vppinfra: more AVX2 and AVX512 inlines
Change-Id: I81bd967a580ae3b476dfd731e9933a9898568a91
Signed-off-by: Damjan Marion <damarion@cisco.com>
diff --git a/src/vppinfra/vector_avx2.h b/src/vppinfra/vector_avx2.h
index c857ad4..482bdd5 100644
--- a/src/vppinfra/vector_avx2.h
+++ b/src/vppinfra/vector_avx2.h
@@ -132,6 +132,16 @@
#undef _
/* *INDENT-ON* */
+static_always_inline u32x8
+u32x8_byte_swap (u32x8 v)
+{
+ u8x32 swap = {
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12,
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12
+ };
+ return (u32x8) _mm256_shuffle_epi8 ((__m256i) v, (__m256i) swap);
+}
+
static_always_inline u16x16
u16x16_byte_swap (u16x16 v)
{
@@ -188,6 +198,9 @@
return (u32x8) _mm256_cvttps_epi32 ((__m256) v);
}
+#define u32x8_blend(a,b,m) \
+ (u32x8) _mm256_blend_epi32 ((__m256i) a, (__m256i) b, m)
+
#define u16x16_blend(v1, v2, mask) \
(u16x16) _mm256_blend_epi16 ((__m256i) (v1), (__m256i) (v2), mask)
@@ -265,6 +278,19 @@
#define u64x4_permute_lanes(a, b, m) \
(u64x4) _mm256_permute2x128_si256 ((__m256i) a, (__m256i) b, m)
+static_always_inline u32x8
+u32x8_min (u32x8 a, u32x8 b)
+{
+ return (u32x8) _mm256_min_epu32 ((__m256i) a, (__m256i) b);
+}
+
+static_always_inline u32
+u32x8_min_scalar (u32x8 v)
+{
+ return u32x4_min_scalar (u32x4_min (u32x8_extract_lo (v),
+ u32x8_extract_hi (v)));
+}
+
static_always_inline void
u32x8_transpose (u32x8 a[8])
{
diff --git a/src/vppinfra/vector_avx512.h b/src/vppinfra/vector_avx512.h
index a5d6739..6fd27a3 100644
--- a/src/vppinfra/vector_avx512.h
+++ b/src/vppinfra/vector_avx512.h
@@ -69,6 +69,52 @@
return (u32) _mm512_movepi16_mask ((__m512i) v);
}
+static_always_inline u32x16
+u32x16_byte_swap (u32x16 v)
+{
+ u8x64 swap = {
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12,
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12,
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12,
+ 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12
+ };
+ return (u32x16) _mm512_shuffle_epi8 ((__m512i) v, (__m512i) swap);
+}
+
+static_always_inline u16x32
+u16x32_byte_swap (u16x32 v)
+{
+ u8x64 swap = {
+ 1, 0, 3, 2, 5, 4, 7, 6, 9, 8, 11, 10, 13, 12, 15, 14,
+ 1, 0, 3, 2, 5, 4, 7, 6, 9, 8, 11, 10, 13, 12, 15, 14,
+ 1, 0, 3, 2, 5, 4, 7, 6, 9, 8, 11, 10, 13, 12, 15, 14,
+ 1, 0, 3, 2, 5, 4, 7, 6, 9, 8, 11, 10, 13, 12, 15, 14
+ };
+ return (u16x32) _mm512_shuffle_epi8 ((__m512i) v, (__m512i) swap);
+}
+
+static_always_inline u32x8
+u32x16_extract_lo (u32x16 v)
+{
+ return (u32x8) _mm512_extracti64x4_epi64 ((__m512i) v, 0);
+}
+
+static_always_inline u32x8
+u32x16_extract_hi (u32x16 v)
+{
+ return (u32x8) _mm512_extracti64x4_epi64 ((__m512i) v, 1);
+}
+
+static_always_inline u32
+u32x16_min_scalar (u32x16 v)
+{
+ return u32x8_min_scalar (u32x8_min (u32x16_extract_lo (v),
+ u32x16_extract_hi (v)));
+}
+
+
+#define u32x16_ternary_logic(a, b, c, d) \
+ (u32x16) _mm512_ternarylogic_epi32 ((__m512i) a, (__m512i) b, (__m512i) c, d)
static_always_inline void
u32x16_transpose (u32x16 m[16])