[PATCH 1/2] Match: zero_one_valued_p should match 0 constants too

[PATCH 1/2] Match: zero_one_valued_p should match 0 constants too

[Andrew Pinski]

While working on `bool0 ? bool1 : bool2` I noticed that
zero_one_valued_p does not match on the constant zero
as in that case tree_nonzero_bits will return 0 and
that is different from 1.

OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.

gcc/ChangeLog:

	* match.pd (zero_one_valued_p): Match 0 integer constant
	too.
---
 gcc/match.pd | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/gcc/match.pd b/gcc/match.pd
index f9cbd757752..f97ff7ef760 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -1983,11 +1983,16 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
   (bit_not (bit_not @0))
   @0)
 
+/* zero_one_valued_p will match when a value is known to be either
+   0 or 1 including the constant 0. */
 (match zero_one_valued_p
  @0
  (if (INTEGRAL_TYPE_P (type) && tree_nonzero_bits (@0) == 1)))
 (match zero_one_valued_p
  truth_valued_p@0)
+(match zero_one_valued_p
+ integer_zerop@0
+ (if (INTEGRAL_TYPE_P (type))))
 
 /* Transform { 0 or 1 } * { 0 or 1 } into { 0 or 1 } & { 0 or 1 }.  */
 (simplify
-- 
2.31.1

[PATCH 2/2] Add match patterns for `a ? onezero : onezero` where one of the two operands are constant

[Andrew Pinski]

This adds a match pattern that are for boolean values
that optimizes `a ? onezero : 0` to `a & onezero` and
`a ? 1 : onezero` to `a | onezero`.

This was reported a few times and I thought I would finally
add the match pattern for this.

This hits a few times in GCC itself too.

Notes on the testcases:
* phi-opt-2.c: This now is optimized to `a & b` in phiopt rather than ifcombine
* phi-opt-25b.c: The test part that was failing was parity which now gets `x & y` treatment.
* ssa-thread-21.c: there is no longer a threading opportunity, so need to disable phiopt.
  Note PR 109957 is filed for the now missing optimization in that testcase too.

gcc/ChangeLog:

	PR tree-optimization/89263
	PR tree-optimization/99069
	PR tree-optimization/20083
	PR tree-optimization/94898
	* match.pd: Add patterns to optimize `a ? onezero : onezero` with
	one of the operands are constant.

gcc/testsuite/ChangeLog:

	* gcc.dg/tree-ssa/phi-opt-2.c: Adjust the testcase.
	* gcc.dg/tree-ssa/phi-opt-25b.c: Adjust the testcase.
	* gcc.dg/tree-ssa/ssa-thread-21.c: Disable phiopt.
	* gcc.dg/tree-ssa/phi-opt-27.c: New test.
	* gcc.dg/tree-ssa/phi-opt-28.c: New test.
	* gcc.dg/tree-ssa/phi-opt-29.c: New test.
	* gcc.dg/tree-ssa/phi-opt-30.c: New test.
	* gcc.dg/tree-ssa/phi-opt-31.c: New test.
	* gcc.dg/tree-ssa/phi-opt-32.c: New test.
---
 gcc/match.pd                                  | 18 ++++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-2.c     | 12 ++--
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-25b.c   |  8 ++-
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-27.c    | 14 +++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-28.c    | 14 +++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-29.c    | 25 +++++++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-30.c    | 55 +++++++++++++++++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-31.c    | 15 +++++
 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-32.c    | 12 ++++
 gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-21.c |  3 +-
 10 files changed, 165 insertions(+), 11 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-27.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-28.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-29.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-30.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-31.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/phi-opt-32.c

diff --git a/gcc/match.pd b/gcc/match.pd
index f97ff7ef760..dc36927cd0f 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -4721,6 +4721,24 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
  )
 )
 
+(simplify
+ (cond @0 zero_one_valued_p@1 zero_one_valued_p@2)
+ (switch
+  /* bool0 ? bool1 : 0 -> bool0 & bool1 */
+  (if (integer_zerop (@2))
+   (bit_and (convert @0) @1))
+  /* bool0 ? 0 : bool2 -> (bool0^1) & bool2 */
+  (if (integer_zerop (@1))
+   (bit_and (bit_xor (convert @0) { build_one_cst (type); } ) @2))
+  /* bool0 ? 1 : bool2 -> bool0 | bool2 */
+  (if (integer_onep (@1))
+   (bit_ior (convert @0) @2))
+  /* bool0 ? bool1 : 1 -> (bool0^1) | bool1 */
+  (if (integer_onep (@2))
+   (bit_ior (bit_xor (convert @0) @2) @1))
+ )
+)
+
 /* Optimize
    # x_5 in range [cst1, cst2] where cst2 = cst1 + 1
    x_5 ? cstN ? cst4 : cst3
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-2.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-2.c
index 5c7815e2c1a..006e8e83052 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-2.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-2.c
@@ -14,10 +14,8 @@ _Bool f1(_Bool a, _Bool b)
 }
 
 
-/* There should be only one if, the outer one; the inner one
-   should have been changed to straight line code with the
-   value of b (except that we don't fold ! (b != 0) into b
-   which can be fixed in a different patch).
-   Test this only when known to be !LOGICAL_OP_NON_SHORT_CIRCUIT,
-   otherwise ifcombine may convert this into return a & b;.  */
-/* { dg-final { scan-tree-dump-times "if" 1 "optimized" } } */
+/* There should be no if statements and be fold into just return a & b.
+   This can be done without ifcombine but in phiopt where a ? b : 0 is
+   converted into a & b. */
+/* { dg-final { scan-tree-dump-not "if" "optimized" } } */
+/* { dg-final { scan-tree-dump-times " & " 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-25b.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-25b.c
index 698a20f7a56..7298da0c96e 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-25b.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-25b.c
@@ -65,6 +65,8 @@ int test_popcountll(unsigned long long x, unsigned long long y)
   return x ? __builtin_popcountll(y) : 0;
 }
 
-/* 4 types of functions, each with 3 types and there are 2 goto each */
-/* { dg-final { scan-tree-dump-times "goto " 24 "optimized" } } */
-
+/* 3 types of functions (not including parity), each with 3 types and there are 2 goto each */
+/* { dg-final { scan-tree-dump-times "goto " 18 "optimized" } } */
+/* { dg-final { scan-tree-dump-times "x_..D. != 0" 12 "optimized" } } */
+/* parity case will be optimized to x!=0 & parity(y) . */
+/* { dg-final { scan-tree-dump-times " & " 3 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-27.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-27.c
new file mode 100644
index 00000000000..11f85b91644
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-27.c
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O1 -fdump-tree-optimized" } */
+
+_Bool f1(_Bool a, _Bool b)
+{
+  if (a)
+    return b;
+  return 0;
+}
+
+
+/* There should be no if statements and be fold into just return a & b. */
+/* { dg-final { scan-tree-dump-not "if" "optimized" } } */
+/* { dg-final { scan-tree-dump-times " & " 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-28.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-28.c
new file mode 100644
index 00000000000..43dadd9341f
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-28.c
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O1 -fdump-tree-optimized" } */
+
+_Bool f1(_Bool a, _Bool b)
+{
+  if (a)
+    return 1;
+  return b;
+}
+
+
+/* There should be no if statements and be fold into just return a & b. */
+/* { dg-final { scan-tree-dump-not "if" "optimized" } } */
+/* { dg-final { scan-tree-dump-times " \\\| " 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-29.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-29.c
new file mode 100644
index 00000000000..53aa01df102
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-29.c
@@ -0,0 +1,25 @@
+/* PR tree-optimization/89263 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized-raw" } */
+
+_Bool f0(_Bool a, _Bool b)
+{
+  return a ? b : 0;
+}
+_Bool f1(_Bool a, _Bool b)
+{
+  return a ? b : 1;
+}
+_Bool t0(_Bool a, _Bool b)
+{
+  return a ? 0 : b;
+}
+_Bool t1(_Bool a, _Bool b)
+{
+  return a ? 1 : b;
+}
+
+/* { dg-final { scan-tree-dump-not "gimple_cond " "optimized" } } */
+/* { dg-final { scan-tree-dump-times "bit_and_expr," 2 "optimized"  } } */
+/* { dg-final { scan-tree-dump-times "bit_ior_expr," 2 "optimized"  } } */
+/* { dg-final { scan-tree-dump-times "bit_not_expr," 2 "optimized"  } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-30.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-30.c
new file mode 100644
index 00000000000..cdfee269b3d
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-30.c
@@ -0,0 +1,55 @@
+/* PR tree-optimization/20083 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized-raw" } */
+
+
+int f(int i, int j, int l)
+{
+  int k = 0;
+  if (i)
+   k = 1;
+  if (j)
+   k = 1; 
+  if (l)
+    k = 1;
+  return k;
+}
+int f1(int i, int j, int l)
+{
+  int k = 0;
+  if (i)
+   k = 1;
+  else if (j)
+   k = 1; 
+  else if (l)
+   k = 1;
+  return k;
+}
+int f2(int i, int j, int l)
+{
+  return i||j||l;
+}
+int f3(int i, int j, int l)
+{
+  int t = i | j;
+  _Bool t1 = l != 0;
+  _Bool t2 = t ? 1 : t1;
+  return t2;
+}
+
+int f4(int i, int j, int l)
+{
+  int t = i | j;
+  _Bool t1 = l != 0;
+  _Bool t2;
+  if (t)
+   t2 = 1;
+  else
+   t2 = t1;
+  return t2;
+}
+
+/* { dg-final { scan-tree-dump-not "gimple_cond " "optimized" } } */
+/* 2 ior for each function. */
+/* { dg-final { scan-tree-dump-times "bit_ior_expr," 10 "optimized"  } } */
+/* { dg-final { scan-tree-dump-times "ne_expr," 5 "optimized"  } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-31.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-31.c
new file mode 100644
index 00000000000..e312cdbdf52
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-31.c
@@ -0,0 +1,15 @@
+/* PR tree-optimization/94898 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized-raw" } */
+
+
+_Bool f(int x, int y)
+{
+    if (x >= y)
+        return x - y;
+    return 0;
+}
+
+/* { dg-final { scan-tree-dump-not "gimple_cond " "optimized" } } */
+/* { dg-final { scan-tree-dump-times "gt_expr," 1 "optimized"  } } */
+/* { dg-final { scan-tree-dump-not "ne_expr," "optimized"  } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-32.c b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-32.c
new file mode 100644
index 00000000000..f51e79fd066
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/phi-opt-32.c
@@ -0,0 +1,12 @@
+/* PR tree-optimization/99069 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized-raw" } */
+
+_Bool f(_Bool x, _Bool y)
+{
+    return (x ? y : 0) ? x : 0;
+}
+
+
+/* { dg-final { scan-tree-dump-not "gimple_cond " "optimized" } } */
+/* { dg-final { scan-tree-dump-times "bit_and_expr," 1 "optimized"  } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-21.c b/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-21.c
index 16537ccfb61..1123b3ebbdf 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-21.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-21.c
@@ -1,5 +1,5 @@
 /* { dg-do compile } */
-/* { dg-options "-O2 -fdump-tree-thread2-stats -fdump-tree-optimized" } */
+/* { dg-options "-O2 -fdump-tree-thread2-stats -fdump-tree-optimized -fno-ssa-phiopt" } */
 
 long a;
 int b;
@@ -19,6 +19,7 @@ int main() {
   a = b;
 }
 
+/* Disable phiopt as the threading opportunity  goes away. */
 /* We need to perform a non-multi-way branch FSM thread creating an
    irreducible loop in thread2 to allow followup threading to
    remove the call to foo ().  */
-- 
2.31.1

Re: [PATCH 1/2] Match: zero_one_valued_p should match 0 constants too

[Jeff Law]

On 6/6/23 18:17, Andrew Pinski via Gcc-patches wrote:
> While working on `bool0 ? bool1 : bool2` I noticed that
> zero_one_valued_p does not match on the constant zero
> as in that case tree_nonzero_bits will return 0 and
> that is different from 1.
> 
> OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
> 
> gcc/ChangeLog:
> 
> 	* match.pd (zero_one_valued_p): Match 0 integer constant
> 	too.
Presumably "1" is matched by the tree_nonzero_bits (@0) == 1.  So OK.

jeff

Re: [PATCH 2/2] Add match patterns for `a ? onezero : onezero` where one of the two operands are constant

[Jeff Law]

On 6/6/23 18:17, Andrew Pinski via Gcc-patches wrote:
> This adds a match pattern that are for boolean values
> that optimizes `a ? onezero : 0` to `a & onezero` and
> `a ? 1 : onezero` to `a | onezero`.
> 
> This was reported a few times and I thought I would finally
> add the match pattern for this.
> 
> This hits a few times in GCC itself too.
> 
> Notes on the testcases:
> * phi-opt-2.c: This now is optimized to `a & b` in phiopt rather than ifcombine
> * phi-opt-25b.c: The test part that was failing was parity which now gets `x & y` treatment.
> * ssa-thread-21.c: there is no longer a threading opportunity, so need to disable phiopt.
>    Note PR 109957 is filed for the now missing optimization in that testcase too.
> 
> gcc/ChangeLog:
> 
> 	PR tree-optimization/89263
> 	PR tree-optimization/99069
> 	PR tree-optimization/20083
> 	PR tree-optimization/94898
> 	* match.pd: Add patterns to optimize `a ? onezero : onezero` with
> 	one of the operands are constant.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* gcc.dg/tree-ssa/phi-opt-2.c: Adjust the testcase.
> 	* gcc.dg/tree-ssa/phi-opt-25b.c: Adjust the testcase.
> 	* gcc.dg/tree-ssa/ssa-thread-21.c: Disable phiopt.
> 	* gcc.dg/tree-ssa/phi-opt-27.c: New test.
> 	* gcc.dg/tree-ssa/phi-opt-28.c: New test.
> 	* gcc.dg/tree-ssa/phi-opt-29.c: New test.
> 	* gcc.dg/tree-ssa/phi-opt-30.c: New test.
> 	* gcc.dg/tree-ssa/phi-opt-31.c: New test.
> 	* gcc.dg/tree-ssa/phi-opt-32.c: New test.
OK.

tbird rendered the bool0^1 as bool0 to the first power, so I was briefly 
confused before I realized what was going on.

jeff

Re: [PATCH 1/2] Match: zero_one_valued_p should match 0 constants too

[Jakub Jelinek]

On Tue, Jun 06, 2023 at 05:17:05PM -0700, Andrew Pinski via Gcc-patches wrote:
> While working on `bool0 ? bool1 : bool2` I noticed that
> zero_one_valued_p does not match on the constant zero
> as in that case tree_nonzero_bits will return 0 and
> that is different from 1.
> 
> OK? Bootstrapped and tested on x86_64-linux-gnu with no regressions.
> 
> gcc/ChangeLog:
> 
> 	* match.pd (zero_one_valued_p): Match 0 integer constant
> 	too.
> ---
>  gcc/match.pd | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/gcc/match.pd b/gcc/match.pd
> index f9cbd757752..f97ff7ef760 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -1983,11 +1983,16 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
>    (bit_not (bit_not @0))
>    @0)
>  
> +/* zero_one_valued_p will match when a value is known to be either
> +   0 or 1 including the constant 0. */
>  (match zero_one_valued_p
>   @0
>   (if (INTEGRAL_TYPE_P (type) && tree_nonzero_bits (@0) == 1)))

So perhaps instead change this to
&& wi::leu_p (tree_nonzero_bits (@0), 1)
?

>  (match zero_one_valued_p
>   truth_valued_p@0)
> +(match zero_one_valued_p
> + integer_zerop@0
> + (if (INTEGRAL_TYPE_P (type))))
>  
>  /* Transform { 0 or 1 } * { 0 or 1 } into { 0 or 1 } & { 0 or 1 }.  */
>  (simplify

	Jakub

Re: [PATCH 1/2] Match: zero_one_valued_p should match 0 constants too

[Jeff Law]

On 6/7/23 01:12, Jakub Jelinek via Gcc-patches wrote:

>>   
>> +/* zero_one_valued_p will match when a value is known to be either
>> +   0 or 1 including the constant 0. */
>>   (match zero_one_valued_p
>>    @0
>>    (if (INTEGRAL_TYPE_P (type) && tree_nonzero_bits (@0) == 1)))
> 
> So perhaps instead change this to
> && wi::leu_p (tree_nonzero_bits (@0), 1)
I guess that would cover both cases without the extra conditional.  I'm 
fine with that approach too.  Consider it pre-approved if someone wants 
to make that change.

jeff