bihash key compare improvements
Looks like CPU doesn't like overlaping loads.
This new codes in some cases shows 3-4 clock improvements.
Change-Id: Ia1b49976ad95140c573f892fdc0a32eebbfa06c8
Signed-off-by: Damjan Marion <damarion@cisco.com>
diff --git a/src/vppinfra/bihash_24_8.h b/src/vppinfra/bihash_24_8.h
index 0c57e35..dcce2a2 100644
--- a/src/vppinfra/bihash_24_8.h
+++ b/src/vppinfra/bihash_24_8.h
@@ -72,9 +72,8 @@
u64x8 v = u64x8_load_unaligned (a) ^ u64x8_load_unaligned (b);
return (u64x8_is_zero_mask (v) & 0x7) == 0;
#elif defined(CLIB_HAVE_VEC128) && defined(CLIB_HAVE_VEC128_UNALIGNED_LOAD_STORE)
- u64x2 v;
- v = u64x2_load_unaligned (a) ^ u64x2_load_unaligned (b);
- v |= u64x2_load_unaligned (a + 1) ^ u64x2_load_unaligned (b + 1);
+ u64x2 v = { a[2] ^ b[2], 0 };
+ v |= u64x2_load_unaligned (a) ^ u64x2_load_unaligned (b);
return u64x2_is_all_zero (v);
#else
return ((a[0] ^ b[0]) | (a[1] ^ b[1]) | (a[2] ^ b[2])) == 0;