This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PATCH: PR target/40838: gcc shouldn't assume that the stack is aligned


On Mon, Oct 19, 2009 at 6:25 PM, Michael Matz <matz@suse.de> wrote:
> Hi,
>
> On Mon, 19 Oct 2009, H.J. Lu wrote:
>
>> > He tries to solve this by pessimistically assuming that potentially
>> > everything imaginable could go on stack. ?What I don't understand is
>> > why we don't instead track hard_stack_alignment in assign_*_temp
>> > (where we then assume that the stack will be aligned perfectly), and
>> > expand stack realignment code _after_ having expanded everything else
>> > (plus examined local variables for the possibility of generating spill
>> > slots).
>>
>> Vectorizer may not call assign_*_temp at all. Instead, x86 backend may
>> call gen_reg_rtx to generate pseudo registers when expanding vector
>> statement.
>
> If that (and not assign_*_temp) is the problem, then it's obvious that
> fiddling with the vectorizer doesn't solve it. ?The expanders can create
> new pseudos for whatever they see fit. ?For expanding vector statements,
> for expanding block moves, for expanding string compares, for expanding
> additions, for anything. ?Mucking around with the vectorizer won't solve
> the problem.
>

Here is a new patch. I added hard_stack_alignment, moved
update_stack_boundary after RTL expansion and updated
ix86_function_ok_for_sibcall to deal with it.  Any comments?

Thanks.


-- 
H.J.
---
gcc/

2009-10-20  H.J. Lu  <hongjiu.lu@intel.com>

	PR target/40836
	* cfgexpand.c (get_decl_align_unit): Update hard_stack_alignment.
	(expand_one_var): Likewise.
	(gimple_expand_cfg): Initialize hard_stack_alignment to 0. Move
	update_stack_boundary call to ...
	(expand_stack_alignment): Here.

	* emit-rtl.c (gen_reg_rtx): Update hard_stack_alignment.
	* function.c (assign_stack_local_1): Likewise.
	(assign_parms): Likewise.
	(locate_and_pad_parm): Likewise.

	* function.h (rtl_data): Add hard_stack_alignment.

	* config/i386/i386.c (ix86_minimum_incoming_stack_boundary): New.
	(verride_options): Don't check ix86_force_align_arg_pointer here.
	(ix86_function_ok_for_sibcall): Use it.
	(ix86_update_stack_boundary): Likewise.

	* config/i386/i386.h (STACK_REALIGN_DEFAULT): Update comments.

gcc/testsuite/

2009-10-20  H.J. Lu  <hongjiu.lu@intel.com>

	PR target/40838
	* gcc.target/i386/incoming-6.c: New.
	* gcc.target/i386/incoming-7.c: Likewise.
	* gcc.target/i386/incoming-8.c: Likewise.
	* gcc.target/i386/incoming-9.c: Likewise.
	* gcc.target/i386/incoming-10.c: Likewise.
	* gcc.target/i386/incoming-11.c: Likewise.
	* gcc.target/i386/incoming-12.c: Likewise.
	* gcc.target/i386/incoming-13.c: Likewise.
	* gcc.target/i386/incoming-14.c: Likewise.
	* gcc.target/i386/incoming-15.c: Likewise.

Attachment: gcc-pr40838-8.patch
Description: Text document


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]