Dodji Seketeli [Mon, 12 Nov 2012 15:53:25 +0000 (15:53 +0000)]
Instrument built-in memory access function calls
This patch instruments many memory access patterns through builtins.
Basically, for a call like:
__builtin_memset (from, 0, n_bytes);
the patch would only instrument the accesses at the beginning and at
the end of the memory region [from, from + n_bytes]. This is the
strategy used by the llvm implementation of asan.
This instrumentation is done for all the memory access builtin
functions that expose a well specified memory region -- one that
explicitly states the number of bytes accessed in the region.
A special treatment is used for __builtin_strlen. The patch
instruments the access to the first byte of its argument, as well as
the access to the byte (of the argument) at the offset returned by
strlen.
For the __sync_* and __atomic* calls the patch instruments the access
to the bytes pointed to by the argument.
While doing this, I have added a new parameter to build_check_stmt to
decide whether to insert the instrumentation code before or after the
statement iterator. This allows us to do away with the
gsi_{next,prev} dance we were doing in the callers of this function.
Tested by running cc1 -fasan on variations of simple programs like:
int
foo ()
{
char foo[10] = {0};
foo[0] = 't';
foo[1] = 'e';
foo[2] = 's';
foo[3] = 't';
int l = __builtin_strlen (foo);
int n = sizeof (foo);
__builtin_memset (&foo[4], 0, n - 4);
__sync_fetch_and_add (&foo[11], 1);
return l;
}
and by starring at the gimple output which for this function is:
;; Function foo (foo, funcdef_no=0, decl_uid=1714, cgraph_uid=0)
foo ()
{
int n;
int l;
char foo[10];
int D.1725;
char * D.1724;
int D.1723;
long unsigned int D.1722;
int D.1721;
long unsigned int D.1720;
long unsigned int _1;
int _4;
long unsigned int _5;
int _6;
char * _7;
int _8;
char * _9;
unsigned long _10;
unsigned long _11;
unsigned long _12;
signed char * _13;
signed char _14;
_Bool _15;
unsigned long _16;
signed char _17;
_Bool _18;
_Bool _19;
char * _20;
unsigned long _21;
unsigned long _22;
unsigned long _23;
signed char * _24;
signed char _25;
_Bool _26;
unsigned long _27;
signed char _28;
_Bool _29;
_Bool _30;
char * _31;
unsigned long _32;
unsigned long _33;
unsigned long _34;
signed char * _35;
signed char _36;
_Bool _37;
unsigned long _38;
signed char _39;
_Bool _40;
_Bool _41;
char * _42;
unsigned long _43;
unsigned long _44;
unsigned long _45;
signed char * _46;
signed char _47;
_Bool _48;
unsigned long _49;
signed char _50;
_Bool _51;
_Bool _52;
char * _53;
unsigned long _54;
unsigned long _55;
unsigned long _56;
signed char * _57;
signed char _58;
_Bool _59;
unsigned long _60;
signed char _61;
_Bool _62;
_Bool _63;
char[10] * _64;
unsigned long _65;
unsigned long _66;
unsigned long _67;
signed char * _68;
signed char _69;
_Bool _70;
unsigned long _71;
signed char _72;
_Bool _73;
_Bool _74;
unsigned long _75;
unsigned long _76;
unsigned long _77;
signed char * _78;
signed char _79;
_Bool _80;
unsigned long _81;
signed char _82;
_Bool _83;
_Bool _84;
long unsigned int _85;
long unsigned int _86;
char * _87;
char * _88;
unsigned long _89;
unsigned long _90;
unsigned long _91;
signed char * _92;
signed char _93;
_Bool _94;
unsigned long _95;
signed char _96;
_Bool _97;
_Bool _98;
char * _99;
unsigned long _100;
unsigned long _101;
unsigned long _102;
signed char * _103;
signed char _104;
_Bool _105;
unsigned long _106;
signed char _107;
_Bool _108;
_Bool _109;
gcc/
* gimple.h (is_gimple_builtin_call): Declare ...
* gimple.c (is_gimple_builtin_call): ... New public function.
* asan.c (insert_if_then_before_iter, instrument_mem_region_access,
instrument_strlen_call, maybe_instrument_builtin_call,
instrument_call): New static functions.
(create_cond_insert_point): Renamed
create_cond_insert_point_before_iter into this. Add a new
parameter to decide whether to insert the condition before or
after the statement iterator.
(build_check_stmt): Adjust for the new create_cond_insert_point.
Add a new parameter to decide whether to add the instrumentation
code before or after the statement iterator.
(instrument_assignment): Factorize from ...
(transform_statements): ... here. Use maybe_instrument_call to
instrument builtin function calls as well.
(instrument_derefs): Adjust for the new parameter of
build_check_stmt. Fix detection of bit-field access.
Jakub Jelinek [Mon, 12 Nov 2012 15:52:42 +0000 (16:52 +0100)]
Implement protection of global variables
This patch implements the protection of global variables. See the
comments appended to the beginning of the asan.c file.
* varasm.c: Include asan.h.
(assemble_noswitch_variable): Grow size by asan_red_zone_size
if decl is asan protected.
(place_block_symbol): Likewise.
(assemble_variable): If decl is asan protected, increase
DECL_ALIGN if needed, and for decls emitted using
assemble_variable_contents append padding zeros after it.
* Makefile.in (varasm.o): Depend on asan.h.
* asan.c: Include output.h.
(asan_pp, asan_pp_initialized, asan_ctor_statements): New variables.
(asan_pp_initialize, asan_pp_string): New functions.
(asan_emit_stack_protection): Use asan_pp{,_initialized}
instead of local pp{,_initialized} vars, use asan_pp_initialize
and asan_pp_string helpers.
(asan_needs_local_alias, asan_protect_global,
asan_global_struct, asan_add_global): New functions.
(asan_finish_file): Protect global vars that can be protected. Use
asan_ctor_statements instead of ctor_statements
* asan.h (asan_protect_global): New prototype.
(asan_red_zone_size): New inline function.
Co-Authored-By: Wei Mi <wmi@google.com>
From-SVN: r193437
Jakub Jelinek [Mon, 12 Nov 2012 15:52:26 +0000 (16:52 +0100)]
Implement protection of stack variables
This patch implements the protection of stack variables.
It lays out stack variables as well as the different red zones,
emits some prologue code to populate the shadow memory as to poison
(mark as non-accessible) the regions of the red zones and mark the
regions of stack variables as accessible, and emit some epilogue code
to un-poison (mark as accessible) the regions of red zones right
before the function exits.
* Makefile.in (asan.o): Depend on $(EXPR_H) $(OPTABS_H).
(cfgexpand.o): Depend on asan.h.
* asan.c: Include expr.h and optabs.h.
(asan_shadow_set): New variable.
(asan_shadow_cst, asan_emit_stack_protection): New functions.
(asan_init_shadow_ptr_types): Initialize also asan_shadow_set.
* cfgexpand.c: Include asan.h. Define HOST_WIDE_INT heap vector.
(partition_stack_vars): If i is large alignment and j small
alignment or vice versa, break out of the loop instead of continue,
and put the test earlier. If flag_asan, break out of the loop
if for small alignment size is different.
(struct stack_vars_data): New type.
(expand_stack_vars): Add DATA argument. Change PRED type to
function taking size_t argument instead of tree. Adjust pred
calls. Fill DATA in and add needed padding in between variables
if -faddress-sanitizer.
(defer_stack_allocation): Defer everything for flag_asan.
(stack_protect_decl_phase_1, stack_protect_decl_phase_2): Take
size_t index into stack_vars array instead of the decl directly.
(asan_decl_phase_3): New function.
(expand_used_vars): Return var destruction sequence. Adjust
expand_stack_vars calls, add another one for flag_asan. Call
asan_emit_stack_protection if expand_stack_vars added anything
to the vectors.
(expand_gimple_basic_block): Add disable_tail_calls argument.
(gimple_expand_cfg): Pass true to it if expand_used_vars returned
non-NULL. Emit the sequence returned by expand_used_vars after
return_label.
* asan.h (asan_emit_stack_protection): New prototype.
(asan_shadow_set): New decl.
(ASAN_RED_ZONE_SIZE, ASAN_STACK_MAGIC_LEFT, ASAN_STACK_MAGIC_MIDDLE,
ASAN_STACK_MAGIC_RIGHT, ASAN_STACK_FRAME_MAGIC): Define.
(asan_protect_stack_decl): New inline.
* toplev.c (process_options): Also disable -faddress-sanitizer on
!FRAME_GROWS_DOWNWARDS targets.
Jakub Jelinek [Mon, 12 Nov 2012 15:52:09 +0000 (16:52 +0100)]
Allow asan at -O0
This patch defines a new asan pass gate that is activated at -O0, in
addition to the pass that was initially activated at -O3 level The
patch also does some comment cleanups here and there.
* asan.c (build_check_stmt): Rename join_bb variable to else_bb.
(gate_asan_O0): New function.
(pass_asan_O0): New variable.
* passes.c (init_optimization_passes): Add pass_asan_O0.
* tree-pass.h (pass_asan_O0): New declaration.
Jakub Jelinek [Mon, 12 Nov 2012 15:51:53 +0000 (16:51 +0100)]
Emit GIMPLE directly instead of gimplifying GENERIC.
This patch cleanups the instrumentation code generation by emitting
GIMPLE directly, as opposed to emitting GENERIC tree and then
gimplifying them. It also does some cleanups here and there
* Makefile.in (GTFILES): Add $(srcdir)/asan.c.
(asan.o): Update the dependencies of asan.o.
* asan.c (tm.h, tree.h, tm_p.h, basic-block.h, flags.h
function.h, tree-inline.h, tree-dump.h, diagnostic.h, demangle.h,
langhooks.h, ggc.h, cgraph.h, gimple.h): Remove these unused but
included headers.
(shadow_ptr_types): New variable.
(report_error_func): Change is_store argument to bool, don't append
newline to function name.
(PROB_VERY_UNLIKELY, PROB_ALWAYS): Define.
(build_check_stmt): Change is_store argument to bool. Emit GIMPLE
directly instead of creating trees and gimplifying them. Mark
the error reporting function as very unlikely.
(instrument_derefs): Change is_store argument to bool. Use
int_size_in_bytes to compute size_in_bytes, simplify size check.
Use build_fold_addr_expr instead of build_addr.
(transform_statements): Adjust instrument_derefs caller.
Use gimple_assign_single_p as stmt test. Don't look at MEM refs
in rhs2.
(asan_init_shadow_ptr_types): New function.
(asan_instrument): Don't push/pop gimplify context.
Call asan_init_shadow_ptr_types if not yet initialized.
* asan.h (ASAN_SHADOW_SHIFT): Adjust comment.
Co-Authored-By: Dodji Seketeli <dodji@redhat.com> Co-Authored-By: Xinliang David Li <davidxl@google.com>
From-SVN: r193434
Jakub Jelinek [Mon, 12 Nov 2012 15:51:34 +0000 (16:51 +0100)]
Initial asan cleanups
This patch defines a new asan_shadow_offset target macro, instead of
having a mere macro in the asan.c file. It becomes thus cleaner to
define the target macro for targets that supports asan, namely x86 for
now. The ASAN_SHADOW_SHIFT (which, along with the asan_shadow_offset
constant, is used to compute the address of the shadow memory byte for
a given memory address) is defined in asan.h.
gcc/ChangeLog
* toplev.c (process_options): Warn and turn off
-faddress-sanitizer if not supported by target.
* asan.c: Include target.h.
(asan_scale, asan_offset_log_32, asan_offset_log_64,
asan_offset_log): Removed.
(build_check_stmt): Use ASAN_SHADOW_SHIFT and
targetm.asan_shadow_offset ().
(asan_instrument): Don't initialize asan_offset_log.
* asan.h (ASAN_SHADOW_SHIFT): Define.
* target.def (TARGET_ASAN_SHADOW_OFFSET): New hook.
* doc/tm.texi.in (TARGET_ASAN_SHADOW_OFFSET): Add it.
* doc/tm.texi: Regenerated.
* Makefile.in (asan.o): Depend on $(TARGET_H).
* config/i386/i386.c (ix86_asan_shadow_offset): New function.
(TARGET_ASAN_SHADOW_OFFSET): Define.
Wei Mi [Mon, 12 Nov 2012 15:51:13 +0000 (15:51 +0000)]
Initial import of asan from the Google branch
This patch imports the initial state of asan as it was in the
Google branch.
It provides basic infrastructure for asan to instrument memory
accesses on the heap, at -O3. Note that it supports neither stack nor
global variable protection.
The rest of the patches of the set is intended to further improve this
base.
gcc/ChangeLog
* Makefile.in: Add asan.c and its dependencies.
* common.opt: Add -faddress-sanitizer option.
* invoke.texi: Document the new flag.
* passes.c: Add the asan pass.
* toplev.c (compile_file): Call asan_finish_file.
* asan.c: New file.
* asan.h: New file.
* tree-pass.h: Declare pass_asan.
Co-Authored-By: Diego Novillo <dnovillo@google.com> Co-Authored-By: Dodji Seketeli <dodji@redhat.com>
From-SVN: r193432
Bin Cheng [Mon, 12 Nov 2012 02:17:38 +0000 (02:17 +0000)]
gcse.c (struct bb_data): Add new fields, old_pressure, live_in and backup.
* gcse.c (struct bb_data): Add new fields, old_pressure, live_in
and backup.
(get_regno_pressure_class): Add prototype.
(update_bb_reg_pressure): New.
(should_hoist_expr_to_dom): Add new parameter from.
Monitor the change of reg pressure and use it to drive hoisting.
(hoist_code): Update LIVE and reg pressure information.
(calculate_bb_reg_pressure): Initialize live_in and backup.
Steven Bosscher [Sun, 11 Nov 2012 20:29:11 +0000 (20:29 +0000)]
re PR middle-end/55263 (ICE: pre_and_rev_post_order_compute, at cfganal.c:875 with -O -fgcse-after-reload -fnon-call-exceptions)
gcc/
PR middle-end/55263
* postreload.c (rest_of_handle_postreload): With non-call exceptions,
if edges are purged call cleanup_cfg to remove unreachable blocks.
testsuite/
PR middle-end/55263
* g++.dg/pr55263.C: New test.
gcc/
* doc/extend.texi: Copy-edit to use "built-in function" instead
of "builtin", per GCC coding standards. Also use consistent
capitalization in titles of sections describing built-in functions.
Martin Jambor [Sun, 11 Nov 2012 19:35:27 +0000 (20:35 +0100)]
re PR tree-optimization/55238 (ICE in find_aggregate_values_for_callers_subset, at ipa-cp.c:2908 building zlib)
PR tree-optimization/55238
* ipa-cp.c (agg_pass_through_permissible_p): New function.
(propagate_aggs_accross_jump_function): Use it.
(find_aggregate_values_for_callers_subset): Likewise and relax an
assert.
Uros Bizjak [Sun, 11 Nov 2012 19:17:17 +0000 (20:17 +0100)]
re PR target/47440 (Use LCM for vzeroupper insertion)
PR target/47440
* config/i386/i386.c (check_avx256_stores): Remove.
(ix86_check_avx256_register): New.
(ix86_avx_u128_mode_needed): Use ix86_check_avx256_register.
Check the whole RTX for 256bit registers using for_each_rtx.
(ix86_check_avx_stores): New.
(ix86_avx_u128_mode_after): Change mode of CALL RTX to AVX_U128_CLEAN
if there are no 256bit registers used in the function return register.
(ix86_avx_u128_mode_entry): Use ix86_check_avx256_register.
(ix86_avx_u128_mode_exit): Ditto.
Eric Botcazou [Sat, 10 Nov 2012 15:32:32 +0000 (15:32 +0000)]
expr.c (store_field): Remove TYPE parameter.
* expr.c (store_field): Remove TYPE parameter. Remove block of code
dealing with BLKmode in registers. Reimplement this support using
pseudo-registers and bit-field techniques.
(store_constructor_field): Remove TYPE parameter and adjust calls to
store_field.
(expand_assignment): Adjust calls to store_field. Add comment.
(store_expr): Add comment.
(store_constructor): Adjust calls to store_constructor_field.
(expand_expr_real_2): Adjust call to store_field.
Ed Smith-Rowland [Sat, 10 Nov 2012 00:08:49 +0000 (00:08 +0000)]
Implement a flag -fext-numeric-literals that allows control of whether GNU...
Implement a flag -fext-numeric-literals that allows control of whether GNU
numeric suffix extensions are parsed or passed to C++ as user-defined literals.
re PR rtl-optimization/55154 (ICE: in curr_insn_transform, at lra-constraints.c:2702 with custom flags)
2012-11-09 Vladimir Makarov <vmakarov@redhat.com>
PR tree-optimization/55154
* lra-int.h (LRA_LOSER_COST_FACTOR, LRA_MAX_REJECT): New macros.
* lra.c (setup_operand_alternative): Use them.
* lra-constraints.c (LOSER_COST_FACTOR, MAX_OVERALL_COST_BOUND):
Remove.
(process_alt_operands): Use LRA_LOSER_COST_FACTOR and
LRA_MAX_REJECT. Accumulate reject instead of setting for
non-const.
(curr_insn_transform): Initialize best_losers and best_overall by
INT_MAX.
2012-11-09 Vladimir Makarov <vmakarov@redhat.com>
PR rtl-optimization/55154
* gcc.target/i386/pr55154.c: New test.
Steven Bosscher [Fri, 9 Nov 2012 12:31:32 +0000 (12:31 +0000)]
re PR middle-end/54385 (ICE in pre_and_rev_post_order_compute, at cfganal.c:873)
PR middle-end/54385
* postreload.c (reload_cse_simplify): Return a bool indicating
whether the CFG was changed.
(reload_cse_regs_1): Traverse the CFG instead of the insns chain.
Cleanup the CFG if edges may have been removed.
(reload_cse_regs): Update.
Eric Botcazou [Fri, 9 Nov 2012 11:07:30 +0000 (11:07 +0000)]
trans.c (gnat_to_gnu): Deal with a missing library unit instead of calling gnat_to_gnu on Emtpy.
* gcc-interface/trans.c (gnat_to_gnu) <N_Subprogram_Body_Stub>: Deal
with a missing library unit instead of calling gnat_to_gnu on Emtpy.
Remove last bits of support for SCIL nodes. Add cases for nodes that
should only be present when annotating types. Add default abort.
Jakub Jelinek [Thu, 8 Nov 2012 20:20:10 +0000 (21:20 +0100)]
re PR debug/53145 (gcc.dg/pch/save-temps-1.c)
PR debug/53145
* dwarf2out.c (gen_compile_unit_die): Don't call gen_producer_string
here, instead add "" if producer_string is NULL.
(dwarf2out_finish): Call gen_producer_string here, unconditionally
decrease refcount of the old indirect string and set val_str to
find_AT_string result.
Jan Hubicka [Thu, 8 Nov 2012 16:46:18 +0000 (17:46 +0100)]
re PR fortran/48636 (Enable more inlining with -O2 and higher)
PR middle-end/48636
* ipa-inline.c (big_speedup_p): New function.
(want_inline_small_function_p): Use it.
(edge_badness): Dump it.
* params.def (inline-min-speedup): New parameter.
* doc/invoke.texi (inline-min-speedup): Document.
David Edelsohn [Wed, 7 Nov 2012 21:03:08 +0000 (21:03 +0000)]
pr46728-[1234578].c: Tighten regex to ignore word powerpc.
* gcc.target/powerpc/pr46728-[1234578].c: Tighten regex to ignore
word powerpc.
* gcc.target/powerpc/vsx-mass-1.c: Escape [ and ] in regex.
* gcc.target/powerpc/loop_align.c: Skip on AIX.
* gcc.target/powerpc/ppc-pow.c: Allow dot symbols in branch.
* gcc.target/powerpc/tfmode_off.c: Skip on AIX.
Uros Bizjak [Wed, 7 Nov 2012 19:35:25 +0000 (20:35 +0100)]
re PR target/55224 (FAIL: gcc.target/i386/tailcall-1.c scan-assembler jmp)
PR target/55224
* config/i386/i386.c (ix86_function_ok_for_sibcall): Put back exception
to make a sibcall if one of the functions has void return type.
Uros Bizjak [Wed, 7 Nov 2012 19:30:48 +0000 (20:30 +0100)]
re PR target/55224 (FAIL: gcc.target/i386/tailcall-1.c scan-assembler jmp)
PR target/55224
* config/i386/i386.c (ix86_function_ok_for_sibcall): Put back exception
to make a sibcall if one of the functions has void return type.
re PR c/53063 (encode group options in the .opt files)
c-family/
2012-11-07 Manuel López-Ibáñez <manu@gcc.gnu.org>
PR c/53063
* c.opt(Warray-bounds,Wdelete-non-virtual-dtor,Wenum-compare,
Wmain,Woverlength-strings, Wunknown-pragmas,Wunused-macros):
Use LangEnabledBy.
(Wswitch,Wswitch-default,Wswitch-enum): Likewise. Move here from
common.opt.
(Wvariadic-macros): Init(1).
* c-opts.c (c_common_handle_option): Do not handle them
explicitly.
(c_common_post_options): Likewise.
(sanitize_cpp_opts): warn_unused_macros is now
cpp_warn_unused_macros.
(push_command_line_include): Likewise.
* c-common.c (warn_unknown_pragmas): Do not define.
* c-common.h (warn_unknown_pragmas): Do not declare.
gcc/
2012-11-07 Manuel López-Ibáñez <manu@gcc.gnu.org>
PR c/53063
* common.opt (Wswitch,Wswitch-default,Wswitch-enum): Move to c.opt.
fortran/
2012-11-07 Manuel López-Ibáñez <manu@gcc.gnu.org>
PR c/53063
* options.c (set_Wall): Do not set warn_switch here.
Martin Jambor [Wed, 7 Nov 2012 15:55:54 +0000 (16:55 +0100)]
re PR tree-optimization/53787 (Possible IPA-SRA / IPA-CP improvement)
2012-11-07 Martin Jambor <mjambor@suse.cz>
PR tree-optimization/53787
* ipa-cp.c (ipcp_value_source): New field offset.
(ipcp_agg_lattice): New type.
(ipcp_param_lattices): Likewise, move virt_call from ipcp_lattice here.
(ipcp_agg_lattice_pool): New variable.
(ipa_get_parm_lattices): New function.
(ipa_get_lattice): Turned into ipa_get_scalar_lat, use the above.
Adjusted all callers.
(print_lattice): New function.
(print_all_lattices): Use the above, also print aggregate lattices.
(set_agg_lats_to_bottom): New function.
(set_agg_lats_contain_variable): Likewise.
(set_all_contains_variable): Likewise.
(initialize_node_lattices): Also handle aggregate lattices, set
virt_call in ipcp_param_lattices.
(add_value_source): Handle offsets.
(add_value_to_lattice): Likewise.
(add_scalar_value_to_lattice): New function.
(propagate_vals_accross_pass_through): Use add_scalar_value_to_lattice.
(propagate_vals_accross_ancestor): Likewise.
(propagate_accross_jump_function): Renamed to
propagate_scalar_accross_jump_function, use
add_scalar_value_to_lattice.
(set_check_aggs_by_ref): New function.
(merge_agg_lats_step): Likewise.
(set_chain_of_aglats_contains_variable): Likewise.
(merge_aggregate_lattices): Likewise.
(propagate_constants_accross_call): Also handle aggregate lattices.
(hint_time_bonus): New function.
(context_independent_aggregate_values): Likewise.
(gather_context_independent_values): Also handle agggregate values.
(agg_jmp_p_vec_for_t_vec): New function.
(estimate_local_effects): Also handle agggregate values.
(add_all_node_vals_to_toposort): Likewise.
(ipcp_propagate_stage): Use struct ipcp_param_lattices.
(get_clone_agg_value): New function.
(cgraph_edge_brings_value_p): Also handle agggregate values.
(create_specialized_node): Likewise.
(find_more_values_for_callers_subset): Rename to
find_more_scalar_values_for_callers_subset. Modify dump.
(copy_plats_to_inter): New function.
(intersect_with_plats): Likewise.
(agg_replacements_to_vector): Likewise.
(intersect_with_agg_replacements): Likewise.
(find_aggregate_values_for_callers_subset): Likewise.
(known_aggs_to_agg_replacement_list): Likewise.
(cgraph_edge_brings_all_scalars_for_node): Likewise.
(cgraph_edge_brings_all_agg_vals_for_node): Likewise.
(perhaps_add_new_callers): Old functionality moved to
cgraph_edge_brings_all_scalars_for_node, call it and
cgraph_edge_brings_all_agg_vals_for_node.
(ipcp_val_in_agg_replacements_p): New function.
(decide_about_value): New function.
(decide_whether_version_node): A lot of functionality moved to
decide_about_value. Also handle agggregate values.
(ipcp_driver): Also allocate ipcp_agg_lattice_pool.
(pass_ipa_cp): Fill in new entries.
* ipa-prop.c (ipa_node_agg_replacements): New variable.
(free_parms_ainfo): New function.
(ipa_analyze_node): Use free_parms_ainfo to free stuff.
(ipa_find_agg_cst_for_param): Do not rely on offset ordering.
(ipa_set_node_agg_value_chain): New function.
(ipa_node_removal_hook): Also handle ipa_node_agg_replacements.
(ipa_node_duplication_hook): Likewise.
(ipa_free_all_structures_after_ipa_cp): Also free ipcp_agg_lattice_pool.
(ipa_free_all_structures_after_iinln): Likewise.
(ipa_dump_agg_replacement_values): New function.
(write_agg_replacement_chain): Likewise.
(read_agg_replacement_chain): Likewise.
(ipa_prop_write_all_agg_replacement): Likewise.
(read_replacements_section): Likewise.
(ipa_prop_read_all_agg_replacement): Likewise.
(adjust_agg_replacement_values): Likewise.
(ipcp_transform_function): Likewise.
* ipa-prop.h: Also define heap vector of ipa_agg_jf_item_t and of
ipa_agg_jump_function_t.
(ipa_node_params): Make lattices an array of ipcp_param_lattices.
(ipa_agg_replacement_value): New type and its vector.
(ipa_set_node_agg_value_chain) Declare.
(ipa_node_agg_replacements): Likewise.
(ipa_get_agg_replacements_for_node): New function.
(ipcp_agg_lattice_pool): Declare.
(ipa_dump_agg_replacement_values): Likewise.
(ipa_prop_write_all_agg_replacement): Likewise.
(ipa_prop_read_all_agg_replacement): Likewise.
(ipcp_transform_function): Likewise.
* ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time): Pass around
known aggregates and hints.
* ipa-inline.h: include ipa-prop.h.
(estimate_ipcp_clone_size_and_time): Adjust declaration.
* lto-streamer.h (lto_section_type): New item
LTO_section_ipcp_transform.
* lto-section-in.c (lto_section_name): New element ipcp_trans.
* params.def (PARAM_IPA_CP_LOOP_HINT_BONUS): New parameter.
* Makefile.in (IPA_INLINE_H): New. Use everywhee instead of
ipa-inline.h.