vppinfra: add 128 and 256 bit vector scatter/gather inlines

Change-Id: If6c65f16c6fba8beb90e189c1443c3d7d67ee02c
Signed-off-by: Damjan Marion <damarion@cisco.com>

2 files changed, 102 insertions, 0 deletions

diff --git a/src/vppinfra/vector_avx2.h b/src/vppinfra/vector_avx2.h
index bd27db11ef9..51625618823 100644
--- a/src/vppinfra/vector_avx2.h
+++ b/src/vppinfra/vector_avx2.h
@@ -188,6 +188,65 @@ u32x8_from_f32x8 (f32x8 v)
   return (u32x8) _mm256_cvttps_epi32 ((__m256) v);
 }
 
+#define u16x16_blend(v1, v2, mask) \
+  (u16x16) _mm256_blend_epi16 ((__m256i) (v1), (__m256i) (v2), mask)
+
+static_always_inline u64x4
+u64x4_gather (void *p0, void *p1, void *p2, void *p3)
+{
+  u64x4 r = {
+    *(u64 *) p0, *(u64 *) p1, *(u64 *) p2, *(u64 *) p3
+  };
+  return r;
+}
+
+static_always_inline u32x8
+u32x8_gather (void *p0, void *p1, void *p2, void *p3, void *p4, void *p5,
+	      void *p6, void *p7)
+{
+  u32x8 r = {
+    *(u32 *) p0, *(u32 *) p1, *(u32 *) p2, *(u32 *) p3,
+    *(u32 *) p4, *(u32 *) p5, *(u32 *) p6, *(u32 *) p7,
+  };
+  return r;
+}
+
+
+static_always_inline void
+u64x4_scatter (u64x4 r, void *p0, void *p1, void *p2, void *p3)
+{
+  *(u64 *) p0 = r[0];
+  *(u64 *) p1 = r[1];
+  *(u64 *) p2 = r[2];
+  *(u64 *) p3 = r[3];
+}
+
+static_always_inline void
+u32x8_scatter (u32x8 r, void *p0, void *p1, void *p2, void *p3, void *p4,
+	       void *p5, void *p6, void *p7)
+{
+  *(u32 *) p0 = r[0];
+  *(u32 *) p1 = r[1];
+  *(u32 *) p2 = r[2];
+  *(u32 *) p3 = r[3];
+  *(u32 *) p4 = r[4];
+  *(u32 *) p5 = r[5];
+  *(u32 *) p6 = r[6];
+  *(u32 *) p7 = r[7];
+}
+
+static_always_inline void
+u64x4_scatter_one (u64x4 r, int index, void *p)
+{
+  *(u64 *) p = r[index];
+}
+
+static_always_inline void
+u32x8_scatter_one (u32x8 r, int index, void *p)
+{
+  *(u32 *) p = r[index];
+}
+
 #endif /* included_vector_avx2_h */
 
 /*
diff --git a/src/vppinfra/vector_sse42.h b/src/vppinfra/vector_sse42.h
index 2b6246e97f8..5d6a47d3915 100644
--- a/src/vppinfra/vector_sse42.h
+++ b/src/vppinfra/vector_sse42.h
@@ -648,6 +648,49 @@ _(i32x4, i64x2, epi32_epi64)
 #undef _
 /* *INDENT-ON* */
 
+static_always_inline u64x2
+u64x2_gather (void *p0, void *p1)
+{
+  u64x2 r = { *(u64 *) p0, *(u64 *) p1 };
+  return r;
+}
+
+static_always_inline u32x4
+u32x4_gather (void *p0, void *p1, void *p2, void *p3, void *p4)
+{
+  u32x4 r = { *(u32 *) p0, *(u32 *) p1, *(u32 *) p2, *(u32 *) p3 };
+  return r;
+}
+
+
+static_always_inline void
+u64x2_scatter (u64x2 r, void *p0, void *p1)
+{
+  *(u64 *) p0 = r[0];
+  *(u64 *) p1 = r[1];
+}
+
+static_always_inline void
+u32x4_scatter (u32x4 r, void *p0, void *p1, void *p2, void *p3)
+{
+  *(u32 *) p0 = r[0];
+  *(u32 *) p1 = r[1];
+  *(u32 *) p2 = r[2];
+  *(u32 *) p3 = r[3];
+}
+
+static_always_inline void
+u64x2_scatter_one (u64x2 r, int index, void *p)
+{
+  *(u64 *) p = r[index];
+}
+
+static_always_inline void
+u32x4_scatter_one (u32x4 r, int index, void *p)
+{
+  *(u32 *) p = r[index];
+}
+
 
 #endif /* included_vector_sse2_h */