While looking into bitwise optimizations, I noticed that
get_known_nonzero_bits_1 does `bm.value () & ~bm.mask ()` which
is ok except it creates a temporary wide_int. Instead if we
use wi::bit_and_not, we can avoid the temporary and on some
targets use the andn/bic instruction.
Bootstrapped and tested on x86_64-linux-gnu.
gcc/ChangeLog:
* tree-ssanames.cc (get_known_nonzero_bits_1): Use
wi::bit_and_not instead of `a & ~b`.
Signed-off-by: Andrew Pinski <quic_apin...@quicinc.com>
---
gcc/tree-ssanames.cc | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gcc/tree-ssanames.cc b/gcc/tree-ssanames.cc
index d7865f29f0b..de7b9b79f94 100644
--- a/gcc/tree-ssanames.cc
+++ b/gcc/tree-ssanames.cc
@@ -576,7 +576,7 @@ get_known_nonzero_bits_1 (const_tree name)
if (tmp.undefined_p ())
return wi::shwi (0, precision);
irange_bitmask bm = tmp.get_bitmask ();
- return bm.value () & ~bm.mask ();
+ return wi::bit_and_not (bm.value (), bm.mask ());
}
/* Return a wide_int with known non-zero bits in SSA_NAME
--
2.43.0
