2 files changed, 16 insertions, 0 deletions
diff --git a/src/opts/SkNx_neon.h b/src/opts/SkNx_neon.h
index b114f8f28a..8d1a249530 100644
--- a/src/opts/SkNx_neon.h
+++ b/src/opts/SkNx_neon.h
@@ -34,6 +34,15 @@ public:
     AI static SkNx Load(const void* ptr) { return vld1_f32((const float*)ptr); }
     AI void store(void* ptr) const { vst1_f32((float*)ptr, fVec); }
 
+    AI static void Store3(void* dst, const SkNx& a, const SkNx& b, const SkNx& c) {
+        float32x2x3_t abc = {{
+            a.fVec,
+            b.fVec,
+            c.fVec,
+        }};
+        vst3_f32((float*) dst, abc);
+    }
+
     AI SkNx invert() const {
         float32x2_t est0 = vrecpe_f32(fVec),
                     est1 = vmul_f32(vrecps_f32(est0, fVec), est0);
diff --git a/src/opts/SkNx_sse.h b/src/opts/SkNx_sse.h
index a5432495eb..dc9a5944da 100644
--- a/src/opts/SkNx_sse.h
+++ b/src/opts/SkNx_sse.h
@@ -29,6 +29,13 @@ public:
 
     AI void store(void* ptr) const { _mm_storel_pi((__m64*)ptr, fVec); }
 
+    AI static void Store3(void* dst, const SkNx& a, const SkNx& b, const SkNx& c) {
+        auto lo = _mm_setr_ps(a[0], b[0], c[0], a[1]),
+             hi = _mm_setr_ps(b[1], c[1],    0,    0);
+        _mm_storeu_ps((float*)dst, lo);
+        _mm_storel_pi(((__m64*)dst) + 2, hi);
+    }
+
     AI SkNx operator - () const { return _mm_xor_ps(_mm_set1_ps(-0.0f), fVec); }
 
     AI SkNx operator + (const SkNx& o) const { return _mm_add_ps(fVec, o.fVec); }