| Message ID | 20200926082020.GB627@kam.mff.cuni.cz |
|---|---|
| State | New |
| Headers | show |
| Series | Implement iterative dataflow in modref to track parameters | expand |
On September 26, 2020 10:20:20 AM GMT+02:00, Jan Hubicka <hubicka@ucw.cz> wrote: >Hi, >this patchs finishes the parameter tracking by implementing the >iterative >dataflow in propagation stage. This is necessary since we now can >propagate how the pointers are passed around recursive calls (as done >in >a testcase) and thus it is no-longer safe to simply merge all summaries >in one SCC component of the call-graph. > >cc1plus stats are now: > >Alias oracle query stats: > refs_may_alias_p: 62971744 disambiguations, 73160711 queries > ref_maybe_used_by_call_p: 141176 disambiguations, 63867883 queries > call_may_clobber_ref_p: 23573 disambiguations, 29322 queries > nonoverlapping_component_refs_p: 0 disambiguations, 37720 queries >nonoverlapping_refs_since_match_p: 19432 disambiguations, 55659 must >overlaps, 75860 queries > aliasing_component_refs_p: 54724 disambiguations, 753570 queries > TBAA oracle: 24124230 disambiguations 56228428 queries > 16058141 are in alias set 0 > 10338303 queries asked about the same object > 125 queries asked about the same alias set > 0 access volatile > 3919230 are dependent in the DAG > 1788399 are aritificially in conflict with void * > >Modref stats: > modref use: 10408 disambiguations, 46993 queries > modref clobber: 1418549 disambiguations, 1951251 queries > 4898707 tbaa queries (2.510547 per modref query) > 396878 base compares (0.203397 per modref query) > >PTA query stats: > pt_solution_includes: 975364 disambiguations, 13604284 queries > pt_solutions_intersect: 1026606 disambiguations, 13181198 queries > >So compared to >https://gcc.gnu.org/pipermail/gcc-patches/2020-September/554692.html we >get 25% >use disambiguations and 91% more clobber disambiguations. > >Tramp3d is > >Alias oracle query stats: > refs_may_alias_p: 2056905 disambiguations, 2317461 queries > ref_maybe_used_by_call_p: 7137 disambiguations, 2093762 queries > call_may_clobber_ref_p: 234 disambiguations, 234 queries > nonoverlapping_component_refs_p: 0 disambiguations, 4313 queries >nonoverlapping_refs_since_match_p: 329 disambiguations, 10200 must >overlaps, 10616 queries > aliasing_component_refs_p: 858 disambiguations, 34600 queries > TBAA oracle: 894996 disambiguations 1695991 queries > 138346 are in alias set 0 > 470668 queries asked about the same object > 0 queries asked about the same alias set > 0 access volatile > 191666 are dependent in the DAG > 315 are aritificially in conflict with void * > >Modref stats: > modref use: 842 disambiguations, 2265 queries > modref clobber: 14833 disambiguations, 28900 queries > 34884 tbaa queries (1.207059 per modref query) > 5041 base compares (0.174429 per modref query) > >PTA query stats: > pt_solution_includes: 313372 disambiguations, 525724 queries > pt_solutions_intersect: 130374 disambiguations, 415138 queries > >So about twice many use and 40% clobber disambiguations. > >Bootstrapped/regtested x86_64-linux, I plan to commit it later today >after >more testing. Any compile time figures for this? Firefox? > >2020-09-26 Jan Hubicka <hubicka@ucw.cz> > > * ipa-inline-transform.c: Include ipa-modref-tree.h and ipa-modref.h. > (inline_call): Call ipa_merge_modref_summary_after_inlining. > * ipa-inline.c (ipa_inline): Do not free summaries. > * ipa-modref.c (dump_records): Fix formating. > (merge_call_side_effects): Break out from ... > (analyze_call): ... here; record recursive calls. > (analyze_stmt): Add new parameter RECURSIVE_CALLS. > (analyze_function): Do iterative dataflow on recursive calls. > (compute_parm_map): New function. > (ipa_merge_modref_summary_after_inlining): New function. > (collapse_loads): New function. > (modref_propagate_in_scc): Break out from ... > (pass_ipa_modref::execute): ... here; Do iterative dataflow. > * ipa-modref.h (ipa_merge_modref_summary_after_inlining): Declare. > >gcc/testsuite/ChangeLog: > >2020-09-26 Jan Hubicka <hubicka@ucw.cz> > > * gcc.dg/lto/modref-1_0.c: New test. > * gcc.dg/lto/modref-1_1.c: New test. > * gcc.dg/tree-ssa/modref-2.c: New test. > >diff --git a/gcc/ipa-inline-transform.c b/gcc/ipa-inline-transform.c >index 5e37e612bfd..af2c2856aaa 100644 >--- a/gcc/ipa-inline-transform.c >+++ b/gcc/ipa-inline-transform.c >@@ -48,6 +48,8 @@ along with GCC; see the file COPYING3. If not see > #include "cfg.h" > #include "basic-block.h" > #include "ipa-utils.h" >+#include "ipa-modref-tree.h" >+#include "ipa-modref.h" > > int ncalls_inlined; > int nfunctions_inlined; >@@ -487,6 +489,7 @@ inline_call (struct cgraph_edge *e, bool >update_original, > gcc_assert (curr->callee->inlined_to == to); > > old_size = ipa_size_summaries->get (to)->size; >+ ipa_merge_modref_summary_after_inlining (e); > ipa_merge_fn_summary_after_inlining (e); > if (e->in_polymorphic_cdtor) > mark_all_inlined_calls_cdtor (e->callee); >diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c >index c667de2a97c..225a0140725 100644 >--- a/gcc/ipa-inline.c >+++ b/gcc/ipa-inline.c >@@ -2770,9 +2770,6 @@ ipa_inline (void) > } > } > >- /* Free ipa-prop structures if they are no longer needed. */ >- ipa_free_all_structures_after_iinln (); >- > if (dump_enabled_p ()) > dump_printf (MSG_NOTE, > "\nInlined %i calls, eliminated %i functions\n\n", >diff --git a/gcc/ipa-modref.c b/gcc/ipa-modref.c >index 44b844b90db..385b0318442 100644 >--- a/gcc/ipa-modref.c >+++ b/gcc/ipa-modref.c >@@ -175,7 +175,7 @@ dump_records (modref_records *tt, FILE *out) > fprintf (out, " Ref %i: alias set %i\n", (int)j, r->ref); > if (r->every_access) > { >- fprintf (out, " Every access\n"); >+ fprintf (out, " Every access\n"); > continue; > } > size_t k; >@@ -437,11 +437,70 @@ ignore_stores_p (tree caller, int flags) > return false; > } > >-/* Analyze function call STMT in function F. */ >+/* Merge side effects of call STMT to function with CALLEE_SUMMARY >+ int CUR_SUMMARY. Return true if something changed. >+ If IGNORE_STORES is true, do not merge stores. */ >+ >+bool >+merge_call_side_effects (modref_summary *cur_summary, >+ gimple *stmt, modref_summary *callee_summary, >+ bool ignore_stores) >+{ >+ auto_vec <int, 32> parm_map; >+ bool changed = false; >+ >+ parm_map.safe_grow (gimple_call_num_args (stmt)); >+ for (unsigned i = 0; i < gimple_call_num_args (stmt); i++) >+ { >+ tree op = gimple_call_arg (stmt, i); >+ STRIP_NOPS (op); >+ if (TREE_CODE (op) == SSA_NAME >+ && SSA_NAME_IS_DEFAULT_DEF (op) >+ && TREE_CODE (SSA_NAME_VAR (op)) == PARM_DECL) >+ { >+ int index = 0; >+ for (tree t = DECL_ARGUMENTS (current_function_decl); >+ t != SSA_NAME_VAR (op); t = DECL_CHAIN (t)) >+ { >+ if (!t) >+ { >+ index = -1; >+ break; >+ } >+ index++; >+ } >+ parm_map[i] = index; >+ } >+ else if (points_to_local_or_readonly_memory_p (op)) >+ parm_map[i] = -2; >+ else >+ parm_map[i] = -1; >+ } >+ >+ /* Merge with callee's summary. */ >+ if (cur_summary->loads) >+ changed |= cur_summary->loads->merge (callee_summary->loads, >&parm_map); >+ if (cur_summary->loads_lto) >+ changed |= cur_summary->loads_lto->merge >(callee_summary->loads_lto, >+ &parm_map); >+ if (!ignore_stores) >+ { >+ if (cur_summary->stores) >+ changed |= cur_summary->stores->merge (callee_summary->stores, >+ &parm_map); >+ if (cur_summary->stores_lto) >+ changed |= cur_summary->stores_lto->merge >(callee_summary->stores_lto, >+ &parm_map); >+ } >+ return changed; >+} >+ >+/* Analyze function call STMT in function F. >+ Remember recursive calls in RECURSIVE_CALLS. */ > > static bool > analyze_call (modref_summary *cur_summary, >- gimple *stmt) >+ gimple *stmt, vec <gimple *> *recursive_calls) > { >/* Check flags on the function call. In certain cases, analysis can be > simplified. */ >@@ -505,6 +564,7 @@ analyze_call (modref_summary *cur_summary, > there's nothing to do. */ > if (recursive_call_p (current_function_decl, callee)) > { >+ recursive_calls->safe_push (stmt); > if (dump_file) > fprintf (dump_file, " - Skipping recursive call.\n"); > return true; >@@ -550,48 +610,7 @@ analyze_call (modref_summary *cur_summary, > return false; > } > >- auto_vec <int, 32> parm_map; >- >- parm_map.safe_grow (gimple_call_num_args (stmt)); >- for (unsigned i = 0; i < gimple_call_num_args (stmt); i++) >- { >- tree op = gimple_call_arg (stmt, i); >- STRIP_NOPS (op); >- if (TREE_CODE (op) == SSA_NAME >- && SSA_NAME_IS_DEFAULT_DEF (op) >- && TREE_CODE (SSA_NAME_VAR (op)) == PARM_DECL) >- { >- int index = 0; >- for (tree t = DECL_ARGUMENTS (current_function_decl); >- t != SSA_NAME_VAR (op); t = DECL_CHAIN (t)) >- { >- if (!t) >- { >- index = -1; >- break; >- } >- index++; >- } >- parm_map[i] = index; >- } >- else if (points_to_local_or_readonly_memory_p (op)) >- parm_map[i] = -2; >- else >- parm_map[i] = -1; >- } >- >- /* Merge with callee's summary. */ >- if (cur_summary->loads) >- cur_summary->loads->merge (callee_summary->loads, &parm_map); >- if (cur_summary->loads_lto) >- cur_summary->loads_lto->merge (callee_summary->loads_lto, >&parm_map); >- if (!ignore_stores) >- { >- if (cur_summary->stores) >- cur_summary->stores->merge (callee_summary->stores, &parm_map); >- if (cur_summary->stores_lto) >- cur_summary->stores_lto->merge (callee_summary->stores_lto, >&parm_map); >- } >+ merge_call_side_effects (cur_summary, stmt, callee_summary, >ignore_stores); > > return true; > } >@@ -654,7 +673,8 @@ analyze_store (gimple *, tree, tree op, void *data) > If IPA is true do not merge in side effects of calls. */ > > static bool >-analyze_stmt (modref_summary *summary, gimple *stmt, bool ipa) >+analyze_stmt (modref_summary *summary, gimple *stmt, bool ipa, >+ vec <gimple *> *recursive_calls) > { > /* There is no need to record clobbers. */ > if (gimple_clobber_p (stmt)) >@@ -677,7 +697,7 @@ analyze_stmt (modref_summary *summary, gimple >*stmt, bool ipa) > return false; > case GIMPLE_CALL: > if (!ipa) >- return analyze_call (summary, stmt); >+ return analyze_call (summary, stmt, recursive_calls); > return true; > default: > /* Nothing to do for other types of statements. */ >@@ -750,6 +770,7 @@ analyze_function (function *f, bool ipa) > } > summary->finished = false; > int ecf_flags = flags_from_decl_or_type (current_function_decl); >+ auto_vec <gimple *, 32> recursive_calls; > > /* Analyze each statement in each basic block of the function. If the >statement cannot be analyzed (for any reason), the entire function >cannot >@@ -760,7 +781,7 @@ analyze_function (function *f, bool ipa) > gimple_stmt_iterator si; > for (si = gsi_after_labels (bb); !gsi_end_p (si); gsi_next (&si)) > { >- if (!analyze_stmt (summary, gsi_stmt (si), ipa) >+ if (!analyze_stmt (summary, gsi_stmt (si), ipa, &recursive_calls) > || !summary->useful_p (ecf_flags)) > { > cgraph_node *fnode = cgraph_node::get (current_function_decl); >@@ -773,6 +794,34 @@ analyze_function (function *f, bool ipa) > } > } > >+ /* In non-IPA mode we need to perform iterative datafow on recursive >calls. >+ This needs to be done after all other side effects are computed. >*/ >+ if (!ipa) >+ { >+ bool changed = true; >+ while (changed) >+ { >+ changed = false; >+ for (unsigned i = 0; i < recursive_calls.length (); i++) >+ { >+ changed |= merge_call_side_effects >+ (summary, recursive_calls[i], summary, >+ ignore_stores_p (current_function_decl, >+ gimple_call_flags >+ (recursive_calls[i]))); >+ if (!summary->useful_p (ecf_flags)) >+ { >+ cgraph_node *fnode = cgraph_node::get (current_function_decl); >+ summaries->remove (fnode); >+ if (dump_file) >+ fprintf (dump_file, >+ " - modref done with result: not tracked.\n"); >+ return; >+ } >+ } >+ } >+ } >+ > if (!ipa) > summary->finished = true; > >@@ -1276,71 +1325,176 @@ ignore_edge (struct cgraph_edge *e) > & (ECF_CONST | ECF_NOVOPS)); > } > >-/* Run the IPA pass. This will take a function's summaries and calls >and >- construct new summaries which represent a transitive closure. So >that >- summary of an analyzed function contains information about the >loads and >- stores that the function or any function that it calls does. */ >+/* Compute parm_map for CALLE_EDGE. */ > >-unsigned int pass_ipa_modref::execute (function *) >+static void >+compute_parm_map (cgraph_edge *callee_edge, vec<int> *parm_map) >+{ >+ class ipa_edge_args *args; >+ if (ipa_node_params_sum >+ && !callee_edge->call_stmt_cannot_inline_p >+ && (args = IPA_EDGE_REF (callee_edge)) != NULL) >+ { >+ int i, count = ipa_get_cs_argument_count (args); >+ class ipa_node_params *caller_parms_info, *callee_pi; >+ class ipa_call_summary *es >+ = ipa_call_summaries->get (callee_edge); >+ cgraph_node *callee >+ = callee_edge->callee->function_or_virtual_thunk_symbol >+ (NULL, callee_edge->caller); >+ >+ caller_parms_info = IPA_NODE_REF >(callee_edge->caller->inlined_to >+ ? callee_edge->caller->inlined_to >+ : callee_edge->caller); >+ callee_pi = IPA_NODE_REF (callee); >+ >+ (*parm_map).safe_grow (count); >+ >+ for (i = 0; i < count; i++) >+ { >+ if (es && es->param[i].points_to_local_or_readonly_memory) >+ { >+ (*parm_map)[i] = -2; >+ continue; >+ } >+ >+ struct ipa_jump_func *jf >+ = ipa_get_ith_jump_func (args, i); >+ if (jf) >+ { >+ tree cst = ipa_value_from_jfunc (caller_parms_info, >+ jf, >+ ipa_get_type >+ (callee_pi, i)); >+ if (cst && points_to_local_or_readonly_memory_p (cst)) >+ { >+ (*parm_map)[i] = -2; >+ continue; >+ } >+ } >+ if (jf && jf->type == IPA_JF_PASS_THROUGH) >+ { >+ (*parm_map)[i] >+ = ipa_get_jf_pass_through_formal_id (jf); >+ continue; >+ } >+ if (jf && jf->type == IPA_JF_ANCESTOR) >+ (*parm_map)[i] = ipa_get_jf_ancestor_formal_id (jf); >+ else >+ (*parm_map)[i] = -1; >+ } >+ if (dump_file) >+ { >+ fprintf (dump_file, " Parm map: "); >+ for (i = 0; i < count; i++) >+ fprintf (dump_file, " %i", (*parm_map)[i]); >+ fprintf (dump_file, "\n"); >+ } >+ } >+} >+ >+/* Call EDGE was inlined; merge summary from callee to the caller. */ >+ >+void >+ipa_merge_modref_summary_after_inlining (cgraph_edge *edge) > { > if (!summaries) >- return 0; >+ return; > >- struct cgraph_node **order = XCNEWVEC (struct cgraph_node *, >- symtab->cgraph_count); >- int order_pos; >- order_pos = ipa_reduced_postorder (order, true, ignore_edge); >- int i; >+ struct cgraph_node *to = (edge->caller->inlined_to >+ ? edge->caller->inlined_to : edge->caller); >+ class modref_summary *to_info = summaries->get (to); > >- /* Iterate over all strongly connected components in post-order. */ >- for (i = 0; i < order_pos; i++) >+ if (!to_info) >+ return; >+ >+ class modref_summary *callee_info = summaries->get (edge->callee); >+ int flags = flags_from_decl_or_type (edge->callee->decl); >+ >+ if (!callee_info) > { >- bool its_hopeless = false; >- modref_records *loads = NULL; >- modref_records *stores = NULL; >- modref_records_lto *loads_lto = NULL; >- modref_records_lto *stores_lto = NULL; >+ if (ignore_stores_p (edge->callee->decl, flags)) >+ { >+ if (to_info->loads) >+ to_info->loads->collapse (); >+ if (to_info->loads_lto) >+ to_info->loads_lto->collapse (); >+ } >+ else >+ { >+ summaries->remove (to); >+ summaries->remove (edge->callee); >+ return; >+ } >+ } >+ else >+ { >+ auto_vec <int, 32> parm_map+ >+ compute_parm_map (edge, &parm_map); >+ >+ if (to_info->loads) >+ to_info->loads->merge (callee_info->loads, &parm_map); >+ if (to_info->stores) >+ to_info->stores->merge (callee_info->stores, &parm_map); >+ if (to_info->loads_lto) >+ to_info->loads_lto->merge (callee_info->loads_lto, &parm_map); >+ if (to_info->stores_lto) >+ to_info->stores_lto->merge (callee_info->stores_lto, &parm_map); >+ } >+ if (!to_info->useful_p (flags)) >+ summaries->remove (to); >+ summaries->remove (edge->callee); >+ return; >+} > >- /* Get the component's representative. That's just any node in >the >- component from which we can traverse the entire component. */ >- struct cgraph_node *component_node = order[i]; >- cgraph_node *first = NULL; >+/* Collapse loads and return true if something changed. */ > >- if (dump_file) >- fprintf (dump_file, "Start of SCC component\n"); >+bool >+collapse_loads (modref_summary *cur_summary) >+{ >+ bool changed = false; >+ >+ if (cur_summary->loads && !cur_summary->loads->every_base) >+ { >+ cur_summary->loads->collapse (); >+ changed = true; >+ } >+ if (cur_summary->loads_lto >+ && !cur_summary->loads_lto->every_base) >+ { >+ cur_summary->loads_lto->collapse (); >+ changed = true; >+ } >+ return changed; >+} > >- /* Walk the component. CUR is the current node of the component >that's >- being processed. */ >- for (struct cgraph_node *cur = component_node; cur && >!its_hopeless; >+/* Perform iterative dataflow on SCC component starting in >COMPONENT_NODE. */ >+ >+static void >+modref_propagate_in_scc (cgraph_node *component_node) >+{ >+ bool changed = true; >+ int iteration = 0; >+ >+ while (changed) >+ { >+ changed = false; >+ for (struct cgraph_node *cur = component_node; cur; > cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) > { >- /* Merge in summaries from CUR. */ >- modref_summary *cur_summary = summaries->get (cur); >- >- if (dump_file) >- fprintf (dump_file, " Processing %s\n", >- cur->dump_name ()); >+ cgraph_node *node = cur->inlined_to ? cur->inlined_to : cur; >+ modref_summary *cur_summary = summaries->get (node); > >- /* We don't know anything about CUR, hence we cannot tell anything >- about the entire component. */ > if (!cur_summary) >- { >- if (dump_file) >- fprintf (dump_file, " No summary\n"); >- its_hopeless = true; >- break; >- } >+ continue; >+ >+ if (dump_file) >+ fprintf (dump_file, " Processing %s%s%s\n", >+ cur->dump_name (), >+ TREE_READONLY (cur->decl) ? " (const)" : "", >+ DECL_PURE_P (cur->decl) ? " (pure)" : ""); > >- /* Summaries are all going to be same, pick first ones and merge >- everything in. */ >- if (!first) >- { >- first = cur; >- loads = cur_summary->loads; >- stores = cur_summary->stores; >- loads_lto = cur_summary->loads_lto; >- stores_lto = cur_summary->stores_lto; >- } > for (cgraph_edge *e = cur->indirect_calls; e; e = e->next_callee) > { > if (e->indirect_info->ecf_flags & (ECF_CONST | ECF_NOVOPS)) >@@ -1350,20 +1504,22 @@ unsigned int pass_ipa_modref::execute (function >*) > if (dump_file) > fprintf (dump_file, " Indirect call: " > "collapsing loads\n"); >- if (loads) >- loads->collapse (); >- if (loads_lto) >- loads_lto->collapse (); >+ changed |= collapse_loads (cur_summary); > } > else > { > if (dump_file) > fprintf (dump_file, " Indirect call: giving up\n"); >- its_hopeless = true; >+ summaries->remove (node); >+ changed = true; >+ cur_summary = NULL; >+ break; > } > } > >- /* Walk every function that CUR calls and merge its summary. */ >+ if (!cur_summary) >+ continue; >+ > for (cgraph_edge *callee_edge = cur->callees; callee_edge; > callee_edge = callee_edge->next_callee) > { >@@ -1371,42 +1527,26 @@ unsigned int pass_ipa_modref::execute (function >*) > modref_summary *callee_summary; > struct cgraph_node *callee; > >- if (flags & (ECF_CONST | ECF_NOVOPS)) >+ if (flags & (ECF_CONST | ECF_NOVOPS) >+ || !callee_edge->inline_failed) > continue; > >- if (dump_file) >- fprintf (dump_file, " Call to %s\n", >- callee_edge->callee->dump_name ()); >- >- /* We can not safely optimize based on summary of callee if it >- does not always bind to current def: it is possible that >- memory load was optimized out earlier which may not happen in >- the interposed variant. */ >- if (!callee_edge->binds_to_current_def_p ()) >- { >- if (loads) >- loads->collapse (); >- if (loads_lto) >- loads_lto->collapse (); >- if (dump_file) >- fprintf (dump_file, " May not bind local;" >- " collapsing loads\n"); >- } >- > /* Get the callee and its summary. */ > enum availability avail; > callee = callee_edge->callee->function_or_virtual_thunk_symbol > (&avail, cur); > >- /* See if we can derive something from ECF flags. Be careful >on >- not skipping calls within the SCC component: we must merge >- all their summaries. >- If we switch to iterative dataflow that may be necessary >- for future improvements this may go away. */ >- if (callee->aux >- && ((struct ipa_dfs_info *)cur->aux)->scc_no >- == ((struct ipa_dfs_info *)callee->aux)->scc_no) >- flags = 0; >+ /* It is not necessary to re-process calls outside of the >+ SCC component. */ >+ if (iteration > 0 >+ && (!callee->aux >+ || ((struct ipa_dfs_info *)cur->aux)->scc_no >+ != ((struct ipa_dfs_info *)callee->aux)->scc_no)) >+ continue; >+ >+ if (dump_file) >+ fprintf (dump_file, " Call to %s\n", >+ callee_edge->callee->dump_name ()); > > bool ignore_stores = ignore_stores_p (cur->decl, flags); > >@@ -1418,101 +1558,128 @@ unsigned int pass_ipa_modref::execute >(function *) > { > if (!ignore_stores) > { >- its_hopeless = true; > if (dump_file && avail <= AVAIL_INTERPOSABLE) > fprintf (dump_file, " Call target interposable" > " or not available\n"); > else if (dump_file) > fprintf (dump_file, " No call target summary\n"); >+ >+ summaries->remove (node); >+ changed = true; > break; > } > else > { >- if (loads) >- loads->collapse (); >- if (loads_lto) >- loads_lto->collapse (); > if (dump_file && avail <= AVAIL_INTERPOSABLE) > fprintf (dump_file, " Call target interposable" >- "or not available; collapsing loads\n"); >+ " or not available; collapsing loads\n"); > else if (dump_file) > fprintf (dump_file, " No call target summary;" > " collapsing loads\n"); >+ >+ changed |= collapse_loads (cur_summary); > continue; > } > } > >+ /* We can not safely optimize based on summary of callee if it >+ does not always bind to current def: it is possible that >+ memory load was optimized out earlier which may not happen in >+ the interposed variant. */ >+ if (!callee_edge->binds_to_current_def_p ()) >+ { >+ changed |= collapse_loads (cur_summary); >+ if (dump_file) >+ fprintf (dump_file, " May not bind local;" >+ " collapsing loads\n"); >+ } >+ >+ > auto_vec <int, 32> parm_map; >- /* TODO: compute parm_map. */ >+ >+ compute_parm_map (callee_edge, &parm_map); > > /* Merge in callee's information. */ >- if (callee_summary->loads >- && callee_summary->loads != loads) >- loads->merge (callee_summary->loads, &parm_map); >- if (callee_summary->stores >- && callee_summary->stores != stores) >- stores->merge (callee_summary->stores, &parm_map); >- if (callee_summary->loads_lto >- && callee_summary->loads_lto != loads_lto) >- loads_lto->merge (callee_summary->loads_lto, &parm_map); >- if (callee_summary->stores_lto >- && callee_summary->stores_lto != stores_lto) >- stores_lto->merge (callee_summary->stores_lto, &parm_map); >+ if (callee_summary->loads) >+ changed |= cur_summary->loads->merge >+ (callee_summary->loads, &parm_map); >+ if (callee_summary->stores) >+ changed |= cur_summary->stores->merge >+ (callee_summary->stores, &parm_map); >+ if (callee_summary->loads_lto) >+ changed |= cur_summary->loads_lto->merge >+ (callee_summary->loads_lto, &parm_map); >+ if (callee_summary->stores_lto) >+ changed |= cur_summary->stores_lto->merge >+ (callee_summary->stores_lto, &parm_map); >+ if (dump_file && changed) >+ cur_summary->dump (dump_file); > } > } >- >- /* At this time, ipa_loads and ipa_stores contain information >- about all loads and stores done by any of the component's nodes >and >- all functions that any of the nodes calls. We will now propagate >- this information to all nodes in the component. Therefore, we >will >- walk the component one more time to do it. */ >- for (struct cgraph_node *cur = component_node; cur; >+ iteration++; >+ } >+ for (struct cgraph_node *cur = component_node; cur; >+ cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) >+ { >+ modref_summary *cur_summary = summaries->get (cur); >+ if (cur_summary) >+ cur_summary->finished = true; >+ } >+ if (dump_file) >+ { >+ fprintf (dump_file, >+ "Propagation finished in %i iterations\n", iteration); >+ for (struct cgraph_node *cur = component_node; cur; > cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) >- { >- modref_summary *cur_summary = summaries->get (cur); >- if (!cur_summary) >- { >- /* The function doesn't have a summary. We must have noticed >- that during the first pass and the hopeless flag must >- therefore be set. Skip the function. */ >- gcc_assert (its_hopeless); >- } >- else if (its_hopeless) >- { >- if (dump_file) >- fprintf (dump_file, "Cleared modref info for %s\n", >- cur->dump_name ()); >- summaries->remove (cur); >- } >- else >- { >- if (cur == first) >- ; >- else >- { >- if (loads) >- cur_summary->loads->merge (loads, NULL); >- if (stores) >- cur_summary->stores->merge (stores, NULL); >- if (loads_lto) >- cur_summary->loads_lto->merge (loads_lto, NULL); >- if (stores_lto) >- cur_summary->stores_lto->merge (stores_lto, NULL); >- } >- cur_summary->finished = true; >- if (dump_file) >- { >- fprintf (dump_file, "Propagated modref for %s%s%s\n", >- cur->dump_name (), >- TREE_READONLY (cur->decl) ? " (const)" : "", >- DECL_PURE_P (cur->decl) ? " (pure)" : ""); >- cur_summary->dump (dump_file); >- } >- } >- } >+ if (!cur->inlined_to) >+ { >+ modref_summary *cur_summary = summaries->get (cur); >+ >+ fprintf (dump_file, "Propagated modref for %s%s%s\n", >+ cur->dump_name (), >+ TREE_READONLY (cur->decl) ? " (const)" : "", >+ DECL_PURE_P (cur->decl) ? " (pure)" : ""); >+ if (cur_summary) >+ cur_summary->dump (dump_file); >+ else >+ fprintf (dump_file, " Not tracked\n"); >+ } >+ } >+} >+ >+/* Run the IPA pass. This will take a function's summaries and calls >and >+ construct new summaries which represent a transitive closure. So >that >+ summary of an analyzed function contains information about the >loads and >+ stores that the function or any function that it calls does. */ >+ >+unsigned int >+pass_ipa_modref::execute (function *) >+{ >+ if (!summaries) >+ return 0; >+ >+ struct cgraph_node **order = XCNEWVEC (struct cgraph_node *, >+ symtab->cgraph_count); >+ int order_pos; >+ order_pos = ipa_reduced_postorder (order, true, ignore_edge); >+ int i; >+ >+ /* Iterate over all strongly connected components in post-order. */ >+ for (i = 0; i < order_pos; i++) >+ { >+ /* Get the component's representative. That's just any node in >the >+ component from which we can traverse the entire component. */ >+ struct cgraph_node *component_node = order[i]; >+ >+ if (dump_file) >+ fprintf (dump_file, "\n\nStart of SCC component\n"); >+ >+ modref_propagate_in_scc (component_node); > } > ((modref_summaries *)summaries)->ipa = false; > ipa_free_postorder_info (); >+ /* Free ipa-prop structures if they are no longer needed. */ >+ ipa_free_all_structures_after_iinln (); > return 0; > } > >diff --git a/gcc/ipa-modref.h b/gcc/ipa-modref.h >index 152e7154aed..b6621b498f0 100644 >--- a/gcc/ipa-modref.h >+++ b/gcc/ipa-modref.h >@@ -47,5 +47,6 @@ struct GTY(()) modref_summary > > modref_summary *get_modref_function_summary (cgraph_node *func); > void ipa_modref_c_finalize (); >+void ipa_merge_modref_summary_after_inlining (cgraph_edge *e); > > #endif >diff --git a/gcc/testsuite/gcc.dg/lto/modref-1_0.c >b/gcc/testsuite/gcc.dg/lto/modref-1_0.c >new file mode 100644 >index 00000000000..8fcb9ec52f1 >--- /dev/null >+++ b/gcc/testsuite/gcc.dg/lto/modref-1_0.c >@@ -0,0 +1,14 @@ >+/* { dg-lto-do run } */ >+/* { dg-lto-options {"-O2 -flto-partition=max -flto"} } */ >+extern void recursive (int *a, int *b, int *c, int level); >+int >+main() >+{ >+ int x = 123, y=124, z=125; >+ recursive (&x,&y,&z,1); >+ if (y) >+ __builtin_abort (); >+ if (!__builtin_constant_p (z)) >+ __builtin_abort (); >+ return 0; >+} >diff --git a/gcc/testsuite/gcc.dg/lto/modref-1_1.c >b/gcc/testsuite/gcc.dg/lto/modref-1_1.c >new file mode 100644 >index 00000000000..c7c0eaec971 >--- /dev/null >+++ b/gcc/testsuite/gcc.dg/lto/modref-1_1.c >@@ -0,0 +1,13 @@ >+short aa; >+void >+__attribute__ ((noinline, noclone)) >+recursive (int *a, int *b, int *c, int level) >+{ >+ if (level && c) >+ { >+ recursive (b,a,c,0); >+ aa++; >+ } >+ else >+ *a=0; >+} >diff --git a/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c >b/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c >new file mode 100644 >index 00000000000..9999d37369d >--- /dev/null >+++ b/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c >@@ -0,0 +1,26 @@ >+/* { dg-do run } */ >+/* { dg-options "-O2" } */ >+short aa; >+void >+__attribute__ ((noinline, noclone)) >+recursive (int *a, int *b, int *c, int level) >+{ >+ if (level && c) >+ { >+ recursive (b,a,c,0); >+ aa++; >+ } >+ else >+ *a=0; >+} >+int >+main() >+{ >+ int x = 123, y=124, z=125; >+ recursive (&x,&y,&z,1); >+ if (y) >+ __builtin_abort (); >+ if (!__builtin_constant_p (z)) >+ __builtin_abort (); >+ return 0; >+}
Hello. The patch caused: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97235 Martin
On 9/29/20 10:13 AM, Martin Liška wrote: > Hello. > > The patch caused: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97235 > > Martin And these 2 PRs: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97243 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97244 Thanks, Martin
diff --git a/gcc/ipa-inline-transform.c b/gcc/ipa-inline-transform.c index 5e37e612bfd..af2c2856aaa 100644 --- a/gcc/ipa-inline-transform.c +++ b/gcc/ipa-inline-transform.c @@ -48,6 +48,8 @@ along with GCC; see the file COPYING3. If not see #include "cfg.h" #include "basic-block.h" #include "ipa-utils.h" +#include "ipa-modref-tree.h" +#include "ipa-modref.h" int ncalls_inlined; int nfunctions_inlined; @@ -487,6 +489,7 @@ inline_call (struct cgraph_edge *e, bool update_original, gcc_assert (curr->callee->inlined_to == to); old_size = ipa_size_summaries->get (to)->size; + ipa_merge_modref_summary_after_inlining (e); ipa_merge_fn_summary_after_inlining (e); if (e->in_polymorphic_cdtor) mark_all_inlined_calls_cdtor (e->callee); diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c index c667de2a97c..225a0140725 100644 --- a/gcc/ipa-inline.c +++ b/gcc/ipa-inline.c @@ -2770,9 +2770,6 @@ ipa_inline (void) } } - /* Free ipa-prop structures if they are no longer needed. */ - ipa_free_all_structures_after_iinln (); - if (dump_enabled_p ()) dump_printf (MSG_NOTE, "\nInlined %i calls, eliminated %i functions\n\n", diff --git a/gcc/ipa-modref.c b/gcc/ipa-modref.c index 44b844b90db..385b0318442 100644 --- a/gcc/ipa-modref.c +++ b/gcc/ipa-modref.c @@ -175,7 +175,7 @@ dump_records (modref_records *tt, FILE *out) fprintf (out, " Ref %i: alias set %i\n", (int)j, r->ref); if (r->every_access) { - fprintf (out, " Every access\n"); + fprintf (out, " Every access\n"); continue; } size_t k; @@ -437,11 +437,70 @@ ignore_stores_p (tree caller, int flags) return false; } -/* Analyze function call STMT in function F. */ +/* Merge side effects of call STMT to function with CALLEE_SUMMARY + int CUR_SUMMARY. Return true if something changed. + If IGNORE_STORES is true, do not merge stores. */ + +bool +merge_call_side_effects (modref_summary *cur_summary, + gimple *stmt, modref_summary *callee_summary, + bool ignore_stores) +{ + auto_vec <int, 32> parm_map; + bool changed = false; + + parm_map.safe_grow (gimple_call_num_args (stmt)); + for (unsigned i = 0; i < gimple_call_num_args (stmt); i++) + { + tree op = gimple_call_arg (stmt, i); + STRIP_NOPS (op); + if (TREE_CODE (op) == SSA_NAME + && SSA_NAME_IS_DEFAULT_DEF (op) + && TREE_CODE (SSA_NAME_VAR (op)) == PARM_DECL) + { + int index = 0; + for (tree t = DECL_ARGUMENTS (current_function_decl); + t != SSA_NAME_VAR (op); t = DECL_CHAIN (t)) + { + if (!t) + { + index = -1; + break; + } + index++; + } + parm_map[i] = index; + } + else if (points_to_local_or_readonly_memory_p (op)) + parm_map[i] = -2; + else + parm_map[i] = -1; + } + + /* Merge with callee's summary. */ + if (cur_summary->loads) + changed |= cur_summary->loads->merge (callee_summary->loads, &parm_map); + if (cur_summary->loads_lto) + changed |= cur_summary->loads_lto->merge (callee_summary->loads_lto, + &parm_map); + if (!ignore_stores) + { + if (cur_summary->stores) + changed |= cur_summary->stores->merge (callee_summary->stores, + &parm_map); + if (cur_summary->stores_lto) + changed |= cur_summary->stores_lto->merge (callee_summary->stores_lto, + &parm_map); + } + return changed; +} + +/* Analyze function call STMT in function F. + Remember recursive calls in RECURSIVE_CALLS. */ static bool analyze_call (modref_summary *cur_summary, - gimple *stmt) + gimple *stmt, vec <gimple *> *recursive_calls) { /* Check flags on the function call. In certain cases, analysis can be simplified. */ @@ -505,6 +564,7 @@ analyze_call (modref_summary *cur_summary, there's nothing to do. */ if (recursive_call_p (current_function_decl, callee)) { + recursive_calls->safe_push (stmt); if (dump_file) fprintf (dump_file, " - Skipping recursive call.\n"); return true; @@ -550,48 +610,7 @@ analyze_call (modref_summary *cur_summary, return false; } - auto_vec <int, 32> parm_map; - - parm_map.safe_grow (gimple_call_num_args (stmt)); - for (unsigned i = 0; i < gimple_call_num_args (stmt); i++) - { - tree op = gimple_call_arg (stmt, i); - STRIP_NOPS (op); - if (TREE_CODE (op) == SSA_NAME - && SSA_NAME_IS_DEFAULT_DEF (op) - && TREE_CODE (SSA_NAME_VAR (op)) == PARM_DECL) - { - int index = 0; - for (tree t = DECL_ARGUMENTS (current_function_decl); - t != SSA_NAME_VAR (op); t = DECL_CHAIN (t)) - { - if (!t) - { - index = -1; - break; - } - index++; - } - parm_map[i] = index; - } - else if (points_to_local_or_readonly_memory_p (op)) - parm_map[i] = -2; - else - parm_map[i] = -1; - } - - /* Merge with callee's summary. */ - if (cur_summary->loads) - cur_summary->loads->merge (callee_summary->loads, &parm_map); - if (cur_summary->loads_lto) - cur_summary->loads_lto->merge (callee_summary->loads_lto, &parm_map); - if (!ignore_stores) - { - if (cur_summary->stores) - cur_summary->stores->merge (callee_summary->stores, &parm_map); - if (cur_summary->stores_lto) - cur_summary->stores_lto->merge (callee_summary->stores_lto, &parm_map); - } + merge_call_side_effects (cur_summary, stmt, callee_summary, ignore_stores); return true; } @@ -654,7 +673,8 @@ analyze_store (gimple *, tree, tree op, void *data) If IPA is true do not merge in side effects of calls. */ static bool -analyze_stmt (modref_summary *summary, gimple *stmt, bool ipa) +analyze_stmt (modref_summary *summary, gimple *stmt, bool ipa, + vec <gimple *> *recursive_calls) { /* There is no need to record clobbers. */ if (gimple_clobber_p (stmt)) @@ -677,7 +697,7 @@ analyze_stmt (modref_summary *summary, gimple *stmt, bool ipa) return false; case GIMPLE_CALL: if (!ipa) - return analyze_call (summary, stmt); + return analyze_call (summary, stmt, recursive_calls); return true; default: /* Nothing to do for other types of statements. */ @@ -750,6 +770,7 @@ analyze_function (function *f, bool ipa) } summary->finished = false; int ecf_flags = flags_from_decl_or_type (current_function_decl); + auto_vec <gimple *, 32> recursive_calls; /* Analyze each statement in each basic block of the function. If the statement cannot be analyzed (for any reason), the entire function cannot @@ -760,7 +781,7 @@ analyze_function (function *f, bool ipa) gimple_stmt_iterator si; for (si = gsi_after_labels (bb); !gsi_end_p (si); gsi_next (&si)) { - if (!analyze_stmt (summary, gsi_stmt (si), ipa) + if (!analyze_stmt (summary, gsi_stmt (si), ipa, &recursive_calls) || !summary->useful_p (ecf_flags)) { cgraph_node *fnode = cgraph_node::get (current_function_decl); @@ -773,6 +794,34 @@ analyze_function (function *f, bool ipa) } } + /* In non-IPA mode we need to perform iterative datafow on recursive calls. + This needs to be done after all other side effects are computed. */ + if (!ipa) + { + bool changed = true; + while (changed) + { + changed = false; + for (unsigned i = 0; i < recursive_calls.length (); i++) + { + changed |= merge_call_side_effects + (summary, recursive_calls[i], summary, + ignore_stores_p (current_function_decl, + gimple_call_flags + (recursive_calls[i]))); + if (!summary->useful_p (ecf_flags)) + { + cgraph_node *fnode = cgraph_node::get (current_function_decl); + summaries->remove (fnode); + if (dump_file) + fprintf (dump_file, + " - modref done with result: not tracked.\n"); + return; + } + } + } + } + if (!ipa) summary->finished = true; @@ -1276,71 +1325,176 @@ ignore_edge (struct cgraph_edge *e) & (ECF_CONST | ECF_NOVOPS)); } -/* Run the IPA pass. This will take a function's summaries and calls and - construct new summaries which represent a transitive closure. So that - summary of an analyzed function contains information about the loads and - stores that the function or any function that it calls does. */ +/* Compute parm_map for CALLE_EDGE. */ -unsigned int pass_ipa_modref::execute (function *) +static void +compute_parm_map (cgraph_edge *callee_edge, vec<int> *parm_map) +{ + class ipa_edge_args *args; + if (ipa_node_params_sum + && !callee_edge->call_stmt_cannot_inline_p + && (args = IPA_EDGE_REF (callee_edge)) != NULL) + { + int i, count = ipa_get_cs_argument_count (args); + class ipa_node_params *caller_parms_info, *callee_pi; + class ipa_call_summary *es + = ipa_call_summaries->get (callee_edge); + cgraph_node *callee + = callee_edge->callee->function_or_virtual_thunk_symbol + (NULL, callee_edge->caller); + + caller_parms_info = IPA_NODE_REF (callee_edge->caller->inlined_to + ? callee_edge->caller->inlined_to + : callee_edge->caller); + callee_pi = IPA_NODE_REF (callee); + + (*parm_map).safe_grow (count); + + for (i = 0; i < count; i++) + { + if (es && es->param[i].points_to_local_or_readonly_memory) + { + (*parm_map)[i] = -2; + continue; + } + + struct ipa_jump_func *jf + = ipa_get_ith_jump_func (args, i); + if (jf) + { + tree cst = ipa_value_from_jfunc (caller_parms_info, + jf, + ipa_get_type + (callee_pi, i)); + if (cst && points_to_local_or_readonly_memory_p (cst)) + { + (*parm_map)[i] = -2; + continue; + } + } + if (jf && jf->type == IPA_JF_PASS_THROUGH) + { + (*parm_map)[i] + = ipa_get_jf_pass_through_formal_id (jf); + continue; + } + if (jf && jf->type == IPA_JF_ANCESTOR) + (*parm_map)[i] = ipa_get_jf_ancestor_formal_id (jf); + else + (*parm_map)[i] = -1; + } + if (dump_file) + { + fprintf (dump_file, " Parm map: "); + for (i = 0; i < count; i++) + fprintf (dump_file, " %i", (*parm_map)[i]); + fprintf (dump_file, "\n"); + } + } +} + +/* Call EDGE was inlined; merge summary from callee to the caller. */ + +void +ipa_merge_modref_summary_after_inlining (cgraph_edge *edge) { if (!summaries) - return 0; + return; - struct cgraph_node **order = XCNEWVEC (struct cgraph_node *, - symtab->cgraph_count); - int order_pos; - order_pos = ipa_reduced_postorder (order, true, ignore_edge); - int i; + struct cgraph_node *to = (edge->caller->inlined_to + ? edge->caller->inlined_to : edge->caller); + class modref_summary *to_info = summaries->get (to); - /* Iterate over all strongly connected components in post-order. */ - for (i = 0; i < order_pos; i++) + if (!to_info) + return; + + class modref_summary *callee_info = summaries->get (edge->callee); + int flags = flags_from_decl_or_type (edge->callee->decl); + + if (!callee_info) { - bool its_hopeless = false; - modref_records *loads = NULL; - modref_records *stores = NULL; - modref_records_lto *loads_lto = NULL; - modref_records_lto *stores_lto = NULL; + if (ignore_stores_p (edge->callee->decl, flags)) + { + if (to_info->loads) + to_info->loads->collapse (); + if (to_info->loads_lto) + to_info->loads_lto->collapse (); + } + else + { + summaries->remove (to); + summaries->remove (edge->callee); + return; + } + } + else + { + auto_vec <int, 32> parm_map+ + compute_parm_map (edge, &parm_map); + + if (to_info->loads) + to_info->loads->merge (callee_info->loads, &parm_map); + if (to_info->stores) + to_info->stores->merge (callee_info->stores, &parm_map); + if (to_info->loads_lto) + to_info->loads_lto->merge (callee_info->loads_lto, &parm_map); + if (to_info->stores_lto) + to_info->stores_lto->merge (callee_info->stores_lto, &parm_map); + } + if (!to_info->useful_p (flags)) + summaries->remove (to); + summaries->remove (edge->callee); + return; +} - /* Get the component's representative. That's just any node in the - component from which we can traverse the entire component. */ - struct cgraph_node *component_node = order[i]; - cgraph_node *first = NULL; +/* Collapse loads and return true if something changed. */ - if (dump_file) - fprintf (dump_file, "Start of SCC component\n"); +bool +collapse_loads (modref_summary *cur_summary) +{ + bool changed = false; + + if (cur_summary->loads && !cur_summary->loads->every_base) + { + cur_summary->loads->collapse (); + changed = true; + } + if (cur_summary->loads_lto + && !cur_summary->loads_lto->every_base) + { + cur_summary->loads_lto->collapse (); + changed = true; + } + return changed; +} - /* Walk the component. CUR is the current node of the component that's - being processed. */ - for (struct cgraph_node *cur = component_node; cur && !its_hopeless; +/* Perform iterative dataflow on SCC component starting in COMPONENT_NODE. */ + +static void +modref_propagate_in_scc (cgraph_node *component_node) +{ + bool changed = true; + int iteration = 0; + + while (changed) + { + changed = false; + for (struct cgraph_node *cur = component_node; cur; cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) { - /* Merge in summaries from CUR. */ - modref_summary *cur_summary = summaries->get (cur); - - if (dump_file) - fprintf (dump_file, " Processing %s\n", - cur->dump_name ()); + cgraph_node *node = cur->inlined_to ? cur->inlined_to : cur; + modref_summary *cur_summary = summaries->get (node); - /* We don't know anything about CUR, hence we cannot tell anything - about the entire component. */ if (!cur_summary) - { - if (dump_file) - fprintf (dump_file, " No summary\n"); - its_hopeless = true; - break; - } + continue; + + if (dump_file) + fprintf (dump_file, " Processing %s%s%s\n", + cur->dump_name (), + TREE_READONLY (cur->decl) ? " (const)" : "", + DECL_PURE_P (cur->decl) ? " (pure)" : ""); - /* Summaries are all going to be same, pick first ones and merge - everything in. */ - if (!first) - { - first = cur; - loads = cur_summary->loads; - stores = cur_summary->stores; - loads_lto = cur_summary->loads_lto; - stores_lto = cur_summary->stores_lto; - } for (cgraph_edge *e = cur->indirect_calls; e; e = e->next_callee) { if (e->indirect_info->ecf_flags & (ECF_CONST | ECF_NOVOPS)) @@ -1350,20 +1504,22 @@ unsigned int pass_ipa_modref::execute (function *) if (dump_file) fprintf (dump_file, " Indirect call: " "collapsing loads\n"); - if (loads) - loads->collapse (); - if (loads_lto) - loads_lto->collapse (); + changed |= collapse_loads (cur_summary); } else { if (dump_file) fprintf (dump_file, " Indirect call: giving up\n"); - its_hopeless = true; + summaries->remove (node); + changed = true; + cur_summary = NULL; + break; } } - /* Walk every function that CUR calls and merge its summary. */ + if (!cur_summary) + continue; + for (cgraph_edge *callee_edge = cur->callees; callee_edge; callee_edge = callee_edge->next_callee) { @@ -1371,42 +1527,26 @@ unsigned int pass_ipa_modref::execute (function *) modref_summary *callee_summary; struct cgraph_node *callee; - if (flags & (ECF_CONST | ECF_NOVOPS)) + if (flags & (ECF_CONST | ECF_NOVOPS) + || !callee_edge->inline_failed) continue; - if (dump_file) - fprintf (dump_file, " Call to %s\n", - callee_edge->callee->dump_name ()); - - /* We can not safely optimize based on summary of callee if it - does not always bind to current def: it is possible that - memory load was optimized out earlier which may not happen in - the interposed variant. */ - if (!callee_edge->binds_to_current_def_p ()) - { - if (loads) - loads->collapse (); - if (loads_lto) - loads_lto->collapse (); - if (dump_file) - fprintf (dump_file, " May not bind local;" - " collapsing loads\n"); - } - /* Get the callee and its summary. */ enum availability avail; callee = callee_edge->callee->function_or_virtual_thunk_symbol (&avail, cur); - /* See if we can derive something from ECF flags. Be careful on - not skipping calls within the SCC component: we must merge - all their summaries. - If we switch to iterative dataflow that may be necessary - for future improvements this may go away. */ - if (callee->aux - && ((struct ipa_dfs_info *)cur->aux)->scc_no - == ((struct ipa_dfs_info *)callee->aux)->scc_no) - flags = 0; + /* It is not necessary to re-process calls outside of the + SCC component. */ + if (iteration > 0 + && (!callee->aux + || ((struct ipa_dfs_info *)cur->aux)->scc_no + != ((struct ipa_dfs_info *)callee->aux)->scc_no)) + continue; + + if (dump_file) + fprintf (dump_file, " Call to %s\n", + callee_edge->callee->dump_name ()); bool ignore_stores = ignore_stores_p (cur->decl, flags); @@ -1418,101 +1558,128 @@ unsigned int pass_ipa_modref::execute (function *) { if (!ignore_stores) { - its_hopeless = true; if (dump_file && avail <= AVAIL_INTERPOSABLE) fprintf (dump_file, " Call target interposable" " or not available\n"); else if (dump_file) fprintf (dump_file, " No call target summary\n"); + + summaries->remove (node); + changed = true; break; } else { - if (loads) - loads->collapse (); - if (loads_lto) - loads_lto->collapse (); if (dump_file && avail <= AVAIL_INTERPOSABLE) fprintf (dump_file, " Call target interposable" - "or not available; collapsing loads\n"); + " or not available; collapsing loads\n"); else if (dump_file) fprintf (dump_file, " No call target summary;" " collapsing loads\n"); + + changed |= collapse_loads (cur_summary); continue; } } + /* We can not safely optimize based on summary of callee if it + does not always bind to current def: it is possible that + memory load was optimized out earlier which may not happen in + the interposed variant. */ + if (!callee_edge->binds_to_current_def_p ()) + { + changed |= collapse_loads (cur_summary); + if (dump_file) + fprintf (dump_file, " May not bind local;" + " collapsing loads\n"); + } + + auto_vec <int, 32> parm_map; - /* TODO: compute parm_map. */ + + compute_parm_map (callee_edge, &parm_map); /* Merge in callee's information. */ - if (callee_summary->loads - && callee_summary->loads != loads) - loads->merge (callee_summary->loads, &parm_map); - if (callee_summary->stores - && callee_summary->stores != stores) - stores->merge (callee_summary->stores, &parm_map); - if (callee_summary->loads_lto - && callee_summary->loads_lto != loads_lto) - loads_lto->merge (callee_summary->loads_lto, &parm_map); - if (callee_summary->stores_lto - && callee_summary->stores_lto != stores_lto) - stores_lto->merge (callee_summary->stores_lto, &parm_map); + if (callee_summary->loads) + changed |= cur_summary->loads->merge + (callee_summary->loads, &parm_map); + if (callee_summary->stores) + changed |= cur_summary->stores->merge + (callee_summary->stores, &parm_map); + if (callee_summary->loads_lto) + changed |= cur_summary->loads_lto->merge + (callee_summary->loads_lto, &parm_map); + if (callee_summary->stores_lto) + changed |= cur_summary->stores_lto->merge + (callee_summary->stores_lto, &parm_map); + if (dump_file && changed) + cur_summary->dump (dump_file); } } - - /* At this time, ipa_loads and ipa_stores contain information - about all loads and stores done by any of the component's nodes and - all functions that any of the nodes calls. We will now propagate - this information to all nodes in the component. Therefore, we will - walk the component one more time to do it. */ - for (struct cgraph_node *cur = component_node; cur; + iteration++; + } + for (struct cgraph_node *cur = component_node; cur; + cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) + { + modref_summary *cur_summary = summaries->get (cur); + if (cur_summary) + cur_summary->finished = true; + } + if (dump_file) + { + fprintf (dump_file, + "Propagation finished in %i iterations\n", iteration); + for (struct cgraph_node *cur = component_node; cur; cur = ((struct ipa_dfs_info *) cur->aux)->next_cycle) - { - modref_summary *cur_summary = summaries->get (cur); - if (!cur_summary) - { - /* The function doesn't have a summary. We must have noticed - that during the first pass and the hopeless flag must - therefore be set. Skip the function. */ - gcc_assert (its_hopeless); - } - else if (its_hopeless) - { - if (dump_file) - fprintf (dump_file, "Cleared modref info for %s\n", - cur->dump_name ()); - summaries->remove (cur); - } - else - { - if (cur == first) - ; - else - { - if (loads) - cur_summary->loads->merge (loads, NULL); - if (stores) - cur_summary->stores->merge (stores, NULL); - if (loads_lto) - cur_summary->loads_lto->merge (loads_lto, NULL); - if (stores_lto) - cur_summary->stores_lto->merge (stores_lto, NULL); - } - cur_summary->finished = true; - if (dump_file) - { - fprintf (dump_file, "Propagated modref for %s%s%s\n", - cur->dump_name (), - TREE_READONLY (cur->decl) ? " (const)" : "", - DECL_PURE_P (cur->decl) ? " (pure)" : ""); - cur_summary->dump (dump_file); - } - } - } + if (!cur->inlined_to) + { + modref_summary *cur_summary = summaries->get (cur); + + fprintf (dump_file, "Propagated modref for %s%s%s\n", + cur->dump_name (), + TREE_READONLY (cur->decl) ? " (const)" : "", + DECL_PURE_P (cur->decl) ? " (pure)" : ""); + if (cur_summary) + cur_summary->dump (dump_file); + else + fprintf (dump_file, " Not tracked\n"); + } + } +} + +/* Run the IPA pass. This will take a function's summaries and calls and + construct new summaries which represent a transitive closure. So that + summary of an analyzed function contains information about the loads and + stores that the function or any function that it calls does. */ + +unsigned int +pass_ipa_modref::execute (function *) +{ + if (!summaries) + return 0; + + struct cgraph_node **order = XCNEWVEC (struct cgraph_node *, + symtab->cgraph_count); + int order_pos; + order_pos = ipa_reduced_postorder (order, true, ignore_edge); + int i; + + /* Iterate over all strongly connected components in post-order. */ + for (i = 0; i < order_pos; i++) + { + /* Get the component's representative. That's just any node in the + component from which we can traverse the entire component. */ + struct cgraph_node *component_node = order[i]; + + if (dump_file) + fprintf (dump_file, "\n\nStart of SCC component\n"); + + modref_propagate_in_scc (component_node); } ((modref_summaries *)summaries)->ipa = false; ipa_free_postorder_info (); + /* Free ipa-prop structures if they are no longer needed. */ + ipa_free_all_structures_after_iinln (); return 0; } diff --git a/gcc/ipa-modref.h b/gcc/ipa-modref.h index 152e7154aed..b6621b498f0 100644 --- a/gcc/ipa-modref.h +++ b/gcc/ipa-modref.h @@ -47,5 +47,6 @@ struct GTY(()) modref_summary modref_summary *get_modref_function_summary (cgraph_node *func); void ipa_modref_c_finalize (); +void ipa_merge_modref_summary_after_inlining (cgraph_edge *e); #endif diff --git a/gcc/testsuite/gcc.dg/lto/modref-1_0.c b/gcc/testsuite/gcc.dg/lto/modref-1_0.c new file mode 100644 index 00000000000..8fcb9ec52f1 --- /dev/null +++ b/gcc/testsuite/gcc.dg/lto/modref-1_0.c @@ -0,0 +1,14 @@ +/* { dg-lto-do run } */ +/* { dg-lto-options {"-O2 -flto-partition=max -flto"} } */ +extern void recursive (int *a, int *b, int *c, int level); +int +main() +{ + int x = 123, y=124, z=125; + recursive (&x,&y,&z,1); + if (y) + __builtin_abort (); + if (!__builtin_constant_p (z)) + __builtin_abort (); + return 0; +} diff --git a/gcc/testsuite/gcc.dg/lto/modref-1_1.c b/gcc/testsuite/gcc.dg/lto/modref-1_1.c new file mode 100644 index 00000000000..c7c0eaec971 --- /dev/null +++ b/gcc/testsuite/gcc.dg/lto/modref-1_1.c @@ -0,0 +1,13 @@ +short aa; +void +__attribute__ ((noinline, noclone)) +recursive (int *a, int *b, int *c, int level) +{ + if (level && c) + { + recursive (b,a,c,0); + aa++; + } + else + *a=0; +} diff --git a/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c b/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c new file mode 100644 index 00000000000..9999d37369d --- /dev/null +++ b/gcc/testsuite/gcc.dg/tree-ssa/modref-2.c @@ -0,0 +1,26 @@ +/* { dg-do run } */ +/* { dg-options "-O2" } */ +short aa; +void +__attribute__ ((noinline, noclone)) +recursive (int *a, int *b, int *c, int level) +{ + if (level && c) + { + recursive (b,a,c,0); + aa++; + } + else + *a=0; +} +int +main() +{ + int x = 123, y=124, z=125; + recursive (&x,&y,&z,1); + if (y) + __builtin_abort (); + if (!__builtin_constant_p (z)) + __builtin_abort (); + return 0; +}
Hi, this patchs finishes the parameter tracking by implementing the iterative dataflow in propagation stage. This is necessary since we now can propagate how the pointers are passed around recursive calls (as done in a testcase) and thus it is no-longer safe to simply merge all summaries in one SCC component of the call-graph. cc1plus stats are now: Alias oracle query stats: refs_may_alias_p: 62971744 disambiguations, 73160711 queries ref_maybe_used_by_call_p: 141176 disambiguations, 63867883 queries call_may_clobber_ref_p: 23573 disambiguations, 29322 queries nonoverlapping_component_refs_p: 0 disambiguations, 37720 queries nonoverlapping_refs_since_match_p: 19432 disambiguations, 55659 must overlaps, 75860 queries aliasing_component_refs_p: 54724 disambiguations, 753570 queries TBAA oracle: 24124230 disambiguations 56228428 queries 16058141 are in alias set 0 10338303 queries asked about the same object 125 queries asked about the same alias set 0 access volatile 3919230 are dependent in the DAG 1788399 are aritificially in conflict with void * Modref stats: modref use: 10408 disambiguations, 46993 queries modref clobber: 1418549 disambiguations, 1951251 queries 4898707 tbaa queries (2.510547 per modref query) 396878 base compares (0.203397 per modref query) PTA query stats: pt_solution_includes: 975364 disambiguations, 13604284 queries pt_solutions_intersect: 1026606 disambiguations, 13181198 queries So compared to https://gcc.gnu.org/pipermail/gcc-patches/2020-September/554692.html we get 25% use disambiguations and 91% more clobber disambiguations. Tramp3d is Alias oracle query stats: refs_may_alias_p: 2056905 disambiguations, 2317461 queries ref_maybe_used_by_call_p: 7137 disambiguations, 2093762 queries call_may_clobber_ref_p: 234 disambiguations, 234 queries nonoverlapping_component_refs_p: 0 disambiguations, 4313 queries nonoverlapping_refs_since_match_p: 329 disambiguations, 10200 must overlaps, 10616 queries aliasing_component_refs_p: 858 disambiguations, 34600 queries TBAA oracle: 894996 disambiguations 1695991 queries 138346 are in alias set 0 470668 queries asked about the same object 0 queries asked about the same alias set 0 access volatile 191666 are dependent in the DAG 315 are aritificially in conflict with void * Modref stats: modref use: 842 disambiguations, 2265 queries modref clobber: 14833 disambiguations, 28900 queries 34884 tbaa queries (1.207059 per modref query) 5041 base compares (0.174429 per modref query) PTA query stats: pt_solution_includes: 313372 disambiguations, 525724 queries pt_solutions_intersect: 130374 disambiguations, 415138 queries So about twice many use and 40% clobber disambiguations. Bootstrapped/regtested x86_64-linux, I plan to commit it later today after more testing. 2020-09-26 Jan Hubicka <hubicka@ucw.cz> * ipa-inline-transform.c: Include ipa-modref-tree.h and ipa-modref.h. (inline_call): Call ipa_merge_modref_summary_after_inlining. * ipa-inline.c (ipa_inline): Do not free summaries. * ipa-modref.c (dump_records): Fix formating. (merge_call_side_effects): Break out from ... (analyze_call): ... here; record recursive calls. (analyze_stmt): Add new parameter RECURSIVE_CALLS. (analyze_function): Do iterative dataflow on recursive calls. (compute_parm_map): New function. (ipa_merge_modref_summary_after_inlining): New function. (collapse_loads): New function. (modref_propagate_in_scc): Break out from ... (pass_ipa_modref::execute): ... here; Do iterative dataflow. * ipa-modref.h (ipa_merge_modref_summary_after_inlining): Declare. gcc/testsuite/ChangeLog: 2020-09-26 Jan Hubicka <hubicka@ucw.cz> * gcc.dg/lto/modref-1_0.c: New test. * gcc.dg/lto/modref-1_1.c: New test. * gcc.dg/tree-ssa/modref-2.c: New test.