Updated DW_OP_GNU_entry_value/DW_TAG_GNU_call_site patch

On Tue, Mar 15, 2011 at 4:19 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> Hi!
>
> Now that we are back in stage 1, I'd like move on with the
> entry_value/call_site debug info extensions.
>
> Here is the http://gcc.gnu.org/ml/gcc-patches/2010-12/msg01793.html
> patch updated to current trunk, bootstrapped/regtested on x86_64-linux
> and i686-linux.  Ok for trunk?
>
> 2011-03-15  Jakub Jelinek  <jakub@redhat.com>
>
>        * final.c (final_scan_insn): Handle NOTE_INSN_CALL_ARG_LOCATION.
>        Call var_location debug hook even on CALL_INSNs.
>        (rest_of_clean_state): Don't print NOTE_INSN_CALL_ARG_LOCATION.
>        * rtl.def (ENTRY_VALUE): New.
>        * dwarf2out.c: Include cfglayout.h.
>        (dwarf_stack_op_name, size_of_loc_descr, output_loc_operands,
>        output_loc_operands_raw): Handle DW_OP_GNU_entry_value.
>        (struct call_arg_loc_node): New type.
>        (call_arg_locations, call_arg_loc_last, block_map, call_site_count,
>        tail_call_site_count): New variables.
>        (dwarf_tag_name): Handle DW_TAG_GNU_call_site and
>        DW_TAG_GNU_call_site_parameter.
>        (dwarf_attr_name): Handle DW_AT_GNU_call_site_value,
>        DW_AT_GNU_call_site_data_value, DW_AT_GNU_call_site_target,
>        DW_AT_GNU_call_site_target_clobbered, DW_AT_GNU_tail_call,
>        DW_AT_GNU_all_tail_call_sites, DW_AT_GNU_all_call_sites
>        and DW_AT_GNU_all_source_call_sites.
>        (mem_loc_descriptor): Handle ENTRY_VALUE.
>        (add_src_coords_attributes): Don't add enything if
>        DECL_SOURCE_LOCATION is UNKNOWN_LOCATION.
>        (dwarf2out_abstract_function): Save and clear call_arg_location,
>        call_site_count and tail_call_site_count around dwarf2out_decl call.
>        (gen_call_site_die): New function.
>        (gen_subprogram_die): Emit DW_TAG_GNU_call_site DIEs for call sites.
>        (gen_lexical_block_die, gen_inlined_subroutine_die): Update block_map.
>        (dwarf2out_function_decl): Clear call_arg_locations,
>        call_arg_loc_last, set call_site_count and tail_call_site_count
>        to -1 and free block_map.
>        (dwarf2out_var_location): Handle NOTE_INSN_CALL_ARG_LOCATION and
>        CALL_INSNs.  Add NOTE_DURING_CALL_P var location notes even when not
>        followed by any real instructions.
>        (dwarf2out_begin_function): Set call_site_count and
>        tail_call_site_count to 0.
>        (resolve_addr): If DW_AT_abstract_origin of DW_TAG_GNU_call_site
>        is dw_val_class_addr, attempt to look it up again, for DECL_EXTERNAL
>        attempt to force a DIE for it and worst case remove the attribute.
>        (resolve_one_addr): For TREE_CONSTANT_POOL_ADDRESS_P SYMBOL_REFs
>        check TREE_ASM_WRITTEN of DECL_INITIAL of the decl instead of
>        the decl itself.
>        * var-tracking.c: Include tm_p.h.
>        (vt_stack_adjustments): For calls call note_register_arguments.
>        (argument_reg_set): New variable.
>        (add_stores): For MO_VAL_SET of non-tracked regs from argument_reg_set
>        ensure the VALUE is resolved.
>        (call_arguments): New variable.
>        (prepare_call_arguments): New function.
>        (add_with_sets): For MO_CALL set u.loc from call_arguments and clear it.
>        (struct expand_loc_callback_data): Add ignore_cur_loc field.
>        (vt_expand_loc_callback): If ignore_cur_loc, don't look at cur_loc and
>        always use the best expression.
>        (vt_expand_loc): Add ignore_cur_loc argument.
>        (vt_expand_loc_dummy): Clear ignore_cur_loc field.
>        (emit_note_insn_var_location): Adjust vt_expand_loc callers.
>        (emit_notes_in_bb) <case MO_CALL>: Add NOTE_INSN_CALL_ARG_LOCATION
>        note for all calls.
>        (vt_add_function_parameter): Use cselib_lookup_from_insn.
>        If dv is a VALUE, enter into hash table also ENTRY_VALUE for the
>        argument.  Don't call cselib_preserve_only_values and
>        cselib_reset_table.
>        (note_register_arguments): New function.
>        (vt_initialize): Compute argument_reg_set.  Call
>        vt_add_function_parameters before processing basic blocks instead of
>        afterwards.  For calls call prepare_call_arguments before calling
>        cselib_process_insn.
>        * print-rtl.c (print_rtx): Handle NOTE_INSN_CALL_ARG_LOCATION.
>        * Makefile.in (dwarf2out.o): Depend on $(CFGLAYOUT_H).
>        (var-tracking.o): Depend on $(TM_P_H).
>        * cfglayout.h (insn_scope): New prototype.
>        * gengtype.c (adjust_field_rtx_def): Handle NOTE_INSN_CALL_ARG_LOCATION.
>        * cfglayout.c (insn_scope): No longer static.
>        * insn-notes.def (CALL_ARG_LOCATION): New.
>        * calls.c (expand_call, emit_library_call_value_1): Put USEs for
>        MEM arguments into CALL_INSN_FUNCTION_USAGE unconditionally.
>        * integrate.c (set_block_origin_self, set_block_abstract_flags): Do
>        nothing for DECL_EXTERNAL BLOCK_VARS.
> cp/
>        * cp-objcp-common.c (cp_function_decl_explicit_p): Don't crash if
>        DECL_LANG_SPECIFIC is NULL.
> include/
>        * dwarf2.h (DW_TAG_GNU_call_site, DW_TAG_GNU_call_site_parameter,
>        DW_AT_GNU_call_site_value, DW_AT_GNU_call_site_data_value,
>        DW_AT_GNU_call_site_target, DW_AT_GNU_call_site_target_clobbered,
>        DW_AT_GNU_tail_call, DW_AT_GNU_all_tail_call_sites,
>        DW_AT_GNU_all_call_sites,, DW_AT_GNU_all_source_call_sites,
>        DW_OP_GNU_entry_value): New.
>

It caused:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48160

Here is a patch.  OK for trunk?

Thanks.

On Wed, Mar 16, 2011 at 03:38:25PM -0700, H.J. Lu wrote:

IMNSHO you really should reconsider using Pmode != ptr_mode for your port.

That said, the patch looks good to me, though I can't approve it.

> diff --git a/gcc/ChangeLog.x32 b/gcc/ChangeLog.x32
> index 764e3de..45db256 100644
> --- a/gcc/ChangeLog.x32
> +++ b/gcc/ChangeLog.x32
> @@ -1,5 +1,9 @@
>  2011-03-16  H.J. Lu  <hongjiu.lu@intel.com>
>  
> +	* var-tracking.c (prepare_call_arguments): Check SUBREG.

Just mention PR debug/48160 in the ChangeLog entry...

> +
> +2011-03-16  H.J. Lu  <hongjiu.lu@intel.com>
> +
>  	PR rtl-optimization/48155
>  	* reload1.c (gen_reload): Check if XEXP (in, 0/1) is SUBREG of
>  	REG in PLUS.
> diff --git a/gcc/var-tracking.c b/gcc/var-tracking.c
> index 266f11f..cfa74a6 100644
> --- a/gcc/var-tracking.c
> +++ b/gcc/var-tracking.c
> @@ -5763,7 +5763,11 @@ prepare_call_arguments (basic_block bb, rtx insn)
>  
>  		    /* Try harder, when passing address of a constant
>  		       pool integer it can be easily read back.  */
> -		    val = CSELIB_VAL_PTR (XEXP (item, 1));
> +		    item = XEXP (item, 1);
> +		    if (GET_CODE (item) == SUBREG)
> +		      item = SUBREG_REG (item);
> +		    gcc_assert (GET_CODE (item) == VALUE);
> +		    val = CSELIB_VAL_PTR (item);
>  		    for (l = val->locs; l; l = l->next)
>  		      if (GET_CODE (l->loc) == SYMBOL_REF
>  			  && TREE_CONSTANT_POOL_ADDRESS_P (l->loc)


	Jakub

On Wed, Mar 16, 2011 at 4:28 PM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Wed, Mar 16, 2011 at 03:38:25PM -0700, H.J. Lu wrote:
>
> IMNSHO you really should reconsider using Pmode != ptr_mode for your port.

Someone can give it a try.

My main concern is the x32 process is running in 64bit mode.  Its
stack operation is 64bit.

> That said, the patch looks good to me, though I can't approve it.

Richard, Jason, is this patch OK with PR debug/48160 in the ChangeLog entry.?

>> diff --git a/gcc/ChangeLog.x32 b/gcc/ChangeLog.x32
>> index 764e3de..45db256 100644
>> --- a/gcc/ChangeLog.x32
>> +++ b/gcc/ChangeLog.x32
>> @@ -1,5 +1,9 @@
>>  2011-03-16  H.J. Lu  <hongjiu.lu@intel.com>
>>
>> +     * var-tracking.c (prepare_call_arguments): Check SUBREG.
>
> Just mention PR debug/48160 in the ChangeLog entry...
>

Done.

Thanks.