Patchwork [tree-optimization] : Improve handling of conditional-branches on targets with high branch costs

login
register
mail settings
Submitter Kai Tietz
Date Oct. 13, 2011, 11:25 a.m.
Message ID <CAEwic4YdvkdhsqnJo0czDUGSGuVB_dDU=b+C+85yTvd-ifZ2RQ@mail.gmail.com>
Download mbox | patch
Permalink /patch/119446/
State New
Headers show

Comments

Kai Tietz - Oct. 13, 2011, 11:25 a.m.
Hello,

this new version addresses the comments from Michael and additional fixes
an latent issue shown up by this rewrite in fold-const.
On gimplify.c's gimple_boolify we didn't handled the case that operands
for TRUTH-expressions need to have same operand-size for transformation to
bitwise operation.  This shows up for Fortran, as here are more then one
boolean-kind type with different mode-sizes.  I added a testcase for this,

ChangeLog

2011-10-13  Kai Tietz  <ktietz@redhat.com>

	* fold-const.c (simple_operand_p_2): New function.
	(fold_truthop): Rename to
	(fold_truth_andor_1): function name.
	Additionally remove branching creation for logical and/or.
	(fold_truth_andor): Handle branching creation for logical and/or here.
	* gimplify.c (gimple_boolify): Take care that for bitwise-binary
	transformation the operands have compatible types.

2011-10-13  Kai Tietz  <ktietz@redhat.com>

	* gfortran.fortran-torture/compile/logical-2.f90: New test.

Bootstrapped and regression-tested for all languages plus Ada and
Obj-C++ on x86_64-pc-linux-gnu.
Ok for apply?

Regards,
Kai
Richard Guenther - Oct. 13, 2011, 11:49 a.m.
On Thu, Oct 13, 2011 at 1:25 PM, Kai Tietz <ktietz70@googlemail.com> wrote:
> Hello,
>
> this new version addresses the comments from Michael and additional fixes
> an latent issue shown up by this rewrite in fold-const.
> On gimplify.c's gimple_boolify we didn't handled the case that operands
> for TRUTH-expressions need to have same operand-size for transformation to
> bitwise operation.  This shows up for Fortran, as here are more then one
> boolean-kind type with different mode-sizes.  I added a testcase for this,
>
> ChangeLog
>
> 2011-10-13  Kai Tietz  <ktietz@redhat.com>
>
>        * fold-const.c (simple_operand_p_2): New function.
>        (fold_truthop): Rename to
>        (fold_truth_andor_1): function name.
>        Additionally remove branching creation for logical and/or.
>        (fold_truth_andor): Handle branching creation for logical and/or here.
>        * gimplify.c (gimple_boolify): Take care that for bitwise-binary
>        transformation the operands have compatible types.
>
> 2011-10-13  Kai Tietz  <ktietz@redhat.com>
>
>        * gfortran.fortran-torture/compile/logical-2.f90: New test.
>
> Bootstrapped and regression-tested for all languages plus Ada and
> Obj-C++ on x86_64-pc-linux-gnu.
> Ok for apply?
>
> Regards,
> Kai
>
> Index: gcc/gcc/fold-const.c
> ===================================================================
> --- gcc.orig/gcc/fold-const.c
> +++ gcc/gcc/fold-const.c
> @@ -112,13 +112,13 @@ static tree decode_field_reference (loca
>  static int all_ones_mask_p (const_tree, int);
>  static tree sign_bit_p (tree, const_tree);
>  static int simple_operand_p (const_tree);
> +static bool simple_operand_p_2 (tree);
>  static tree range_binop (enum tree_code, tree, tree, int, tree, int);
>  static tree range_predecessor (tree);
>  static tree range_successor (tree);
>  static tree fold_range_test (location_t, enum tree_code, tree, tree, tree);
>  static tree fold_cond_expr_with_comparison (location_t, tree, tree,
> tree, tree);
>  static tree unextend (tree, int, int, tree);
> -static tree fold_truthop (location_t, enum tree_code, tree, tree, tree);
>  static tree optimize_minmax_comparison (location_t, enum tree_code,
>                                        tree, tree, tree);
>  static tree extract_muldiv (tree, tree, enum tree_code, tree, bool *);
> @@ -3500,7 +3500,7 @@ optimize_bit_field_compare (location_t l
>   return lhs;
>  }
>
> -/* Subroutine for fold_truthop: decode a field reference.
> +/* Subroutine for fold_truth_andor_1: decode a field reference.
>
>    If EXP is a comparison reference, we return the innermost reference.
>
> @@ -3668,7 +3668,7 @@ sign_bit_p (tree exp, const_tree val)
>   return NULL_TREE;
>  }
>
> -/* Subroutine for fold_truthop: determine if an operand is simple enough
> +/* Subroutine for fold_truth_andor_1: determine if an operand is simple enough
>    to be evaluated unconditionally.  */
>
>  static int
> @@ -3678,7 +3678,7 @@ simple_operand_p (const_tree exp)
>   STRIP_NOPS (exp);
>
>   return (CONSTANT_CLASS_P (exp)
> -         || TREE_CODE (exp) == SSA_NAME
> +         || TREE_CODE (exp) == SSA_NAME
>          || (DECL_P (exp)
>              && ! TREE_ADDRESSABLE (exp)
>              && ! TREE_THIS_VOLATILE (exp)
> @@ -3692,6 +3692,46 @@ simple_operand_p (const_tree exp)
>                 registers aren't expensive.  */
>              && (! TREE_STATIC (exp) || DECL_REGISTER (exp))));
>  }
> +
> +/* Subroutine for fold_truth_andor: determine if an operand is simple enough
> +   to be evaluated unconditionally.
> +   I addition to simple_operand_p, we assume that comparisons and logic-not
> +   operations are simple, if their operands are simple, too.  */
> +
> +static bool
> +simple_operand_p_2 (tree exp)
> +{
> +  enum tree_code code;
> +
> +  /* Strip any conversions that don't change the machine mode.  */
> +  STRIP_NOPS (exp);
> +
> +  code = TREE_CODE (exp);
> +
> +  if (TREE_CODE_CLASS (code) == tcc_comparison)
> +    return (!tree_could_trap_p (exp)
> +           && simple_operand_p_2 (TREE_OPERAND (exp, 0))
> +           && simple_operand_p_2 (TREE_OPERAND (exp, 1)));
> +
> +  if (TREE_SIDE_EFFECTS (exp)
> +      || tree_could_trap_p (exp))
> +    return false;
> +
> +  switch (code)
> +    {
> +    case SSA_NAME:
> +      return true;
> +    case TRUTH_NOT_EXPR:
> +      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
> +    case BIT_NOT_EXPR:
> +      if (TREE_CODE (TREE_TYPE (exp)) != BOOLEAN_TYPE)
> +       return false;
> +      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
> +    default:
> +      return simple_operand_p (exp);
> +    }
> +}
> +
>
>  /* The following functions are subroutines to fold_range_test and allow it to
>    try to change a logical combination of comparisons into a range test.
> @@ -4888,7 +4928,7 @@ fold_range_test (location_t loc, enum tr
>   return 0;
>  }
>
> -/* Subroutine for fold_truthop: C is an INTEGER_CST interpreted as a P
> +/* Subroutine for fold_truth_andor_1: C is an INTEGER_CST interpreted as a P
>    bit value.  Arrange things so the extra bits will be set to zero if and
>    only if C is signed-extended to its full width.  If MASK is nonzero,
>    it is an INTEGER_CST that should be AND'ed with the extra bits.  */
> @@ -5025,8 +5065,8 @@ merge_truthop_with_opposite_arm (locatio
>    We return the simplified tree or 0 if no optimization is possible.  */
>
>  static tree
> -fold_truthop (location_t loc, enum tree_code code, tree truth_type,
> -             tree lhs, tree rhs)
> +fold_truth_andor_1 (location_t loc, enum tree_code code, tree truth_type,
> +                   tree lhs, tree rhs)
>  {
>   /* If this is the "or" of two comparisons, we can do something if
>      the comparisons are NE_EXPR.  If this is the "and", we can do something
> @@ -5054,8 +5094,6 @@ fold_truthop (location_t loc, enum tree_
>   tree lntype, rntype, result;
>   HOST_WIDE_INT first_bit, end_bit;
>   int volatilep;
> -  tree orig_lhs = lhs, orig_rhs = rhs;
> -  enum tree_code orig_code = code;
>
>   /* Start by getting the comparison codes.  Fail if anything is volatile.
>      If one operand is a BIT_AND_EXPR with the constant one, treat it as if
> @@ -5119,8 +5157,7 @@ fold_truthop (location_t loc, enum tree_
>   /* If the RHS can be evaluated unconditionally and its operands are
>      simple, it wins to evaluate the RHS unconditionally on machines
>      with expensive branches.  In this case, this isn't a comparison
> -     that can be merged.  Avoid doing this if the RHS is a floating-point
> -     comparison since those can trap.  */
> +     that can be merged.  */
>
>   if (BRANCH_COST (optimize_function_for_speed_p (cfun),
>                   false) >= 2
> @@ -5149,13 +5186,6 @@ fold_truthop (location_t loc, enum tree_
>                           build2 (BIT_IOR_EXPR, TREE_TYPE (ll_arg),
>                                   ll_arg, rl_arg),
>                           build_int_cst (TREE_TYPE (ll_arg), 0));
> -
> -      if (LOGICAL_OP_NON_SHORT_CIRCUIT)
> -       {
> -         if (code != orig_code || lhs != orig_lhs || rhs != orig_rhs)
> -           return build2_loc (loc, code, truth_type, lhs, rhs);
> -         return NULL_TREE;
> -       }
>     }
>
>   /* See if the comparisons can be merged.  Then get all the parameters for
> @@ -8380,13 +8410,49 @@ fold_truth_andor (location_t loc, enum t
>      lhs is another similar operation, try to merge its rhs with our
>      rhs.  Then try to merge our lhs and rhs.  */
>   if (TREE_CODE (arg0) == code
> -      && 0 != (tem = fold_truthop (loc, code, type,
> -                                  TREE_OPERAND (arg0, 1), arg1)))
> +      && 0 != (tem = fold_truth_andor_1 (loc, code, type,
> +                                        TREE_OPERAND (arg0, 1), arg1)))
>     return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0), tem);
>
> -  if ((tem = fold_truthop (loc, code, type, arg0, arg1)) != 0)
> +  if ((tem = fold_truth_andor_1 (loc, code, type, arg0, arg1)) != 0)
>     return tem;
>
> +  if ((code == TRUTH_ANDIF_EXPR || code == TRUTH_ORIF_EXPR)
> +      && (BRANCH_COST (optimize_function_for_speed_p (cfun),
> +                      false) >= 2)
> +      && LOGICAL_OP_NON_SHORT_CIRCUIT
> +      && simple_operand_p_2 (arg1))
> +    {
> +      enum tree_code ncode = (code == TRUTH_ANDIF_EXPR ? TRUTH_AND_EXPR
> +                                                      : TRUTH_OR_EXPR);
> +
> +      /* Transform ((A AND-IF B) AND-IF C) into (A AND-IF (B AND C)),
> +         or ((A OR-IF B) OR-IF C) into (A OR-IF (B OR C))
> +         We don't want to pack more than two leafs to a non-IF AND/OR
> +         expression.
> +         If tree-code of left-hand operand isn't an AND/OR-IF code and not
> +         equal to CODE, then we don't want to add right-hand operand.
> +         If the inner right-hand side of left-hand operand has side-effects,
> +         or isn't simple, then we can't add to it, as otherwise we might
> +         destroy if-sequence.  */
> +      if (TREE_CODE (arg0) == code
> +         /* Needed for sequence points to handle trappings, and
> +            side-effects.  */
> +         && simple_operand_p_2 (TREE_OPERAND (arg0, 1)))
> +       {
> +         tem = fold_build2_loc (loc, ncode, type, TREE_OPERAND (arg0, 1),
> +                               arg1);
> +         return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0),
> +                                tem);
> +       }
> +     /* Transform (A AND-IF B) into (A AND B), or (A OR-IF B)
> +        into (A OR B).
> +        For sequence point consistancy, we need to check for trapping, and
> +        side-effects.  */
> +     else if (simple_operand_p_2 (arg0))
> +       return fold_build2_loc (loc, ncode, type, arg0, arg1);
> +    }
> +
>   return NULL_TREE;
>  }
>
> Index: gcc/gcc/gimplify.c
> ===================================================================
> --- gcc.orig/gcc/gimplify.c
> +++ gcc/gcc/gimplify.c
> @@ -2804,8 +2804,13 @@ tree
>  gimple_boolify (tree expr)
>  {
>   tree type = TREE_TYPE (expr);
> +  tree new_expr_type;
>   location_t loc = EXPR_LOCATION (expr);
>
> +  new_expr_type = type;
> +  if (TREE_CODE (type) != BOOLEAN_TYPE)
> +    new_expr_type = boolean_type_node;
> +
>   if (TREE_CODE (expr) == NE_EXPR
>       && TREE_CODE (TREE_OPERAND (expr, 0)) == CALL_EXPR
>       && integer_zerop (TREE_OPERAND (expr, 1)))
> @@ -2845,11 +2850,26 @@ gimple_boolify (tree expr)
>     case TRUTH_ORIF_EXPR:
>       /* Also boolify the arguments of truth exprs.  */
>       TREE_OPERAND (expr, 1) = gimple_boolify (TREE_OPERAND (expr, 1));
> -      /* FALLTHRU */
> +
> +      /* Make sure that operand type is compatible to final
> +         expression's type.  */
> +      if (!useless_type_conversion_p (new_expr_type,
> +                                     TREE_TYPE (TREE_OPERAND (expr, 1))))
> +        TREE_OPERAND (expr, 1) =
> +          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 1));
> +
> +      /* Fall through.  */
>
>     case TRUTH_NOT_EXPR:
>       TREE_OPERAND (expr, 0) = gimple_boolify (TREE_OPERAND (expr, 0));
>
> +      /* Make sure that operand type is compatible to final
> +         expression's type.  */
> +      if (!useless_type_conversion_p (new_expr_type,
> +                                     TREE_TYPE (TREE_OPERAND (expr, 0))))
> +        TREE_OPERAND (expr, 0) =
> +          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 0));
> +
>       /* These expressions always produce boolean results.  */
>       if (TREE_CODE (type) != BOOLEAN_TYPE)
>        TREE_TYPE (expr) = boolean_type_node;

No, the callers are responsible for this.

> Index: gcc/gcc/testsuite/gfortran.fortran-torture/compile/logical-2.f90
> ===================================================================
> --- /dev/null
> +++ gcc/gcc/testsuite/gfortran.fortran-torture/compile/logical-2.f90
> @@ -0,0 +1,10 @@
> +! Check for operand type validity after gimplification
> +
> +subroutine whatever()
> +logical(kind=1) :: l1
> +logical(kind=2) :: l2
> +logical(kind=4) :: l3
> +if ((l1 .and. l2) .neqv. l3) then
> +   l1 = .true.
> +endif
> +end
>
Kai Tietz - Oct. 13, 2011, 12:08 p.m.
2011/10/13 Richard Guenther <richard.guenther@gmail.com>:
> On Thu, Oct 13, 2011 at 1:25 PM, Kai Tietz <ktietz70@googlemail.com> wrote:
>> Hello,
>>
>> this new version addresses the comments from Michael and additional fixes
>> an latent issue shown up by this rewrite in fold-const.
>> On gimplify.c's gimple_boolify we didn't handled the case that operands
>> for TRUTH-expressions need to have same operand-size for transformation to
>> bitwise operation.  This shows up for Fortran, as here are more then one
>> boolean-kind type with different mode-sizes.  I added a testcase for this,
>>
>> ChangeLog
>>
>> 2011-10-13  Kai Tietz  <ktietz@redhat.com>
>>
>>        * fold-const.c (simple_operand_p_2): New function.
>>        (fold_truthop): Rename to
>>        (fold_truth_andor_1): function name.
>>        Additionally remove branching creation for logical and/or.
>>        (fold_truth_andor): Handle branching creation for logical and/or here.
>>        * gimplify.c (gimple_boolify): Take care that for bitwise-binary
>>        transformation the operands have compatible types.
>>
>> 2011-10-13  Kai Tietz  <ktietz@redhat.com>
>>
>>        * gfortran.fortran-torture/compile/logical-2.f90: New test.
>>
>> Bootstrapped and regression-tested for all languages plus Ada and
>> Obj-C++ on x86_64-pc-linux-gnu.
>> Ok for apply?
>>
>> Regards,
>> Kai
>>
>> Index: gcc/gcc/fold-const.c
>> ===================================================================
>> --- gcc.orig/gcc/fold-const.c
>> +++ gcc/gcc/fold-const.c
>> @@ -112,13 +112,13 @@ static tree decode_field_reference (loca
>>  static int all_ones_mask_p (const_tree, int);
>>  static tree sign_bit_p (tree, const_tree);
>>  static int simple_operand_p (const_tree);
>> +static bool simple_operand_p_2 (tree);
>>  static tree range_binop (enum tree_code, tree, tree, int, tree, int);
>>  static tree range_predecessor (tree);
>>  static tree range_successor (tree);
>>  static tree fold_range_test (location_t, enum tree_code, tree, tree, tree);
>>  static tree fold_cond_expr_with_comparison (location_t, tree, tree,
>> tree, tree);
>>  static tree unextend (tree, int, int, tree);
>> -static tree fold_truthop (location_t, enum tree_code, tree, tree, tree);
>>  static tree optimize_minmax_comparison (location_t, enum tree_code,
>>                                        tree, tree, tree);
>>  static tree extract_muldiv (tree, tree, enum tree_code, tree, bool *);
>> @@ -3500,7 +3500,7 @@ optimize_bit_field_compare (location_t l
>>   return lhs;
>>  }
>>
>> -/* Subroutine for fold_truthop: decode a field reference.
>> +/* Subroutine for fold_truth_andor_1: decode a field reference.
>>
>>    If EXP is a comparison reference, we return the innermost reference.
>>
>> @@ -3668,7 +3668,7 @@ sign_bit_p (tree exp, const_tree val)
>>   return NULL_TREE;
>>  }
>>
>> -/* Subroutine for fold_truthop: determine if an operand is simple enough
>> +/* Subroutine for fold_truth_andor_1: determine if an operand is simple enough
>>    to be evaluated unconditionally.  */
>>
>>  static int
>> @@ -3678,7 +3678,7 @@ simple_operand_p (const_tree exp)
>>   STRIP_NOPS (exp);
>>
>>   return (CONSTANT_CLASS_P (exp)
>> -         || TREE_CODE (exp) == SSA_NAME
>> +         || TREE_CODE (exp) == SSA_NAME
>>          || (DECL_P (exp)
>>              && ! TREE_ADDRESSABLE (exp)
>>              && ! TREE_THIS_VOLATILE (exp)
>> @@ -3692,6 +3692,46 @@ simple_operand_p (const_tree exp)
>>                 registers aren't expensive.  */
>>              && (! TREE_STATIC (exp) || DECL_REGISTER (exp))));
>>  }
>> +
>> +/* Subroutine for fold_truth_andor: determine if an operand is simple enough
>> +   to be evaluated unconditionally.
>> +   I addition to simple_operand_p, we assume that comparisons and logic-not
>> +   operations are simple, if their operands are simple, too.  */
>> +
>> +static bool
>> +simple_operand_p_2 (tree exp)
>> +{
>> +  enum tree_code code;
>> +
>> +  /* Strip any conversions that don't change the machine mode.  */
>> +  STRIP_NOPS (exp);
>> +
>> +  code = TREE_CODE (exp);
>> +
>> +  if (TREE_CODE_CLASS (code) == tcc_comparison)
>> +    return (!tree_could_trap_p (exp)
>> +           && simple_operand_p_2 (TREE_OPERAND (exp, 0))
>> +           && simple_operand_p_2 (TREE_OPERAND (exp, 1)));
>> +
>> +  if (TREE_SIDE_EFFECTS (exp)
>> +      || tree_could_trap_p (exp))
>> +    return false;
>> +
>> +  switch (code)
>> +    {
>> +    case SSA_NAME:
>> +      return true;
>> +    case TRUTH_NOT_EXPR:
>> +      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
>> +    case BIT_NOT_EXPR:
>> +      if (TREE_CODE (TREE_TYPE (exp)) != BOOLEAN_TYPE)
>> +       return false;
>> +      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
>> +    default:
>> +      return simple_operand_p (exp);
>> +    }
>> +}
>> +
>>
>>  /* The following functions are subroutines to fold_range_test and allow it to
>>    try to change a logical combination of comparisons into a range test.
>> @@ -4888,7 +4928,7 @@ fold_range_test (location_t loc, enum tr
>>   return 0;
>>  }
>>
>> -/* Subroutine for fold_truthop: C is an INTEGER_CST interpreted as a P
>> +/* Subroutine for fold_truth_andor_1: C is an INTEGER_CST interpreted as a P
>>    bit value.  Arrange things so the extra bits will be set to zero if and
>>    only if C is signed-extended to its full width.  If MASK is nonzero,
>>    it is an INTEGER_CST that should be AND'ed with the extra bits.  */
>> @@ -5025,8 +5065,8 @@ merge_truthop_with_opposite_arm (locatio
>>    We return the simplified tree or 0 if no optimization is possible.  */
>>
>>  static tree
>> -fold_truthop (location_t loc, enum tree_code code, tree truth_type,
>> -             tree lhs, tree rhs)
>> +fold_truth_andor_1 (location_t loc, enum tree_code code, tree truth_type,
>> +                   tree lhs, tree rhs)
>>  {
>>   /* If this is the "or" of two comparisons, we can do something if
>>      the comparisons are NE_EXPR.  If this is the "and", we can do something
>> @@ -5054,8 +5094,6 @@ fold_truthop (location_t loc, enum tree_
>>   tree lntype, rntype, result;
>>   HOST_WIDE_INT first_bit, end_bit;
>>   int volatilep;
>> -  tree orig_lhs = lhs, orig_rhs = rhs;
>> -  enum tree_code orig_code = code;
>>
>>   /* Start by getting the comparison codes.  Fail if anything is volatile.
>>      If one operand is a BIT_AND_EXPR with the constant one, treat it as if
>> @@ -5119,8 +5157,7 @@ fold_truthop (location_t loc, enum tree_
>>   /* If the RHS can be evaluated unconditionally and its operands are
>>      simple, it wins to evaluate the RHS unconditionally on machines
>>      with expensive branches.  In this case, this isn't a comparison
>> -     that can be merged.  Avoid doing this if the RHS is a floating-point
>> -     comparison since those can trap.  */
>> +     that can be merged.  */
>>
>>   if (BRANCH_COST (optimize_function_for_speed_p (cfun),
>>                   false) >= 2
>> @@ -5149,13 +5186,6 @@ fold_truthop (location_t loc, enum tree_
>>                           build2 (BIT_IOR_EXPR, TREE_TYPE (ll_arg),
>>                                   ll_arg, rl_arg),
>>                           build_int_cst (TREE_TYPE (ll_arg), 0));
>> -
>> -      if (LOGICAL_OP_NON_SHORT_CIRCUIT)
>> -       {
>> -         if (code != orig_code || lhs != orig_lhs || rhs != orig_rhs)
>> -           return build2_loc (loc, code, truth_type, lhs, rhs);
>> -         return NULL_TREE;
>> -       }
>>     }
>>
>>   /* See if the comparisons can be merged.  Then get all the parameters for
>> @@ -8380,13 +8410,49 @@ fold_truth_andor (location_t loc, enum t
>>      lhs is another similar operation, try to merge its rhs with our
>>      rhs.  Then try to merge our lhs and rhs.  */
>>   if (TREE_CODE (arg0) == code
>> -      && 0 != (tem = fold_truthop (loc, code, type,
>> -                                  TREE_OPERAND (arg0, 1), arg1)))
>> +      && 0 != (tem = fold_truth_andor_1 (loc, code, type,
>> +                                        TREE_OPERAND (arg0, 1), arg1)))
>>     return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0), tem);
>>
>> -  if ((tem = fold_truthop (loc, code, type, arg0, arg1)) != 0)
>> +  if ((tem = fold_truth_andor_1 (loc, code, type, arg0, arg1)) != 0)
>>     return tem;
>>
>> +  if ((code == TRUTH_ANDIF_EXPR || code == TRUTH_ORIF_EXPR)
>> +      && (BRANCH_COST (optimize_function_for_speed_p (cfun),
>> +                      false) >= 2)
>> +      && LOGICAL_OP_NON_SHORT_CIRCUIT
>> +      && simple_operand_p_2 (arg1))
>> +    {
>> +      enum tree_code ncode = (code == TRUTH_ANDIF_EXPR ? TRUTH_AND_EXPR
>> +                                                      : TRUTH_OR_EXPR);
>> +
>> +      /* Transform ((A AND-IF B) AND-IF C) into (A AND-IF (B AND C)),
>> +         or ((A OR-IF B) OR-IF C) into (A OR-IF (B OR C))
>> +         We don't want to pack more than two leafs to a non-IF AND/OR
>> +         expression.
>> +         If tree-code of left-hand operand isn't an AND/OR-IF code and not
>> +         equal to CODE, then we don't want to add right-hand operand.
>> +         If the inner right-hand side of left-hand operand has side-effects,
>> +         or isn't simple, then we can't add to it, as otherwise we might
>> +         destroy if-sequence.  */
>> +      if (TREE_CODE (arg0) == code
>> +         /* Needed for sequence points to handle trappings, and
>> +            side-effects.  */
>> +         && simple_operand_p_2 (TREE_OPERAND (arg0, 1)))
>> +       {
>> +         tem = fold_build2_loc (loc, ncode, type, TREE_OPERAND (arg0, 1),
>> +                               arg1);
>> +         return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0),
>> +                                tem);
>> +       }
>> +     /* Transform (A AND-IF B) into (A AND B), or (A OR-IF B)
>> +        into (A OR B).
>> +        For sequence point consistancy, we need to check for trapping, and
>> +        side-effects.  */
>> +     else if (simple_operand_p_2 (arg0))
>> +       return fold_build2_loc (loc, ncode, type, arg0, arg1);
>> +    }
>> +
>>   return NULL_TREE;
>>  }
>>
>> Index: gcc/gcc/gimplify.c
>> ===================================================================
>> --- gcc.orig/gcc/gimplify.c
>> +++ gcc/gcc/gimplify.c
>> @@ -2804,8 +2804,13 @@ tree
>>  gimple_boolify (tree expr)
>>  {
>>   tree type = TREE_TYPE (expr);
>> +  tree new_expr_type;
>>   location_t loc = EXPR_LOCATION (expr);
>>
>> +  new_expr_type = type;
>> +  if (TREE_CODE (type) != BOOLEAN_TYPE)
>> +    new_expr_type = boolean_type_node;
>> +
>>   if (TREE_CODE (expr) == NE_EXPR
>>       && TREE_CODE (TREE_OPERAND (expr, 0)) == CALL_EXPR
>>       && integer_zerop (TREE_OPERAND (expr, 1)))
>> @@ -2845,11 +2850,26 @@ gimple_boolify (tree expr)
>>     case TRUTH_ORIF_EXPR:
>>       /* Also boolify the arguments of truth exprs.  */
>>       TREE_OPERAND (expr, 1) = gimple_boolify (TREE_OPERAND (expr, 1));
>> -      /* FALLTHRU */
>> +
>> +      /* Make sure that operand type is compatible to final
>> +         expression's type.  */
>> +      if (!useless_type_conversion_p (new_expr_type,
>> +                                     TREE_TYPE (TREE_OPERAND (expr, 1))))
>> +        TREE_OPERAND (expr, 1) =
>> +          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 1));
>> +
>> +      /* Fall through.  */
>>
>>     case TRUTH_NOT_EXPR:
>>       TREE_OPERAND (expr, 0) = gimple_boolify (TREE_OPERAND (expr, 0));
>>
>> +      /* Make sure that operand type is compatible to final
>> +         expression's type.  */
>> +      if (!useless_type_conversion_p (new_expr_type,
>> +                                     TREE_TYPE (TREE_OPERAND (expr, 0))))
>> +        TREE_OPERAND (expr, 0) =
>> +          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 0));
>> +
>>       /* These expressions always produce boolean results.  */
>>       if (TREE_CODE (type) != BOOLEAN_TYPE)
>>        TREE_TYPE (expr) = boolean_type_node;
>
> No, the callers are responsible for this.

Hmm, well, this is a bit inconsistant IMHO and leads to
code-duplications.  Also we need the same check for final expression
type at those places, which makes logical flow a bit harder to read.
The caller of gimple_boolify isn't responsible to check for other
inner expressions operand consistancy, too.  Its current use is more
"please make this expression and its operands boolean-typed".
Well, but if you insists that I duplicate for all places
TRUTH-expressions are boolified that operands match final typel, then
I will do so.

This issue occuress only for TRUTH-stuff from C-tree, as here operands
can have different types than the expression itself.

Regards,
Kai
Michael Matz - Oct. 13, 2011, 1:41 p.m.
Hi,

On Thu, 13 Oct 2011, Kai Tietz wrote:

> this new version addresses the comments from Michael and additional fixes
> an latent issue shown up by this rewrite in fold-const.
> On gimplify.c's gimple_boolify we didn't handled the case that operands
> for TRUTH-expressions need to have same operand-size for transformation to
> bitwise operation.

The requirement comes from BIT_AND_EXPR, not from any of the TRUTH_* 
expressions, hence the point of generating the BIT_AND_EXPR is the point 
to fixup the types.  Similar to this (fixes your testcase):

Index: gimplify.c
===================================================================
--- gimplify.c  (revision 179855)
+++ gimplify.c  (working copy)
   /* See if any simplifications can be done based on what the RHS is.  */
@@ -7257,6 +7264,18 @@ gimplify_expr (tree *expr_p, gimple_seq
          {
            tree orig_type = TREE_TYPE (*expr_p);
            *expr_p = gimple_boolify (*expr_p);
+           /* We are going to transform this into BIT operations,
+              which have stricter requirements on the operand types.  */
+           if (!useless_type_conversion_p
+                (orig_type, TREE_TYPE (TREE_OPERAND (*expr_p, 0))))
+             TREE_OPERAND (*expr_p, 0)
+               = fold_convert_loc (input_location, orig_type,
+                                   TREE_OPERAND (*expr_p, 0));
+           if (!useless_type_conversion_p
+                (orig_type, TREE_TYPE (TREE_OPERAND (*expr_p, 1))))
+             TREE_OPERAND (*expr_p, 1)
+               = fold_convert_loc (input_location, orig_type,
+                                   TREE_OPERAND (*expr_p, 1));
            if (!useless_type_conversion_p (orig_type, TREE_TYPE (*expr_p)))
              {
                *expr_p = fold_convert_loc (input_location, orig_type, *expr_p);


Ciao,
Michael.
Kai Tietz - Oct. 13, 2011, 1:51 p.m.
Yes, I have already sent an patch with Richard's wish.  Indeed we need
only to do this type-casting for operands on transcription of
TRUTH_(AND|OR|XOR)_EXPR to BIT_(AND|OR|XOR)_EXPR.

Cheers,
Kai

Patch

Index: gcc/gcc/fold-const.c
===================================================================
--- gcc.orig/gcc/fold-const.c
+++ gcc/gcc/fold-const.c
@@ -112,13 +112,13 @@  static tree decode_field_reference (loca
 static int all_ones_mask_p (const_tree, int);
 static tree sign_bit_p (tree, const_tree);
 static int simple_operand_p (const_tree);
+static bool simple_operand_p_2 (tree);
 static tree range_binop (enum tree_code, tree, tree, int, tree, int);
 static tree range_predecessor (tree);
 static tree range_successor (tree);
 static tree fold_range_test (location_t, enum tree_code, tree, tree, tree);
 static tree fold_cond_expr_with_comparison (location_t, tree, tree,
tree, tree);
 static tree unextend (tree, int, int, tree);
-static tree fold_truthop (location_t, enum tree_code, tree, tree, tree);
 static tree optimize_minmax_comparison (location_t, enum tree_code,
 					tree, tree, tree);
 static tree extract_muldiv (tree, tree, enum tree_code, tree, bool *);
@@ -3500,7 +3500,7 @@  optimize_bit_field_compare (location_t l
   return lhs;
 }
 
-/* Subroutine for fold_truthop: decode a field reference.
+/* Subroutine for fold_truth_andor_1: decode a field reference.

    If EXP is a comparison reference, we return the innermost reference.

@@ -3668,7 +3668,7 @@  sign_bit_p (tree exp, const_tree val)
   return NULL_TREE;
 }

-/* Subroutine for fold_truthop: determine if an operand is simple enough
+/* Subroutine for fold_truth_andor_1: determine if an operand is simple enough
    to be evaluated unconditionally.  */

 static int
@@ -3678,7 +3678,7 @@  simple_operand_p (const_tree exp)
   STRIP_NOPS (exp);

   return (CONSTANT_CLASS_P (exp)
-	  || TREE_CODE (exp) == SSA_NAME
+  	  || TREE_CODE (exp) == SSA_NAME
 	  || (DECL_P (exp)
 	      && ! TREE_ADDRESSABLE (exp)
 	      && ! TREE_THIS_VOLATILE (exp)
@@ -3692,6 +3692,46 @@  simple_operand_p (const_tree exp)
 		 registers aren't expensive.  */
 	      && (! TREE_STATIC (exp) || DECL_REGISTER (exp))));
 }
+
+/* Subroutine for fold_truth_andor: determine if an operand is simple enough
+   to be evaluated unconditionally.
+   I addition to simple_operand_p, we assume that comparisons and logic-not
+   operations are simple, if their operands are simple, too.  */
+
+static bool
+simple_operand_p_2 (tree exp)
+{
+  enum tree_code code;
+
+  /* Strip any conversions that don't change the machine mode.  */
+  STRIP_NOPS (exp);
+
+  code = TREE_CODE (exp);
+
+  if (TREE_CODE_CLASS (code) == tcc_comparison)
+    return (!tree_could_trap_p (exp)
+	    && simple_operand_p_2 (TREE_OPERAND (exp, 0))
+	    && simple_operand_p_2 (TREE_OPERAND (exp, 1)));
+
+  if (TREE_SIDE_EFFECTS (exp)
+      || tree_could_trap_p (exp))
+    return false;
+
+  switch (code)
+    {
+    case SSA_NAME:
+      return true;
+    case TRUTH_NOT_EXPR:
+      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
+    case BIT_NOT_EXPR:
+      if (TREE_CODE (TREE_TYPE (exp)) != BOOLEAN_TYPE)
+	return false;
+      return simple_operand_p_2 (TREE_OPERAND (exp, 0));
+    default:
+      return simple_operand_p (exp);
+    }
+}
+
 
 /* The following functions are subroutines to fold_range_test and allow it to
    try to change a logical combination of comparisons into a range test.
@@ -4888,7 +4928,7 @@  fold_range_test (location_t loc, enum tr
   return 0;
 }
 
-/* Subroutine for fold_truthop: C is an INTEGER_CST interpreted as a P
+/* Subroutine for fold_truth_andor_1: C is an INTEGER_CST interpreted as a P
    bit value.  Arrange things so the extra bits will be set to zero if and
    only if C is signed-extended to its full width.  If MASK is nonzero,
    it is an INTEGER_CST that should be AND'ed with the extra bits.  */
@@ -5025,8 +5065,8 @@  merge_truthop_with_opposite_arm (locatio
    We return the simplified tree or 0 if no optimization is possible.  */

 static tree
-fold_truthop (location_t loc, enum tree_code code, tree truth_type,
-	      tree lhs, tree rhs)
+fold_truth_andor_1 (location_t loc, enum tree_code code, tree truth_type,
+		    tree lhs, tree rhs)
 {
   /* If this is the "or" of two comparisons, we can do something if
      the comparisons are NE_EXPR.  If this is the "and", we can do something
@@ -5054,8 +5094,6 @@  fold_truthop (location_t loc, enum tree_
   tree lntype, rntype, result;
   HOST_WIDE_INT first_bit, end_bit;
   int volatilep;
-  tree orig_lhs = lhs, orig_rhs = rhs;
-  enum tree_code orig_code = code;

   /* Start by getting the comparison codes.  Fail if anything is volatile.
      If one operand is a BIT_AND_EXPR with the constant one, treat it as if
@@ -5119,8 +5157,7 @@  fold_truthop (location_t loc, enum tree_
   /* If the RHS can be evaluated unconditionally and its operands are
      simple, it wins to evaluate the RHS unconditionally on machines
      with expensive branches.  In this case, this isn't a comparison
-     that can be merged.  Avoid doing this if the RHS is a floating-point
-     comparison since those can trap.  */
+     that can be merged.  */

   if (BRANCH_COST (optimize_function_for_speed_p (cfun),
 		   false) >= 2
@@ -5149,13 +5186,6 @@  fold_truthop (location_t loc, enum tree_
 			   build2 (BIT_IOR_EXPR, TREE_TYPE (ll_arg),
 				   ll_arg, rl_arg),
 			   build_int_cst (TREE_TYPE (ll_arg), 0));
-
-      if (LOGICAL_OP_NON_SHORT_CIRCUIT)
-	{
-	  if (code != orig_code || lhs != orig_lhs || rhs != orig_rhs)
-	    return build2_loc (loc, code, truth_type, lhs, rhs);
-	  return NULL_TREE;
-	}
     }

   /* See if the comparisons can be merged.  Then get all the parameters for
@@ -8380,13 +8410,49 @@  fold_truth_andor (location_t loc, enum t
      lhs is another similar operation, try to merge its rhs with our
      rhs.  Then try to merge our lhs and rhs.  */
   if (TREE_CODE (arg0) == code
-      && 0 != (tem = fold_truthop (loc, code, type,
-				   TREE_OPERAND (arg0, 1), arg1)))
+      && 0 != (tem = fold_truth_andor_1 (loc, code, type,
+					 TREE_OPERAND (arg0, 1), arg1)))
     return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0), tem);

-  if ((tem = fold_truthop (loc, code, type, arg0, arg1)) != 0)
+  if ((tem = fold_truth_andor_1 (loc, code, type, arg0, arg1)) != 0)
     return tem;

+  if ((code == TRUTH_ANDIF_EXPR || code == TRUTH_ORIF_EXPR)
+      && (BRANCH_COST (optimize_function_for_speed_p (cfun),
+		       false) >= 2)
+      && LOGICAL_OP_NON_SHORT_CIRCUIT
+      && simple_operand_p_2 (arg1))
+    {
+      enum tree_code ncode = (code == TRUTH_ANDIF_EXPR ? TRUTH_AND_EXPR
+						       : TRUTH_OR_EXPR);
+
+      /* Transform ((A AND-IF B) AND-IF C) into (A AND-IF (B AND C)),
+         or ((A OR-IF B) OR-IF C) into (A OR-IF (B OR C))
+         We don't want to pack more than two leafs to a non-IF AND/OR
+         expression.
+         If tree-code of left-hand operand isn't an AND/OR-IF code and not
+         equal to CODE, then we don't want to add right-hand operand.
+         If the inner right-hand side of left-hand operand has side-effects,
+         or isn't simple, then we can't add to it, as otherwise we might
+         destroy if-sequence.  */
+      if (TREE_CODE (arg0) == code
+      	  /* Needed for sequence points to handle trappings, and
+      	     side-effects.  */
+      	  && simple_operand_p_2 (TREE_OPERAND (arg0, 1)))
+       {
+         tem = fold_build2_loc (loc, ncode, type, TREE_OPERAND (arg0, 1),
+				arg1);
+         return fold_build2_loc (loc, code, type, TREE_OPERAND (arg0, 0),
+				 tem);
+       }
+     /* Transform (A AND-IF B) into (A AND B), or (A OR-IF B)
+        into (A OR B).
+        For sequence point consistancy, we need to check for trapping, and
+        side-effects.  */
+     else if (simple_operand_p_2 (arg0))
+       return fold_build2_loc (loc, ncode, type, arg0, arg1);
+    }
+
   return NULL_TREE;
 }

Index: gcc/gcc/gimplify.c
===================================================================
--- gcc.orig/gcc/gimplify.c
+++ gcc/gcc/gimplify.c
@@ -2804,8 +2804,13 @@  tree
 gimple_boolify (tree expr)
 {
   tree type = TREE_TYPE (expr);
+  tree new_expr_type;
   location_t loc = EXPR_LOCATION (expr);

+  new_expr_type = type;
+  if (TREE_CODE (type) != BOOLEAN_TYPE)
+    new_expr_type = boolean_type_node;
+
   if (TREE_CODE (expr) == NE_EXPR
       && TREE_CODE (TREE_OPERAND (expr, 0)) == CALL_EXPR
       && integer_zerop (TREE_OPERAND (expr, 1)))
@@ -2845,11 +2850,26 @@  gimple_boolify (tree expr)
     case TRUTH_ORIF_EXPR:
       /* Also boolify the arguments of truth exprs.  */
       TREE_OPERAND (expr, 1) = gimple_boolify (TREE_OPERAND (expr, 1));
-      /* FALLTHRU */
+
+      /* Make sure that operand type is compatible to final
+         expression's type.  */
+      if (!useless_type_conversion_p (new_expr_type,
+      				      TREE_TYPE (TREE_OPERAND (expr, 1))))
+        TREE_OPERAND (expr, 1) =
+          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 1));
+
+      /* Fall through.  */

     case TRUTH_NOT_EXPR:
       TREE_OPERAND (expr, 0) = gimple_boolify (TREE_OPERAND (expr, 0));

+      /* Make sure that operand type is compatible to final
+         expression's type.  */
+      if (!useless_type_conversion_p (new_expr_type,
+      				      TREE_TYPE (TREE_OPERAND (expr, 0))))
+        TREE_OPERAND (expr, 0) =
+          fold_convert_loc (loc, new_expr_type, TREE_OPERAND (expr, 0));
+
       /* These expressions always produce boolean results.  */
       if (TREE_CODE (type) != BOOLEAN_TYPE)
 	TREE_TYPE (expr) = boolean_type_node;
Index: gcc/gcc/testsuite/gfortran.fortran-torture/compile/logical-2.f90
===================================================================
--- /dev/null
+++ gcc/gcc/testsuite/gfortran.fortran-torture/compile/logical-2.f90
@@ -0,0 +1,10 @@ 
+! Check for operand type validity after gimplification
+
+subroutine whatever()
+logical(kind=1) :: l1
+logical(kind=2) :: l2
+logical(kind=4) :: l3
+if ((l1 .and. l2) .neqv. l3) then
+   l1 = .true.
+endif
+end