@@ -15,12 +15,22 @@ when analysis is incomplete. The symbolic execution includes constant
15
15
folding so loops that have a constant number of iterations will be
16
16
handled completely (assuming the unwinding limit is sufficient).
17
17
18
- The output of the symbolic execution is a system of equations; an object
19
- containing a list of ` symex_target_elements ` , each of which are
20
- equalities between ` expr ` expressions. See ` symex_target_equation.h ` .
21
- The output is in static, single assignment (SSA) form, which is * not*
22
- the case for goto-programs.
18
+ The output of symbolic execution is a system of equations; an object
19
+ of type ` symex_target_equationt ` , containing a list of
20
+ ` symex_target_equationt::SSA_stept ` , each of which are equalities
21
+ between ` exprt ` expressions. This list is constructed incrementally as
22
+ the symbolic execution engine walks through the goto-program using the
23
+ ` symex_targett ` interface. This interface (implemented by
24
+ ` symex_target_equationt ` ) exposes functions that append SSA steps into
25
+ the aforementioned list while transforming expressions into
26
+ Static Single Assignment (SSA) form. For more details on this process
27
+ see ` symex_target_equation.h ` , for an overview of SSA see
28
+ \ref static-single-assignment.
29
+
30
+ At a later stage, BMC will convert the generated SSA steps into an
31
+ equation that can be passed to the solver.
23
32
33
+ ---
24
34
\section symbolic-execution Symbolic Execution
25
35
26
36
In the \ref goto-symex directory.
@@ -95,47 +105,80 @@ execution run does not add any paths to the workqueue but rather merges
95
105
all the paths together, so the additional path-exploration loop is
96
106
skipped over.
97
107
98
- \subsection ssa-renaming SSA renaming levels
99
-
100
- In goto-programs, variable names get a prefix to indicate their scope
101
- (like ` main::1::%foo ` or whatever). At symbolic execution level, variables
102
- also get a _ suffix_ because we’re doing single-static assignment. There
103
- are three “levels” of renaming. At Level 2, variables are renamed every
104
- time they are encountered in a function. At Level 1, variables are
105
- renamed every time the functions that contain those variables are
106
- called. At Level 0, variables are renamed every time a new thread
107
- containing those variables are spawned. We can inspect the renamed
108
- variable names with the –show-vcc flag, which prints a string with the
109
- following format: ` %%s!%%d@%%d#%%d ` . The string field is the variable name,
110
- and the numbers after the !, @, and # are the L0, L1, and L2 suffixes
111
- respectively. The following examples illustrate Level 1 and 2 renaming:
108
+ ---
109
+ \section static-single-assignment Static Single Assignment (SSA) Form
110
+
111
+ ** Key classes:**
112
+ * \ref symex_targett
113
+ * \ref symex_target_equationt
114
+
115
+ * Static Single Assignment (SSA)* form is an intermediate
116
+ representation that satisfies the following properties:
117
+
118
+ 1 . Every variable is * assigned exactly once* .
119
+ 2 . Every variable must be * defined* before use.
120
+
121
+ In-order to convert a goto-program to SSA form all variables are
122
+ indexed (renamed) through the addition of a _ suffix_ .
123
+
124
+ There are three “levels” of indexing:
125
+
126
+ ** Level 2 (L2):** variables are indexed every time they are
127
+ encountered in a function.
128
+
129
+ ** Level 1 (L1):** variables are indexed every time the functions that
130
+ contain those variables are called.
131
+
132
+ ** Level 0 (L0):** variables are indexed every time a new thread
133
+ containing those variables are spawned.
134
+
135
+ We can inspect the indexed variable names with the ** --show-vcc** or
136
+ ** --program-only** flags. Variables in SSA form are printed in the
137
+ following format: ` %%s!%%d@%%d#%%d ` . Where the string field is the
138
+ variable name, and the numbers after the !, @, and # are the L0, L1,
139
+ and L2 suffixes respectively.
140
+
141
+ > Note: ** --program-only** prints all the SSA steps in-order. In
142
+ > contrast, ** --show-vcc** will for each assertion print the SSA steps
143
+ > (assumes, assignments and constraints only) that synthetically
144
+ > precede the assertion. In the presence of multiple threads it will
145
+ > also print SSA steps that succeed the assertion.
146
+
147
+ \subsection L1-L2 Level 1 and level 2 indexing
148
+
149
+ The following examples illustrate Level 1 and 2 indexing.
112
150
113
151
$ cat l1.c
114
- int main() {
152
+ int main()
153
+ {
115
154
int x=7;
116
155
x=8;
117
156
assert(0);
118
157
}
158
+
119
159
$ cbmc --show-vcc l1.c
120
160
...
121
161
{-12} x!0@1#2 == 7
122
162
{-13} x!0@1#3 == 8
123
163
124
- That is, the L0 names for both xs are 0; the L1 names for both xs are 1;
125
- but each occurrence of x within main() gets a fresh L2 suffix (2 and 3,
126
- respectively).
164
+ That is, the L0 names for both x's are 0; the L1 names for both x's
165
+ are 1; but each occurrence of x within main() gets a fresh L2 suffix
166
+ (2 and 3, respectively).
127
167
128
168
$ cat l2.c
129
- void foo(){
169
+ void foo()
170
+ {
130
171
int x=7;
131
172
x=8;
132
173
x=9;
133
174
}
134
- int main(){
175
+ int main()
176
+ {
135
177
foo();
136
178
foo();
137
179
assert(0);
138
180
}
181
+
139
182
$ cbmc --show-vcc l2.c
140
183
...
141
184
{-12} x!0@1#2 == 7
@@ -146,21 +189,42 @@ respectively).
146
189
{-17} x!0@2#4 == 9
147
190
148
191
That is, every time we enter function foo, the L1 counter of x is
149
- incremented (from 1 to 2) and the L0 counter is reset (back to 2, after
150
- having been incremented up to 4). The L0 counter then increases every
151
- time we access x as we walk through the function.
192
+ incremented (from 1 to 2) and the L2 counter is reset (back to 2,
193
+ after having been incremented up to 4). The L2 counter then increases
194
+ every time we access x as we walk through the function.
195
+
196
+ \subsection L0 Level 0 indexing (threads only)
197
+
198
+ TODO: describe and give a concrete example
199
+
200
+ \subsection PL Relevant Primary Literature
201
+
202
+ Thread indexing is based on the following paper:
203
+
204
+ > Lee, J., Midkiff, S.P. and Padua, D.A., 1997, August. Concurrent
205
+ > static single assignment form and constant propagation for
206
+ > explicitly parallel programs. In International Workshop on Languages
207
+ > and Compilers for Parallel Computing (pp. 114-130). Springer,
208
+ > Berlin, Heidelberg.
209
+
210
+ Seminal paper on SSA:
211
+
212
+ > Rosen, B.K., Wegman, M.N. and Zadeck, F.K., 1988, January. Global
213
+ > value numbers and redundant computations. In Proceedings of the 15th
214
+ > ACM SIGPLAN-SIGACT symposium on Principles of programming languages
215
+ > (pp. 12-27). ACM.
152
216
153
217
---
154
218
\section counter-example-production Counter Example Production
155
219
156
220
In the \ref goto-symex directory.
157
221
158
222
** Key classes:**
159
- * symex_target_equationt
160
- * prop_convt
223
+ * \ref symex_target_equationt
224
+ * \ref prop_convt
161
225
* \ref bmct
162
- * fault_localizationt
163
- * counterexample_beautificationt
226
+ * \ref fault_localizationt
227
+ * \ref counterexample_beautificationt
164
228
165
229
\dot
166
230
digraph G {
0 commit comments