The instructions LSR(AND(x, M), S) are already simplified into
AND(LSR(x, S), (M >> 12)) but only if AND(x, M) has a single user.
However, if (M >> S) == (-1 >> S), the AND part is redundant and the
whole can always directly be simplified into LSR(x, S).
For example, code like:
unsigned foo(unsigned x)
{
unsigned t = (x & 0xfffff000);
return ((t >> 12) ^ (x >> 12)) & t;
}
is now optimized into:
foo:
ret.32 $0
because (t >> 12) is simplified into (x >> 12).
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@xxxxxxxxx>
---
simplify.c | 2 ++
validation/optim/lsr-and1.c | 1 -
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/simplify.c b/simplify.c
index fe264417e..8ace8a5da 100644
--- a/simplify.c
+++ b/simplify.c
@@ -654,6 +654,8 @@ static int simplify_shift(struct instruction *insn, pseudo_t pseudo, long long v
nmask = omask & mask;
if (nmask == 0)
return replace_with_pseudo(insn, value_pseudo(0));
+ if (nmask == mask)
+ return replace_pseudo(insn, &insn->src1, def->src1);
if (nbr_users(pseudo) > 1)
break;
def->opcode = OP_LSR;
diff --git a/validation/optim/lsr-and1.c b/validation/optim/lsr-and1.c
index 906d9946f..ad249848b 100644
--- a/validation/optim/lsr-and1.c
+++ b/validation/optim/lsr-and1.c
@@ -11,7 +11,6 @@ unsigned foo(unsigned x)
/*
* check-name: lsr-and1
* check-command: test-linearize -Wno-decl $file
- * check-known-to-fail
*
* check-output-ignore
* check-output-contains: ret\\..*\\$0$
--
2.18.0
