Hi!
The following patch adds some clz simplifications. If
clz is 0, then the MSB of the argument is set, and if clz is prec-1, then
the argument is 1.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2020-12-31 Jakub Jelinek <ja...@redhat.com>
PR tree-optimization/94802
* match.pd (clz(X) == 0 -> (int)X < 0): New simplification.
(clz(X) == (prec-1) -> X == 1): Likewise.
* gcc.dg/tree-ssa/pr94802-1.c: New test.
--- gcc/match.pd.jj 2020-12-30 12:00:00.924301784 +0100
+++ gcc/match.pd 2020-12-30 17:04:27.340815135 +0100
@@ -6226,6 +6226,39 @@ (define_operator_list COND_TERNARY
(if (single_use (@3))
(IFN_FMA @0 @1 @2))))
+/* CLZ simplifications. */
+(for clz (CLZ)
+ (for op (eq ne)
+ cmp (lt ge)
+ (simplify
+ (op (clz:s @0) INTEGER_CST@1)
+ (if (integer_zerop (@1))
+ /* clz(X) == 0 is (int)X < 0 and clz(X) != 0 is (int)X >= 0. */
+ (with { tree stype = signed_type_for (TREE_TYPE (@0));
+ HOST_WIDE_INT val = 0;
+#ifdef CLZ_DEFINED_VALUE_AT_ZERO
+ /* Punt on hypothetical weird targets. */
+ if (CLZ_DEFINED_VALUE_AT_ZERO (TYPE_MODE (TREE_TYPE (@0)),
+ val) == 2
+ && val == 0)
+ stype = NULL_TREE;
+#endif
+ }
+ (if (stype)
+ (cmp (convert:stype @0) { build_zero_cst (stype); })))
+ /* clz(X) == (prec-1) is X == 1 and clz(X) != (prec-1) is X != 1. */
+ (with { bool ok = true;
+#ifdef CLZ_DEFINED_VALUE_AT_ZERO
+ /* Punt on hypothetical weird targets. */
+ if (CLZ_DEFINED_VALUE_AT_ZERO (TYPE_MODE (TREE_TYPE (@0)),
+ val) == 2
+ && val == TYPE_PRECISION (TREE_TYPE (@0)) - 1)
+ ok = false;
+#endif
+ }
+ (if (ok && wi::to_wide (@1) == (TYPE_PRECISION (TREE_TYPE (@0)) - 1))
+ (op @0 { build_one_cst (TREE_TYPE (@0)); })))))))
+
/* POPCOUNT simplifications. */
/* popcount(X) + popcount(Y) is popcount(X|Y) when X&Y must be zero. */
(simplify
--- gcc/testsuite/gcc.dg/tree-ssa/pr94802-1.c.jj 2020-12-30
17:25:40.173364488 +0100
+++ gcc/testsuite/gcc.dg/tree-ssa/pr94802-1.c 2020-12-30 17:26:06.687063794
+0100
@@ -0,0 +1,68 @@
+/* PR tree-optimization/94802 */
+/* { dg-do run } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+/* { dg-final { scan-tree-dump-not " = __builtin_clz " "optimized" } } */
+
+__attribute__((noipa)) int
+f1 (int a, int b)
+{
+ return __builtin_clz (a - b) != 0;
+}
+
+__attribute__((noipa)) int
+f2 (int x)
+{
+ return __builtin_clz (x) == 0;
+}
+
+__attribute__((noipa)) int
+f3 (int x)
+{
+ return __builtin_clz (x) != 0;
+}
+
+__attribute__((noipa)) int
+f4 (int a, int b)
+{
+ return __builtin_clz (a - b) == sizeof (int) * __CHAR_BIT__ - 1;
+}
+
+__attribute__((noipa)) int
+f5 (int x)
+{
+ return __builtin_clz (x) == sizeof (int) * __CHAR_BIT__ - 1;
+}
+
+__attribute__((noipa)) int
+f6 (int x)
+{
+ return __builtin_clz (x) != sizeof (int) * __CHAR_BIT__ - 1;
+}
+
+int
+main ()
+{
+ if (f1 (5, 7) != 0
+ || f1 (7, 5) != 1
+ || f2 (1) != 0
+ || f2 (137) != 0
+ || f2 (-1) != 1
+ || f2 (-137) != 1
+ || f3 (1) != 1
+ || f3 (137) != 1
+ || f3 (-1) != 0
+ || f3 (-137) != 0
+ || f4 (5, 4) != 1
+ || f4 (6, 4) != 0
+ || f4 (4, 5) != 0
+ || f5 (1) != 1
+ || f5 (17) != 0
+ || f5 (-1) != 0
+ || f5 (-17) != 0
+ || f6 (1) != 0
+ || f6 (17) != 1
+ || f6 (-1) != 1
+ || f6 (-17) != 1)
+ __builtin_abort ();
+ return 0;
+}
Jakub
