2004-09-06 Paolo Bonzini <bonzini@gnu.org> Unify the management of RTL and tree-level dump files. * cfgexpand.c (tree_expand_cfg): Fix incorrect comment. Don't print function name to the dump file, the pass manager would do this for us. Add code from the top of rest_of_compilation, up to the initial RTL dump. * passes.c (rest_of_handle_jump): Call fixup_tail_calls and close the DFI_sibling dump file. (rest_of_compilation): Don't do that here. Remove code up to the initial RTL dump. (init_optimization_passes): Remove. (pass_rest_of_compilation): Change pass name to NULL. * toplev.c (lang_dependent_init): Do not use an empty dump file prefix. Do not call init_optimization_passes. * toplev.h (init_optimization_passes): Remove. * graph.c (print_rtl_graph_with_bb, clean_graph_dump_file, finish_graph_dump_file): Remove SUFFIX parameter. * graph.h (print_rtl_graph_with_bb, clean_graph_dump_file, finish_graph_dump_file): Likewise. * tree-pass.h (struct tree_opt_pass): Add `letter' field. * cfgexpand.c (pass_expand): Adjust. * gimple-low.c (pass_lower_cf, pass_remove_useless_vars): Adjust. * passes.c (pass_rest_of_compilation): Adjust. * predict.c (pass_profile): Adjust. * tree-alias-common.c (pass_build_pta, pass_del_pta): Adjust. * tree-cfg.c (pass_build_cfg, pass_remove_useless_stmts, pass_split_crit_edges, pass_warn_function_return): Adjust. * tree-complex.c (pass_lower_vector_ssa, pass_pre_expand): Adjust. * tree-dfa.c (pass_referenced_vars): Adjust. * tree-eh.c (pass_lower_eh): Adjust. * tree-if-conv.c (pass_build_ssa): Adjust. * tree-into-ssa.c (pass_build_ssa): Adjust. * tree-mudflap.c (pass_mudflap_1, pass_mudflap_2): Adjust. * tree-nomudflap.c (pass_mudflap_1, pass_mudflap_2): Adjust. * tree-nrv.c (pass_nrv): Adjust. * tree-optimize.c (pass_gimple, pass_all_optimizations, pass_cleanup_cfg_post_optimizing, pass_free_datastructures, pass_init_datastructures): Adjust. * tree-outof-ssa.c (pass_del_ssa): Adjust. * tree-profile.c (pass_tree_profile): Adjust. * tree-sra.c (pass_sra): Adjust. * tree-ssa-alias.c (pass_may_alias): Adjust. * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Adjust. * tree-ssa-copyrename.c (pass_rename_ssa_copies): Adjust. * tree-ssa-dce.c (pass_dce, pass_cd_dce): Adjust. * tree-ssa-dom.c (pass_dominator): Adjust. * tree-ssa-dse.c (pass_dse): Adjust. * tree-ssa-forwprop.c (pass_forwprop): Adjust. * tree-ssa-if-conv.c (pass_if_conversion): Adjust. * tree-ssa-loop-ch.c (pass_ch): Adjust. * tree-ssa-loop.c (pass_loop, pass_loop_init, pass_lim, pass_loop_done, pass_complete_unroll, pass_iv_canon, pass_iv_optimize, pass_vectorize): Adjust. * tree-ssa-phiopt.c (pass_phiopt): Adjust. * tree-ssa-pre.c (pass_pre, pass_fre): Adjust. * tree-ssa.c (pass_redundant_phi, pass_early_warn_uninitialized, pass_late_warn_uninnitialized): Adjust. * tree-tailcall.c (pass_tail_recursion, pass_tail_calls): Adjust. * Makefile.in (tree-dump.o): Add new dependencies. * cgraph.c (cgraph_remove_node): TDF_all -> TDF_tree_all. * cgraphunit.c (cgraph_preserve_function_body_p, cgraph_optimize): Likewise. * toplev.c (dump_file_name): New. * tree-dump.c (dump_enable_all): Add LETTER parameter. (struct dump_file_info): Add NUM and LETTER fields. (dump_files): Adjust and add RTL dump files. (dump_register): Add NUM and LETTER fields. (get_dump_file_name, dump_initialized_p, enable_rtl_dump_file): New. (dump_begin): Use get_dump_file_name. (dump_switch_p_1): Adjust call to dump_enable_all. * tree-dump.h (dump_register): Adjust prototype. * tree-optimize.c (register_one_dump_file): Take dump file index. Support flags for RTL dumps. (register_dump_files): Fill in NUM field of struct dump_file_info. Track properties both when the gate is executed and when it is not. (execute_todo): Dump RTL. Add PROPERTIES parameter. (execute_one_pass): Pass properties to execute_todo. Handle VCG dumps of RTL. * tree-pass.h (dump_file_name): New. * tree.h (TDF_TREE, TDF_RTL, get_dump_file_name, dump_initialized_p): New. * Makefile.in (passes.o): Add new dependencies. * passes.c (struct dump_file_info, enum dump_file_index, dump_file_tbl, init_optimization_passes): Remove. (open_dump_file, close_dump_file): Use tree-dumping infrastructure. (rest_of_handle_new_regalloc, rest_of_handle_old_regalloc): Use dump_enabled_p. (finish_optimization_passes): Update finish_graph_dump_file loop. (enable_rtl_dump_file): Remove. * tree-dump.c (dump_files): Adjust and add RTL dump files. (enable_rtl_dump_file): Add here. * tree.h (enum tree_dump_index): Add RTL dump file indices. * doc/invoke.texi (Debugging options): Document new RTL debugging options. Update. From-SVN: r87113
220 lines
6.9 KiB
C
220 lines
6.9 KiB
C
/* Language independent return value optimizations
|
|
Copyright (C) 2004 Free Software Foundation, Inc.
|
|
|
|
This file is part of GCC.
|
|
|
|
GCC is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 2, or (at your option)
|
|
any later version.
|
|
|
|
GCC is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with GCC; see the file COPYING. If not, write to
|
|
the Free Software Foundation, 59 Temple Place - Suite 330,
|
|
Boston, MA 02111-1307, USA. */
|
|
|
|
#include "config.h"
|
|
#include "system.h"
|
|
#include "coretypes.h"
|
|
#include "tm.h"
|
|
#include "tree.h"
|
|
#include "rtl.h"
|
|
#include "function.h"
|
|
#include "basic-block.h"
|
|
#include "expr.h"
|
|
#include "diagnostic.h"
|
|
#include "tree-flow.h"
|
|
#include "timevar.h"
|
|
#include "tree-dump.h"
|
|
#include "tree-pass.h"
|
|
#include "langhooks.h"
|
|
|
|
/* This file implements return value optimizations for functions which
|
|
return aggregate types.
|
|
|
|
Basically this pass searches the function for return statements which
|
|
return a local aggregate. When converted to RTL such statements will
|
|
generate a copy from the local aggregate to final return value destination
|
|
mandated by the target's ABI.
|
|
|
|
That copy can often be avoided by directly constructing the return value
|
|
into the final destination mandated by the target's ABI.
|
|
|
|
This is basically a generic equivalent to the C++ front-end's
|
|
Named Return Value optimization. */
|
|
|
|
struct nrv_data
|
|
{
|
|
/* This is the temporary (a VAR_DECL) which appears in all of
|
|
this function's RETURN_EXPR statements. */
|
|
tree var;
|
|
|
|
/* This is the function's RESULT_DECL. We will replace all occurrences
|
|
of VAR with RESULT_DECL when we apply this optimization. */
|
|
tree result;
|
|
};
|
|
|
|
static tree finalize_nrv_r (tree *, int *, void *);
|
|
|
|
/* Callback for the tree walker.
|
|
|
|
If TP refers to a RETURN_EXPR, then set the expression being returned
|
|
to nrv_data->result.
|
|
|
|
If TP refers to nrv_data->var, then replace nrv_data->var with
|
|
nrv_data->result.
|
|
|
|
If we reach a node where we know all the subtrees are uninteresting,
|
|
then set *WALK_SUBTREES to zero. */
|
|
|
|
static tree
|
|
finalize_nrv_r (tree *tp, int *walk_subtrees, void *data)
|
|
{
|
|
struct nrv_data *dp = (struct nrv_data *)data;
|
|
|
|
/* No need to walk into types. */
|
|
if (TYPE_P (*tp))
|
|
*walk_subtrees = 0;
|
|
|
|
/* If this is a RETURN_EXPR, set the expression being returned to RESULT. */
|
|
else if (TREE_CODE (*tp) == RETURN_EXPR)
|
|
TREE_OPERAND (*tp, 0) = dp->result;
|
|
|
|
/* Otherwise replace all occurrences of VAR with RESULT. */
|
|
else if (*tp == dp->var)
|
|
*tp = dp->result;
|
|
|
|
/* Keep iterating. */
|
|
return NULL_TREE;
|
|
}
|
|
|
|
/* Main entry point for return value optimizations.
|
|
|
|
If this function always returns the same local variable, and that
|
|
local variable is an aggregate type, then replace the variable with
|
|
the function's DECL_RESULT.
|
|
|
|
This is the equivalent of the C++ named return value optimization
|
|
applied to optimized trees in a language independent form. If we
|
|
ever encounter languages which prevent this kind of optimization,
|
|
then we could either have the languages register the optimization or
|
|
we could change the gating function to check the current language. */
|
|
|
|
static void
|
|
tree_nrv (void)
|
|
{
|
|
tree result = DECL_RESULT (current_function_decl);
|
|
tree result_type = TREE_TYPE (result);
|
|
tree found = NULL;
|
|
basic_block bb;
|
|
struct nrv_data data;
|
|
|
|
/* If this function does not return an aggregate type in memory, then
|
|
there is nothing to do. */
|
|
if (!aggregate_value_p (result, current_function_decl))
|
|
return;
|
|
|
|
/* Look through each block for suitable return expressions. RETURN_EXPRs
|
|
end basic blocks, so we only have to look at the last statement in
|
|
each block. That makes this very fast. */
|
|
FOR_EACH_BB (bb)
|
|
{
|
|
tree stmt = last_stmt (bb);
|
|
|
|
if (stmt && TREE_CODE (stmt) == RETURN_EXPR)
|
|
{
|
|
tree ret_expr = TREE_OPERAND (stmt, 0);
|
|
|
|
/* This probably should not happen, but just to be safe do
|
|
not perform NRV optimizations if only some of the return
|
|
statement return a value. */
|
|
if (!ret_expr
|
|
|| TREE_CODE (ret_expr) != MODIFY_EXPR
|
|
|| TREE_CODE (TREE_OPERAND (ret_expr, 0)) != RESULT_DECL)
|
|
return;
|
|
|
|
/* Now verify that this return statement uses the same value
|
|
as any previously encountered return statement. */
|
|
if (found != NULL)
|
|
{
|
|
/* If we found a return statement using a different variable
|
|
than previous return statements, then we can not perform
|
|
NRV optimizations. */
|
|
if (found != TREE_OPERAND (ret_expr, 1))
|
|
return;
|
|
}
|
|
else
|
|
found = TREE_OPERAND (ret_expr, 1);
|
|
|
|
/* The returned value must be a local automatic variable of the
|
|
same type and alignment as the function's result. */
|
|
if (TREE_CODE (found) != VAR_DECL
|
|
|| DECL_CONTEXT (found) != current_function_decl
|
|
|| TREE_STATIC (found)
|
|
|| TREE_ADDRESSABLE (found)
|
|
|| DECL_ALIGN (found) > DECL_ALIGN (result)
|
|
|| !lang_hooks.types_compatible_p (TREE_TYPE (found),
|
|
result_type))
|
|
return;
|
|
}
|
|
}
|
|
|
|
if (!found)
|
|
return;
|
|
|
|
/* If dumping details, then note once and only the NRV replacement. */
|
|
if (dump_file && (dump_flags & TDF_DETAILS))
|
|
{
|
|
fprintf (dump_file, "NRV Replaced: ");
|
|
print_generic_expr (dump_file, found, dump_flags);
|
|
fprintf (dump_file, " with: ");
|
|
print_generic_expr (dump_file, result, dump_flags);
|
|
fprintf (dump_file, "\n");
|
|
}
|
|
|
|
/* At this point we know that all the return statements return the
|
|
same local which has suitable attributes for NRV. Copy debugging
|
|
information from FOUND to RESULT. */
|
|
DECL_NAME (result) = DECL_NAME (found);
|
|
DECL_SOURCE_LOCATION (result) = DECL_SOURCE_LOCATION (found);
|
|
DECL_ABSTRACT_ORIGIN (result) = DECL_ABSTRACT_ORIGIN (found);
|
|
TREE_ADDRESSABLE (result) = TREE_ADDRESSABLE (found);
|
|
|
|
/* Now walk through the function changing all references to VAR to be
|
|
RESULT. */
|
|
data.var = found;
|
|
data.result = result;
|
|
FOR_EACH_BB (bb)
|
|
{
|
|
block_stmt_iterator bsi;
|
|
|
|
for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi))
|
|
walk_tree (bsi_stmt_ptr (bsi), finalize_nrv_r, &data, 0);
|
|
}
|
|
|
|
/* FOUND is no longer used. Ensure it gets removed. */
|
|
var_ann (found)->used = 0;
|
|
}
|
|
|
|
struct tree_opt_pass pass_nrv =
|
|
{
|
|
"nrv", /* name */
|
|
NULL, /* gate */
|
|
tree_nrv, /* execute */
|
|
NULL, /* sub */
|
|
NULL, /* next */
|
|
0, /* static_pass_number */
|
|
TV_TREE_NRV, /* tv_id */
|
|
PROP_cfg, /* properties_required */
|
|
0, /* properties_provided */
|
|
0, /* properties_destroyed */
|
|
0, /* todo_flags_start */
|
|
TODO_dump_func | TODO_ggc_collect, /* todo_flags_finish */
|
|
0 /* letter */
|
|
};
|