": 0
}, {
"name": "y=x*2;w=y",
"x": 300,
"y": 100
}, {
"name": "y=x-3",
"x": 500,
"y": 100
}, {
"name": "w=x-y;z=x+y",
"x": 400,
"y": 200
}],
"links": [{
"source": "x=5;x=x-3;x<3?",
"target": "y=x*2;w=y"
}, {
"source": "x=5;x=x-3;x<3?",
"target": "y=x-3"
}, {
"source": "y=x-3",
"target": "w=x-y;z=x+y"
}, {
"source": "y=x*2;w=y",
"target": "w=x-y;z=x+y"
}],
"lineStyle": {
"opacity": 0.9,
"width": 2,
"curveness": 0
}
}]
}
全域数值编号 + 静态单一分配形式 转化代码
{
"title": {
"text": "Graph 简单示例"
},
"tooltip": {},
"animationDurationUpdate": 1500,
"animationEasingUpdate": "quinticInOut",
"series": [{
"type": "graph",
"layout": "none",
"symbolSize": 50,
"roam": true,
"label": {
"show": true
},
"edgeSymbol": ["circle", "arrow"],
"edgeSymbolSize": [4, 10],
"edgeLabel": {
"fontSize": 20
},
"data": [{
"name": "x1=5;x2=x1-3;x2<3?",
"x": 400,
"y": 0
}, {
"name": "y1=x2*2;w1=y1",
"x": 300,
"y": 100
}, {
"name": "y2=x2-3",
"x": 500,
"y": 100
}, {
"name": "w2=x2-y?;z1=x2+y?",
"x": 400,
"y": 200
}],
"links": [{
"source": "x1=5;x2=x1-3;x2<3?",
"target": "y1=x2*2;w1=y1"
}, {
"source": "x1=5;x2=x1-3;x2<3?",
"target": "y2=x2-3"
}, {
"source": "y1=x2*2;w1=y1",
"target": "w2=x2-y?;z1=x2+y?"
}, {
"source": "y2=x2-3",
"target": "w2=x2-y?;z1=x2+y?"
}],
"lineStyle": {
"opacity": 0.9,
"width": 2,
"curveness": 0
}
}]
}
全域数值编号把每个静态量都编号了，也保证每个静态量一定是静态的是单一的，只被赋值一次单最后出现了 y 的编号不知道是多少的问题
所以在最后一步加入了
Φ (Phi) function.
y3 = Φ(y1,y2)
所以转换以后为
{
"title": {
"text": "Graph 简单示例"
},
"tooltip": {},
"animationDurationUpdate": 1500,
"animationEasingUpdate": "quinticInOut",
"series": [{
"type": "graph",
"layout": "none",
"symbolSize": 50,
"roam": true,
"label": {
"show": true
},
"edgeSymbol": ["circle", "arrow"],
"edgeSymbolSize": [4, 10],
"edgeLabel": {
"fontSize": 20
},
"data": [{
"name": "x1=5;x2=x1-3;x2<3?",
"x": 400,
"y": 0
}, {
"name": "y1=x2*2;w1=y1",
"x": 300,
"y": 100
}, {
"name": "y2=x2-3",
"x": 500,
"y": 100
}, {
"name": "y3 = Φ(y1,y2);w2=x2-y3;z1=x2+y3",
"x": 400,
"y": 200
}],
"links": [{
"source": "x1=5;x2=x1-3;x2<3?",
"target": "y1=x2*2;w1=y1"
}, {
"source": "x1=5;x2=x1-3;x2<3?",
"target": "y2=x2-3"
}, {
"source": "y1=x2*2;w1=y1",
"target": "y3 = Φ(y1,y2);w2=x2-y3;z1=x2+y3"
}, {
"source": "y2=x2-3",
"target": "y3 = Φ(y1,y2);w2=x2-y3;z1=x2+y3"
}],
"lineStyle": {
"opacity": 0.9,
"width": 2,
"curveness": 0
}
}]
}
SSA 的主要作用是对代码进行优化，所以它是编译器后端的一部分
欢迎来到这里！
我们正在构建一个小众社区，大家在这里相互信任，以平等 • 自由 • 奔放的价值观进行分享交流。最终，希望大家能够找到与自己志同道合的伙伴，共同成长。
注册 关于