Fix an SVE failure in the Fortran matmul* tests
authorRichard Sandiford <richard.sandiford@linaro.org>
Mon, 18 Sep 2017 15:42:08 +0000 (15:42 +0000)
committerRichard Sandiford <rsandifo@gcc.gnu.org>
Mon, 18 Sep 2017 15:42:08 +0000 (15:42 +0000)
commitac8936b4677fa10b676e5b12aa682b9d2d42c1e5
tree59d050791a2f94e34d70ce4be586d4fc30974f7d
parentb161f2c927bb7fb70dc0c6d4e9ab22cdba29db6d
Fix an SVE failure in the Fortran matmul* tests

The vectoriser was calling vect_get_smallest_scalar_type without
having proven that the type actually is a scalar.  This seems to
be the intended behaviour: the ultimate test of whether the type
is interesting (and hence scalar) is whether an associated vector
type exists, but this is only tested later.

The patch simply makes the function cope gracefully with non-scalar
inputs.

2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
    Alan Hayward  <alan.hayward@arm.com>
    David Sherwood  <david.sherwood@arm.com>

gcc/
* tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
with types that aren't in fact scalar.

Co-Authored-By: Alan Hayward <alan.hayward@arm.com>
Co-Authored-By: David Sherwood <david.sherwood@arm.com>
From-SVN: r252934
gcc/ChangeLog
gcc/tree-vect-data-refs.c