middle-end: check that both sides of complex expression is a mul.

Message ID patch-15095-tamar@arm.com
State Committed
Commit 4f0a2f5a3ddb1024b885c066a18caae4d733bb6c
Headers
Series middle-end: check that both sides of complex expression is a mul. |

Commit Message

Tamar Christina Nov. 18, 2021, 4:57 p.m. UTC
  Hi All,

Both sides of the VEC_PERM_EXPR need to be a MULT but the check
was accidentally checking if both sides are a mul.

The FMS case would be handled by the validate_multiplication but
this makes the requirement more explicit and we exit earlier.

Bootstrapped Regtested on aarch64-none-linux-gnu and no issues.

Ok for master?

Thanks,
Tamar


gcc/ChangeLog:

	PR tree-optimizaion/103311
	* tree-vect-slp-patterns.c (complex_mul_pattern::matches,
	complex_fms_pattern::matches): Check for multiplications.

gcc/testsuite/ChangeLog:

	PR tree-optimizaion/103311
	* gcc.target/aarch64/pr103311.c: New test.

--- inline copy of patch -- 
diff --git a/gcc/testsuite/gcc.target/aarch64/pr103311.c b/gcc/testsuite/gcc.target/aarch64/pr103311.c
new file mode 100644
index 0000000000000000000000000000000000000000..7bcc7db79be8f1b4dc1f95c6cf10f5292191a892


--
  

Comments

Richard Biener Nov. 18, 2021, 5:07 p.m. UTC | #1
On November 18, 2021 5:57:45 PM GMT+01:00, Tamar Christina <tamar.christina@arm.com> wrote:
>Hi All,
>
>Both sides of the VEC_PERM_EXPR need to be a MULT but the check
>was accidentally checking if both sides are a mul.
>
>The FMS case would be handled by the validate_multiplication but
>this makes the requirement more explicit and we exit earlier.
>
>Bootstrapped Regtested on aarch64-none-linux-gnu and no issues.
>
>Ok for master?

Ok. 

Richard. 

>Thanks,
>Tamar
>
>
>gcc/ChangeLog:
>
>	PR tree-optimizaion/103311
>	* tree-vect-slp-patterns.c (complex_mul_pattern::matches,
>	complex_fms_pattern::matches): Check for multiplications.
>
>gcc/testsuite/ChangeLog:
>
>	PR tree-optimizaion/103311
>	* gcc.target/aarch64/pr103311.c: New test.
>
>--- inline copy of patch -- 
>diff --git a/gcc/testsuite/gcc.target/aarch64/pr103311.c b/gcc/testsuite/gcc.target/aarch64/pr103311.c
>new file mode 100644
>index 0000000000000000000000000000000000000000..7bcc7db79be8f1b4dc1f95c6cf10f5292191a892
>--- /dev/null
>+++ b/gcc/testsuite/gcc.target/aarch64/pr103311.c
>@@ -0,0 +1,16 @@
>+/* { dg-do compile } */
>+/* { dg-require-effective-target arm_v8_3a_complex_neon_ok } */
>+/* { dg-add-options arm_v8_3a_complex_neon } */
>+/* { dg-additional-options "-std=c99" } */
>+
>+_Complex double y;
>+
>+void
>+cbknu (_Complex double f)
>+{
>+  const _Complex double cone = 1.0e0;
>+
>+  f = f * cone;
>+  y = f * cone;
>+}
>+
>diff --git a/gcc/tree-vect-slp-patterns.c b/gcc/tree-vect-slp-patterns.c
>index 53fbe5185f543298b27d084a93c30a1c161d3d32..d916fc9cef9a86a0d797caaf2b2f253685e9b7dd 100644
>--- a/gcc/tree-vect-slp-patterns.c
>+++ b/gcc/tree-vect-slp-patterns.c
>@@ -949,7 +949,7 @@ complex_mul_pattern::matches (complex_operation_t op,
> 
>   bool mul0 = vect_match_expression_p (l0node[0], MULT_EXPR);
>   bool mul1 = vect_match_expression_p (l0node[1], MULT_EXPR);
>-  if (!mul0 && !mul1)
>+  if (!mul0 || !mul1)
>     return IFN_LAST;
> 
>   /* Now operand2+4 may lead to another expression.  */
>@@ -1204,7 +1204,9 @@ complex_fms_pattern::matches (complex_operation_t op,
> 
>   /* If these nodes don't have any children then they're
>      not ones we're interested in.  */
>-  if (left_op.length () != 2 || right_op.length () != 2)
>+  if (left_op.length () != 2
>+      || right_op.length () != 2
>+      || !vect_match_expression_p (l0node[1], MULT_EXPR))
>     return IFN_LAST;
> 
>   bool is_neg = vect_normalize_conj_loc (left_op);
>
>
  

Patch

diff --git a/gcc/testsuite/gcc.target/aarch64/pr103311.c b/gcc/testsuite/gcc.target/aarch64/pr103311.c
new file mode 100644
index 0000000000000000000000000000000000000000..7bcc7db79be8f1b4dc1f95c6cf10f5292191a892
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/pr103311.c
@@ -0,0 +1,16 @@ 
+/* { dg-do compile } */
+/* { dg-require-effective-target arm_v8_3a_complex_neon_ok } */
+/* { dg-add-options arm_v8_3a_complex_neon } */
+/* { dg-additional-options "-std=c99" } */
+
+_Complex double y;
+
+void
+cbknu (_Complex double f)
+{
+  const _Complex double cone = 1.0e0;
+
+  f = f * cone;
+  y = f * cone;
+}
+
diff --git a/gcc/tree-vect-slp-patterns.c b/gcc/tree-vect-slp-patterns.c
index 53fbe5185f543298b27d084a93c30a1c161d3d32..d916fc9cef9a86a0d797caaf2b2f253685e9b7dd 100644
--- a/gcc/tree-vect-slp-patterns.c
+++ b/gcc/tree-vect-slp-patterns.c
@@ -949,7 +949,7 @@  complex_mul_pattern::matches (complex_operation_t op,
 
   bool mul0 = vect_match_expression_p (l0node[0], MULT_EXPR);
   bool mul1 = vect_match_expression_p (l0node[1], MULT_EXPR);
-  if (!mul0 && !mul1)
+  if (!mul0 || !mul1)
     return IFN_LAST;
 
   /* Now operand2+4 may lead to another expression.  */
@@ -1204,7 +1204,9 @@  complex_fms_pattern::matches (complex_operation_t op,
 
   /* If these nodes don't have any children then they're
      not ones we're interested in.  */
-  if (left_op.length () != 2 || right_op.length () != 2)
+  if (left_op.length () != 2
+      || right_op.length () != 2
+      || !vect_match_expression_p (l0node[1], MULT_EXPR))
     return IFN_LAST;
 
   bool is_neg = vect_normalize_conj_loc (left_op);