1 | /* Conditional Dead Call Elimination pass for the GNU compiler. |
2 | Copyright (C) 2008-2024 Free Software Foundation, Inc. |
3 | Contributed by Xinliang David Li <davidxl@google.com> |
4 | |
5 | This file is part of GCC. |
6 | |
7 | GCC is free software; you can redistribute it and/or modify it |
8 | under the terms of the GNU General Public License as published by the |
9 | Free Software Foundation; either version 3, or (at your option) any |
10 | later version. |
11 | |
12 | GCC is distributed in the hope that it will be useful, but WITHOUT |
13 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
14 | FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
15 | for more details. |
16 | |
17 | You should have received a copy of the GNU General Public License |
18 | along with GCC; see the file COPYING3. If not see |
19 | <http://www.gnu.org/licenses/>. */ |
20 | |
21 | #include "config.h" |
22 | #include "system.h" |
23 | #include "coretypes.h" |
24 | #include "backend.h" |
25 | #include "tree.h" |
26 | #include "gimple.h" |
27 | #include "cfghooks.h" |
28 | #include "tree-pass.h" |
29 | #include "ssa.h" |
30 | #include "gimple-pretty-print.h" |
31 | #include "fold-const.h" |
32 | #include "stor-layout.h" |
33 | #include "gimple-iterator.h" |
34 | #include "tree-cfg.h" |
35 | #include "tree-into-ssa.h" |
36 | #include "builtins.h" |
37 | #include "internal-fn.h" |
38 | #include "tree-dfa.h" |
39 | |
40 | |
41 | /* This pass serves two closely-related purposes: |
42 | |
43 | 1. It conditionally executes calls that set errno if (a) the result of |
44 | the call is unused and (b) a simple range check on the arguments can |
45 | detect most cases where errno does not need to be set. |
46 | |
47 | This is the "conditional dead-code elimination" that gave the pass |
48 | its original name, since the call is dead for most argument values. |
49 | The calls for which it helps are usually part of the C++ abstraction |
50 | penalty exposed after inlining. |
51 | |
52 | 2. It looks for calls to built-in functions that set errno and whose |
53 | result is used. It checks whether there is an associated internal |
54 | function that doesn't set errno and whether the target supports |
55 | that internal function. If so, the pass uses the internal function |
56 | to compute the result of the built-in function but still arranges |
57 | for errno to be set when necessary. There are two ways of setting |
58 | errno: |
59 | |
60 | a. by protecting the original call with the same argument checks as (1) |
61 | |
62 | b. by protecting the original call with a check that the result |
63 | of the internal function is not equal to itself (i.e. is NaN). |
64 | |
65 | (b) requires that NaNs are the only erroneous results. It is not |
66 | appropriate for functions like log, which returns ERANGE for zero |
67 | arguments. (b) is also likely to perform worse than (a) because it |
68 | requires the result to be calculated first. The pass therefore uses |
69 | (a) when it can and uses (b) as a fallback. |
70 | |
71 | For (b) the pass can replace the original call with a call to |
72 | IFN_SET_EDOM, if the target supports direct assignments to errno. |
73 | |
74 | In both cases, arguments that require errno to be set should occur |
75 | rarely in practice. Checks of the errno result should also be rare, |
76 | but the compiler would need powerful interprocedural analysis to |
77 | prove that errno is not checked. It's much easier to add argument |
78 | checks or result checks instead. |
79 | |
80 | An example of (1) is: |
81 | |
82 | log (x); // Mostly dead call |
83 | ==> |
84 | if (__builtin_islessequal (x, 0)) |
85 | log (x); |
86 | |
87 | With this change, call to log (x) is effectively eliminated, as |
88 | in the majority of the cases, log won't be called with x out of |
89 | range. The branch is totally predictable, so the branch cost |
90 | is low. |
91 | |
92 | An example of (2) is: |
93 | |
94 | y = sqrt (x); |
95 | ==> |
96 | if (__builtin_isless (x, 0)) |
97 | y = sqrt (x); |
98 | else |
99 | y = IFN_SQRT (x); |
100 | In the vast majority of cases we should then never need to call sqrt. |
101 | |
102 | Note that library functions are not supposed to clear errno to zero without |
103 | error. See IEEE Std 1003.1, section 2.3 Error Numbers, and section 7.5:3 of |
104 | ISO/IEC 9899 (C99). |
105 | |
106 | The condition wrapping the builtin call is conservatively set to avoid too |
107 | aggressive (wrong) shrink wrapping. */ |
108 | |
109 | |
110 | /* A structure for representing input domain of |
111 | a function argument in integer. If the lower |
112 | bound is -inf, has_lb is set to false. If the |
113 | upper bound is +inf, has_ub is false. |
114 | is_lb_inclusive and is_ub_inclusive are flags |
115 | to indicate if lb and ub value are inclusive |
116 | respectively. */ |
117 | |
118 | struct inp_domain |
119 | { |
120 | int lb; |
121 | int ub; |
122 | bool has_lb; |
123 | bool has_ub; |
124 | bool is_lb_inclusive; |
125 | bool is_ub_inclusive; |
126 | }; |
127 | |
128 | /* A helper function to construct and return an input |
129 | domain object. LB is the lower bound, HAS_LB is |
130 | a boolean flag indicating if the lower bound exists, |
131 | and LB_INCLUSIVE is a boolean flag indicating if the |
132 | lower bound is inclusive or not. UB, HAS_UB, and |
133 | UB_INCLUSIVE have the same meaning, but for upper |
134 | bound of the domain. */ |
135 | |
136 | static inp_domain |
137 | get_domain (int lb, bool has_lb, bool lb_inclusive, |
138 | int ub, bool has_ub, bool ub_inclusive) |
139 | { |
140 | inp_domain domain; |
141 | domain.lb = lb; |
142 | domain.has_lb = has_lb; |
143 | domain.is_lb_inclusive = lb_inclusive; |
144 | domain.ub = ub; |
145 | domain.has_ub = has_ub; |
146 | domain.is_ub_inclusive = ub_inclusive; |
147 | return domain; |
148 | } |
149 | |
150 | /* A helper function to check the target format for the |
151 | argument type. In this implementation, only IEEE formats |
152 | are supported. ARG is the call argument to be checked. |
153 | Returns true if the format is supported. To support other |
154 | target formats, function get_no_error_domain needs to be |
155 | enhanced to have range bounds properly computed. Since |
156 | the check is cheap (very small number of candidates |
157 | to be checked), the result is not cached for each float type. */ |
158 | |
159 | static bool |
160 | check_target_format (tree arg) |
161 | { |
162 | tree type; |
163 | machine_mode mode; |
164 | const struct real_format *rfmt; |
165 | |
166 | type = TREE_TYPE (arg); |
167 | mode = TYPE_MODE (type); |
168 | rfmt = REAL_MODE_FORMAT (mode); |
169 | if ((mode == SFmode |
170 | && (rfmt == &ieee_single_format || rfmt == &mips_single_format |
171 | || rfmt == &motorola_single_format)) |
172 | || (mode == DFmode |
173 | && (rfmt == &ieee_double_format || rfmt == &mips_double_format |
174 | || rfmt == &motorola_double_format)) |
175 | /* For long double, we cannot really check XFmode |
176 | which is only defined on intel platforms. |
177 | Candidate pre-selection using builtin function |
178 | code guarantees that we are checking formats |
179 | for long double modes: double, quad, and extended. */ |
180 | || (mode != SFmode && mode != DFmode |
181 | && (rfmt == &ieee_quad_format |
182 | || rfmt == &mips_quad_format |
183 | || rfmt == &ieee_extended_motorola_format |
184 | || rfmt == &ieee_extended_intel_96_format |
185 | || rfmt == &ieee_extended_intel_128_format |
186 | || rfmt == &ieee_extended_intel_96_round_53_format))) |
187 | return true; |
188 | |
189 | return false; |
190 | } |
191 | |
192 | |
193 | /* A helper function to help select calls to pow that are suitable for |
194 | conditional DCE transformation. It looks for pow calls that can be |
195 | guided with simple conditions. Such calls either have constant base |
196 | values or base values converted from integers. Returns true if |
197 | the pow call POW_CALL is a candidate. */ |
198 | |
199 | /* The maximum integer bit size for base argument of a pow call |
200 | that is suitable for shrink-wrapping transformation. */ |
201 | #define MAX_BASE_INT_BIT_SIZE 32 |
202 | |
203 | static bool |
204 | check_pow (gcall *pow_call) |
205 | { |
206 | tree base, expn; |
207 | enum tree_code bc, ec; |
208 | |
209 | if (gimple_call_num_args (gs: pow_call) != 2) |
210 | return false; |
211 | |
212 | base = gimple_call_arg (gs: pow_call, index: 0); |
213 | expn = gimple_call_arg (gs: pow_call, index: 1); |
214 | |
215 | if (!check_target_format (arg: expn)) |
216 | return false; |
217 | |
218 | bc = TREE_CODE (base); |
219 | ec = TREE_CODE (expn); |
220 | |
221 | /* Folding candidates are not interesting. |
222 | Can actually assert that it is already folded. */ |
223 | if (ec == REAL_CST && bc == REAL_CST) |
224 | return false; |
225 | |
226 | if (bc == REAL_CST) |
227 | { |
228 | /* Only handle a fixed range of constant. */ |
229 | REAL_VALUE_TYPE mv; |
230 | REAL_VALUE_TYPE bcv = TREE_REAL_CST (base); |
231 | if (real_equal (&bcv, &dconst1)) |
232 | return false; |
233 | if (real_less (&bcv, &dconst1)) |
234 | return false; |
235 | real_from_integer (&mv, TYPE_MODE (TREE_TYPE (base)), 256, UNSIGNED); |
236 | if (real_less (&mv, &bcv)) |
237 | return false; |
238 | return true; |
239 | } |
240 | else if (bc == SSA_NAME) |
241 | { |
242 | tree base_val0, type; |
243 | gimple *base_def; |
244 | int bit_sz; |
245 | |
246 | /* Only handles cases where base value is converted |
247 | from integer values. */ |
248 | base_def = SSA_NAME_DEF_STMT (base); |
249 | if (gimple_code (g: base_def) != GIMPLE_ASSIGN) |
250 | return false; |
251 | |
252 | if (gimple_assign_rhs_code (gs: base_def) != FLOAT_EXPR) |
253 | return false; |
254 | base_val0 = gimple_assign_rhs1 (gs: base_def); |
255 | |
256 | type = TREE_TYPE (base_val0); |
257 | if (TREE_CODE (type) != INTEGER_TYPE) |
258 | return false; |
259 | bit_sz = TYPE_PRECISION (type); |
260 | /* If the type of the base is too wide, |
261 | the resulting shrink wrapping condition |
262 | will be too conservative. */ |
263 | if (bit_sz > MAX_BASE_INT_BIT_SIZE) |
264 | return false; |
265 | |
266 | return true; |
267 | } |
268 | else |
269 | return false; |
270 | } |
271 | |
272 | /* A helper function to help select candidate function calls that are |
273 | suitable for conditional DCE. Candidate functions must have single |
274 | valid input domain in this implementation except for pow (see check_pow). |
275 | Returns true if the function call is a candidate. */ |
276 | |
277 | static bool |
278 | check_builtin_call (gcall *bcall) |
279 | { |
280 | tree arg; |
281 | |
282 | arg = gimple_call_arg (gs: bcall, index: 0); |
283 | return check_target_format (arg); |
284 | } |
285 | |
286 | /* Return true if built-in function call CALL calls a math function |
287 | and if we know how to test the range of its arguments to detect _most_ |
288 | situations in which errno is not set. The test must err on the side |
289 | of treating non-erroneous values as potentially erroneous. */ |
290 | |
291 | static bool |
292 | can_test_argument_range (gcall *call) |
293 | { |
294 | switch (DECL_FUNCTION_CODE (decl: gimple_call_fndecl (gs: call))) |
295 | { |
296 | /* Trig functions. */ |
297 | CASE_FLT_FN (BUILT_IN_ACOS): |
298 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ACOS): |
299 | CASE_FLT_FN (BUILT_IN_ASIN): |
300 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ASIN): |
301 | /* Hyperbolic functions. */ |
302 | CASE_FLT_FN (BUILT_IN_ACOSH): |
303 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ACOSH): |
304 | CASE_FLT_FN (BUILT_IN_ATANH): |
305 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ATANH): |
306 | CASE_FLT_FN (BUILT_IN_COSH): |
307 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_COSH): |
308 | CASE_FLT_FN (BUILT_IN_SINH): |
309 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_SINH): |
310 | /* Log functions. */ |
311 | CASE_FLT_FN (BUILT_IN_LOG): |
312 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG): |
313 | CASE_FLT_FN (BUILT_IN_LOG2): |
314 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG2): |
315 | CASE_FLT_FN (BUILT_IN_LOG10): |
316 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG10): |
317 | CASE_FLT_FN (BUILT_IN_LOG1P): |
318 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG1P): |
319 | /* Exp functions. */ |
320 | CASE_FLT_FN (BUILT_IN_EXP): |
321 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_EXP): |
322 | CASE_FLT_FN (BUILT_IN_EXP2): |
323 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_EXP2): |
324 | CASE_FLT_FN (BUILT_IN_EXP10): |
325 | CASE_FLT_FN (BUILT_IN_EXPM1): |
326 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_EXPM1): |
327 | CASE_FLT_FN (BUILT_IN_POW10): |
328 | /* Sqrt. */ |
329 | CASE_FLT_FN (BUILT_IN_SQRT): |
330 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_SQRT): |
331 | return check_builtin_call (bcall: call); |
332 | /* Special one: two argument pow. */ |
333 | case BUILT_IN_POW: |
334 | return check_pow (pow_call: call); |
335 | default: |
336 | break; |
337 | } |
338 | |
339 | return false; |
340 | } |
341 | |
342 | /* Return true if CALL can produce a domain error (EDOM) but can never |
343 | produce a pole, range overflow or range underflow error (all ERANGE). |
344 | This means that we can tell whether a function would have set errno |
345 | by testing whether the result is a NaN. */ |
346 | |
347 | static bool |
348 | edom_only_function (gcall *call) |
349 | { |
350 | switch (DECL_FUNCTION_CODE (decl: gimple_call_fndecl (gs: call))) |
351 | { |
352 | CASE_FLT_FN (BUILT_IN_ACOS): |
353 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ACOS): |
354 | CASE_FLT_FN (BUILT_IN_ASIN): |
355 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ASIN): |
356 | CASE_FLT_FN (BUILT_IN_ATAN): |
357 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ATAN): |
358 | CASE_FLT_FN (BUILT_IN_COS): |
359 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_COS): |
360 | CASE_FLT_FN (BUILT_IN_SIGNIFICAND): |
361 | CASE_FLT_FN (BUILT_IN_SIN): |
362 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_SIN): |
363 | CASE_FLT_FN (BUILT_IN_SQRT): |
364 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_SQRT): |
365 | CASE_FLT_FN (BUILT_IN_FMOD): |
366 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_FMOD): |
367 | CASE_FLT_FN (BUILT_IN_REMAINDER): |
368 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_REMAINDER): |
369 | return true; |
370 | |
371 | default: |
372 | return false; |
373 | } |
374 | } |
375 | |
376 | /* Return true if it is structurally possible to guard CALL. */ |
377 | |
378 | static bool |
379 | can_guard_call_p (gimple *call) |
380 | { |
381 | return (!stmt_ends_bb_p (call) |
382 | || find_fallthru_edge (edges: gimple_bb (g: call)->succs)); |
383 | } |
384 | |
385 | /* For a comparison code return the comparison code we should use if we don't |
386 | HONOR_NANS. */ |
387 | |
388 | static enum tree_code |
389 | comparison_code_if_no_nans (tree_code code) |
390 | { |
391 | switch (code) |
392 | { |
393 | case UNLT_EXPR: |
394 | return LT_EXPR; |
395 | case UNGT_EXPR: |
396 | return GT_EXPR; |
397 | case UNLE_EXPR: |
398 | return LE_EXPR; |
399 | case UNGE_EXPR: |
400 | return GE_EXPR; |
401 | case UNEQ_EXPR: |
402 | return EQ_EXPR; |
403 | case LTGT_EXPR: |
404 | return NE_EXPR; |
405 | |
406 | case LT_EXPR: |
407 | case GT_EXPR: |
408 | case LE_EXPR: |
409 | case GE_EXPR: |
410 | case EQ_EXPR: |
411 | case NE_EXPR: |
412 | return code; |
413 | |
414 | default: |
415 | gcc_unreachable (); |
416 | } |
417 | } |
418 | |
419 | /* A helper function to generate gimple statements for one bound |
420 | comparison, so that the built-in function is called whenever |
421 | TCODE <ARG, LBUB> is *false*. TEMP_NAME1/TEMP_NAME2 are names |
422 | of the temporaries, CONDS is a vector holding the produced GIMPLE |
423 | statements, and NCONDS points to the variable holding the number of |
424 | logical comparisons. CONDS is either empty or a list ended with a |
425 | null tree. */ |
426 | |
427 | static void |
428 | gen_one_condition (tree arg, int lbub, |
429 | enum tree_code tcode, |
430 | const char *temp_name1, |
431 | const char *temp_name2, |
432 | vec<gimple *> conds, |
433 | unsigned *nconds) |
434 | { |
435 | if (!HONOR_NANS (arg)) |
436 | tcode = comparison_code_if_no_nans (code: tcode); |
437 | |
438 | tree lbub_real_cst, lbub_cst, float_type; |
439 | tree temp, tempn, tempc, tempcn; |
440 | gassign *stmt1; |
441 | gassign *stmt2; |
442 | gcond *stmt3; |
443 | |
444 | float_type = TREE_TYPE (arg); |
445 | lbub_cst = build_int_cst (integer_type_node, lbub); |
446 | lbub_real_cst = build_real_from_int_cst (float_type, lbub_cst); |
447 | |
448 | temp = create_tmp_var (float_type, temp_name1); |
449 | stmt1 = gimple_build_assign (temp, arg); |
450 | tempn = make_ssa_name (var: temp, stmt: stmt1); |
451 | gimple_assign_set_lhs (gs: stmt1, lhs: tempn); |
452 | |
453 | tempc = create_tmp_var (boolean_type_node, temp_name2); |
454 | stmt2 = gimple_build_assign (tempc, |
455 | fold_build2 (tcode, |
456 | boolean_type_node, |
457 | tempn, lbub_real_cst)); |
458 | tempcn = make_ssa_name (var: tempc, stmt: stmt2); |
459 | gimple_assign_set_lhs (gs: stmt2, lhs: tempcn); |
460 | |
461 | stmt3 = gimple_build_cond_from_tree (tempcn, NULL_TREE, NULL_TREE); |
462 | conds.quick_push (obj: stmt1); |
463 | conds.quick_push (obj: stmt2); |
464 | conds.quick_push (obj: stmt3); |
465 | (*nconds)++; |
466 | } |
467 | |
468 | /* A helper function to generate GIMPLE statements for |
469 | out of input domain check. ARG is the call argument |
470 | to be runtime checked, DOMAIN holds the valid domain |
471 | for the given function, CONDS points to the vector |
472 | holding the result GIMPLE statements. *NCONDS is |
473 | the number of logical comparisons. This function |
474 | produces no more than two logical comparisons, one |
475 | for lower bound check, one for upper bound check. */ |
476 | |
477 | static void |
478 | gen_conditions_for_domain (tree arg, inp_domain domain, |
479 | vec<gimple *> conds, |
480 | unsigned *nconds) |
481 | { |
482 | if (domain.has_lb) |
483 | gen_one_condition (arg, lbub: domain.lb, |
484 | tcode: (domain.is_lb_inclusive |
485 | ? UNGE_EXPR : UNGT_EXPR), |
486 | temp_name1: "DCE_COND_LB" , temp_name2: "DCE_COND_LB_TEST" , |
487 | conds, nconds); |
488 | |
489 | if (domain.has_ub) |
490 | { |
491 | /* Now push a separator. */ |
492 | if (domain.has_lb) |
493 | conds.quick_push (NULL); |
494 | |
495 | gen_one_condition (arg, lbub: domain.ub, |
496 | tcode: (domain.is_ub_inclusive |
497 | ? UNLE_EXPR : UNLT_EXPR), |
498 | temp_name1: "DCE_COND_UB" , temp_name2: "DCE_COND_UB_TEST" , |
499 | conds, nconds); |
500 | } |
501 | } |
502 | |
503 | |
504 | /* A helper function to generate condition |
505 | code for the y argument in call pow (some_const, y). |
506 | See candidate selection in check_pow. Since the |
507 | candidates' base values have a limited range, |
508 | the guarded code generated for y are simple: |
509 | if (__builtin_isgreater (y, max_y)) |
510 | pow (const, y); |
511 | Note max_y can be computed separately for each |
512 | const base, but in this implementation, we |
513 | choose to compute it using the max base |
514 | in the allowed range for the purpose of |
515 | simplicity. BASE is the constant base value, |
516 | EXPN is the expression for the exponent argument, |
517 | *CONDS is the vector to hold resulting statements, |
518 | and *NCONDS is the number of logical conditions. */ |
519 | |
520 | static void |
521 | gen_conditions_for_pow_cst_base (tree base, tree expn, |
522 | vec<gimple *> conds, |
523 | unsigned *nconds) |
524 | { |
525 | inp_domain exp_domain; |
526 | /* Validate the range of the base constant to make |
527 | sure it is consistent with check_pow. */ |
528 | REAL_VALUE_TYPE mv; |
529 | REAL_VALUE_TYPE bcv = TREE_REAL_CST (base); |
530 | gcc_assert (!real_equal (&bcv, &dconst1) |
531 | && !real_less (&bcv, &dconst1)); |
532 | real_from_integer (&mv, TYPE_MODE (TREE_TYPE (base)), 256, UNSIGNED); |
533 | gcc_assert (!real_less (&mv, &bcv)); |
534 | |
535 | exp_domain = get_domain (lb: 0, has_lb: false, lb_inclusive: false, |
536 | ub: 127, has_ub: true, ub_inclusive: false); |
537 | |
538 | gen_conditions_for_domain (arg: expn, domain: exp_domain, |
539 | conds, nconds); |
540 | } |
541 | |
542 | /* Generate error condition code for pow calls with |
543 | non constant base values. The candidates selected |
544 | have their base argument value converted from |
545 | integer (see check_pow) value (1, 2, 4 bytes), and |
546 | the max exp value is computed based on the size |
547 | of the integer type (i.e. max possible base value). |
548 | The resulting input domain for exp argument is thus |
549 | conservative (smaller than the max value allowed by |
550 | the runtime value of the base). BASE is the integer |
551 | base value, EXPN is the expression for the exponent |
552 | argument, *CONDS is the vector to hold resulting |
553 | statements, and *NCONDS is the number of logical |
554 | conditions. */ |
555 | |
556 | static void |
557 | gen_conditions_for_pow_int_base (tree base, tree expn, |
558 | vec<gimple *> conds, |
559 | unsigned *nconds) |
560 | { |
561 | gimple *base_def; |
562 | tree base_val0; |
563 | tree int_type; |
564 | tree temp, tempn; |
565 | tree cst0; |
566 | gimple *stmt1, *stmt2; |
567 | int bit_sz, max_exp; |
568 | inp_domain exp_domain; |
569 | |
570 | base_def = SSA_NAME_DEF_STMT (base); |
571 | base_val0 = gimple_assign_rhs1 (gs: base_def); |
572 | int_type = TREE_TYPE (base_val0); |
573 | bit_sz = TYPE_PRECISION (int_type); |
574 | gcc_assert (bit_sz > 0 |
575 | && bit_sz <= MAX_BASE_INT_BIT_SIZE); |
576 | |
577 | /* Determine the max exp argument value according to |
578 | the size of the base integer. The max exp value |
579 | is conservatively estimated assuming IEEE754 double |
580 | precision format. */ |
581 | if (bit_sz == 8) |
582 | max_exp = 128; |
583 | else if (bit_sz == 16) |
584 | max_exp = 64; |
585 | else |
586 | { |
587 | gcc_assert (bit_sz == MAX_BASE_INT_BIT_SIZE); |
588 | max_exp = 32; |
589 | } |
590 | |
591 | /* For pow ((double)x, y), generate the following conditions: |
592 | cond 1: |
593 | temp1 = x; |
594 | if (__builtin_islessequal (temp1, 0)) |
595 | |
596 | cond 2: |
597 | temp2 = y; |
598 | if (__builtin_isgreater (temp2, max_exp_real_cst)) */ |
599 | |
600 | /* Generate condition in reverse order -- first |
601 | the condition for the exp argument. */ |
602 | |
603 | exp_domain = get_domain (lb: 0, has_lb: false, lb_inclusive: false, |
604 | ub: max_exp, has_ub: true, ub_inclusive: true); |
605 | |
606 | gen_conditions_for_domain (arg: expn, domain: exp_domain, |
607 | conds, nconds); |
608 | |
609 | /* Now generate condition for the base argument. |
610 | Note it does not use the helper function |
611 | gen_conditions_for_domain because the base |
612 | type is integer. */ |
613 | |
614 | /* Push a separator. */ |
615 | conds.quick_push (NULL); |
616 | |
617 | temp = create_tmp_var (int_type, "DCE_COND1" ); |
618 | cst0 = build_int_cst (int_type, 0); |
619 | stmt1 = gimple_build_assign (temp, base_val0); |
620 | tempn = make_ssa_name (var: temp, stmt: stmt1); |
621 | gimple_assign_set_lhs (gs: stmt1, lhs: tempn); |
622 | stmt2 = gimple_build_cond (GT_EXPR, tempn, cst0, NULL_TREE, NULL_TREE); |
623 | |
624 | conds.quick_push (obj: stmt1); |
625 | conds.quick_push (obj: stmt2); |
626 | (*nconds)++; |
627 | } |
628 | |
629 | /* Method to generate conditional statements for guarding conditionally |
630 | dead calls to pow. One or more statements can be generated for |
631 | each logical condition. Statement groups of different conditions |
632 | are separated by a NULL tree and they are stored in the vec |
633 | conds. The number of logical conditions are stored in *nconds. |
634 | |
635 | See C99 standard, 7.12.7.4:2, for description of pow (x, y). |
636 | The precise condition for domain errors are complex. In this |
637 | implementation, a simplified (but conservative) valid domain |
638 | for x and y are used: x is positive to avoid dom errors, while |
639 | y is smaller than a upper bound (depending on x) to avoid range |
640 | errors. Runtime code is generated to check x (if not constant) |
641 | and y against the valid domain. If it is out, jump to the call, |
642 | otherwise the call is bypassed. POW_CALL is the call statement, |
643 | *CONDS is a vector holding the resulting condition statements, |
644 | and *NCONDS is the number of logical conditions. */ |
645 | |
646 | static void |
647 | gen_conditions_for_pow (gcall *pow_call, vec<gimple *> conds, |
648 | unsigned *nconds) |
649 | { |
650 | tree base, expn; |
651 | enum tree_code bc; |
652 | |
653 | gcc_checking_assert (check_pow (pow_call)); |
654 | |
655 | *nconds = 0; |
656 | |
657 | base = gimple_call_arg (gs: pow_call, index: 0); |
658 | expn = gimple_call_arg (gs: pow_call, index: 1); |
659 | |
660 | bc = TREE_CODE (base); |
661 | |
662 | if (bc == REAL_CST) |
663 | gen_conditions_for_pow_cst_base (base, expn, conds, nconds); |
664 | else if (bc == SSA_NAME) |
665 | gen_conditions_for_pow_int_base (base, expn, conds, nconds); |
666 | else |
667 | gcc_unreachable (); |
668 | } |
669 | |
670 | /* A helper routine to help computing the valid input domain |
671 | for a builtin function. See C99 7.12.7 for details. In this |
672 | implementation, we only handle single region domain. The |
673 | resulting region can be conservative (smaller) than the actual |
674 | one and rounded to integers. Some of the bounds are documented |
675 | in the standard, while other limit constants are computed |
676 | assuming IEEE floating point format (for SF and DF modes). |
677 | Since IEEE only sets minimum requirements for long double format, |
678 | different long double formats exist under different implementations |
679 | (e.g, 64 bit double precision (DF), 80 bit double-extended |
680 | precision (XF), and 128 bit quad precision (TF) ). For simplicity, |
681 | in this implementation, the computed bounds for long double assume |
682 | 64 bit format (DF) except when it is IEEE quad or extended with the same |
683 | emax, and are therefore sometimes conservative. Another assumption is |
684 | that single precision float type is always SF mode, and double type is DF |
685 | mode. This function is quite implementation specific, so it may not be |
686 | suitable to be part of builtins.cc. This needs to be revisited later |
687 | to see if it can be leveraged in x87 assembly expansion. */ |
688 | |
689 | static inp_domain |
690 | get_no_error_domain (enum built_in_function fnc) |
691 | { |
692 | switch (fnc) |
693 | { |
694 | /* Trig functions: return [-1, +1] */ |
695 | CASE_FLT_FN (BUILT_IN_ACOS): |
696 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ACOS): |
697 | CASE_FLT_FN (BUILT_IN_ASIN): |
698 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ASIN): |
699 | return get_domain (lb: -1, has_lb: true, lb_inclusive: true, |
700 | ub: 1, has_ub: true, ub_inclusive: true); |
701 | /* Hyperbolic functions. */ |
702 | CASE_FLT_FN (BUILT_IN_ACOSH): |
703 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ACOSH): |
704 | /* acosh: [1, +inf) */ |
705 | return get_domain (lb: 1, has_lb: true, lb_inclusive: true, |
706 | ub: 1, has_ub: false, ub_inclusive: false); |
707 | CASE_FLT_FN (BUILT_IN_ATANH): |
708 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_ATANH): |
709 | /* atanh: (-1, +1) */ |
710 | return get_domain (lb: -1, has_lb: true, lb_inclusive: false, |
711 | ub: 1, has_ub: true, ub_inclusive: false); |
712 | case BUILT_IN_COSHF16: |
713 | case BUILT_IN_SINHF16: |
714 | /* coshf16: (-11, +11) */ |
715 | return get_domain (lb: -11, has_lb: true, lb_inclusive: false, |
716 | ub: 11, has_ub: true, ub_inclusive: false); |
717 | case BUILT_IN_COSHF: |
718 | case BUILT_IN_SINHF: |
719 | case BUILT_IN_COSHF32: |
720 | case BUILT_IN_SINHF32: |
721 | /* coshf: (-89, +89) */ |
722 | return get_domain (lb: -89, has_lb: true, lb_inclusive: false, |
723 | ub: 89, has_ub: true, ub_inclusive: false); |
724 | case BUILT_IN_COSH: |
725 | case BUILT_IN_SINH: |
726 | case BUILT_IN_COSHF64: |
727 | case BUILT_IN_SINHF64: |
728 | case BUILT_IN_COSHF32X: |
729 | case BUILT_IN_SINHF32X: |
730 | /* cosh: (-710, +710) */ |
731 | return get_domain (lb: -710, has_lb: true, lb_inclusive: false, |
732 | ub: 710, has_ub: true, ub_inclusive: false); |
733 | case BUILT_IN_COSHF128: |
734 | case BUILT_IN_SINHF128: |
735 | /* coshf128: (-11357, +11357) */ |
736 | return get_domain (lb: -11357, has_lb: true, lb_inclusive: false, |
737 | ub: 11357, has_ub: true, ub_inclusive: false); |
738 | case BUILT_IN_COSHL: |
739 | case BUILT_IN_SINHL: |
740 | if (REAL_MODE_FORMAT (TYPE_MODE (long_double_type_node))->emax == 16384) |
741 | return get_no_error_domain (fnc: BUILT_IN_COSHF128); |
742 | return get_no_error_domain (fnc: BUILT_IN_COSH); |
743 | case BUILT_IN_COSHF64X: |
744 | case BUILT_IN_SINHF64X: |
745 | if (REAL_MODE_FORMAT (TYPE_MODE (float64x_type_node))->emax == 16384) |
746 | return get_no_error_domain (fnc: BUILT_IN_COSHF128); |
747 | return get_no_error_domain (fnc: BUILT_IN_COSH); |
748 | /* Log functions: (0, +inf) */ |
749 | CASE_FLT_FN (BUILT_IN_LOG): |
750 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG): |
751 | CASE_FLT_FN (BUILT_IN_LOG2): |
752 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG2): |
753 | CASE_FLT_FN (BUILT_IN_LOG10): |
754 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG10): |
755 | return get_domain (lb: 0, has_lb: true, lb_inclusive: false, |
756 | ub: 0, has_ub: false, ub_inclusive: false); |
757 | CASE_FLT_FN (BUILT_IN_LOG1P): |
758 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_LOG1P): |
759 | return get_domain (lb: -1, has_lb: true, lb_inclusive: false, |
760 | ub: 0, has_ub: false, ub_inclusive: false); |
761 | /* Exp functions. */ |
762 | case BUILT_IN_EXPF16: |
763 | case BUILT_IN_EXPM1F16: |
764 | /* expf16: (-inf, 11) */ |
765 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
766 | ub: 11, has_ub: true, ub_inclusive: false); |
767 | case BUILT_IN_EXPF: |
768 | case BUILT_IN_EXPM1F: |
769 | case BUILT_IN_EXPF32: |
770 | case BUILT_IN_EXPM1F32: |
771 | /* expf: (-inf, 88) */ |
772 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
773 | ub: 88, has_ub: true, ub_inclusive: false); |
774 | case BUILT_IN_EXP: |
775 | case BUILT_IN_EXPM1: |
776 | case BUILT_IN_EXPF64: |
777 | case BUILT_IN_EXPM1F64: |
778 | case BUILT_IN_EXPF32X: |
779 | case BUILT_IN_EXPM1F32X: |
780 | /* exp: (-inf, 709) */ |
781 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
782 | ub: 709, has_ub: true, ub_inclusive: false); |
783 | case BUILT_IN_EXPF128: |
784 | case BUILT_IN_EXPM1F128: |
785 | /* expf128: (-inf, 11356) */ |
786 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
787 | ub: 11356, has_ub: true, ub_inclusive: false); |
788 | case BUILT_IN_EXPL: |
789 | case BUILT_IN_EXPM1L: |
790 | if (REAL_MODE_FORMAT (TYPE_MODE (long_double_type_node))->emax == 16384) |
791 | return get_no_error_domain (fnc: BUILT_IN_EXPF128); |
792 | return get_no_error_domain (fnc: BUILT_IN_EXP); |
793 | case BUILT_IN_EXPF64X: |
794 | case BUILT_IN_EXPM1F64X: |
795 | if (REAL_MODE_FORMAT (TYPE_MODE (float64x_type_node))->emax == 16384) |
796 | return get_no_error_domain (fnc: BUILT_IN_EXPF128); |
797 | return get_no_error_domain (fnc: BUILT_IN_EXP); |
798 | case BUILT_IN_EXP2F16: |
799 | /* exp2f16: (-inf, 16) */ |
800 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
801 | ub: 16, has_ub: true, ub_inclusive: false); |
802 | case BUILT_IN_EXP2F: |
803 | case BUILT_IN_EXP2F32: |
804 | /* exp2f: (-inf, 128) */ |
805 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
806 | ub: 128, has_ub: true, ub_inclusive: false); |
807 | case BUILT_IN_EXP2: |
808 | case BUILT_IN_EXP2F64: |
809 | case BUILT_IN_EXP2F32X: |
810 | /* exp2: (-inf, 1024) */ |
811 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
812 | ub: 1024, has_ub: true, ub_inclusive: false); |
813 | case BUILT_IN_EXP2F128: |
814 | /* exp2f128: (-inf, 16384) */ |
815 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
816 | ub: 16384, has_ub: true, ub_inclusive: false); |
817 | case BUILT_IN_EXP2L: |
818 | if (REAL_MODE_FORMAT (TYPE_MODE (long_double_type_node))->emax == 16384) |
819 | return get_no_error_domain (fnc: BUILT_IN_EXP2F128); |
820 | return get_no_error_domain (fnc: BUILT_IN_EXP2); |
821 | case BUILT_IN_EXP2F64X: |
822 | if (REAL_MODE_FORMAT (TYPE_MODE (float64x_type_node))->emax == 16384) |
823 | return get_no_error_domain (fnc: BUILT_IN_EXP2F128); |
824 | return get_no_error_domain (fnc: BUILT_IN_EXP2); |
825 | case BUILT_IN_EXP10F: |
826 | case BUILT_IN_POW10F: |
827 | /* exp10f: (-inf, 38) */ |
828 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
829 | ub: 38, has_ub: true, ub_inclusive: false); |
830 | case BUILT_IN_EXP10: |
831 | case BUILT_IN_POW10: |
832 | /* exp10: (-inf, 308) */ |
833 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
834 | ub: 308, has_ub: true, ub_inclusive: false); |
835 | case BUILT_IN_EXP10L: |
836 | case BUILT_IN_POW10L: |
837 | if (REAL_MODE_FORMAT (TYPE_MODE (long_double_type_node))->emax == 16384) |
838 | /* exp10l: (-inf, 4932) */ |
839 | return get_domain (lb: -1, has_lb: false, lb_inclusive: false, |
840 | ub: 4932, has_ub: true, ub_inclusive: false); |
841 | return get_no_error_domain (fnc: BUILT_IN_EXP10); |
842 | /* sqrt: [0, +inf) */ |
843 | CASE_FLT_FN (BUILT_IN_SQRT): |
844 | CASE_FLT_FN_FLOATN_NX (BUILT_IN_SQRT): |
845 | return get_domain (lb: 0, has_lb: true, lb_inclusive: true, |
846 | ub: 0, has_ub: false, ub_inclusive: false); |
847 | default: |
848 | gcc_unreachable (); |
849 | } |
850 | |
851 | gcc_unreachable (); |
852 | } |
853 | |
854 | /* The function to generate shrink wrap conditions for a partially |
855 | dead builtin call whose return value is not used anywhere, |
856 | but has to be kept live due to potential error condition. |
857 | BI_CALL is the builtin call, CONDS is the vector of statements |
858 | for condition code, NCODES is the pointer to the number of |
859 | logical conditions. Statements belonging to different logical |
860 | condition are separated by NULL tree in the vector. */ |
861 | |
862 | static void |
863 | gen_shrink_wrap_conditions (gcall *bi_call, const vec<gimple *> &conds, |
864 | unsigned int *nconds) |
865 | { |
866 | gcall *call; |
867 | tree fn; |
868 | enum built_in_function fnc; |
869 | |
870 | gcc_assert (nconds && conds.exists ()); |
871 | gcc_assert (conds.length () == 0); |
872 | gcc_assert (is_gimple_call (bi_call)); |
873 | |
874 | call = bi_call; |
875 | fn = gimple_call_fndecl (gs: call); |
876 | gcc_assert (fn && fndecl_built_in_p (fn)); |
877 | fnc = DECL_FUNCTION_CODE (decl: fn); |
878 | *nconds = 0; |
879 | |
880 | if (fnc == BUILT_IN_POW) |
881 | gen_conditions_for_pow (pow_call: call, conds, nconds); |
882 | else |
883 | { |
884 | tree arg; |
885 | inp_domain domain = get_no_error_domain (fnc); |
886 | *nconds = 0; |
887 | arg = gimple_call_arg (gs: bi_call, index: 0); |
888 | gen_conditions_for_domain (arg, domain, conds, nconds); |
889 | } |
890 | |
891 | return; |
892 | } |
893 | |
894 | /* Shrink-wrap BI_CALL so that it is only called when one of the NCONDS |
895 | conditions in CONDS is false. Also move BI_NEWCALL to a new basic block |
896 | when it is non-null, it is called while all of the CONDS are true. */ |
897 | |
898 | static void |
899 | shrink_wrap_one_built_in_call_with_conds (gcall *bi_call, |
900 | const vec <gimple *> &conds, |
901 | unsigned int nconds, |
902 | gcall *bi_newcall = NULL) |
903 | { |
904 | gimple_stmt_iterator bi_call_bsi; |
905 | basic_block bi_call_bb, bi_newcall_bb, join_tgt_bb, guard_bb; |
906 | edge join_tgt_in_edge_from_call, join_tgt_in_edge_fall_thru; |
907 | edge bi_call_in_edge0, guard_bb_in_edge; |
908 | unsigned tn_cond_stmts; |
909 | unsigned ci; |
910 | gimple *cond_expr = NULL; |
911 | gimple *cond_expr_start; |
912 | |
913 | /* The cfg we want to create looks like this: |
914 | [guard n-1] <- guard_bb (old block) |
915 | | \ |
916 | | [guard n-2] } |
917 | | / \ } |
918 | | / ... } new blocks |
919 | | / [guard 0] } |
920 | | / / | } |
921 | [call] | <- bi_call_bb } |
922 | \ [newcall] <-bi_newcall_bb} |
923 | \ | |
924 | [join] <- join_tgt_bb (old iff call must end bb) |
925 | possible EH edges (only if [join] is old) |
926 | |
927 | When [join] is new, the immediate dominators for these blocks are: |
928 | |
929 | 1. [guard n-1]: unchanged |
930 | 2. [call]: [guard n-1] |
931 | 3. [newcall]: [guard 0] |
932 | 4. [guard m]: [guard m+1] for 0 <= m <= n-2 |
933 | 5. [join]: [guard n-1] |
934 | |
935 | We punt for the more complex case of [join] being old and |
936 | simply free the dominance info. We also punt on postdominators, |
937 | which aren't expected to be available at this point anyway. */ |
938 | bi_call_bb = gimple_bb (g: bi_call); |
939 | |
940 | /* Now find the join target bb -- split bi_call_bb if needed. */ |
941 | if (stmt_ends_bb_p (bi_call)) |
942 | { |
943 | /* We checked that there was a fallthrough edge in |
944 | can_guard_call_p. */ |
945 | join_tgt_in_edge_from_call = find_fallthru_edge (edges: bi_call_bb->succs); |
946 | gcc_assert (join_tgt_in_edge_from_call); |
947 | /* We don't want to handle PHIs. */ |
948 | if (EDGE_COUNT (join_tgt_in_edge_from_call->dest->preds) > 1) |
949 | join_tgt_bb = split_edge (join_tgt_in_edge_from_call); |
950 | else |
951 | { |
952 | join_tgt_bb = join_tgt_in_edge_from_call->dest; |
953 | /* We may have degenerate PHIs in the destination. Propagate |
954 | those out. */ |
955 | for (gphi_iterator i = gsi_start_phis (join_tgt_bb); !gsi_end_p (i);) |
956 | { |
957 | gphi *phi = i.phi (); |
958 | replace_uses_by (gimple_phi_result (gs: phi), |
959 | gimple_phi_arg_def (gs: phi, index: 0)); |
960 | remove_phi_node (&i, true); |
961 | } |
962 | } |
963 | } |
964 | else |
965 | { |
966 | join_tgt_in_edge_from_call = split_block (bi_call_bb, bi_call); |
967 | join_tgt_bb = join_tgt_in_edge_from_call->dest; |
968 | } |
969 | |
970 | bi_call_bsi = gsi_for_stmt (bi_call); |
971 | |
972 | /* Now it is time to insert the first conditional expression |
973 | into bi_call_bb and split this bb so that bi_call is |
974 | shrink-wrapped. */ |
975 | tn_cond_stmts = conds.length (); |
976 | cond_expr = NULL; |
977 | cond_expr_start = conds[0]; |
978 | for (ci = 0; ci < tn_cond_stmts; ci++) |
979 | { |
980 | gimple *c = conds[ci]; |
981 | gcc_assert (c || ci != 0); |
982 | if (!c) |
983 | break; |
984 | gsi_insert_before (&bi_call_bsi, c, GSI_SAME_STMT); |
985 | cond_expr = c; |
986 | } |
987 | ci++; |
988 | gcc_assert (cond_expr && gimple_code (cond_expr) == GIMPLE_COND); |
989 | |
990 | typedef std::pair<edge, edge> edge_pair; |
991 | auto_vec<edge_pair, 8> edges; |
992 | |
993 | bi_call_in_edge0 = split_block (bi_call_bb, cond_expr); |
994 | bi_call_in_edge0->flags &= ~EDGE_FALLTHRU; |
995 | bi_call_in_edge0->flags |= EDGE_FALSE_VALUE; |
996 | guard_bb = bi_call_bb; |
997 | bi_call_bb = bi_call_in_edge0->dest; |
998 | join_tgt_in_edge_fall_thru = make_edge (guard_bb, join_tgt_bb, |
999 | EDGE_TRUE_VALUE); |
1000 | |
1001 | edges.reserve (nelems: nconds); |
1002 | edges.quick_push (obj: edge_pair (bi_call_in_edge0, join_tgt_in_edge_fall_thru)); |
1003 | |
1004 | /* Code generation for the rest of the conditions */ |
1005 | for (unsigned int i = 1; i < nconds; ++i) |
1006 | { |
1007 | unsigned ci0; |
1008 | edge bi_call_in_edge; |
1009 | gimple_stmt_iterator guard_bsi = gsi_for_stmt (cond_expr_start); |
1010 | ci0 = ci; |
1011 | cond_expr_start = conds[ci0]; |
1012 | for (; ci < tn_cond_stmts; ci++) |
1013 | { |
1014 | gimple *c = conds[ci]; |
1015 | gcc_assert (c || ci != ci0); |
1016 | if (!c) |
1017 | break; |
1018 | gsi_insert_before (&guard_bsi, c, GSI_SAME_STMT); |
1019 | cond_expr = c; |
1020 | } |
1021 | ci++; |
1022 | gcc_assert (cond_expr && gimple_code (cond_expr) == GIMPLE_COND); |
1023 | guard_bb_in_edge = split_block (guard_bb, cond_expr); |
1024 | guard_bb_in_edge->flags &= ~EDGE_FALLTHRU; |
1025 | guard_bb_in_edge->flags |= EDGE_TRUE_VALUE; |
1026 | |
1027 | bi_call_in_edge = make_edge (guard_bb, bi_call_bb, EDGE_FALSE_VALUE); |
1028 | edges.quick_push (obj: edge_pair (bi_call_in_edge, guard_bb_in_edge)); |
1029 | } |
1030 | |
1031 | /* Move BI_NEWCALL to new basic block when it is non-null. */ |
1032 | if (bi_newcall) |
1033 | { |
1034 | /* Get bi_newcall_bb by split join_tgt_in_edge_fall_thru edge, |
1035 | and move BI_NEWCALL to bi_newcall_bb. */ |
1036 | bi_newcall_bb = split_edge (join_tgt_in_edge_fall_thru); |
1037 | gimple_stmt_iterator to_gsi = gsi_start_bb (bb: bi_newcall_bb); |
1038 | gimple_stmt_iterator from_gsi = gsi_for_stmt (bi_newcall); |
1039 | gsi_move_before (&from_gsi, &to_gsi); |
1040 | join_tgt_in_edge_fall_thru = EDGE_SUCC (bi_newcall_bb, 0); |
1041 | join_tgt_bb = join_tgt_in_edge_fall_thru->dest; |
1042 | |
1043 | tree bi_newcall_lhs = gimple_call_lhs (gs: bi_newcall); |
1044 | tree bi_call_lhs = gimple_call_lhs (gs: bi_call); |
1045 | if (!bi_call_lhs) |
1046 | { |
1047 | bi_call_lhs = copy_ssa_name (var: bi_newcall_lhs); |
1048 | gimple_call_set_lhs (gs: bi_call, lhs: bi_call_lhs); |
1049 | SSA_NAME_DEF_STMT (bi_call_lhs) = bi_call; |
1050 | } |
1051 | |
1052 | /* Create phi node for lhs of BI_CALL and BI_NEWCALL. */ |
1053 | gphi *new_phi = create_phi_node (copy_ssa_name (var: bi_newcall_lhs), |
1054 | join_tgt_bb); |
1055 | SSA_NAME_OCCURS_IN_ABNORMAL_PHI (PHI_RESULT (new_phi)) |
1056 | = SSA_NAME_OCCURS_IN_ABNORMAL_PHI (bi_newcall_lhs); |
1057 | add_phi_arg (new_phi, bi_call_lhs, join_tgt_in_edge_from_call, |
1058 | gimple_location (g: bi_call)); |
1059 | add_phi_arg (new_phi, bi_newcall_lhs, join_tgt_in_edge_fall_thru, |
1060 | gimple_location (g: bi_newcall)); |
1061 | |
1062 | /* Replace all use of original return value with result of phi node. */ |
1063 | use_operand_p use_p; |
1064 | gimple *use_stmt; |
1065 | imm_use_iterator iterator; |
1066 | FOR_EACH_IMM_USE_STMT (use_stmt, iterator, bi_newcall_lhs) |
1067 | if (use_stmt != new_phi) |
1068 | FOR_EACH_IMM_USE_ON_STMT (use_p, iterator) |
1069 | SET_USE (use_p, PHI_RESULT (new_phi)); |
1070 | } |
1071 | |
1072 | /* Now update the probability and profile information, processing the |
1073 | guards in order of execution. |
1074 | |
1075 | There are two approaches we could take here. On the one hand we |
1076 | could assign a probability of X to the call block and distribute |
1077 | that probability among its incoming edges. On the other hand we |
1078 | could assign a probability of X to each individual call edge. |
1079 | |
1080 | The choice only affects calls that have more than one condition. |
1081 | In those cases, the second approach would give the call block |
1082 | a greater probability than the first. However, the difference |
1083 | is only small, and our chosen X is a pure guess anyway. |
1084 | |
1085 | Here we take the second approach because it's slightly simpler |
1086 | and because it's easy to see that it doesn't lose profile counts. */ |
1087 | bi_call_bb->count = profile_count::zero (); |
1088 | while (!edges.is_empty ()) |
1089 | { |
1090 | edge_pair e = edges.pop (); |
1091 | edge call_edge = e.first; |
1092 | edge nocall_edge = e.second; |
1093 | basic_block src_bb = call_edge->src; |
1094 | gcc_assert (src_bb == nocall_edge->src); |
1095 | |
1096 | call_edge->probability = profile_probability::very_unlikely (); |
1097 | nocall_edge->probability = profile_probability::always () |
1098 | - call_edge->probability; |
1099 | |
1100 | bi_call_bb->count += call_edge->count (); |
1101 | |
1102 | if (nocall_edge->dest != join_tgt_bb) |
1103 | nocall_edge->dest->count = src_bb->count - bi_call_bb->count; |
1104 | } |
1105 | |
1106 | if (dom_info_available_p (CDI_DOMINATORS)) |
1107 | { |
1108 | /* The split_blocks leave [guard 0] as the immediate dominator |
1109 | of [call] and [call] as the immediate dominator of [join]. |
1110 | Fix them up. */ |
1111 | set_immediate_dominator (CDI_DOMINATORS, bi_call_bb, guard_bb); |
1112 | set_immediate_dominator (CDI_DOMINATORS, join_tgt_bb, guard_bb); |
1113 | } |
1114 | |
1115 | if (dump_file && (dump_flags & TDF_DETAILS)) |
1116 | { |
1117 | location_t loc; |
1118 | loc = gimple_location (g: bi_call); |
1119 | fprintf (stream: dump_file, |
1120 | format: "%s:%d: note: function call is shrink-wrapped" |
1121 | " into error conditions.\n" , |
1122 | LOCATION_FILE (loc), LOCATION_LINE (loc)); |
1123 | } |
1124 | } |
1125 | |
1126 | /* Shrink-wrap BI_CALL so that it is only called when it might set errno |
1127 | (but is always called if it would set errno). */ |
1128 | |
1129 | static void |
1130 | shrink_wrap_one_built_in_call (gcall *bi_call) |
1131 | { |
1132 | unsigned nconds = 0; |
1133 | auto_vec<gimple *, 12> conds; |
1134 | gen_shrink_wrap_conditions (bi_call, conds, nconds: &nconds); |
1135 | gcc_assert (nconds != 0); |
1136 | shrink_wrap_one_built_in_call_with_conds (bi_call, conds, nconds); |
1137 | } |
1138 | |
1139 | /* Return true if built-in function call CALL could be implemented using |
1140 | a combination of an internal function to compute the result and a |
1141 | separate call to set errno. */ |
1142 | |
1143 | static bool |
1144 | can_use_internal_fn (gcall *call) |
1145 | { |
1146 | /* Only replace calls that set errno. */ |
1147 | if (!gimple_vdef (g: call)) |
1148 | return false; |
1149 | |
1150 | /* See whether there is an internal function for this built-in. */ |
1151 | if (replacement_internal_fn (call) == IFN_LAST) |
1152 | return false; |
1153 | |
1154 | /* See whether we can catch all cases where errno would be set, |
1155 | while still avoiding the call in most cases. */ |
1156 | if (!can_test_argument_range (call) |
1157 | && !edom_only_function (call)) |
1158 | return false; |
1159 | |
1160 | return true; |
1161 | } |
1162 | |
1163 | /* Implement built-in function call CALL using an internal function. */ |
1164 | |
1165 | static void |
1166 | use_internal_fn (gcall *call) |
1167 | { |
1168 | /* We'll be inserting another call with the same arguments after the |
1169 | lhs has been set, so prevent any possible coalescing failure from |
1170 | having both values live at once. See PR 71020. */ |
1171 | replace_abnormal_ssa_names (call); |
1172 | |
1173 | unsigned nconds = 0; |
1174 | auto_vec<gimple *, 12> conds; |
1175 | bool is_arg_conds = false; |
1176 | if (can_test_argument_range (call)) |
1177 | { |
1178 | gen_shrink_wrap_conditions (bi_call: call, conds, nconds: &nconds); |
1179 | is_arg_conds = true; |
1180 | gcc_assert (nconds != 0); |
1181 | } |
1182 | else |
1183 | gcc_assert (edom_only_function (call)); |
1184 | |
1185 | internal_fn ifn = replacement_internal_fn (call); |
1186 | gcc_assert (ifn != IFN_LAST); |
1187 | |
1188 | /* Construct the new call, with the same arguments as the original one. */ |
1189 | auto_vec <tree, 16> args; |
1190 | unsigned int nargs = gimple_call_num_args (gs: call); |
1191 | for (unsigned int i = 0; i < nargs; ++i) |
1192 | args.safe_push (obj: gimple_call_arg (gs: call, index: i)); |
1193 | gcall *new_call = gimple_build_call_internal_vec (ifn, args); |
1194 | gimple_set_location (g: new_call, location: gimple_location (g: call)); |
1195 | gimple_call_set_nothrow (s: new_call, nothrow_p: gimple_call_nothrow_p (s: call)); |
1196 | |
1197 | /* Transfer the LHS to the new call. */ |
1198 | tree lhs = gimple_call_lhs (gs: call); |
1199 | gimple_call_set_lhs (gs: new_call, lhs); |
1200 | gimple_call_set_lhs (gs: call, NULL_TREE); |
1201 | SSA_NAME_DEF_STMT (lhs) = new_call; |
1202 | |
1203 | /* Insert the new call. */ |
1204 | gimple_stmt_iterator gsi = gsi_for_stmt (call); |
1205 | gsi_insert_before (&gsi, new_call, GSI_SAME_STMT); |
1206 | |
1207 | if (nconds == 0) |
1208 | { |
1209 | /* Skip the call if LHS == LHS. If we reach here, EDOM is the only |
1210 | valid errno value and it is used iff the result is NaN. */ |
1211 | conds.quick_push (obj: gimple_build_cond (EQ_EXPR, lhs, lhs, |
1212 | NULL_TREE, NULL_TREE)); |
1213 | nconds++; |
1214 | |
1215 | /* Try replacing the original call with a direct assignment to |
1216 | errno, via an internal function. */ |
1217 | if (set_edom_supported_p () && !stmt_ends_bb_p (call)) |
1218 | { |
1219 | gimple_stmt_iterator gsi = gsi_for_stmt (call); |
1220 | gcall *new_call = gimple_build_call_internal (IFN_SET_EDOM, 0); |
1221 | gimple_move_vops (new_call, call); |
1222 | gimple_set_location (g: new_call, location: gimple_location (g: call)); |
1223 | gsi_replace (&gsi, new_call, false); |
1224 | call = new_call; |
1225 | } |
1226 | } |
1227 | shrink_wrap_one_built_in_call_with_conds (bi_call: call, conds, nconds, |
1228 | bi_newcall: is_arg_conds ? new_call : NULL); |
1229 | } |
1230 | |
1231 | /* The top level function for conditional dead code shrink |
1232 | wrapping transformation. */ |
1233 | |
1234 | static void |
1235 | shrink_wrap_conditional_dead_built_in_calls (const vec<gcall *> &calls) |
1236 | { |
1237 | unsigned i = 0; |
1238 | |
1239 | unsigned n = calls.length (); |
1240 | for (; i < n ; i++) |
1241 | { |
1242 | gcall *bi_call = calls[i]; |
1243 | if (gimple_call_lhs (gs: bi_call)) |
1244 | use_internal_fn (call: bi_call); |
1245 | else |
1246 | shrink_wrap_one_built_in_call (bi_call); |
1247 | } |
1248 | } |
1249 | |
1250 | namespace { |
1251 | |
1252 | const pass_data pass_data_call_cdce = |
1253 | { |
1254 | .type: GIMPLE_PASS, /* type */ |
1255 | .name: "cdce" , /* name */ |
1256 | .optinfo_flags: OPTGROUP_NONE, /* optinfo_flags */ |
1257 | .tv_id: TV_TREE_CALL_CDCE, /* tv_id */ |
1258 | .properties_required: ( PROP_cfg | PROP_ssa ), /* properties_required */ |
1259 | .properties_provided: 0, /* properties_provided */ |
1260 | .properties_destroyed: 0, /* properties_destroyed */ |
1261 | .todo_flags_start: 0, /* todo_flags_start */ |
1262 | .todo_flags_finish: 0, /* todo_flags_finish */ |
1263 | }; |
1264 | |
1265 | class pass_call_cdce : public gimple_opt_pass |
1266 | { |
1267 | public: |
1268 | pass_call_cdce (gcc::context *ctxt) |
1269 | : gimple_opt_pass (pass_data_call_cdce, ctxt) |
1270 | {} |
1271 | |
1272 | /* opt_pass methods: */ |
1273 | bool gate (function *) final override |
1274 | { |
1275 | /* The limit constants used in the implementation |
1276 | assume IEEE floating point format. Other formats |
1277 | can be supported in the future if needed. */ |
1278 | return flag_tree_builtin_call_dce != 0; |
1279 | } |
1280 | |
1281 | unsigned int execute (function *) final override; |
1282 | |
1283 | }; // class pass_call_cdce |
1284 | |
1285 | unsigned int |
1286 | pass_call_cdce::execute (function *fun) |
1287 | { |
1288 | basic_block bb; |
1289 | gimple_stmt_iterator i; |
1290 | auto_vec<gcall *> cond_dead_built_in_calls; |
1291 | FOR_EACH_BB_FN (bb, fun) |
1292 | { |
1293 | /* Skip blocks that are being optimized for size, since our |
1294 | transformation always increases code size. */ |
1295 | if (optimize_bb_for_size_p (bb)) |
1296 | continue; |
1297 | |
1298 | /* Collect dead call candidates. */ |
1299 | for (i = gsi_start_bb (bb); !gsi_end_p (i); gsi_next (i: &i)) |
1300 | { |
1301 | gcall *stmt = dyn_cast <gcall *> (p: gsi_stmt (i)); |
1302 | if (stmt |
1303 | && gimple_call_builtin_p (stmt, BUILT_IN_NORMAL) |
1304 | && (gimple_call_lhs (gs: stmt) |
1305 | ? can_use_internal_fn (call: stmt) |
1306 | : can_test_argument_range (call: stmt)) |
1307 | && can_guard_call_p (call: stmt)) |
1308 | { |
1309 | if (dump_file && (dump_flags & TDF_DETAILS)) |
1310 | { |
1311 | fprintf (stream: dump_file, format: "Found conditional dead call: " ); |
1312 | print_gimple_stmt (dump_file, stmt, 0, TDF_SLIM); |
1313 | fprintf (stream: dump_file, format: "\n" ); |
1314 | } |
1315 | if (!cond_dead_built_in_calls.exists ()) |
1316 | cond_dead_built_in_calls.create (nelems: 64); |
1317 | cond_dead_built_in_calls.safe_push (obj: stmt); |
1318 | } |
1319 | } |
1320 | } |
1321 | |
1322 | if (!cond_dead_built_in_calls.exists ()) |
1323 | return 0; |
1324 | |
1325 | shrink_wrap_conditional_dead_built_in_calls (calls: cond_dead_built_in_calls); |
1326 | free_dominance_info (CDI_POST_DOMINATORS); |
1327 | /* As we introduced new control-flow we need to insert PHI-nodes |
1328 | for the call-clobbers of the remaining call. */ |
1329 | mark_virtual_operands_for_renaming (fun); |
1330 | return TODO_update_ssa; |
1331 | } |
1332 | |
1333 | } // anon namespace |
1334 | |
1335 | gimple_opt_pass * |
1336 | make_pass_call_cdce (gcc::context *ctxt) |
1337 | { |
1338 | return new pass_call_cdce (ctxt); |
1339 | } |
1340 | |