diff options
| author |   Damjan Marion <damarion@cisco.com> | 2018-05-19 00:04:23 +0200 |
|---|---|---|
| committer | Damjan Marion <damarion@cisco.com> | 2018-05-20 14:52:39 +0200 |
| commit | a52e1668c9976bd5cdd20d02b668df41ea41f16f (patch) | |
| tree | 75ed44c2df7a88e0e3af39ae1c4de4cc49f36ad7 /src/vppinfra/vector_avx512.h | |
| parent | 3b854a5cb876b5af4e69c56028e2c2824100f4b2 (diff) | |
vector functions cleanup and improvements
Remove functions which have native C equivalent (i.e. _is_equal can be
replaced with ==, _add with +)
Add SSE4.2, AVX-512 implementations of splat, load_unaligned, store_unaligned,
is_all_zero, is_equal, is_all_equal
Change-Id: Ie80b0e482e7a76248ad79399c2576468532354cd
Signed-off-by: Damjan Marion <damarion@cisco.com>
Diffstat (limited to 'src/vppinfra/vector_avx512.h')
| -rw-r--r-- | src/vppinfra/vector_avx512.h | 16 |
1 files changed, 14 insertions, 2 deletions
diff --git a/src/vppinfra/vector_avx512.h b/src/vppinfra/vector_avx512.h index c1b7c42a260..c2903e2aa1a 100644 --- a/src/vppinfra/vector_avx512.h +++ b/src/vppinfra/vector_avx512.h @@ -19,6 +19,7 @@ #include <vppinfra/clib.h> #include <x86intrin.h> +/* *INDENT-OFF* */ #define foreach_avx512_vec512i \ _(i,8,64,epi8) _(i,16,32,epi16) _(i,32,16,epi32) _(i,64,8,epi64) #define foreach_avx512_vec512u \ @@ -26,8 +27,8 @@ #define foreach_avx512_vec512f \ _(f,32,8,ps) _(f,64,4,pd) -/* splat, load_unaligned, store_unaligned */ -/* *INDENT-OFF* */ +/* splat, load_unaligned, store_unaligned, is_all_zero, is_equal, + is_all_equal */ #define _(t, s, c, i) \ static_always_inline t##s##x##c \ t##s##x##c##_splat (t##s x) \ @@ -41,6 +42,17 @@ static_always_inline void \ t##s##x##c##_store_unaligned (t##s##x##c v, void *p) \ { _mm512_storeu_si512 ((__m512i *) p, (__m512i) v); } \ \ +static_always_inline int \ +t##s##x##c##_is_all_zero (t##s##x##c v) \ +{ return (_mm512_test_epi64_mask ((__m512i) v, (__m512i) v) == 0); } \ +\ +static_always_inline int \ +t##s##x##c##_is_equal (t##s##x##c a, t##s##x##c b) \ +{ return t##s##x##c##_is_all_zero (a ^b); } \ +\ +static_always_inline int \ +t##s##x##c##_is_all_equal (t##s##x##c v, t##s x) \ +{ return t##s##x##c##_is_equal (v, t##s##x##c##_splat (x)); } \ foreach_avx512_vec512i foreach_avx512_vec512u #undef _ |