@@ -22,17 +22,147 @@ Tasks of the compiler
2222
23234 . Translate function calls into machine-level calls
2424
25- Infrastructure of the compiler
26- ------------------------------
25+ ### Ad 1. Variable names to addresses
2726
28- Signature (global symbol table):
29- - name and type of functions suitable for Jasmin
30- - Jasmin types:
27+ Jasmin types:
3128 * ` I ` : ` int `
3229 * ` D ` : ` double `
3330 * ` V ` : ` void `
3431 * ` Z ` : ` boolean `
3532
33+ ``` c
34+ void foo // size=0 max=0
35+ ( double x // x:0 size=2 max=2
36+ , int y) { // y:2 size=3 max=3
37+ int i; // i:3 size=4 max=4
38+ { // newBlock
39+ double y; // y:4 size=6 max=6
40+ bool b; // b:6 size=7 max=7
41+ } // popBlock size=4 max=7
42+ int j; // j:4 size=5 max=7
43+ }
44+ ```
45+ ``` jasmin
46+ .method public static foo(DI)V
47+ .limit locals 7
48+ .end method
49+ ```
50+
51+ ### Ad 2. Expressions trees to stack instructions
52+
53+ Example:
54+ ``` c
55+ int main () {
56+ int i; // i:0
57+ double x = 3 * (i = 42) + i++; // x:1
58+ }
59+ ```
60+ Translates to JVM:
61+ ``` scheme
62+ .method public static main()I
63+ .limit stack 4
64+
65+ ;; instr ;; expr ;; stack (S) ;; store (V)
66+ iconst_3 ;; 3 ;; 3
67+ bipush 42 ;; 42 ;; 3 . 42
68+ istore_0 ;; i = ;; 3 ;; 0:42
69+ iload_0 ;; i = 42 ;; 3 . 42
70+ imul ;; 3 * (i = 42) ;; 126
71+ iload_0 ;; i ;; 126 . 42
72+ dup ;; ;; 126 . 42 . 42
73+ iconst_1 ;; 1 ;; 126 . 42 . 42 . 1
74+ iadd ;; i + 1 ;; 126 . 42 . 43
75+ istore_0 ;; i++ ;; 126 . 42 ;; 0:43
76+ iadd ;; 3 * (i = 42) + i++ ;; 168
77+ i2d ;; (double) ;; 168.0
78+ dstore_1 ;; x = 3 * (i = 42) + i++ ;; ;; 0:43, 1:168.0
79+ ```
80+ Maximum stack size: 4 (32bit words)
81+
82+ JVM instruction small-step semantics (without jumps):
83+ ```
84+ I : ⟨V,S⟩ → ⟨V',S'⟩
85+
86+ I : JVM instruction (or instruction sequence)
87+ V / V' : variable store before/after
88+ S / S' : stack before/after
89+ ```
90+
91+ Some jump-free instructions:
92+
93+ | I | V | S | V' | S' |
94+ | --------------| -----| ---------| ----------| -----------------|
95+ | ` iconst ` _ i_ | _ V_ | _ S_ | _ V_ | _ S.i_ |
96+ | ` bipush ` _ i_ | _ V_ | _ S_ | _ V_ | _ S.i_ |
97+ | ` iadd ` | _ V_ | _ S.i.j_ | _ V_ | _ S.(i+j)_ |
98+ | ` imul ` | _ V_ | _ S.i.j_ | _ V_ | _ S.(i* j)_ |
99+ | ` i2d ` | _ V_ | _ S.i_ | _ V_ | _ S.(double)i_ |
100+ | ` dup ` | _ V_ | _ S.i_ | _ V_ | _ S.i.i_ |
101+ | ` istore ` _ a_ | _ V_ | _ S.i_ | _ V[ a: i ] _ | _ S_ |
102+ | ` dstore ` _ a_ | _ V_ | _ S.d_ | _ V[ a: d ] _ | _ S_ |
103+ | ` iload ` _ a_ | _ V_ | _ S_ | _ V_ | _ S.V[ a] _ |
104+ | ` dload ` _ a_ | _ V_ | _ S_ | _ V_ | _ S.V[ a] _ |
105+
106+
107+ Ad 3. Booleans and jumps
108+ ------------------------
109+
110+ Example:
111+ ``` c
112+ bool inside (int i, int j, int k) {
113+ return i <= j && j < k;
114+ }
115+ ```
116+ Machines typically do not have boolean operations; use jumps instead.
117+ ```scheme
118+ .method public static inside(III)Z
119+ .limit locals 3
120+ .limit stack 2
121+
122+ iload_0 ;; i
123+ iload_1 ;; j
124+ if_icmpgt Lfalse ;; i <= j ;; false if i > j
125+ iload_1 ;; j
126+ iload_2 ;; k
127+ if_icmpge Lfalse ;; j < k ;; false if j >= k
128+
129+ iconst_1 ;; true
130+ goto Ldone
131+
132+ Lfalse:
133+ iconst_0 ;; false
134+
135+ Ldone:
136+ ireturn ;; return
137+
138+ .end method
139+ ```
140+
141+ JVM instruction small-step semantics (with jumps):
142+ ```
143+ I : ⟨P,V,S⟩ → ⟨P',V',S'⟩
144+
145+ P / P' : code position (program counter) before/after
146+ ```
147+
148+ | I | P | V | S | P' | V' | S' | Condition |
149+ | -----------------| -----| -----| ---------| -------| -----| -----| -----------|
150+ | ` goto ` _ L_ | _ P_ | _ V_ | _ S_ | _ L_ | _ V_ | _ S_ | |
151+ | ` ifeq ` _ L_ | _ P_ | _ V_ | _ S.i_ | _ L_ | _ V_ | _ S_ | _ i = 0_ |
152+ | ` ifeq ` _ L_ | _ P_ | _ V_ | _ S.i_ | _ P+1_ | _ V_ | _ S_ | _ i ≠ 0_ |
153+ | ` ifne ` _ L_ | _ P_ | _ V_ | _ S.i_ | _ L_ | _ V_ | _ S_ | _ i ≠ 0_ |
154+ | ` ifne ` _ L_ | _ P_ | _ V_ | _ S.i_ | _ P+1_ | _ V_ | _ S_ | _ i = 0_ |
155+ | ` if_icmpgt ` _ L_ | _ P_ | _ V_ | _ S.i.j_ | _ L_ | _ V_ | _ S_ | _ i > j_ |
156+ | ` if_icmpgt ` _ L_ | _ P_ | _ V_ | _ S.i.j_ | _ P+1_ | _ V_ | _ S_ | _ i ≤ j_ |
157+
158+
159+
160+ Infrastructure of the compiler
161+ ------------------------------
162+
163+ Signature (global symbol table):
164+ - name and type of functions suitable for Jasmin
165+
36166Context (local symbol table):
371671 . allocate local variables (` newLocal ` )
381682 . resolve variable names to addresses (` lookupVar ` )
@@ -90,7 +220,6 @@ rest of the stack unchanged.
90220 emit(t- store a) -- either istore or dstore
91221 -- Problem here
92222```
93- (We omit ` emit ` in the following.)
94223
95224## Compiler correctness
96225
@@ -130,17 +259,30 @@ stack is nil (no change).
130259 compileStm(SInit t x e):
131260 a <- newLocal t x
132261 compileExp(e)
133- t- store a -- either istore or dstore
262+ emit( t- store a) -- either istore or dstore
134263
135264 compileStm(SExp t e):
136265 compileExp(e)
137- t- pop -- either pop or pop2
266+ emit( t- pop) -- either pop or pop2
138267
139268 compileStm(SBlock ss):
140269 newBlock
141270 for (s : ss)
142271 compileStm(s)
143272 popBlock
273+
274+ compileStm(SWhile e s): -- s is a SBlock
275+ LStart , LEnd <- newLabel
276+
277+ emit(LStart : )
278+
279+ compileExp(e) -- executing e leaves boolean on stack
280+ emit(ifeq LEnd ) -- if "false" end loop
281+
282+ compileStm(s) -- body of loop
283+ emit(goto LStart ) -- recheck loop condition
284+
285+ emit(LEnd : )
144286```
145287
146288Correctness statement (simplified):
@@ -150,6 +292,8 @@ Correctness statement (simplified):
150292 then compileStm(s) : ⟨V,S⟩ →* ⟨V',S⟩
151293 such that γ' ~ V'.
152294
295+ (We omit ` emit ` in the following.)
296+
153297Compiling booleans
154298------------------
155299
0 commit comments