6 files changed, 77 insertions, 6 deletions
diff --git a/src/cpu/kernels/add/generic/neon/fp16.cpp b/src/cpu/kernels/add/generic/neon/fp16.cpp
index 12d4a467b7..bb6636af1e 100644
--- a/src/cpu/kernels/add/generic/neon/fp16.cpp
+++ b/src/cpu/kernels/add/generic/neon/fp16.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -33,6 +33,11 @@ void add_fp16_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const
 {
     return add_same_neon<float16_t>(src0, src1, dst, policy, window);
 }
+
+void add_fp16_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    return add_same_neon_as_1d_array<float16_t>(src0, src1, dst, policy, window);
+}
 }
 } // namespace arm_compute
 #endif /* (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */
diff --git a/src/cpu/kernels/add/generic/neon/fp32.cpp b/src/cpu/kernels/add/generic/neon/fp32.cpp
index 3563162fce..1d313a191d 100644
--- a/src/cpu/kernels/add/generic/neon/fp32.cpp
+++ b/src/cpu/kernels/add/generic/neon/fp32.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -32,5 +32,10 @@ void add_fp32_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const
 {
     return add_same_neon<float>(src0, src1, dst, policy, window);
 }
+
+void add_fp32_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    return add_same_neon_as_1d_array<float>(src0, src1, dst, policy, window);
+}
 }
 } // namespace arm_compute
diff --git a/src/cpu/kernels/add/generic/neon/impl.cpp b/src/cpu/kernels/add/generic/neon/impl.cpp
index ad3e445ab0..67985c985e 100644
--- a/src/cpu/kernels/add/generic/neon/impl.cpp
+++ b/src/cpu/kernels/add/generic/neon/impl.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2021 Arm Limited.
+ * Copyright (c) 2020-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -128,6 +128,35 @@ void add_same_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const
     }
 }
 
+template <typename ScalarType>
+void add_same_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    const ScalarType *src0_ptr = reinterpret_cast<const ScalarType *>(src0->buffer());
+    const ScalarType *src1_ptr = reinterpret_cast<const ScalarType *>(src1->buffer());
+    ScalarType       *dst_ptr  = reinterpret_cast<ScalarType *>(dst->buffer());
+
+    constexpr int window_step_x  = 16 / sizeof(ScalarType);
+    const auto    window_start_x = static_cast<int>(window.x().start());
+    const auto    window_end_x   = static_cast<int>(window.x().end());
+
+    int x = window_start_x;
+    for(; x <= (window_end_x - window_step_x); x += window_step_x)
+    {
+        const auto val1 = wrapper::vloadq(src0_ptr + x);
+        const auto val2 = wrapper::vloadq(src1_ptr + x);
+        const auto res  = (policy == ConvertPolicy::SATURATE) ? wrapper::vqadd(val1, val2) : wrapper::vadd(val1, val2);
+        wrapper::vstore(dst_ptr + x, res);
+    }
+
+    // Compute left-over elements
+    for(; x < window_end_x; ++x)
+    {
+        const auto val1 = *(src0_ptr + x);
+        const auto val2 = *(src1_ptr + x);
+        *(dst_ptr + x)  = (policy == ConvertPolicy::SATURATE) ? wrapper::add_sat(val1, val2) : val1 + val2;
+    }
+}
+
 template void add_same_neon<float>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
 template void add_same_neon<uint8_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
 template void add_same_neon<int32_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
@@ -137,5 +166,14 @@ template void add_same_neon<int16_t>(const ITensor *src0, const ITensor *src1, I
 template void add_same_neon<float16_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
 #endif /* (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */
 
+template void add_same_neon_as_1d_array<float>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+template void add_same_neon_as_1d_array<uint8_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+template void add_same_neon_as_1d_array<int32_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+template void add_same_neon_as_1d_array<int16_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+
+#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS)
+template void add_same_neon_as_1d_array<float16_t>(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+#endif /* (__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */
+
 } // namespace cpu
 } // namespace arm_compute
diff --git a/src/cpu/kernels/add/generic/neon/impl.h b/src/cpu/kernels/add/generic/neon/impl.h
index 07afdda225..f8f0f517b0 100644
--- a/src/cpu/kernels/add/generic/neon/impl.h
+++ b/src/cpu/kernels/add/generic/neon/impl.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -32,6 +32,9 @@ namespace cpu
 {
 template <typename ScalarType>
 void add_same_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
+
+template <typename ScalarType>
+void add_same_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window);
 } // namespace cpu
 } // namespace arm_compute
 #endif // SRC_CORE_NEON_KERNELS_ADD_IMPL_H
 \ No newline at end of file
diff --git a/src/cpu/kernels/add/generic/neon/integer.cpp b/src/cpu/kernels/add/generic/neon/integer.cpp
index 62c19e66b1..ffead03474 100644
--- a/src/cpu/kernels/add/generic/neon/integer.cpp
+++ b/src/cpu/kernels/add/generic/neon/integer.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2021 Arm Limited.
+ * Copyright (c) 2021-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -42,5 +42,20 @@ void add_s32_neon(const ITensor *src0, const ITensor *src1, ITensor *dst, const
 {
     return add_same_neon<int32_t>(src0, src1, dst, policy, window);
 }
+
+void add_u8_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    return add_same_neon_as_1d_array<uint8_t>(src0, src1, dst, policy, window);
+}
+
+void add_s16_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    return add_same_neon_as_1d_array<int16_t>(src0, src1, dst, policy, window);
+}
+
+void add_s32_neon_as_1d_array(const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window)
+{
+    return add_same_neon_as_1d_array<int32_t>(src0, src1, dst, policy, window);
+}
 }
 } // namespace arm_compute
diff --git a/src/cpu/kernels/add/list.h b/src/cpu/kernels/add/list.h
index 9d7c9a67ff..0285b231e0 100644
--- a/src/cpu/kernels/add/list.h
+++ b/src/cpu/kernels/add/list.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020-2021 Arm Limited.
+ * Copyright (c) 2020-2022 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -38,10 +38,15 @@ DECLARE_ADD_KERNEL(add_qasymm8_neon);
 DECLARE_ADD_KERNEL(add_qasymm8_signed_neon);
 DECLARE_ADD_KERNEL(add_qsymm16_neon);
 DECLARE_ADD_KERNEL(add_fp32_neon);
+DECLARE_ADD_KERNEL(add_fp32_neon_as_1d_array);
 DECLARE_ADD_KERNEL(add_fp16_neon);
+DECLARE_ADD_KERNEL(add_fp16_neon_as_1d_array);
 DECLARE_ADD_KERNEL(add_u8_neon);
+DECLARE_ADD_KERNEL(add_u8_neon_as_1d_array);
 DECLARE_ADD_KERNEL(add_s16_neon);
+DECLARE_ADD_KERNEL(add_s16_neon_as_1d_array);
 DECLARE_ADD_KERNEL(add_s32_neon);
+DECLARE_ADD_KERNEL(add_s32_neon_as_1d_array);
 DECLARE_ADD_KERNEL(add_fp32_sve);
 DECLARE_ADD_KERNEL(add_fp16_sve);
 DECLARE_ADD_KERNEL(add_u8_sve);