This is done by RTL. Do it earlier in Gimple to potentially allow more
optimizations opportunities.
Bootstrapped and regression tested in x86_64.
PR 126644
gcc/ChangeLog:
* match.pd:
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/pr126644.c: New test.
---
gcc/match.pd | 19 +++++++++++++++++++
gcc/testsuite/gcc.dg/tree-ssa/pr126644.c | 18 ++++++++++++++++++
2 files changed, 37 insertions(+)
create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/pr126644.c
diff --git a/gcc/match.pd b/gcc/match.pd
index 2d1161b860b..06bbe32c029 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -1561,6 +1561,25 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
{ wide_int_to_tree (type,
wi::lshift (wone, wi::to_wide (@2))); }))))
+#if GIMPLE
+/* PR126644: A << CST >> CST -> A
+ As long as the upper CST bits of A are known to be zero. */
+(simplify
+ (rshift (lshift@2 @0 INTEGER_CST@1) @1)
+ (if (INTEGRAL_TYPE_P (type)
+ && INTEGRAL_TYPE_P (TREE_TYPE (@0))
+ && tree_fits_uhwi_p (@1)
+ && tree_to_uhwi (@1) < TYPE_PRECISION (type)
+ && single_use (@2))
+ (with {
+ unsigned int prec = TYPE_PRECISION (type);
+ wide_int mask = wi::mask (prec - tree_to_uhwi (@1),
+ true, prec);
+ }
+ (if ((tree_nonzero_bits (@0) & mask) == 0)
+ @0))))
+#endif
+
/* Fold (-x >> C) into -(x > 0) where C = precision(type) - 1. */
(for cst (INTEGER_CST VECTOR_CST)
(simplify
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr126644.c
b/gcc/testsuite/gcc.dg/tree-ssa/pr126644.c
new file mode 100644
index 00000000000..73286cd5a1e
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr126644.c
@@ -0,0 +1,18 @@
+/* { dg-do compile } */
+/* { dg-options "-O1 -fdump-tree-optimized" } */
+
+int f1 (int a)
+{
+ if ((a & 0xE0000000) != 0) __builtin_unreachable ();
+
+ return a << 3 >> 3;
+}
+
+/* Can't be optimized. */
+int f2 (int a, int b)
+{
+ return a << 3 >> 3;
+}
+
+/* { dg-final { scan-tree-dump-times " << " 1 "optimized" } } */
+/* { dg-final { scan-tree-dump-times " >> " 1 "optimized" } } */
--
2.43.0
