Fix folding of vector mask EQ/NE expressions
authorRichard Sandiford <richard.sandiford@linaro.org>
Sat, 13 Jan 2018 17:57:17 +0000 (17:57 +0000)
committerRichard Sandiford <rsandifo@gcc.gnu.org>
Sat, 13 Jan 2018 17:57:17 +0000 (17:57 +0000)
commit779fed5fdb6098e67213a82dfd27f5b326a75e88
treec7d81773fba7399e72265ef818081c1fb946baa9
parentdbc3af4fc620aeb6fbf088e961fe8658bbd959c7
Fix folding of vector mask EQ/NE expressions

fold_binary_loc assumed that if the type of the result wasn't a vector,
the operands wouldn't be either.  This isn't necessarily true for
EQ_EXPR and NE_EXPR of vector masks, which can return a single scalar
for the mask as a whole.

2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
    Alan Hayward  <alan.hayward@arm.com>
    David Sherwood  <david.sherwood@arm.com>

gcc/
* fold-const.c (fold_binary_loc): Check the argument types
rather than the result type when testing for a vector operation.

gcc/testsuite/
* gcc.target/aarch64/sve/vec_bool_cmp_1.c: New test.
* gcc.target/aarch64/sve/vec_bool_cmp_1_run.c: Likweise.

Co-Authored-By: Alan Hayward <alan.hayward@arm.com>
Co-Authored-By: David Sherwood <david.sherwood@arm.com>
From-SVN: r256616
gcc/ChangeLog
gcc/fold-const.c
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.target/aarch64/sve/vec_bool_cmp_1.c [new file with mode: 0644]
gcc/testsuite/gcc.target/aarch64/sve/vec_bool_cmp_1_run.c [new file with mode: 0644]