diff options
| author | Jakub Jelinek <jakub@redhat.com> | 2020-04-07 14:39:24 +0200 |
|---|---|---|
| committer | Jakub Jelinek <jakub@redhat.com> | 2020-04-07 14:39:24 +0200 |
| commit | d51af82b4cf4c95c4a7451df2180cc6ebb44856b (patch) | |
| tree | 83cf14d47f41bf298b86f5e1d8960ede45a9ee9d /gcc/config/i386/i386-expand.c | |
| parent | 4df50a059fbd4d4a1cb067bd43caccdfca0327a8 (diff) | |
i386: Fix V{64QI,32HI}mode constant permutations [PR94509]
The following testcases are miscompiled, because expand_vec_perm_pshufb
incorrectly thinks it can use vpshufb instruction for the permutations
when it can't.
The
if (vmode == V32QImode)
{
/* vpshufb only works intra lanes, it is not
possible to shuffle bytes in between the lanes. */
for (i = 0; i < nelt; ++i)
if ((d->perm[i] ^ i) & (nelt / 2))
return false;
}
intra-lane check which is correct has been copied and adjusted for 64-byte
modes into:
if (vmode == V64QImode)
{
/* vpshufb only works intra lanes, it is not
possible to shuffle bytes in between the lanes. */
for (i = 0; i < nelt; ++i)
if ((d->perm[i] ^ i) & (nelt / 4))
return false;
}
which is not correct, because 64-byte modes have 4 lanes rather than just
two and the above is only testing that the permutation grabs even lane elts
from even lanes and odd lane elts from odd lanes, but not that they are
from the same 256-bit half.
The following patch fixes it by using 3 * nelt / 4 instead of nelt / 4,
so we actually check the most significant 2 bits rather than just one.
2020-04-07 Jakub Jelinek <jakub@redhat.com>
PR target/94509
* config/i386/i386-expand.c (expand_vec_perm_pshufb): Fix the check
for inter-lane permutation for 64-byte modes.
* gcc.target/i386/avx512bw-pr94509-1.c: New test.
* gcc.target/i386/avx512bw-pr94509-2.c: New test.
Diffstat (limited to 'gcc/config/i386/i386-expand.c')
| -rw-r--r-- | gcc/config/i386/i386-expand.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c index 066de99e660..48f00c5fcfc 100644 --- a/gcc/config/i386/i386-expand.c +++ b/gcc/config/i386/i386-expand.c @@ -16781,7 +16781,7 @@ expand_vec_perm_pshufb (struct expand_vec_perm_d *d) /* vpshufb only works intra lanes, it is not possible to shuffle bytes in between the lanes. */ for (i = 0; i < nelt; ++i) - if ((d->perm[i] ^ i) & (nelt / 4)) + if ((d->perm[i] ^ i) & (3 * nelt / 4)) return false; } } |