Re: [2/2] PR96463 -- changes to type checking vec_perm_expr in middle end

[Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>]

On Mon, 9 May 2022 at 21:21, Prathamesh Kulkarni
<prathamesh.kulkarni@linaro.org> wrote:
>
> On Mon, 9 May 2022 at 19:22, Richard Sandiford
> <richard.sandiford@arm.com> wrote:
> >
> > Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> > > On Tue, 3 May 2022 at 18:25, Richard Sandiford
> > > <richard.sandiford@arm.com> wrote:
> > >>
> > >> Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> > >> > On Tue, 4 Jan 2022 at 19:12, Richard Sandiford
> > >> > <richard.sandiford@arm.com> wrote:
> > >> >>
> > >> >> Richard Biener <rguenther@suse.de> writes:
> > >> >> > On Tue, 4 Jan 2022, Richard Sandiford wrote:
> > >> >> >
> > >> >> >> Richard Biener <rguenther@suse.de> writes:
> > >> >> >> > On Fri, 17 Dec 2021, Richard Sandiford wrote:
> > >> >> >> >
> > >> >> >> >> Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org> writes:
> > >> >> >> >> > Hi,
> > >> >> >> >> > The attached patch rearranges order of type-check for vec_perm_expr
> > >> >> >> >> > and relaxes type checking for
> > >> >> >> >> > lhs = vec_perm_expr<rhs1, rhs2, mask>
> > >> >> >> >> >
> > >> >> >> >> > when:
> > >> >> >> >> > rhs1 == rhs2,
> > >> >> >> >> > lhs is variable length vector,
> > >> >> >> >> > rhs1 is fixed length vector,
> > >> >> >> >> > TREE_TYPE (lhs) == TREE_TYPE (rhs1)
> > >> >> >> >> >
> > >> >> >> >> > I am not sure tho if this check is correct ? My intent was to capture
> > >> >> >> >> > case when vec_perm_expr is used to "extend" fixed length vector to
> > >> >> >> >> > it's VLA equivalent.
> > >> >> >> >>
> > >> >> >> >> VLAness isn't really the issue.  We want the same thing to work for
> > >> >> >> >> -msve-vector-bits=256, -msve-vector-bits=512, etc., even though the
> > >> >> >> >> vectors are fixed-length in that case.
> > >> >> >> >>
> > >> >> >> >> The principle is that for:
> > >> >> >> >>
> > >> >> >> >>   A = VEC_PERM_EXPR <B, C, D>;
> > >> >> >> >>
> > >> >> >> >> the requirements are:
> > >> >> >> >>
> > >> >> >> >> - A, B, C and D must be vectors
> > >> >> >> >> - A, B and C must have the same element type
> > >> >> >> >> - D must have an integer element type
> > >> >> >> >> - A and D must have the same number of elements (NA)
> > >> >> >> >> - B and C must have the same number of elements (NB)
> > >> >> >> >>
> > >> >> >> >> The semantics are that we create a joined vector BC (all elements of B
> > >> >> >> >> followed by all element of C) and that:
> > >> >> >> >>
> > >> >> >> >>   A[i] = BC[D[i] % (NB+NB)]
> > >> >> >> >>
> > >> >> >> >> for 0 ≤ i < NA.
> > >> >> >> >>
> > >> >> >> >> This operation makes sense even if NA != NB.
> > >> >> >> >
> > >> >> >> > But note that we don't currently expect NA != NB and the optab just
> > >> >> >> > has a single mode.
> > >> >> >>
> > >> >> >> True, but we only need this for constant permutes.  They are already
> > >> >> >> special in that they allow the index elements to be wider than the data
> > >> >> >> elements.
> > >> >> >
> > >> >> > OK, then we should reflect this in the stmt verification and only relax
> > >> >> > the constant permute vector case and also amend the
> > >> >> > TARGET_VECTORIZE_VEC_PERM_CONST accordingly.
> > >> >>
> > >> >> Sounds good.
> > >> >>
> > >> >> > For non-constant permutes the docs say the mode of vec_perm is
> > >> >> > the common mode of operands 1 and 2 whilst the mode of operand 0
> > >> >> > is unspecified - even unconstrained by the docs.  I'm not sure
> > >> >> > if vec_perm expansion is expected to eventually FAIL.  Updating the
> > >> >> > docs of vec_perm would be appreciated as well.
> > >> >>
> > >> >> Yeah, I guess de facto operand 0 has to be the same mode as operands
> > >> >> 1 and 2.  Maybe that was just an oversight, or maybe it seemed obvious
> > >> >> or self-explanatory at the time. :-)
> > >> >>
> > >> >> > As said I prefer to not mangle the existing stmt checking too much
> > >> >> > at this stage so minimal adjustment is prefered there.
> > >> >>
> > >> >> The PR is only an enhancement request rather than a bug, so I think the
> > >> >> patch would need to wait for GCC 13 whatever happens.
> > >> > Hi,
> > >> > In attached patch, the type checking is relaxed only if mask is constant.
> > >> > Does this look OK ?
> > >> >
> > >> > Thanks,
> > >> > Prathamesh
> > >> >>
> > >> >> Thanks,
> > >> >> Richard
> > >> >
> > >> > diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
> > >> > index e321d929fd0..02b88f67855 100644
> > >> > --- a/gcc/tree-cfg.cc
> > >> > +++ b/gcc/tree-cfg.cc
> > >> > @@ -4307,6 +4307,24 @@ verify_gimple_assign_ternary (gassign *stmt)
> > >> >        break;
> > >> >
> > >> >      case VEC_PERM_EXPR:
> > >> > +      /* If permute is constant, then we allow for lhs and rhs
> > >> > +      to have different vector types, provided:
> > >> > +      (1) lhs, rhs1, rhs2, and rhs3 have same element type.
> > >>
> > >> This isn't a requirement for rhs3.
> > >>
> > >> > +      (2) rhs3 vector has integer element type.
> > >> > +      (3) len(lhs) == len(rhs3) && len(rhs1) == len(rhs2).  */
> > >> > +
> > >> > +      if (TREE_CONSTANT (rhs3)
> > >> > +       && VECTOR_TYPE_P (lhs_type)
> > >> > +       && VECTOR_TYPE_P (rhs1_type)
> > >> > +       && VECTOR_TYPE_P (rhs2_type)
> > >> > +       && VECTOR_TYPE_P (rhs3_type)
> > >> > +       && TREE_TYPE (lhs_type) == TREE_TYPE (rhs1_type)
> > >> > +       && TREE_TYPE (lhs_type) == TREE_TYPE (rhs2_type)
> > >> > +       && INTEGRAL_TYPE_P (TREE_TYPE (rhs3_type))
> > >> > +       && known_eq (TYPE_VECTOR_SUBPARTS (lhs_type), TYPE_VECTOR_SUBPARTS (rhs3_type))
> > >> > +       && known_eq (TYPE_VECTOR_SUBPARTS (rhs1_type), TYPE_VECTOR_SUBPARTS (rhs2_type)))
> > >> > +     return false;
> > >> > +
> > >>
> > >> I think this should be integrated into the existing conditions
> > >> rather than done as an initial special case.
> > >>
> > >> It might make sense to start with:
> > >>
> > >>       if (TREE_CODE (rhs1_type) != VECTOR_TYPE
> > >>           || TREE_CODE (rhs2_type) != VECTOR_TYPE
> > >>           || TREE_CODE (rhs3_type) != VECTOR_TYPE)
> > >>         {
> > >>
> > >> but expanded to test lhs_type too.  Then the other parts of the new test
> > >> should be distributed across the existing conditions.
> > >>
> > >> The type tests should use useless_type_conversion_p rather than ==.
> > > Hi Richard,
> > > Thanks for the suggestions. In the attached patch, I tried to
> > > distribute the checks
> > > across existing conditions, does it look OK ?
> > > I am not sure how to write tests for the type checks tho, does
> > > gimple-fe support vec_perm_expr ?
> > > I did a quick testsuite run for vect.exp and the patch doesn't seem to
> > > cause any unexpected failures.
> > >
> > > Thanks,
> > > Prathamesh
> > >>
> > >> Thanks,
> > >> Richard
> > >>
> > >>
> > >>
> > >> >        if (!useless_type_conversion_p (lhs_type, rhs1_type)
> > >> >         || !useless_type_conversion_p (lhs_type, rhs2_type))
> > >> >       {
> > >
> > > diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
> > > index e321d929fd0..a845c7fff93 100644
> > > --- a/gcc/tree-cfg.cc
> > > +++ b/gcc/tree-cfg.cc
> > > @@ -4307,18 +4307,14 @@ verify_gimple_assign_ternary (gassign *stmt)
> > >        break;
> > >
> > >      case VEC_PERM_EXPR:
> > > -      if (!useless_type_conversion_p (lhs_type, rhs1_type)
> > > -       || !useless_type_conversion_p (lhs_type, rhs2_type))
> > > -     {
> > > -       error ("type mismatch in %qs", code_name);
> > > -       debug_generic_expr (lhs_type);
> > > -       debug_generic_expr (rhs1_type);
> > > -       debug_generic_expr (rhs2_type);
> > > -       debug_generic_expr (rhs3_type);
> > > -       return true;
> > > -     }
> > > +      /* If permute is constant, then we allow for lhs and rhs
> > > +      to have different vector types, provided:
> > > +      (1) lhs, rhs1, rhs2 have same element type.
> > > +      (2) rhs3 vector has integer element type.
> > > +      (3) len(lhs) == len(rhs3) && len(rhs1) == len(rhs2).  */
> > >
> > > -      if (TREE_CODE (rhs1_type) != VECTOR_TYPE
> > > +      if (TREE_CODE (lhs_type) != VECTOR_TYPE
> > > +       || TREE_CODE (rhs1_type) != VECTOR_TYPE
> > >         || TREE_CODE (rhs2_type) != VECTOR_TYPE
> > >         || TREE_CODE (rhs3_type) != VECTOR_TYPE)
> > >       {
> > > @@ -4330,10 +4326,29 @@ verify_gimple_assign_ternary (gassign *stmt)
> > >         return true;
> > >       }
> > >
> > > +      /* If lhs, rhs1, and rhs2 are different vector types,
> > > +      then relax the check if rhs3 is constant and lhs, rhs1, and rhs2
> > > +      have same element types.  */
> > > +      if ((!useless_type_conversion_p (lhs_type, rhs1_type)
> > > +        || !useless_type_conversion_p (lhs_type, rhs2_type))
> > > +       && (!TREE_CONSTANT (rhs3)
> > > +           || TREE_TYPE (lhs_type) != TREE_TYPE (rhs1_type)
> > > +           || TREE_TYPE (lhs_type) != TREE_TYPE (rhs2_type)))
> >
> > These TREE_TYPE tests should use !useless_type_conversion_p too,
> > instead of !=.  Maybe it would be easier to follow as:
> >
> >   if (TREE_CONSTANT (rhs3)
> >       ? ...
> >       : ...)
> >
> > so that we don't have doubled useless_type_conversion_p checks
> > for the TREE_CONSTANT case.
> >
> > > +     {
> > > +         error ("type mismatch in %qs", code_name);
> > > +         debug_generic_expr (lhs_type);
> > > +         debug_generic_expr (rhs1_type);
> > > +         debug_generic_expr (rhs2_type);
> > > +         debug_generic_expr (rhs3_type);
> > > +         return true;
> > > +     }
> > > +
> > > +      /* If rhs3 is constant, relax the check len(rhs2) == len(rhs3).  */
> > >        if (maybe_ne (TYPE_VECTOR_SUBPARTS (rhs1_type),
> > >                   TYPE_VECTOR_SUBPARTS (rhs2_type))
> > > -       || maybe_ne (TYPE_VECTOR_SUBPARTS (rhs2_type),
> > > +       || (maybe_ne (TYPE_VECTOR_SUBPARTS (rhs2_type),
> > >                      TYPE_VECTOR_SUBPARTS (rhs3_type))
> > > +           && !TREE_CONSTANT (rhs3))
> >
> > Very minor, but I think this reads better with the !TREE_CONSTANT first
> > in the && rather than second.  There's no reason to compare the lengths
> > for TREE_CONSTANT.
> >
> > Otherwise it looks good to me, but Richard should have the final say.
> Thanks, I addressed the above suggestions in the attached patch.
> Does it look OK ?
ping https://gcc.gnu.org/pipermail/gcc-patches/2022-May/594349.html
Patch passes bootstrap+test on x86_64-linux-gnu.
On aarch64-linux-gnu, the patch passes bootstrap, but I am seeing
these discrepancies in test results:

New tests that FAIL (2 tests):

g++.dg/modules/macloc-1_c.C -std=c++17  dg-regexp 13 not found:
"[^\\n]*macloc-1_c.C:8:7: error: too many arguments to function 'int
you@agnes\\(\\)'\\nIn module agnes, imported at
[^\\n]*macloc-1_b.C:8,\\nof module edith, imported at
[^\\n]*macloc-1_c.C:3:\\n[^\\n]*macloc-1_a.C:12:14: note: declared
here\\n[^\\n]*macloc-1_a.C:9:22: note: in definition of macro
'KEVIN'\\n"
g++.dg/modules/macloc-1_c.C -std=c++17 (test for excess errors)

Old tests that failed, that have disappeared (2 tests): (Eeek!)

c-c++-common/missing-header-3.c  -Wc++-compat   expected multiline
pattern lines 5-7 not found: "\\s*#import
<this-file-does-not-exist\\.h>[^\\n\\r]*\\n
\\^~~~~~~~~~~~~~~~~~~~~~~~~~~~\\ncompilation
terminated\\.[^\\n\\r]*\\n"
c-c++-common/missing-header-3.c  -Wc++-compat  (test for excess errors)

I am not sure if these seem related to patch tho ?

Thanks,
Prathamesh
>
> Thanks,
> Prathamesh
> >
> > Thanks,
> > Richard
> >
> > >         || maybe_ne (TYPE_VECTOR_SUBPARTS (rhs3_type),
> > >                      TYPE_VECTOR_SUBPARTS (lhs_type)))
> > >       {