[PATCH 2/8 v9]middle-end slp: fix is_linear_load_p to prevent multiple answers
Tamar Christina
tamar.christina@arm.com
Mon Dec 28 13:36:19 GMT 2020
Hi All,
This fixes an issue where is_linear_load_p could return the incorrect
permutation kind because it is singe pass.
This arranges the candidates in such a way that there won't be any ambiguity so
that the function can still be linear but give correct values.
Bootstrapped Regtested on aarch64-none-linux-gnu, x86_64-pc-linux-gnu
and no issues.
Ok for master?
Thanks,
Tamar
gcc/ChangeLog:
* tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
--- inline copy of patch --
diff --git a/gcc/tree-vect-slp-patterns.c b/gcc/tree-vect-slp-patterns.c
index fede88923af8521ee4954c8ae27b0e589f975610..7fd79d91c6ba4ccdbf361307a6105fb7e46aa961 100644
--- a/gcc/tree-vect-slp-patterns.c
+++ b/gcc/tree-vect-slp-patterns.c
@@ -140,32 +140,32 @@ is_linear_load_p (load_permutation_t loads)
unsigned load, i;
complex_perm_kinds_t candidates[4]
- = { PERM_EVENODD
- , PERM_ODDEVEN
- , PERM_ODDODD
+ = { PERM_ODDODD
, PERM_EVENEVEN
+ , PERM_EVENODD
+ , PERM_ODDEVEN
};
int valid_patterns = 4;
- FOR_EACH_VEC_ELT_FROM (loads, i, load, 1)
+ FOR_EACH_VEC_ELT (loads, i, load)
{
- if (candidates[0] != PERM_UNKNOWN && load != i)
+ if (candidates[0] != PERM_UNKNOWN && load != 1)
{
candidates[0] = PERM_UNKNOWN;
valid_patterns--;
}
- if (candidates[1] != PERM_UNKNOWN
- && load != (i % 2 == 0 ? i + 1 : i - 1))
+ if (candidates[1] != PERM_UNKNOWN && load != 0)
{
candidates[1] = PERM_UNKNOWN;
valid_patterns--;
}
- if (candidates[2] != PERM_UNKNOWN && load != 1)
+ if (candidates[2] != PERM_UNKNOWN && load != i)
{
candidates[2] = PERM_UNKNOWN;
valid_patterns--;
}
- if (candidates[3] != PERM_UNKNOWN && load != 0)
+ if (candidates[3] != PERM_UNKNOWN
+ && load != (i % 2 == 0 ? i + 1 : i - 1))
{
candidates[3] = PERM_UNKNOWN;
valid_patterns--;
--
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rb13957.patch
Type: text/x-diff
Size: 1402 bytes
Desc: not available
URL: <https://gcc.gnu.org/pipermail/gcc-patches/attachments/20201228/74df6d4b/attachment-0001.bin>
More information about the Gcc-patches
mailing list