Jakub Jelinek <[email protected]> writes: > On Tue, Sep 30, 2025 at 01:02:25PM +0000, Kyrylo Tkachov wrote: >> > - /* At this point we invalidate porfile confistency until >> > IFN_LOOP_VECTORIZED >> > + /* At this point we invalidate profile confistency until >> > IFN_LOOP_VECTORIZED >> >> By sheer coincidence there’s another typo in this line in “consistency” :) > > On Tue, Sep 30, 2025 at 02:11:48PM +0100, Sam James wrote: >> > -/* Generate IFN_ASAN_MARK call that poisons shadow of a for DECL >> > variable. */ >> > +/* Generate IFN_ASAN_MARK call that poisons shadow for DECL variable. */ >> >> This reads a bit odd still. Looking at asan_poison_variable, something >> like: >> >> "Generate IFN_ASAN_MARK call that poisons shadow memory of the DECL >> variable." > > Here is a new version with those two changes incorporated. > > Ok for trunk?
Looks good/obvious now. > > 2025-09-30 Jakub Jelinek <[email protected]> > > gcc/ > * auto-profile.h (maybe_hot_afdo_count_p): Fix comment typos, > possiby -> possibly and ture -> true. > * gimplify.cc (build_asan_poison_call_expr): Change "of a for" > to "memory of the" in a comment. > * ipa-devirt.cc (add_type_duplicate): Fix comment typo, > mangles -> mangled. > * auto-profile.cc: Fix comment typo, -fauto-profile-inlinig > -> -fauto-profile-inlining. > (maybe_hot_afdo_count_p): Fix comment typos, possiby -> possibly > and ture -> true. > (function_instance::removed_icall_target): Fix comment typo, > Reutrn -> Return. > (function_instance::in_worklist_): Fix comment typo, Ture -> True. > (function_instance::offline): Fix comment typo, tolevel -> toplevel. > (function_instance::match): Fix comment typo, craeate_gcov -> > create_gcov. > (autofdo_source_profile::offline_external_functions): Fix comment > typos, tolevel -> toplevel and porfile -> profile. > (autofdo_source_profile::get_function_instance_by_inline_stack): Fix > comment typo, chekcing -> checking. > (struct scale): Fix comment typo, scalle -> scale. > * gimple.h (currently_expanding_gimple_stmt): Fix comment typo, > comminucating -> communicating. > * tree.h (canonical_type_used_p): Fix comment typo, ture -> true. > * tree-ssa-alias.cc (types_equal_for_same_type_for_tbaa_p): Likewise. > * ipa-profile.cc (contains_hot_call_p): Likewise. > * cfgexpand.cc (add_scope_conflicts_2): > * ipa-param-manipulation.cc (currently_expanding_gimple_stmt): Fix > comment typo, comminucating -> communicating. > * ipa-prop.cc (ipa_cst_ref_desc::refcount): Fix comment typo, > if -> is. > * tree-if-conv.cc (version_loop_for_if_conversion): Fix comment typos, > porfile -> profile and confistency -> consistency. > * fold-const.cc: Change size_int_wide in comment to size_int as > size_int_wide doesn't exit for 21 years. > gcc/testsuite/ > * gcc.dg/vect/tsvc/vect-tsvc-s1244.c (s1244): Fix comment typo, > ture -> true. > * gcc.dg/vect/tsvc/vect-tsvc-s2244.c (s2244): Likewise. > > --- gcc/auto-profile.h.jj 2025-07-04 09:01:47.195520958 +0200 > +++ gcc/auto-profile.h 2025-09-30 14:00:07.339413394 +0200 > @@ -39,7 +39,7 @@ extern void remove_afdo_speculative_targ > considered cols. */ > extern gcov_type afdo_hot_bb_threshold; > > -/* Return ture if COUNT is possiby hot. */ > +/* Return true if COUNT is possibly hot. */ > extern bool maybe_hot_afdo_count_p (profile_count count); > > #endif /* AUTO_PROFILE_H */ > --- gcc/gimplify.cc.jj 2025-09-29 15:01:09.144838185 +0200 > +++ gcc/gimplify.cc 2025-09-30 13:45:15.648693114 +0200 > @@ -1280,7 +1280,8 @@ build_stack_save_restore (gcall **save, > 1, tmp_var); > } > > -/* Generate IFN_ASAN_MARK call that poisons shadow of a for DECL variable. > */ > +/* Generate IFN_ASAN_MARK call that poisons shadow memory of the DECL > + variable. */ > > static tree > build_asan_poison_call_expr (tree decl) > --- gcc/ipa-devirt.cc.jj 2025-09-23 10:26:42.906779714 +0200 > +++ gcc/ipa-devirt.cc 2025-09-30 14:12:54.302846046 +0200 > @@ -1801,7 +1801,7 @@ add_type_duplicate (odr_type val, tree t > before we can pass them to odr_types_equivalent_p (PR lto/83121). */ > if (lto_location_cache::current_cache) > lto_location_cache::current_cache->apply_location_cache (); > - /* As a special case we stream mangles names of integer types so we can see > + /* As a special case we stream mangled names of integer types so we can see > if they are believed to be same even though they have different > representation. Avoid bogus warning on mismatches in these. */ > if (TREE_CODE (type) != INTEGER_TYPE > --- gcc/auto-profile.cc.jj 2025-09-30 13:37:48.356854676 +0200 > +++ gcc/auto-profile.cc 2025-09-30 14:10:48.542578787 +0200 > @@ -100,7 +100,7 @@ along with GCC; see the file COPYING3. > significantly form one inline instance to another and from the > offline version. > > - This is controlled by -fauto-profile-inlinig and is independent > + This is controlled by -fauto-profile-inlining and is independent > of -fearly-inlining. > > Phase 4: In AFDO pass. > @@ -128,7 +128,7 @@ along with GCC; see the file COPYING3. > considered cols. */ > gcov_type afdo_hot_bb_threshod = -1; > > -/* Return ture if COUNT is possiby hot. */ > +/* Return true if COUNT is possibly hot. */ > bool > maybe_hot_afdo_count_p (profile_count count) > { > @@ -372,7 +372,7 @@ public: > removed_icall_target_ = true; > } > > - /* Reutrn true if function is removed from indir target list. */ > + /* Return true if function is removed from indir target list. */ > bool > removed_icall_target () > { > @@ -506,7 +506,7 @@ private: > to it. */ > bool realized_; > > - /* Ture if function is in worklist for merging/offlining. */ > + /* True if function is in worklist for merging/offlining. */ > bool in_worklist_; > > /* Pointer to outer function instance or NULL if this > @@ -1090,7 +1090,7 @@ function_instance::merge (function_insta > } > > /* Make inline function FN offline. > - If tolevel function of same name already exists, then merge profiles. > + If toplevel function of same name already exists, then merge profiles. > Otherwise turn FN toplevel. Return true if new toplevel function > was introduced. > If new toplevel functions are created and NEW_FUNCTIONS != NULL, > @@ -1638,7 +1638,7 @@ function_instance::match (cgraph_node *n > unsigned int start_location = get_combined_location > (DECL_STRUCT_FUNCTION (node->decl)->function_start_locus, node->decl); > /* When outputting code to builtins location we use line number 0. > - craeate_gcov is stupid and hapilly computes offsets across files. > + create_gcov is stupid and happily computes offsets across files. > Silently ignore it. */ > unsigned int zero_location > = ((unsigned)(1-DECL_SOURCE_LINE (node->decl))) << 16; > @@ -2090,11 +2090,11 @@ autofdo_source_profile::offline_external > seen.add (iter.first); > } > > - /* Now process all tolevel (offline) function instances. > + /* Now process all toplevel (offline) function instances. > > If instance has no definition in this translation unit, > first offline all inlined functions which are defined here > - (so we do not lose porfile due to cross-module inlining > + (so we do not lose profile due to cross-module inlining > done by link-time optimizers). > > If instance has a definition, look into all inlined functions > @@ -2694,7 +2694,7 @@ autofdo_source_profile::get_function_ins > if (s == NULL) > { > /* afdo inliner extends the stack by last entry with unknown > - location while chekcing if function was inlined during train run. > + location while checking if function was inlined during train run. > We do not want to print diagnostics about every function > which is not inlined. */ > if (s && dump_enabled_p () && stack[i].location != UNKNOWN_LOCATION) > @@ -3438,7 +3438,7 @@ cmp (const void *a, const void *b) > return 0; > } > > -/* To scalle a connected component of graph we collect desired scales of > +/* To scale a connected component of graph we collect desired scales of > basic blocks on the boundary and then compute a robust average. */ > > struct scale > --- gcc/gimple.h.jj 2025-08-06 10:41:32.349075826 +0200 > +++ gcc/gimple.h 2025-09-30 14:25:18.821542260 +0200 > @@ -1545,7 +1545,7 @@ extern size_t const gimple_ops_offset_[] > extern enum gimple_statement_structure_enum const gss_for_code_[]; > > /* This variable holds the currently expanded gimple statement for purposes > - of comminucating the profile info to the builtin expanders. */ > + of communicating the profile info to the builtin expanders. */ > extern gimple *currently_expanding_gimple_stmt; > > size_t gimple_size (enum gimple_code code, unsigned num_ops = 0); > --- gcc/tree.h.jj 2025-09-29 15:01:29.937552088 +0200 > +++ gcc/tree.h 2025-09-30 14:00:07.341413366 +0200 > @@ -5960,7 +5960,7 @@ tree_code_for_canonical_type_merging (en > return code; > } > > -/* Return ture if get_alias_set care about TYPE_CANONICAL of given type. > +/* Return true if get_alias_set care about TYPE_CANONICAL of given type. > We don't define the types for pointers, arrays and vectors. The reason is > that pointers are handled specially: ptr_type_node accesses conflict with > accesses to all other pointers. This is done by alias.cc. > --- gcc/tree-ssa-alias.cc.jj 2025-08-01 08:36:20.707714629 +0200 > +++ gcc/tree-ssa-alias.cc 2025-09-30 14:00:07.340413380 +0200 > @@ -4167,7 +4167,7 @@ attr_fnspec::verify () > } > } > > -/* Return ture if TYPE1 and TYPE2 will always give the same answer > +/* Return true if TYPE1 and TYPE2 will always give the same answer > when compared with other types using same_type_for_tbaa. */ > > static bool > @@ -4191,7 +4191,7 @@ types_equal_for_same_type_for_tbaa_p (tr > return TYPE_CANONICAL (type1) == TYPE_CANONICAL (type2); > } > > -/* Return ture if TYPE1 and TYPE2 will always give the same answer > +/* Return true if TYPE1 and TYPE2 will always give the same answer > when compared with other types using same_type_for_tbaa. */ > > bool > --- gcc/ipa-profile.cc.jj 2025-09-30 13:37:48.364854566 +0200 > +++ gcc/ipa-profile.cc 2025-09-30 14:00:07.339413394 +0200 > @@ -625,7 +625,7 @@ ipa_propagate_frequency_1 (struct cgraph > return edge != NULL; > } > > -/* Return ture if NODE contains hot calls. */ > +/* Return true if NODE contains hot calls. */ > > bool > contains_hot_call_p (struct cgraph_node *node) > --- gcc/cfgexpand.cc.jj 2025-07-27 23:31:09.197013784 +0200 > +++ gcc/cfgexpand.cc 2025-09-30 14:27:15.685921218 +0200 > @@ -90,7 +90,7 @@ along with GCC; see the file COPYING3. > struct ssaexpand SA; > > /* This variable holds the currently expanded gimple statement for purposes > - of comminucating the profile info to the builtin expanders. */ > + of communicating the profile info to the builtin expanders. */ > gimple *currently_expanding_gimple_stmt; > > static rtx expand_debug_expr (tree); > @@ -834,8 +834,8 @@ add_scope_conflicts_2 (vars_ssa_cache &c > { > gcc_assert (TREE_CODE (name) == SSA_NAME); > > - /* Querry the cache for the mapping of addresses that are referendd by > - ssa name NAME. Querrying it will fill in it. */ > + /* Query the cache for the mapping of addresses that are referenced by > + ssa name NAME. Querying it will fill in it. */ > bitmap_iterator bi; > unsigned i; > const_bitmap bmap = cache (name); > --- gcc/ipa-param-manipulation.cc.jj 2025-04-08 14:08:51.865273318 +0200 > +++ gcc/ipa-param-manipulation.cc 2025-09-30 14:16:02.748249620 +0200 > @@ -1672,7 +1672,7 @@ ipa_param_body_adjustments > in ADJUSTMENTS. FNDECL designates the new function clone which is being > modified. OLD_FNDECL is the function of which FNDECL is a clone (and > which > at the time of invocation still share DECL_ARGUMENTS). ID is the > - copy_body_data structure driving the wholy body copying process. VARS is > a > + copy_body_data structure driving the whole body copying process. VARS is > a > pointer to the head of the list of new local variables, TREE_MAP is the > map > that drives tree substitution in the cloning process. */ > > --- gcc/ipa-prop.cc.jj 2025-06-02 11:00:15.000000000 +0200 > +++ gcc/ipa-prop.cc 2025-09-30 14:14:58.009141602 +0200 > @@ -130,7 +130,7 @@ struct ipa_cst_ref_desc > /* Linked list of duplicates created when call graph edges are cloned. */ > struct ipa_cst_ref_desc *next_duplicate; > /* Number of references in IPA structures, IPA_UNDESCRIBED_USE if the value > - if out of control. */ > + is out of control. */ > int refcount; > }; > > --- gcc/tree-if-conv.cc.jj 2025-08-28 10:08:44.868229341 +0200 > +++ gcc/tree-if-conv.cc 2025-09-30 14:05:30.084966542 +0200 > @@ -3444,7 +3444,7 @@ version_loop_for_if_conversion (class lo > } > > initialize_original_copy_tables (); > - /* At this point we invalidate porfile confistency until > IFN_LOOP_VECTORIZED > + /* At this point we invalidate profile consistency until > IFN_LOOP_VECTORIZED > is re-merged in the vectorizer. */ > new_loop = loop_version (loop, cond, &cond_bb, > profile_probability::always (), > --- gcc/fold-const.cc.jj 2025-09-04 10:37:06.172634713 +0200 > +++ gcc/fold-const.cc 2025-09-30 13:46:27.279706672 +0200 > @@ -25,7 +25,7 @@ along with GCC; see the file COPYING3. > @@ This would also make life easier when this technology is used > @@ for cross-compilers. */ > > -/* The entry points in this file are fold, size_int_wide and size_binop. > +/* The entry points in this file are fold, size_int and size_binop. > > fold takes a tree as argument and returns a simplified tree. > > --- gcc/testsuite/gcc.dg/vect/tsvc/vect-tsvc-s1244.c.jj 2021-12-30 > 15:12:43.454147209 +0100 > +++ gcc/testsuite/gcc.dg/vect/tsvc/vect-tsvc-s1244.c 2025-09-30 > 14:00:07.789407202 +0200 > @@ -9,7 +9,7 @@ > real_t s1244(struct args_t * func_args) > { > // node splitting > -// cycle with ture and anti dependency > +// cycle with true and anti dependency > > initialise_arrays(__func__); > > --- gcc/testsuite/gcc.dg/vect/tsvc/vect-tsvc-s2244.c.jj 2021-12-30 > 15:12:43.454147209 +0100 > +++ gcc/testsuite/gcc.dg/vect/tsvc/vect-tsvc-s2244.c 2025-09-30 > 14:00:07.739407891 +0200 > @@ -9,7 +9,7 @@ > real_t s2244(struct args_t * func_args) > { > // node splitting > -// cycle with ture and anti dependency > +// cycle with true and anti dependency > > initialise_arrays(__func__); > > > > Jakub
