On 2019/7/2 4:51 PM, Richard Biener wrote:
On Tue, 2 Jul 2019, Richard Biener wrote:

On Tue, 2 Jul 2019, Li Jia He wrote:



On 2019/7/1 3:30 PM, Richard Biener wrote:
On Fri, 28 Jun 2019, Andrew Pinski wrote:

On Thu, Jun 27, 2019 at 9:55 PM Li Jia He <heli...@linux.ibm.com> wrote:



On 2019/6/27 11:48 PM, Jeff Law wrote:
On 6/27/19 12:11 AM, Li Jia He wrote:
Hi,

According to the optimizable case described by Qi Feng on
issue 88784, we can combine the cases into the following:

1. x >  y  &&  x != XXX_MIN  -->   x > y
2. x >  y  &&  x == XXX_MIN  -->   false
3. x <= y  &&  x == XXX_MIN  -->   x == XXX_MIN

4. x <  y  &&  x != XXX_MAX  -->   x < y
5. x <  y  &&  x == XXX_MAX  -->   false
6. x >= y  &&  x == XXX_MAX  -->   x == XXX_MAX

7. x >  y  ||  x != XXX_MIN  -->   x != XXX_MIN
8. x <= y  ||  x != XXX_MIN  -->   true
9. x <= y  ||  x == XXX_MIN  -->   x <= y

10. x <  y  ||  x != XXX_MAX  -->   x != UXXX_MAX
11. x >= y  ||  x != XXX_MAX  -->   true
12. x >= y  ||  x == XXX_MAX  -->   x >= y

Note: XXX_MIN represents the minimum value of type x.
         XXX_MAX represents the maximum value of type x.

Here we don't need to care about whether the operation is
signed or unsigned.  For example, in the below equation:

'x >  y  &&  x != XXX_MIN  -->   x > y'

If the x type is signed int and XXX_MIN is INT_MIN, we can
optimize it to 'x > y'.  However, if the type of x is unsigned
int and XXX_MIN is 0, we can still optimize it to 'x > y'.

The regression testing for the patch was done on GCC mainline on

       powerpc64le-unknown-linux-gnu (Power 9 LE)

with no regressions.  Is it OK for trunk ?

Thanks,
Lijia He

gcc/ChangeLog

2019-06-27  Li Jia He  <heli...@linux.ibm.com>
           Qi Feng  <ffen...@linux.ibm.com>

       PR middle-end/88784
       * gimple-fold.c (and_comparisons_contain_equal_operands): New
function.
       (and_comparisons_1): Use
and_comparisons_contain_equal_operands.
       (or_comparisons_contain_equal_operands): New function.
       (or_comparisons_1): Use or_comparisons_contain_equal_operands.
Would this be better done via match.pd?  ISTM this transformation
would
be well suited for that framework.

Hi, Jeff

I did this because of the following test case:
`
_Bool comp(unsigned x, unsigned y)
{
     return x > y && x != 0;
}
`
The gimple file dumped on the power platform is:
`
comp (unsigned int x, unsigned int y)
{
     _Bool D.2837;
     int iftmp.0;

     if (x > y) goto <D.2841>; else goto <D.2839>;
     <D.2841>:
     if (x != 0) goto <D.2842>; else goto <D.2839>;
     <D.2842>:
     iftmp.0 = 1;
     goto <D.2840>;
     <D.2839>:
     iftmp.0 = 0;
     <D.2840>:
     D.2837 = (_Bool) iftmp.0;
     return D.2837;
}
`
However, the gimple file dumped on x86 is
`
comp (unsigned int x, unsigned int y)
{
     _Bool D.2837;

     _1 = x > y;
     _2 = x != 0;
     _3 = _1 & _2;
     _4 = (int) _3;
     D.2837 = (_Bool) _4;
     return D.2837;
}
`

The reason for the inconsistency between these two behaviors is param
logical-op-non-short-circuit.  If we add the pattern to the match.pd
file, we can only optimize the situation in which the statement is in
the same basic block (logical-op-non-short-circuit=1, x86).  But for
a cross-basic block (logical-op-non-short-circuit=0, power), match.pd
can't handle this situation.

Another reason is that I found out maybe_fold_and_comparisons and
maybe_fold_or_comparisons are not only called by ifcombine pass but
also by reassoc pass. Using this method can basically unify param
logical-op-non-short-circuit=0 or 1.


As mentioned before ifcombine pass should be using gimple-match
instead of fold_build.  Try converting ifcombine over to gimple-match
infrastructure and add these to match.pd.

Yes, I mentioned that in the PR.  The issue is that at the moment
to combine x > y with x <= y you'd have to build GENERIC trees
for both or temporary GIMPLE assign with a SSA def (and then feed
that into the GENERIC or GIMPLE match.pd path).

Hi,

I did some experimentation using ‘temporary GIMPLE with a SSA def (and then
feed that into the GIMPLE match.pd path’.  Could we consider the code in the
attachment(I did a test and the code took effect)?

No, that's excessive overhead IMHO - the whole point of these functions
originally was to avoid build2 (...) on both conditionals.  If we
now create two SSA names and two GIMPLE statements that's too much.

As said it shouldn't be rocket science to teach genmatch to
auto-generate those functions from match.pd patterns but it certainly
is some work (less so for somebody with genmatch knowledge).
I'll give it a poke...

OK, it's not so easy.  I guess we could lower the cost of building
SSA names / gimple stmts significantly if we allocated them on the
stack like via

gimple *stmt1 = XALLOCAVEC (char, gimple_size (GIMPLE_ASSIGN) + 2 *
sizeof (tree));
memset (stmt1, 0, ...);
gimple_set_code (stmt1, GIMPLE_ASSIGN);
gimple_set_num_ops (stmt1, 3);
gimple_init_singleton (stmt1);

gimple_stmt_iterator gsi = gsi_for_stmt (stmt1);
gimple_assign_set_rhs_with_ops (&gsi, actual operation...);

tree lhs1 = XALLOCA (tree_ssa_name);
memset (lhs1, 0, sizeof (tree_ssa_name));
TREE_SET_CODE (lhs1, SSA_NAME);
TREE_TYPE (lhs1) = ...;

gimple_assing_set_lhs (stmt1, lhs1);

it's all a bit ugly and some factoring in the generic gimple machinery
might easen this kind of hacks, but well...

With the above you could then use

   gimple_match_op op (gimple_match_cond::UNCOND, BIT_AND_EXPR /* or OR */,
                      boolean_type_node, lhs1, lhs2);
   if (gimple_resimplify2 (NULL, &op, follow_all_ssa_edges))
     ... successfully simplified into 'op' ...

with the simplification path then extracting the simplified result
from 'op'.  Note this deliberately leaves out passing a gimple_seq
as storage for more complex simplification results to match
existing behavior.

Your patch has the GC allocation and SSA name (and namespace!)
allocation overhead and most definitely misses releasing the
SSA names you allocated.

Note with the above hack the simplified result has to be checked
for mentions of lhs1 or lhs2 - a case we have to reject because
their definitions are transitional.


Hi,

  I made some changes based on the recommendations. Would you like to
  help me to see it again ? Sorry, it took so long time to provide the
  patch.

  Note: 1. I keep the code for and_comparisons_1 and or_comparisons_1.
        The reason is that I did not found a good way to handle the
        optimization of '((x CODE1 y) AND (x CODE2 y))' in match.pd.
        Maybe I missing some important information about match.pd.
        2. The gimple_resimplify2 function is not used.  Since stmt1,
        stmt2, lhs1 and lhs2 are allocated on the stack, Is there a
        question with the value on the stack as the return value ?
        I may have misunderstood Richard's intention.

Thanks,
Lijia He

Richard.

Richard.

Thanks,
Lijia He


maybe_fold_and/or_comparisons handle two exploded binary expressions
while the current match.pd entries handle at most one exploded one
(the outermost then, either AND or OR).  But it would be definitely
doable to auto-generate maybe_fold_and/or_comparisons from match.pd
patterns which is what I'd ultimatively suggest to do (in some more
generalized form maybe).  Either with a separate genmatch invocation
or as part of the --gimple processing (not sure what is more feasible
here).

I told Li Jia He that I don't expect him to do this work.

Note I didn't review the actual patch yet.

Thanks,
Richard.





From 2699516000abdef60b06b203d703f80b8905cf09 Mon Sep 17 00:00:00 2001
From: Li Jia He <heli...@linux.ibm.com>
Date: Mon, 15 Jul 2019 00:30:25 -0500
Subject: [PATCH 1/2] Auto-generate maybe_fold_and/or_comparisons from match.pd

Hi,

This patch try to auto-generate maybe_fold_and/or_comparisons from match.pd.
We try to build tempory assign with SSA def and then feed that into the GIMPLE
match.pd path.

In order to lower the cost of building SSA names / gimple stmts significantly,
we need to allocated them on the stack.  This will cause the code a bit ugly.

The regression testing for the patch was done on GCC mainline on

        powerpc64le-unknown-linux-gnu (Power 9 LE)

with no regressions.  Is it OK for trunk ?

Thanks,
Lijia He

gcc/ChangeLog

2019-07-16  Li Jia He  <heli...@linux.ibm.com>

        * gimple.h (gimple_set_code): Declare.
        (gimple_size): Likewise.
        * gimple.c (gimple_set_code): Remove static and inline restrictions.
        (gimple_size): Likewise.
        * gimple-fold.c (maybe_fold_comparisons_from_match_pd): New function.
        (maybe_fold_and_comparisons): Modify and_comparisons_1 invocation and
        call maybe_fold_comparisons_from_match_pd.
        (maybe_fold_or_comparisons): Modify or_comparisons_1 invocation and
        call maybe_fold_comparisons_from_match_pd.
---
 gcc/gimple-fold.c | 122 ++++++++++++++++++++++++++++++++++++++++++----
 gcc/gimple.c      |   4 +-
 gcc/gimple.h      |   2 +
 3 files changed, 117 insertions(+), 11 deletions(-)

diff --git a/gcc/gimple-fold.c b/gcc/gimple-fold.c
index dfb31a02078..a5c2790361f 100644
--- a/gcc/gimple-fold.c
+++ b/gcc/gimple-fold.c
@@ -5789,6 +5789,91 @@ and_comparisons_1 (enum tree_code code1, tree op1a, tree 
op1b,
   return NULL_TREE;
 }
 
+/* Helper function for maybe_fold_and_comparisons and maybe_fold_or_comparisons
+   : try to simplify the AND/OR of the ssa variable VAR with the comparison
+   specified by (OP2A CODE2 OP2B) from match.pd.  Return NULL_EXPR if we can't
+   simplify this to a single expression.  As we are going to lower the cost
+   of building SSA names / gimple stmts significantly, we need to allocate
+   them ont the stack.  This will cause the code to be a bit ugly.  */
+
+static tree
+maybe_fold_comparisons_from_match_pd (enum tree_code code, enum tree_code 
code1,
+                                     tree op1a, tree op1b,
+                                     enum tree_code code2, tree op2a,
+                                     tree op2b)
+{
+  /* Allocate gimple stmt1 on the stack.  */
+  gimple *stmt1 = (gimple *) (XALLOCAVEC (char, gimple_size (GIMPLE_ASSIGN)
+                                                 + 2 * sizeof (tree)));
+  memset (stmt1, 0, gimple_size (GIMPLE_ASSIGN) + 2 * sizeof (tree));
+  gimple_set_code (stmt1, GIMPLE_ASSIGN);
+  gimple_set_num_ops (stmt1, 3);
+  gimple_init_singleton (stmt1);
+
+  /* Allocate gimple stmt2 on the stack.  */
+  gimple *stmt2 = (gimple *) (XALLOCAVEC (char, gimple_size (GIMPLE_ASSIGN)
+                                                 + 2 * sizeof (tree)));
+  memset (stmt2, 0, gimple_size (GIMPLE_ASSIGN) + 2 * sizeof (tree));
+  gimple_set_code (stmt2, GIMPLE_ASSIGN);
+  gimple_set_num_ops (stmt2, 3);
+  gimple_init_singleton (stmt2);
+
+  gimple_assign_set_rhs_code (stmt1, code1);
+  gimple_assign_set_rhs1 (stmt1, op1a);
+  gimple_assign_set_rhs2 (stmt1, op1b);
+
+  gimple_assign_set_rhs_code (stmt2, code2);
+  gimple_assign_set_rhs1 (stmt2, op2a);
+  gimple_assign_set_rhs2 (stmt2, op2b);
+
+  /* Allocate SSA names(lhs1) on the stack.  */
+  tree lhs1 = XALLOCA (tree_node);
+  memset (lhs1, 0, sizeof (tree_node));
+  TREE_SET_CODE (lhs1, SSA_NAME);
+  TREE_TYPE (lhs1) = boolean_type_node;
+
+  /* Allocate SSA names(lhs2) on the stack.  */
+  tree lhs2 = XALLOCA (tree_node);
+  memset (lhs2, 0, sizeof (tree_node));
+  TREE_SET_CODE (lhs2, SSA_NAME);
+  TREE_TYPE (lhs2) = boolean_type_node;
+
+  gimple_assign_set_lhs (stmt1, lhs1);
+  gimple_assign_set_lhs (stmt2, lhs2);
+
+  /* Call the interface function of match.pd to simplify the expression.  */
+  tree t = gimple_simplify (code, boolean_type_node, gimple_assign_lhs (stmt1),
+                           gimple_assign_lhs (stmt2), NULL,
+                           follow_all_ssa_edges);
+
+  if (t)
+    {
+      switch (TREE_CODE (t))
+       {
+         case SSA_NAME:
+           {
+             gimple *def = SSA_NAME_DEF_STMT (t);
+
+             if (!is_gimple_assign (def)
+                 || TREE_CODE_CLASS (gimple_assign_rhs_code (def))
+                   != tcc_comparison)
+               return NULL_TREE;
+
+             return fold_build2 (gimple_assign_rhs_code (def),
+                                 boolean_type_node, gimple_assign_rhs1 (def),
+                                 gimple_assign_rhs2 (def));
+         }
+       case INTEGER_CST:
+         /* Fold expression to boolean_true_node or boolean_false_node.  */
+         return t;
+       default:
+         return NULL_TREE;
+       }
+    }
+
+  return NULL_TREE;
+}
+
 /* Try to simplify the AND of two comparisons, specified by
    (OP1A CODE1 OP1B) and (OP2B CODE2 OP2B), respectively.
    If this can be simplified to a single expression (without requiring
@@ -5800,11 +5885,21 @@ tree
 maybe_fold_and_comparisons (enum tree_code code1, tree op1a, tree op1b,
                            enum tree_code code2, tree op2a, tree op2b)
 {
-  tree t = and_comparisons_1 (code1, op1a, op1b, code2, op2a, op2b);
-  if (t)
+  if (tree t = and_comparisons_1 (code1, op1a, op1b, code2, op2a, op2b))
     return t;
-  else
-    return and_comparisons_1 (code2, op2a, op2b, code1, op1a, op1b);
+
+  if (tree t = and_comparisons_1 (code2, op2a, op2b, code1, op1a, op1b))
+    return t;
+
+  if (tree t = maybe_fold_comparisons_from_match_pd (BIT_AND_EXPR, code1, op1a,
+                                                    op1b, code2, op2a, op2b))
+    return t;
+
+  if (tree t = maybe_fold_comparisons_from_match_pd (BIT_AND_EXPR, code2, op2a,
+                                                    op2b, code1, op1a, op1b))
+    return t;
+
+  return NULL_TREE;
 }
 
 /* Helper function for or_comparisons_1:  try to simplify the OR of the
@@ -6264,13 +6359,22 @@ tree
 maybe_fold_or_comparisons (enum tree_code code1, tree op1a, tree op1b,
                           enum tree_code code2, tree op2a, tree op2b)
 {
-  tree t = or_comparisons_1 (code1, op1a, op1b, code2, op2a, op2b);
-  if (t)
+  if (tree t = or_comparisons_1 (code1, op1a, op1b, code2, op2a, op2b))
+    return t;
+
+  if (tree t = or_comparisons_1 (code2, op2a, op2b, code1, op1a, op1b))
+    return t;
+
+  if (tree t = maybe_fold_comparisons_from_match_pd (BIT_IOR_EXPR, code1, op1a,
+                                                    op1b, code2, op2a, op2b))
+    return t;
+
+  if (tree t = maybe_fold_comparisons_from_match_pd (BIT_IOR_EXPR, code2, op2a,
+                                                    op2b, code1, op1a, op1b))
     return t;
-  else
-    return or_comparisons_1 (code2, op2a, op2b, code1, op1a, op1b);
-}
 
+  return NULL_TREE;
+}
 
 /* Fold STMT to a constant using VALUEIZE to valueize SSA names.
 
diff --git a/gcc/gimple.c b/gcc/gimple.c
index 63c8d5e85ae..6769bd9edec 100644
--- a/gcc/gimple.c
+++ b/gcc/gimple.c
@@ -101,7 +101,7 @@ const enum gimple_code gcond::code_;
 
 /* Set the code for statement G to CODE.  */
 
-static inline void
+void
 gimple_set_code (gimple *g, enum gimple_code code)
 {
   g->code = code;
@@ -110,7 +110,7 @@ gimple_set_code (gimple *g, enum gimple_code code)
 /* Return the number of bytes needed to hold a GIMPLE statement with
    code CODE.  */
 
-static inline size_t
+size_t
 gimple_size (enum gimple_code code)
 {
   return gsstruct_code_size[gss_for_code (code)];
diff --git a/gcc/gimple.h b/gcc/gimple.h
index 47070e7f409..a32708ad349 100644
--- a/gcc/gimple.h
+++ b/gcc/gimple.h
@@ -1447,6 +1447,8 @@ extern enum gimple_statement_structure_enum const 
gss_for_code_[];
    of comminucating the profile info to the builtin expanders.  */
 extern gimple *currently_expanding_gimple_stmt;
 
+void gimple_set_code (gimple *g, enum gimple_code code);
+size_t gimple_size (enum gimple_code code);
 gimple *gimple_alloc (enum gimple_code, unsigned CXX_MEM_STAT_INFO);
 greturn *gimple_build_return (tree);
 void gimple_call_reset_alias_info (gcall *);
-- 
2.17.1

From 757bae5a8f3985f8704d846f1701bf345ab0fe43 Mon Sep 17 00:00:00 2001
From: Li Jia He <heli...@linux.ibm.com>
Date: Mon, 15 Jul 2019 03:50:34 -0500
Subject: [PATCH 2/2] Fix PR88784, middle end is missing some optimizations
 about unsigned

Hi,

According to the optimizable case described by Qi Feng on
issue 88784, we can combine the cases into the following:

1. x >  y  &&  x != XXX_MIN  -->   x > y
2. x >  y  &&  x == XXX_MIN  -->   false
3. x <= y  &&  x == XXX_MIN  -->   x == XXX_MIN

4. x <  y  &&  x != XXX_MAX  -->   x < y
5. x <  y  &&  x == XXX_MAX  -->   false
6. x >= y  &&  x == XXX_MAX  -->   x == XXX_MAX

7. x >  y  ||  x != XXX_MIN  -->   x != XXX_MIN
8. x <= y  ||  x != XXX_MIN  -->   true
9. x <= y  ||  x == XXX_MIN  -->   x <= y

10. x <  y  ||  x != XXX_MAX  -->   x != UXXX_MAX
11. x >= y  ||  x != XXX_MAX  -->   true
12. x >= y  ||  x == XXX_MAX  -->   x >= y

Note: XXX_MIN represents the minimum value of type x.
      XXX_MAX represents the maximum value of type x.

Here we don't need to care about whether the operation is
signed or unsigned.  For example, in the below equation:

'x >  y  &&  x != XXX_MIN  -->   x > y'

If the x type is signed int and XXX_MIN is INT_MIN, we can
optimize it to 'x > y'.  However, if the type of x is unsigned
int and XXX_MIN is 0, we can still optimize it to 'x > y'.

The regression testing for the patch was done on GCC mainline on

    powerpc64le-unknown-linux-gnu (Power 9 LE)

with no regressions.  Is it OK for trunk ?

Thanks,
Lijia He

gcc/ChangeLog

2019-07-16  Li Jia He  <heli...@linux.ibm.com>
            Qi Feng  <ffen...@linux.ibm.com>

        PR middle-end/88784
        * match.pd (x >  y  &&  x != XXX_MIN): Optimize into 'x > y'.
        (x >  y  &&  x == XXX_MIN): Optimize into 'false'.
        (x <= y  &&  x == XXX_MIN): Optimize into 'x == XXX_MIN'.
        (x <  y  &&  x != XXX_MAX): Optimize into 'x < y'.
        (x <  y  &&  x == XXX_MAX): Optimize into 'false'.
        (x >= y  &&  x == XXX_MAX): Optimize into 'x == XXX_MAX'.
        (x >  y  ||  x != XXX_MIN): Optimize into 'x != XXX_MIN'.
        (x <= y  ||  x != XXX_MIN): Optimize into 'true'.
        (x <= y  ||  x == XXX_MIN): Optimize into 'x <= y'.
        (x <  y  ||  x != XXX_MAX): Optimize into 'x != XXX_MAX'.
        (x >= y  ||  x != XXX_MAX): Optimize into 'true'.
        (x >= y  ||  x == XXX_MAX): Optimize into 'x >= y'.

gcc/testsuite/ChangeLog

2019-07-16  Li Jia He  <heli...@linux.ibm.com>
            Qi Feng  <ffen...@linux.ibm.com>

        PR middle-end/88784
        * gcc.dg/pr88784-1.c: New testcase.
        * gcc.dg/pr88784-2.c: New testcase.
        * gcc.dg/pr88784-3.c: New testcase.
        * gcc.dg/pr88784-4.c: New testcase.
        * gcc.dg/pr88784-5.c: New testcase.
        * gcc.dg/pr88784-6.c: New testcase.
        * gcc.dg/pr88784-7.c: New testcase.
        * gcc.dg/pr88784-8.c: New testcase.
        * gcc.dg/pr88784-9.c: New testcase.
        * gcc.dg/pr88784-10.c: New testcase.
        * gcc.dg/pr88784-11.c: New testcase.
        * gcc.dg/pr88784-12.c: New testcase.
---
 gcc/match.pd                      | 95 +++++++++++++++++++++++++++++++
 gcc/testsuite/gcc.dg/pr88784-1.c  | 30 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-10.c | 32 +++++++++++
 gcc/testsuite/gcc.dg/pr88784-11.c | 30 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-12.c | 30 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-2.c  | 30 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-3.c  | 32 +++++++++++
 gcc/testsuite/gcc.dg/pr88784-4.c  | 32 +++++++++++
 gcc/testsuite/gcc.dg/pr88784-5.c  | 31 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-6.c  | 31 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-7.c  | 31 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-8.c  | 31 ++++++++++
 gcc/testsuite/gcc.dg/pr88784-9.c  | 32 +++++++++++
 13 files changed, 467 insertions(+)
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-1.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-10.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-11.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-12.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-2.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-3.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-4.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-5.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-6.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-7.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-8.c
 create mode 100644 gcc/testsuite/gcc.dg/pr88784-9.c

diff --git a/gcc/match.pd b/gcc/match.pd
index f8e35e96d22..21a147d0ff1 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -1859,6 +1859,101 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
     { wide_int_to_tree (type, (wi::to_wide (@1)
                               & (bitpos / BITS_PER_UNIT))); }))))
 
+/* x >  y  &&  x != XXX_MIN  -->  x > y  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (gt:c@3 @0 @1) (ne @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2)))))
+    @3)))
+
+/* x >  y  &&  x == XXX_MIN  -->  false  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (gt:c @0 @1) (eq @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2)))))
+    { boolean_false_node; })))
+
+/* x <=  y  &&  x == XXX_MIN  -->  x == XXX_MIN  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (le:c @0 @1) (eq@3 @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2)))))
+    @3)))
+
+/* x <  y  &&  x != XXX_MAX  -->  x < y  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (lt:c@3 @0 @1) (ne @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2)))))
+    @3)))
+
+/* x <  y  &&  x == XXX_MAX  -->  false  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (lt:c @0 @1) (eq @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2)))))
+    { boolean_false_node; })))
+
+/* x >=  y  &&  x == XXX_MAX  -->  x == XXX_MAX  */
+(for and (truth_and bit_and)
+ (simplify
+  (and:c (ge:c @0 @1) (eq@3 @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && (wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2)))))
+    @3)))
+
+/* x >  y  ||  x != XXX_MIN   -->  x != XXX_MIN  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (gt:c @0 @1) (ne@3 @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2))))
+    @3)))
+
+/* x <=  y  ||  x != XXX_MIN   -->  true  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (le:c @0 @1) (ne @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2))))
+    { boolean_true_node; })))
+
+/* x <=  y  ||  x == XXX_MIN   -->  x <= y  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (le:c@3 @0 @1) (eq @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::min_value (TREE_TYPE (@2))))
+    @3)))
+
+/* x <  y  ||  x != XXX_MAX   -->  x != XXX_MAX  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (lt:c @0 @1) (ne@3 @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2))))
+    @3)))
+
+/* x >=  y  ||  x != XXX_MAX   -->  true  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (ge:c @0 @1) (ne @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2))))
+    { boolean_true_node; })))
+
+/* x >=  y  ||  x == XXX_MAX   -->  x >= y  */
+(for or (truth_or bit_ior)
+ (simplify
+  (or:c (ge:c@3 @0 @1) (eq @0 INTEGER_CST@2))
+  (if (INTEGRAL_TYPE_P (TREE_TYPE (@0)) && INTEGRAL_TYPE_P (TREE_TYPE(@1))
+       && wi::eq_p (wi::to_wide (@2), wi::max_value (TREE_TYPE (@2))))
+    @3)))
 
 /* We can't reassociate at all for saturating types.  */
 (if (!TYPE_SATURATING (type))
diff --git a/gcc/testsuite/gcc.dg/pr88784-1.c b/gcc/testsuite/gcc.dg/pr88784-1.c
new file mode 100644
index 00000000000..067d40d0739
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-1.c
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x > y && x != 0 --> x > y */
+  return x > y && x != 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x < y && x != UINT_MAX --> x < y */
+  return x < y && x != UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x > y && x != INT_MIN --> x > y */
+  return x > y && x != INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x < y && x != INT_MAX --> x < y */
+  return x < y && x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " != " "ifcombine" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-10.c 
b/gcc/testsuite/gcc.dg/pr88784-10.c
new file mode 100644
index 00000000000..2218faa3b7c
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-10.c
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-gimple --param 
logical-op-non-short-circuit=1" } */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x <= y || x != 0 --> true */
+  return x <= y || x != 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x >= y || x != UINT_MAX --> true */
+  return x >= y || x != UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x <= y || x != INT_MIN --> true */
+  return x <= y || x != INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x >= y || x != INT_MAX --> true */
+  return x >= y || x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " != " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " <= " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " >= " "gimple" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-11.c 
b/gcc/testsuite/gcc.dg/pr88784-11.c
new file mode 100644
index 00000000000..4465910efbb
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-11.c
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x <= y || x == 0 --> x <= y */
+  return x <= y || x == 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x >= y || x == UINT_MAX --> x >= y */
+  return x >= y || x == UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x <= y || x == INT_MIN --> x <= y */
+  return x <= y || x == INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x >= y || x == INT_MAX --> x >= y */
+  return x >= y || x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " == " "ifcombine" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-12.c 
b/gcc/testsuite/gcc.dg/pr88784-12.c
new file mode 100644
index 00000000000..477bba07821
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-12.c
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-dce3 --param logical-op-non-short-circuit=1" 
} */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x <= y || x == 0 --> x <= y */
+  return x <= y || x == 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x >= y || x == UINT_MAX --> x >= y */
+  return x >= y || x == UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x <= y || x == INT_MIN --> x <= y */
+  return x <= y || x == INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x >= y || x == INT_MAX --> x >= y */
+  return x >= y || x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " == " "dce3" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-2.c b/gcc/testsuite/gcc.dg/pr88784-2.c
new file mode 100644
index 00000000000..02647204726
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-2.c
@@ -0,0 +1,30 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-gimple --param 
logical-op-non-short-circuit=1" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x > y && x != 0 --> x > y */
+  return x > y && x != 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x < y && x != UINT_MAX --> x < y */
+  return x < y && x != UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x > y && x != INT_MIN --> x > y */
+  return x > y && x != INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x < y && x != INT_MAX --> x < y */
+  return x < y && x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " != " "gimple" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-3.c b/gcc/testsuite/gcc.dg/pr88784-3.c
new file mode 100644
index 00000000000..be2ce315e60
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-3.c
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x > y && x == 0 --> false */
+  return x > y && x == 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x < y && x == UINT_MAX --> false */
+  return x < y && x == UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x > y && x == INT_MIN --> false */
+  return x > y && x == INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x < y && x == INT_MAX --> false */
+  return x < y && x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " == " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " > " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " < " "ifcombine" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-4.c b/gcc/testsuite/gcc.dg/pr88784-4.c
new file mode 100644
index 00000000000..d1363659f40
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-4.c
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-gimple --param 
logical-op-non-short-circuit=1" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x > y && x == 0 --> false */
+  return x > y && x == 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x < y && x == UINT_MAX --> false */
+  return x < y && x == UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x > y && x == INT_MIN --> false */
+  return x > y && x == INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x < y && x == INT_MAX --> false */
+  return x < y && x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " == " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " > " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " < " "gimple" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-5.c b/gcc/testsuite/gcc.dg/pr88784-5.c
new file mode 100644
index 00000000000..c6a349d7c75
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-5.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x <= y && x == 0 --> x == 0 */
+  return x <= y && x == 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x >= y && x == UINT_MAX --> x == UINT_MAX */
+  return x >= y && x == UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x <= y && x == INT_MIN --> x == INT_MIN */
+  return x <= y && x == INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x >= y && x == INT_MAX --> x == INT_MAX */
+  return x >= y && x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " <= " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " >= " "ifcombine" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-6.c b/gcc/testsuite/gcc.dg/pr88784-6.c
new file mode 100644
index 00000000000..b0e81e0165e
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-6.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-gimple --param 
logical-op-non-short-circuit=1" } */
+
+#include <limits.h>
+
+_Bool and1(unsigned x, unsigned y)
+{
+  /* x <= y && x == 0 --> x == 0 */
+  return x <= y && x == 0;
+}
+
+_Bool and2(unsigned x, unsigned y)
+{
+  /* x >= y && x == UINT_MAX --> x == UINT_MAX */
+  return x >= y && x == UINT_MAX;
+}
+
+_Bool and3(signed x, signed y)
+{
+  /* x <= y && x == INT_MIN --> x == INT_MIN */
+  return x <= y && x == INT_MIN;
+}
+
+_Bool and4(signed x, signed y)
+{
+  /* x >= y && x == INT_MAX --> x == INT_MAX */
+  return x >= y && x == INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " <= " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " >= " "gimple" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-7.c b/gcc/testsuite/gcc.dg/pr88784-7.c
new file mode 100644
index 00000000000..937d2d26593
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-7.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x > y || x != 0 --> x != 0 */
+  return x > y || x != 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x < y || x != UINT_MAX --> x != UINT_MAX */
+  return x < y || x != UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x > y || x != INT_MIN --> x != INT_MIN */
+  return x > y || x != INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x < y || x != INT_MAX --> x != INT_MAX */
+  return x < y || x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " > " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " < " "ifcombine" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-8.c b/gcc/testsuite/gcc.dg/pr88784-8.c
new file mode 100644
index 00000000000..aa384c305b1
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-8.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-gimple --param 
logical-op-non-short-circuit=1" } */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x > y || x != 0 --> x != 0 */
+  return x > y || x != 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x < y || x != UINT_MAX --> x != UINT_MAX */
+  return x < y || x != UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x > y || x != INT_MIN --> x != INT_MIN */
+  return x > y || x != INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x < y || x != INT_MAX --> x != INT_MAX */
+  return x < y || x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " > " "gimple" } } */
+/* { dg-final { scan-tree-dump-not " < " "gimple" } } */
diff --git a/gcc/testsuite/gcc.dg/pr88784-9.c b/gcc/testsuite/gcc.dg/pr88784-9.c
new file mode 100644
index 00000000000..94f62d94ede
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr88784-9.c
@@ -0,0 +1,32 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ifcombine --param 
logical-op-non-short-circuit=0" } */
+
+#include <limits.h>
+
+_Bool or1(unsigned x, unsigned y)
+{
+  /* x <= y || x != 0 --> true */
+  return x <= y || x != 0;
+}
+
+_Bool or2(unsigned x, unsigned y)
+{
+  /* x >= y || x != UINT_MAX --> true */
+  return x >= y || x != UINT_MAX;
+}
+
+_Bool or3(signed x, signed y)
+{
+  /* x <= y || x != INT_MIN --> true */
+  return x <= y || x != INT_MIN;
+}
+
+_Bool or4(signed x, signed y)
+{
+  /* x >= y || x != INT_MAX --> true */
+  return x >= y || x != INT_MAX;
+}
+
+/* { dg-final { scan-tree-dump-not " != " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " <= " "ifcombine" } } */
+/* { dg-final { scan-tree-dump-not " >= " "ifcombine" } } */
-- 
2.17.1

Reply via email to