comparison gcc/jit/docs/intro/tutorial03.rst @ 111:04ced10e8804

gcc 7
author kono
date Fri, 27 Oct 2017 22:46:09 +0900
parents
children 84e7813d76e9
comparison
equal deleted inserted replaced
68:561a7518be6b 111:04ced10e8804
1 .. Copyright (C) 2014-2017 Free Software Foundation, Inc.
2 Originally contributed by David Malcolm <dmalcolm@redhat.com>
3
4 This is free software: you can redistribute it and/or modify it
5 under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
8
9 This program is distributed in the hope that it will be useful, but
10 WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 General Public License for more details.
13
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see
16 <http://www.gnu.org/licenses/>.
17
18 Tutorial part 3: Loops and variables
19 ------------------------------------
20 Consider this C function:
21
22 .. code-block:: c
23
24 int loop_test (int n)
25 {
26 int sum = 0;
27 for (int i = 0; i < n; i++)
28 sum += i * i;
29 return sum;
30 }
31
32 This example demonstrates some more features of libgccjit, with local
33 variables and a loop.
34
35 To break this down into libgccjit terms, it's usually easier to reword
36 the `for` loop as a `while` loop, giving:
37
38 .. code-block:: c
39
40 int loop_test (int n)
41 {
42 int sum = 0;
43 int i = 0;
44 while (i < n)
45 {
46 sum += i * i;
47 i++;
48 }
49 return sum;
50 }
51
52 Here's what the final control flow graph will look like:
53
54 .. figure:: sum-of-squares.png
55 :alt: image of a control flow graph
56
57 As before, we include the libgccjit header and make a
58 :c:type:`gcc_jit_context *`.
59
60 .. code-block:: c
61
62 #include <libgccjit.h>
63
64 void test (void)
65 {
66 gcc_jit_context *ctxt;
67 ctxt = gcc_jit_context_acquire ();
68
69 The function works with the C `int` type:
70
71 .. code-block:: c
72
73 gcc_jit_type *the_type =
74 gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT);
75 gcc_jit_type *return_type = the_type;
76
77 though we could equally well make it work on, say, `double`:
78
79 .. code-block:: c
80
81 gcc_jit_type *the_type =
82 gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_DOUBLE);
83
84 Let's build the function:
85
86 .. code-block:: c
87
88 gcc_jit_param *n =
89 gcc_jit_context_new_param (ctxt, NULL, the_type, "n");
90 gcc_jit_param *params[1] = {n};
91 gcc_jit_function *func =
92 gcc_jit_context_new_function (ctxt, NULL,
93 GCC_JIT_FUNCTION_EXPORTED,
94 return_type,
95 "loop_test",
96 1, params, 0);
97
98 Expressions: lvalues and rvalues
99 ********************************
100
101 The base class of expression is the :c:type:`gcc_jit_rvalue *`,
102 representing an expression that can be on the *right*-hand side of
103 an assignment: a value that can be computed somehow, and assigned
104 *to* a storage area (such as a variable). It has a specific
105 :c:type:`gcc_jit_type *`.
106
107 Anothe important class is :c:type:`gcc_jit_lvalue *`.
108 A :c:type:`gcc_jit_lvalue *`. is something that can of the *left*-hand
109 side of an assignment: a storage area (such as a variable).
110
111 In other words, every assignment can be thought of as:
112
113 .. code-block:: c
114
115 LVALUE = RVALUE;
116
117 Note that :c:type:`gcc_jit_lvalue *` is a subclass of
118 :c:type:`gcc_jit_rvalue *`, where in an assignment of the form:
119
120 .. code-block:: c
121
122 LVALUE_A = LVALUE_B;
123
124 the `LVALUE_B` implies reading the current value of that storage
125 area, assigning it into the `LVALUE_A`.
126
127 So far the only expressions we've seen are `i * i`:
128
129 .. code-block:: c
130
131 gcc_jit_rvalue *expr =
132 gcc_jit_context_new_binary_op (
133 ctxt, NULL,
134 GCC_JIT_BINARY_OP_MULT, int_type,
135 gcc_jit_param_as_rvalue (param_i),
136 gcc_jit_param_as_rvalue (param_i));
137
138 which is a :c:type:`gcc_jit_rvalue *`, and the various function
139 parameters: `param_i` and `param_n`, instances of
140 :c:type:`gcc_jit_param *`, which is a subclass of
141 :c:type:`gcc_jit_lvalue *` (and, in turn, of :c:type:`gcc_jit_rvalue *`):
142 we can both read from and write to function parameters within the
143 body of a function.
144
145 Our new example has a couple of local variables. We create them by
146 calling :c:func:`gcc_jit_function_new_local`, supplying a type and a
147 name:
148
149 .. code-block:: c
150
151 /* Build locals: */
152 gcc_jit_lvalue *i =
153 gcc_jit_function_new_local (func, NULL, the_type, "i");
154 gcc_jit_lvalue *sum =
155 gcc_jit_function_new_local (func, NULL, the_type, "sum");
156
157 These are instances of :c:type:`gcc_jit_lvalue *` - they can be read from
158 and written to.
159
160 Note that there is no precanned way to create *and* initialize a variable
161 like in C:
162
163 .. code-block:: c
164
165 int i = 0;
166
167 Instead, having added the local to the function, we have to separately add
168 an assignment of `0` to `local_i` at the beginning of the function.
169
170 Control flow
171 ************
172
173 This function has a loop, so we need to build some basic blocks to
174 handle the control flow. In this case, we need 4 blocks:
175
176 1. before the loop (initializing the locals)
177 2. the conditional at the top of the loop (comparing `i < n`)
178 3. the body of the loop
179 4. after the loop terminates (`return sum`)
180
181 so we create these as :c:type:`gcc_jit_block *` instances within the
182 :c:type:`gcc_jit_function *`:
183
184 .. code-block:: c
185
186 gcc_jit_block *b_initial =
187 gcc_jit_function_new_block (func, "initial");
188 gcc_jit_block *b_loop_cond =
189 gcc_jit_function_new_block (func, "loop_cond");
190 gcc_jit_block *b_loop_body =
191 gcc_jit_function_new_block (func, "loop_body");
192 gcc_jit_block *b_after_loop =
193 gcc_jit_function_new_block (func, "after_loop");
194
195 We now populate each block with statements.
196
197 The entry block `b_initial` consists of initializations followed by a jump
198 to the conditional. We assign `0` to `i` and to `sum`, using
199 :c:func:`gcc_jit_block_add_assignment` to add
200 an assignment statement, and using :c:func:`gcc_jit_context_zero` to get
201 the constant value `0` for the relevant type for the right-hand side of
202 the assignment:
203
204 .. code-block:: c
205
206 /* sum = 0; */
207 gcc_jit_block_add_assignment (
208 b_initial, NULL,
209 sum,
210 gcc_jit_context_zero (ctxt, the_type));
211
212 /* i = 0; */
213 gcc_jit_block_add_assignment (
214 b_initial, NULL,
215 i,
216 gcc_jit_context_zero (ctxt, the_type));
217
218 We can then terminate the entry block by jumping to the conditional:
219
220 .. code-block:: c
221
222 gcc_jit_block_end_with_jump (b_initial, NULL, b_loop_cond);
223
224 The conditional block is equivalent to the line `while (i < n)` from our
225 C example. It contains a single statement: a conditional, which jumps to
226 one of two destination blocks depending on a boolean
227 :c:type:`gcc_jit_rvalue *`, in this case the comparison of `i` and `n`.
228 We build the comparison using :c:func:`gcc_jit_context_new_comparison`:
229
230 .. code-block:: c
231
232 /* (i >= n) */
233 gcc_jit_rvalue *guard =
234 gcc_jit_context_new_comparison (
235 ctxt, NULL,
236 GCC_JIT_COMPARISON_GE,
237 gcc_jit_lvalue_as_rvalue (i),
238 gcc_jit_param_as_rvalue (n));
239
240 and can then use this to add `b_loop_cond`'s sole statement, via
241 :c:func:`gcc_jit_block_end_with_conditional`:
242
243 .. code-block:: c
244
245 /* Equivalent to:
246 if (guard)
247 goto after_loop;
248 else
249 goto loop_body; */
250 gcc_jit_block_end_with_conditional (
251 b_loop_cond, NULL,
252 guard,
253 b_after_loop, /* on_true */
254 b_loop_body); /* on_false */
255
256 Next, we populate the body of the loop.
257
258 The C statement `sum += i * i;` is an assignment operation, where an
259 lvalue is modified "in-place". We use
260 :c:func:`gcc_jit_block_add_assignment_op` to handle these operations:
261
262 .. code-block:: c
263
264 /* sum += i * i */
265 gcc_jit_block_add_assignment_op (
266 b_loop_body, NULL,
267 sum,
268 GCC_JIT_BINARY_OP_PLUS,
269 gcc_jit_context_new_binary_op (
270 ctxt, NULL,
271 GCC_JIT_BINARY_OP_MULT, the_type,
272 gcc_jit_lvalue_as_rvalue (i),
273 gcc_jit_lvalue_as_rvalue (i)));
274
275 The `i++` can be thought of as `i += 1`, and can thus be handled in
276 a similar way. We use :c:func:`gcc_jit_context_one` to get the constant
277 value `1` (for the relevant type) for the right-hand side
278 of the assignment.
279
280 .. code-block:: c
281
282 /* i++ */
283 gcc_jit_block_add_assignment_op (
284 b_loop_body, NULL,
285 i,
286 GCC_JIT_BINARY_OP_PLUS,
287 gcc_jit_context_one (ctxt, the_type));
288
289 .. note::
290
291 For numeric constants other than 0 or 1, we could use
292 :c:func:`gcc_jit_context_new_rvalue_from_int` and
293 :c:func:`gcc_jit_context_new_rvalue_from_double`.
294
295 The loop body completes by jumping back to the conditional:
296
297 .. code-block:: c
298
299 gcc_jit_block_end_with_jump (b_loop_body, NULL, b_loop_cond);
300
301 Finally, we populate the `b_after_loop` block, reached when the loop
302 conditional is false. We want to generate the equivalent of:
303
304 .. code-block:: c
305
306 return sum;
307
308 so the block is just one statement:
309
310 .. code-block:: c
311
312 /* return sum */
313 gcc_jit_block_end_with_return (
314 b_after_loop,
315 NULL,
316 gcc_jit_lvalue_as_rvalue (sum));
317
318 .. note::
319
320 You can intermingle block creation with statement creation,
321 but given that the terminator statements generally include references
322 to other blocks, I find it's clearer to create all the blocks,
323 *then* all the statements.
324
325 We've finished populating the function. As before, we can now compile it
326 to machine code:
327
328 .. code-block:: c
329
330 gcc_jit_result *result;
331 result = gcc_jit_context_compile (ctxt);
332
333 typedef int (*loop_test_fn_type) (int);
334 loop_test_fn_type loop_test =
335 (loop_test_fn_type)gcc_jit_result_get_code (result, "loop_test");
336 if (!loop_test)
337 goto error;
338 printf ("result: %d", loop_test (10));
339
340 .. code-block:: bash
341
342 result: 285
343
344
345 Visualizing the control flow graph
346 **********************************
347
348 You can see the control flow graph of a function using
349 :c:func:`gcc_jit_function_dump_to_dot`:
350
351 .. code-block:: c
352
353 gcc_jit_function_dump_to_dot (func, "/tmp/sum-of-squares.dot");
354
355 giving a .dot file in GraphViz format.
356
357 You can convert this to an image using `dot`:
358
359 .. code-block:: bash
360
361 $ dot -Tpng /tmp/sum-of-squares.dot -o /tmp/sum-of-squares.png
362
363 or use a viewer (my preferred one is xdot.py; see
364 https://github.com/jrfonseca/xdot.py; on Fedora you can
365 install it with `yum install python-xdot`):
366
367 .. figure:: sum-of-squares.png
368 :alt: image of a control flow graph
369
370 Full example
371 ************
372
373 .. literalinclude:: ../examples/tut03-sum-of-squares.c
374 :lines: 1-
375 :language: c
376
377 Building and running it:
378
379 .. code-block:: console
380
381 $ gcc \
382 tut03-sum-of-squares.c \
383 -o tut03-sum-of-squares \
384 -lgccjit
385
386 # Run the built program:
387 $ ./tut03-sum-of-squares
388 loop_test returned: 285