Re: [PATCH] mips: Fix overaligned function arguments [PR109435]

[YunQiang Su <wzssyqa@gmail.com>]

Jovan Dmitrovic <Jovan.Dmitrovic@syrmia.com> 于2023年6月6日周二 18:29写道：
>
> I suppose that it is possible to check assembly.
>

Great.

> Following is part of diff before and after my patch:
>
> 29,32c29,32
> <       sd      $6,0($2)
> <       sd      $7,8($2)
> <       sd      $8,16($2)
> <       sd      $9,24($2)
> ---
> >       sd      $5,0($2)
> >       sd      $6,8($2)
> >       sd      $7,16($2)
> >       sd      $8,24($2)
> 63,66c63,66
> <       sd      $6,0($2)
> <       sd      $7,8($2)
> <       sd      $8,16($2)
> <       sd      $9,24($2)
> ---
> >       sd      $5,0($2)
> >       sd      $6,8($2)
> >       sd      $7,16($2)
> >       sd      $8,24($2)
> 138,141c138,141
> <       ld      $6,64($16)
> <       ld      $7,72($16)
> <       ld      $8,80($16)
> <       ld      $9,88($16)
> ---
> >       ld      $5,64($16)
> >       ld      $6,72($16)
> >       ld      $7,80($16)
> >       ld      $8,88($16)
> 148,151c148,151
> <       ld      $6,64($16)
> <       ld      $7,72($16)
> <       ld      $8,80($16)
> <       ld      $9,88($16)
> ---
> >       ld      $5,64($16)
> >       ld      $6,72($16)
> >       ld      $7,80($16)
> >       ld      $8,88($16)
> 167,170c167,170
> <       ld      $6,0($16)
> <       ld      $7,8($16)
> <       ld      $8,16($16)
> <       ld      $9,24($16)
> ---
> >       ld      $5,0($16)
> >       ld      $6,8($16)
> >       ld      $7,16($16)
> >       ld      $8,24($16)
>
> What my patch effectively does it rearranges the data in
> registers when invoking a function. I don't know whether
> writing this testcase as an assembly check would make sense,
> because that would make the testcase much less readable.

I prefer an assembly check, because the test can be used even
for cross building.

It is not required, I guess.

> ________________________________________
> From: YunQiang Su <wzssyqa@gmail.com>
> Sent: Wednesday, May 31, 2023 12:05 PM
> To: Jovan Dmitrovic
> Cc: gcc-patches@gcc.gnu.org; Djordje Todorovic
> Subject: Re: [PATCH] mips: Fix overaligned function arguments [PR109435]
>
> Jovan Dmitrovic <Jovan.Dmitrovic@syrmia.com> 于2023年5月29日周一 19:00写道：
> >
> > This patch changes alignment for typedef types when passed as
> > arguments, making the alignment equal to the alignment of
> > original (aliased) types.
> >
> > This change makes it impossible for a typedef type to have
> > alignment that is less than its size.
> >
> > Signed-off-by: Jovan Dmitrovic <jovan.dmitrovic@syrmia.com>
> >
> > gcc/ChangeLog:
> >         PR target/109435
> >         * config/mips/mips.cc (mips_function_arg_alignment): Returns
> >     the alignment of function argument. In case of typedef type,
> >     it returns the aligment of the aliased type.
> >         (mips_function_arg_boundary): Relocated calculation of the
> >     aligment of function arguments.
> >
> > gcc/testsuite/ChangeLog:
> >         PR target/109435
> >         * gcc.target/mips/align-1.c: New test.
> > ---
> >  gcc/config/mips/mips.cc                 | 18 +++++++++++++-
> >  gcc/testsuite/gcc.target/mips/align-1.c | 33 +++++++++++++++++++++++++
> >  2 files changed, 50 insertions(+), 1 deletion(-)
> >  create mode 100644 gcc/testsuite/gcc.target/mips/align-1.c
> >
> > diff --git a/gcc/config/mips/mips.cc b/gcc/config/mips/mips.cc
> > index ca822758b41..2019b7cd7d9 100644
> > --- a/gcc/config/mips/mips.cc
> > +++ b/gcc/config/mips/mips.cc
> > @@ -6190,6 +6190,22 @@ mips_arg_partial_bytes (cumulative_args_t cum, const function_arg_info &arg)
> >    return info.stack_words > 0 ? info.reg_words * UNITS_PER_WORD : 0;
> >  }
> >
> > +/* Given MODE and TYPE of a function argument, return the alignment in
> > +   bits. In case of typedef, alignment of its original type is
> > +   used.  */
> > +
> > +static unsigned int
> > +mips_function_arg_alignment (machine_mode mode, const_tree type)
> > +{
> > +  if (!type)
> > +    return GET_MODE_ALIGNMENT (mode);
> > +
> > +  if (is_typedef_decl (TYPE_NAME (type)))
> > +    type = DECL_ORIGINAL_TYPE (TYPE_NAME (type));
> > +
> > +  return TYPE_ALIGN (type);
> > +}
> > +
> >  /* Implement TARGET_FUNCTION_ARG_BOUNDARY.  Every parameter gets at
> >     least PARM_BOUNDARY bits of alignment, but will be given anything up
> >     to STACK_BOUNDARY bits if the type requires it.  */
> > @@ -6198,8 +6214,8 @@ static unsigned int
> >  mips_function_arg_boundary (machine_mode mode, const_tree type)
> >  {
> >    unsigned int alignment;
> > +  alignment = mips_function_arg_alignment (mode, type);
> >
> > -  alignment = type ? TYPE_ALIGN (type) : GET_MODE_ALIGNMENT (mode);
> >    if (alignment < PARM_BOUNDARY)
> >      alignment = PARM_BOUNDARY;
> >    if (alignment > STACK_BOUNDARY)
> > diff --git a/gcc/testsuite/gcc.target/mips/align-1.c b/gcc/testsuite/gcc.target/mips/align-1.c
> > new file mode 100644
> > index 00000000000..816751b8099
> > --- /dev/null
> > +++ b/gcc/testsuite/gcc.target/mips/align-1.c
> > @@ -0,0 +1,33 @@
> > +/* Check that typedef alignment does not affect passing of function
> > +   parameters. */
> > +/* { dg-do run { target { "mips*-*-linux*" } } } */
>
> Is it possible to check the result with something like
>    scan-assembler
>    scan-assembler-not
> instead of real running?
>
> > +
> > +#include <assert.h>
> > +
> > +typedef struct ui8
> > +{
> > +  unsigned v[8];
> > +} uint8 __attribute__ ((aligned(64)));
> > +
> > +unsigned
> > +callee (int x, uint8 a)
> > +{
> > +  return a.v[0];
> > +}
> > +
> > +uint8
> > +identity (uint8 in)
> > +{
> > +  return in;
> > +}
> > +
> > +int
> > +main (void)
> > +{
> > +  uint8 vec = {{1, 2, 3, 4, 5, 6, 7, 8}};
> > +  uint8 temp = identity (vec);
> > +  unsigned temp2 = callee (1, identity (vec));
> > +  assert (callee (1, temp) == 1);
> > +  assert (temp2 == 1);
> > +  return 0;
> > +}
> > --
> > 2.34.1
> >



-- 
YunQiang Su