Re: [PATCH] match.pd: build signed low-bit masks in an unsigned type

From b9104974eadf573757d290d942bd4887d667202a Mon Sep 17 00:00:00 2001
From: Kyrylo Tkachov <[email protected]>
Date: Sat, 15 Aug 2026 12:06:01 +0200
Subject: [PATCH] match.pd: build signed low-bit masks in an unsigned type

The PR71636 fold turns

  x & ((1U << b) - 1)

into

  x & ~(~0U << b)

but only when the mask type is unsigned.  Signed source and vector forms keep
the longer expression.

  int f (int x, int b)
  {
    return x & ((1 << b) - 1);
  }

aarch64 -O2 before:

  f:
          mov     w2, 1
          lsl     w2, w2, w1
          sub     w2, w2, #1
          and     w0, w2, w0
          ret

aarch64 -O2 after:

  f:
          mov     w2, -1
          lsl     w2, w2, w1
          bic     w0, w0, w2
          ret

Build a signed mask in the corresponding unsigned type and convert it back.
This makes the all-ones shift defined and exposes the shorter form.

Canonicalize subtraction of an integral vector constant into addition of its
negation when all encoded elements can be negated without overflow.  This
lets the low-bit-mask rule use only the canonical addition of minus one.

The signed mask form is not valid when the source subtraction can trap or is
instrumented for overflow.  It can also remove the signed shift-base check for
the top-bit count.  Keep these cases.  After GIMPLE lowering, an explicit
shift sanitizer check remains visible, so the fold is safe again.  Also keep
a vector subtraction when a constant element is the signed minimum or the
operation is instrumented for overflow.

The canonicalized producer remains visible before dead-code elimination in
one existing forwprop dump.  Adjust its addition count.

Bootstrapped and tested on aarch64-none-linux-gnu.
Tested on x86_64-pc-linux-gnu.

gcc/ChangeLog:

	* match.pd (x & ((1 << b) - 1)): Handle signed scalar and vector
	types.
	(negate_expr_p): Handle non-wrapping integral vector constants.

gcc/testsuite/ChangeLog:

	* gcc.dg/tree-ssa/forwprop-27.c: Adjust the expected addition count.
	* gcc.dg/tree-ssa/pr71636-signed-1.c: New test.
	* gcc.dg/tree-ssa/pr71636-signed-vector-1.c: Likewise.
	* gcc.dg/tree-ssa/pr71636-signed-trap-1.c: Likewise.
	* gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c: Likewise.
	* gcc.dg/tree-ssa/pr71636-signed-shift-ubsan-1.c: Likewise.
	* gcc.dg/tree-ssa/vector-sub-const-1.c: Likewise.
	* gcc.dg/tree-ssa/vector-sub-const-2.c: Likewise.

Signed-off-by: Kyrylo Tkachov <[email protected]>
---
 gcc/match.pd                                  | 45 +++++++++++++++----
 gcc/testsuite/gcc.dg/tree-ssa/forwprop-27.c   |  2 +-
 .../gcc.dg/tree-ssa/pr71636-signed-1.c        | 24 ++++++++++
 .../tree-ssa/pr71636-signed-shift-ubsan-1.c   | 11 +++++
 .../gcc.dg/tree-ssa/pr71636-signed-trap-1.c   | 10 +++++
 .../gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c  | 10 +++++
 .../gcc.dg/tree-ssa/pr71636-signed-vector-1.c | 24 ++++++++++
 .../gcc.dg/tree-ssa/vector-sub-const-1.c      | 26 +++++++++++
 .../gcc.dg/tree-ssa/vector-sub-const-2.c      | 12 +++++
 9 files changed, 155 insertions(+), 9 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-shift-ubsan-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-trap-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-vector-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-1.c
 create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-2.c

diff --git a/gcc/match.pd b/gcc/match.pd
index eae8717bcfe..dc51f79c58c 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -1577,11 +1577,20 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
     (convert @0)
     (convert @1)))))
 
-/* PR71636: Transform x & ((1U << b) - 1) -> x & ~(~0U << b);  */
-(simplify
-  (bit_and:c @0 (plus:s (lshift:s integer_onep @1) integer_minus_onep))
-  (if (TYPE_UNSIGNED (type))
-    (bit_and @0 (bit_not (lshift { build_all_ones_cst (type); } @1)))))
+/* PR71636: Transform x & ((1U << b) - 1) -> x & ~(~0U << b).  For signed
+   types, build the mask in the corresponding unsigned type, where shifting
+   all ones left is defined.  Preserve signed overflow and shift checks.  */
+(simplify
+ (bit_and:c @0
+  (plus:s (lshift:s integer_onep @1) integer_minus_onep))
+ (with { tree etype = VECTOR_TYPE_P (type) ? TREE_TYPE (type) : type; }
+  (if (TYPE_UNSIGNED (etype)
+       || (!TYPE_OVERFLOW_TRAPS (etype)
+	   && !TYPE_OVERFLOW_SANITIZED (etype)
+	   && (GIMPLE || !sanitize_flags_p (SANITIZE_SHIFT_BASE))))
+   (with { tree utype = unsigned_type_for (type); }
+    (bit_and @0 (convert
+		  (bit_not (lshift { build_all_ones_cst (utype); } @1))))))))
 
 /* PR112533: Canonicalize boolean comparisons of masked pow2 bits into
    xor-mask tests.
@@ -2428,11 +2437,31 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
 (match negate_expr_p
  REAL_CST
  (if (REAL_VALUE_NEGATIVE (TREE_REAL_CST (t)))))
-/* VECTOR_CST handling of non-wrapping types would recurse in unsupported
-   ways.  */
+/* A non-wrapping integral vector constant can be negated when none of its
+   encoded elements is the minimum value.  */
 (match negate_expr_p
  VECTOR_CST
- (if (FLOAT_TYPE_P (TREE_TYPE (type)) || TYPE_OVERFLOW_WRAPS (type))))
+ (with
+  {
+    tree etype = TREE_TYPE (type);
+    bool negatable = (FLOAT_TYPE_P (etype) || TYPE_OVERFLOW_WRAPS (type));
+    if (!negatable
+	&& INTEGRAL_TYPE_P (etype)
+	&& !TYPE_OVERFLOW_SANITIZED (etype))
+      {
+	negatable = true;
+	if (!TYPE_UNSIGNED (etype))
+	  for (unsigned int i = 0; i < vector_cst_encoded_nelts (t); ++i)
+	    if (TREE_CODE (VECTOR_CST_ENCODED_ELT (t, i)) != INTEGER_CST
+		|| !may_negate_without_overflow_p
+		     (VECTOR_CST_ENCODED_ELT (t, i)))
+	      {
+		negatable = false;
+		break;
+	      }
+      }
+  }
+  (if (negatable))))
 (match negate_expr_p
  (minus @0 @1)
  (if ((ANY_INTEGRAL_TYPE_P (type) && TYPE_OVERFLOW_WRAPS (type))
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/forwprop-27.c b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-27.c
index 6c71a4fc81c..29d39c67c79 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/forwprop-27.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/forwprop-27.c
@@ -33,7 +33,7 @@ void i (V *v1, V *v2){
   *v2 = (c1-*v2)+c2;
 }
 
-/* { dg-final { scan-tree-dump-times "\\\+" 1 "forwprop1"} } */
+/* { dg-final { scan-tree-dump-times "\\\+" 2 "forwprop1"} } */
 /* { dg-final { scan-tree-dump "{ 0, 4 }" "forwprop1"} } */
 /* { dg-final { scan-tree-dump "{ 37, -5 }" "forwprop1"} } */
 /* { dg-final { scan-tree-dump "{ 27, 23 }" "forwprop1"} } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-1.c
new file mode 100644
index 00000000000..9db533fdf64
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-1.c
@@ -0,0 +1,24 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+int
+f_signed (int x, int b)
+{
+  return x & ((1 << b) - 1);
+}
+
+unsigned int
+f_unsigned (unsigned int x, int b)
+{
+  return x & ((1U << b) - 1U);
+}
+
+long
+f_long (long x, int b)
+{
+  return x & ((1L << b) - 1L);
+}
+
+/* { dg-final { scan-tree-dump-not "1 <<" "optimized" } } */
+/* { dg-final { scan-tree-dump-not " \\+ -1;" "optimized" } } */
+/* { dg-final { scan-tree-dump-times "= ~" 3 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-shift-ubsan-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-shift-ubsan-1.c
new file mode 100644
index 00000000000..3abee74264c
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-shift-ubsan-1.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fsanitize=shift-base -fdump-tree-optimized" } */
+
+int
+f (int x, int b)
+{
+  return x & ((1 << b) - 1);
+}
+
+/* { dg-final { scan-tree-dump-times "__builtin___ubsan_handle_shift_out_of_bounds" 1 "optimized" } } */
+/* { dg-final { scan-tree-dump-times "= ~" 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-trap-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-trap-1.c
new file mode 100644
index 00000000000..13a1a5b4cf2
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-trap-1.c
@@ -0,0 +1,10 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftrapv -fdump-tree-optimized" } */
+
+int
+f (int x, int b)
+{
+  return x & ((1 << b) - 1);
+}
+
+/* { dg-final { scan-tree-dump-times " \\+ -1;" 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c
new file mode 100644
index 00000000000..68c1b92f18b
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-ubsan-1.c
@@ -0,0 +1,10 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fsanitize=signed-integer-overflow -fdump-tree-optimized" } */
+
+int
+f (int x, int b)
+{
+  return x & ((1 << b) - 1);
+}
+
+/* { dg-final { scan-tree-dump-times "\\.UBSAN_CHECK_SUB" 1 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-vector-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-vector-1.c
new file mode 100644
index 00000000000..51ba0b35334
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr71636-signed-vector-1.c
@@ -0,0 +1,24 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-require-effective-target vect_int } */
+/* { dg-require-effective-target vect_var_shift } */
+
+typedef int v4si __attribute__ ((vector_size (16)));
+typedef unsigned int v4ui __attribute__ ((vector_size (16)));
+
+v4si
+f_signed (v4si x, v4si b)
+{
+  v4si one = { 1, 1, 1, 1 };
+  return x & ((one << b) - one);
+}
+
+v4ui
+f_unsigned (v4ui x, v4ui b)
+{
+  v4ui one = { 1, 1, 1, 1 };
+  return x & ((one << b) - one);
+}
+
+/* { dg-final { scan-tree-dump-not "\\{ 1, 1, 1, 1 \\} <<" "optimized" } } */
+/* { dg-final { scan-tree-dump-times "= ~" 2 "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-1.c b/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-1.c
new file mode 100644
index 00000000000..229ca60dde4
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-1.c
@@ -0,0 +1,26 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -Wno-psabi -fdump-tree-forwprop1" } */
+
+typedef __INT32_TYPE__ v4si __attribute__ ((vector_size (16)));
+
+v4si
+sub_splat (v4si x)
+{
+  return x - (v4si) { 1, 1, 1, 1 };
+}
+
+v4si
+sub_mixed (v4si x)
+{
+  return x - (v4si) { 1, -2, 3, -4 };
+}
+
+v4si
+keep_min (v4si x)
+{
+  return x - (v4si) { -__INT32_MAX__ - 1, 1, 1, 1 };
+}
+
+/* { dg-final { scan-tree-dump-times " \\+ \\{ -1, -1, -1, -1 \\}" 1 "forwprop1" } } */
+/* { dg-final { scan-tree-dump-times " \\+ \\{ -1, 2, -3, 4 \\}" 1 "forwprop1" } } */
+/* { dg-final { scan-tree-dump-times " - \\{ -2147483648, 1, 1, 1 \\}" 1 "forwprop1" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-2.c b/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-2.c
new file mode 100644
index 00000000000..27de9e77be2
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/vector-sub-const-2.c
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -Wno-psabi -fsanitize=signed-integer-overflow -fdump-tree-forwprop1" } */
+
+typedef __INT32_TYPE__ v4si __attribute__ ((vector_size (16)));
+
+v4si
+f (v4si x)
+{
+  return x - (v4si) { 1, 1, 1, 1 };
+}
+
+/* { dg-final { scan-tree-dump-times "\\.UBSAN_CHECK_SUB" 1 "forwprop1" } } */
-- 
2.50.1 (Apple Git-155)