On Sun, Dec 17, 2017 at 03:14:08AM +0000, Segher Boessenkool wrote:
> Hi!
>
> On Mon, Dec 11, 2017 at 02:18:53PM +0000, James Greenhalgh wrote:
> >
> > In simplify_set we try transforming the paradoxical subreg expression:
> >
> > (set FOO (subreg:M (mem:N BAR) 0))
> >
> > in to:
> >
> > (set FOO (zero_extend:M (mem:N BAR)))
> >
> > However, this code does not consider the case where M is a vector
> > mode, allowing it to construct (for example):
> >
> > (zero_extend:V4SI (mem:SI))
> >
> > This would clearly have the wrong semantics, but fortunately we fail long
> > before then in expand_compound_operation. As we really don't want a vector
> > zero_extend of a scalar value.
> >
> > We need to explicitly reject vector modes from this transformation.
>
> It does not consider any other modes either. Both modes involved are
> required to be SCALAR_INT_MODE_P (for zero_extend to be valid at all and
> to be equivalent to the subreg); could you test for that instead please?
Makes sense. I've committed this as obvious after an AArch64 bootstrap and
test as revision 255945.
Thanks,
James
---
gcc/
2017-12-21 James Greenhalgh <james.greenha...@arm.com>
* combine.c (simplify_set): Do not transform subregs to zero_extends
if the destination is not a scalar int mode.
diff --git a/gcc/combine.c b/gcc/combine.c
index 9f19ee4..91bed06 100644
--- a/gcc/combine.c
+++ b/gcc/combine.c
@@ -6967,12 +6967,13 @@ simplify_set (rtx x)
/* If we have (set FOO (subreg:M (mem:N BAR) 0)) with M wider than N, this
would require a paradoxical subreg. Replace the subreg with a
zero_extend to avoid the reload that would otherwise be required.
- Don't do this for vector modes, as the transformation is incorrect. */
+ Don't do this unless we have a scalar integer mode, otherwise the
+ transformation is incorrect. */
enum rtx_code extend_op;
if (paradoxical_subreg_p (src)
&& MEM_P (SUBREG_REG (src))
- && !VECTOR_MODE_P (GET_MODE (src))
+ && SCALAR_INT_MODE_P (GET_MODE (src))
&& (extend_op = load_extend_op (GET_MODE (SUBREG_REG (src)))) != UNKNOWN)
{
SUBST (SET_SRC (x),
