[i386] Fix PR 57756

Forgot to add the test case. Patch updated.

Sri

On Fri, Sep 20, 2013 at 7:06 PM, Sriraman Tallam <tmsriram@google.com> wrote:
> On Wed, Aug 14, 2013 at 3:38 AM, Richard Biener
> <richard.guenther@gmail.com> wrote:
>> On Wed, Aug 14, 2013 at 2:02 AM, Sriraman Tallam <tmsriram@google.com>
>> wrote:
>>>
>>> Hi,
>>>
>>>     I have attached a patch to fix PR57756.  Description:  The
>>> following program,
>>>
>>> __attribute__((always_inline,target("sse4.2")))
>>> __inline int callee ()
>>> {
>>>   return 0;
>>> }
>>>
>>> __attribute__((target("sse")))
>>> __inline int caller ()
>>> {
>>>   return callee();
>>> }
>>>
>>> does not generate an error and callee is inlined into caller. This is
>>> because callee has a higher target ISA.  Interchanging the position of
>>> caller and callee will generate the correct error. Also, removing the
>>> target attribute from caller will generate the error.
>>>
>>> The reason for the bug is that when the caller's target options are
>>> processed, global_options contain the ISA flags of the callee
>>> (previously processed) and doing this in i386-common.c, where opts is
>>> global_options:
>>>
>>> opts->x_ix86_isa_flags |= OPTION_MASK_ISA_SSE_SET;
>>>
>>> is not changing global options.  The fix is to reset global_options to
>>> the default each time a new target option needs to be processed.
>>
>>
>> Shouldn't  ix86_valid_target_attribute_tree be refactored to work on a
>> selected
>> opt structure rather than global_options?
>
> I have attached a patch that does the refactoring of
> ix86_valid_target_attribute_tree. In that process I had to refactor
> ix86_valid_target_attribute_inner_p and ix86_option_override_internal.
> This made it a large patch because global_options structure is touched
> by these functions in multiple places and through vvarious aliases
> like ix86_arch_string for instance. In effect, this frees these
> functions from depending on global_options and the code in
> ix86_alid_target_attribute_p shows how the original bug can be fixed
> by just using a different gcc_options structure.
>
> Please review.
>
> Thanks
> Sri
>
>
>
>> Richard.
>>
>>>
>>> Patch ok?
>>>
>>> Thanks
>>> Sri
>>
>>
* testsuite/gcc.target/i386/pr57756.c: New test. 	
	* tree.c (build_optimization_node): New parameter.
	(build_target_option_node): New parameter.
	* tree.h (build_optimization_node): New parameter.
	(build_target_option_node): New parameter.
	* c-family/c-common.c (handle_optimize_attribute): Fix calls to
	build_optimization_node and build_target_option_node.
	* c-family/c-pragma.c (handle_pragma_optimize): Ditto.
	(handle_pragma_push_options): Ditto.
	* toplev.c (process_options): Ditto.
	* config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
	(ix86_pragma_target_parse): Ditto.
	* config/i386/i386-protos.h (ix86_valid_target_attribute_tree): New
	parameters.
	* config/rs6000/rs6000.c (rs6000_option_override_internal): Fix calls
	to  build_optimization_node and build_target_option_node.
	(rs6000_valid_attribute_p): Ditto.
	(rs6000_pragma_target_parse): Ditto.
	* config/i386/i386.c (ix86_option_override_internal): New parameters
	opts and opts_set.
	Change ix86_tune_string to access opts->x_ix86_tune_string.
	Change ix86_isa_flags to access opts->x_ix86_isa_flags.
	Change ix86_arch_string to access opts->x_ix86_arch_string.
	Change ix86_stringop_alg to access opts->x_ix86_stringop_alg.
	Change ix86_pmode to access opts->x_ix86_pmode.
	Change ix86_abi to access opts->x_ix86_abi.
	Change ix86_cmodel to access opts->x_ix86_cmodel.
	Change ix86_asm_dialect to access opts->x_ix86_asm_dialect.
	Change ix86_isa_flags_explicit to access
	opts->x_ix86_isa_flags_explicit.
	Change ix86_dump_tunes to access opts->x_ix86_dump_tunes.
	Change ix86_regparm to access opts->x_ix86_regparm.
	Change ix86_branch_cost to access opts->x_ix86_branch_cost.
	Change ix86_preferred_stack_boundary_arg to access
	opts->x_ix86_preferred_stack_boundary_arg.
	Change ix86_force_align_arg_pointer to access
	opts->x_ix86_force_align_arg_pointer.
	Change ix86_incoming_stack_boundar_arg to access
	opts->x_ix86_incoming_stack_boundar_arg.
	Change ix86_fpmath to access opts->x_ix86_fpmath.
	Change ix86_veclibabi_type to access opts->x_ix86_veclibabi_type.
	Change ix86_recip_name to access opts->x_ix86_recip_name.
	Change ix86_stack_protector_guard to access
	opts->x_ix86_stack_protector_guard.
	Change ix86_tune_memcpy_strategy to access
	opts->x_ix86_tune_memcpy_strategy.
	Change ix86_tune_memset_strategy to access
	opts->x_ix86_tune_memset_strategy.
	Change global_options to access opts.
	Change global_options_set to access opts_set.
	(ix86_valid_target_attribute_inner_p): New parameters.
	Fix recursive call.
	Fix call to ix86_handle_option and set_option.
	(ix86_valid_target_attribute_tree): New parameters.
	Change global_options to access opts.
	Change global_options_set to access opts_set.
	Fix call to ix86_valid_target_attribute_inner_p.
	Change ix86_tune_string to access opts->x_ix86_tune_string.
	Change ix86_arch_string to access opts->x_ix86_arch_string.
	Change ix86_fpmath to access opts->x_ix86_fpmath
	Fix call to ix86_option_override_internal.
	Fix call to ix86_add_new_builtins.
	Fix calls to build_optimization_node and build_target_option_node.
	(ix86_valid_target_attribute_p): Remove access to global_options.
	Use new gcc_options structure func_options.
	Fix call to ix86_valid_target_attribute_tree.
	Fix call to  build_optimization_node.
	(get_builtin_code_for_version):	Fix call to
	ix86_valid_target_attribute_tree.

[i386] Fix PR 57756

Commit Message

Comments

Patch