COMPMID-881: RSH new arm_gemm interface.

Change-Id: I1e2a1a77097d8017c274af3f97eba6964f80f5fa
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/122592
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

7 files changed, 0 insertions, 1108 deletions

diff --git a/src/core/NEON/kernels/arm64/NEGEMMAArch64Kernel.cpp b/src/core/NEON/kernels/arm64/NEGEMMAArch64Kernel.cpp
deleted file mode 100644
index 0eaa9aa39b..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMMAArch64Kernel.cpp
+++ /dev/null
@@ -1,127 +0,0 @@
-/*
- * Copyright (c) 2017 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMMAArch64Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/AccessWindowTranspose.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/NEON/NEFixedPoint.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#include "arm_compute/core/NEON/kernels/assembly/gemm_interleaved.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_sgemm_12x8.hpp"
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-#include <tuple>
-
-namespace arm_compute
-{
-void NEGEMMAArch64Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1, output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input0, input1, output);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info());
-
-    AccessWindowRectangle output_access(output->info(), 0, 0, 12, 8);
-
-    const int input0_access_end = ceil_to_multiple(input0->info()->tensor_shape().x(), 8);
-    const int input1_access_end = ceil_to_multiple(input1->info()->tensor_shape().x(), 12);
-
-    update_window_and_padding(win,
-                              AccessWindowStatic(input0->info(), 0, 0, input0_access_end, input0->info()->tensor_shape().y()),
-                              AccessWindowStatic(input1->info(), 0, 0, input1_access_end, input1->info()->tensor_shape().y()),
-                              output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEGEMMAArch64Kernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-
-    const int lda = _input0->info()->strides_in_bytes().y() / sizeof(float);
-    const int ldb = _input1->info()->strides_in_bytes().y() / sizeof(float);
-    const int ldc = _output->info()->strides_in_bytes().y() / sizeof(float);
-
-    const auto in1_ptr = reinterpret_cast<const float *>(_input1->buffer());
-
-    const int M = std::min(_output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = _output->info()->tensor_shape().x();
-    const int K = _input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(_input0, window);
-    Iterator out(_output, window);
-
-    GemmInterleaved<sgemm_12x8, float, float> gemm(&info.cpu_info, M, N, K, _is_transposed_0, _is_transposed_1);
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *workspace      = _workspace->buffer() + offset;
-    size_t           workspace_size = _workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const float *>(in0.ptr()), lda,
-                     reinterpret_cast<const float *>(in1_ptr), ldb,
-                     reinterpret_cast<float *>(out.ptr()), ldc,
-                     _alpha, _beta, workspace);
-    },
-    in0, out);
-}
-} // namespace arm_compute
diff --git a/src/core/NEON/kernels/arm64/NEGEMMAArch64NativeKernel.cpp b/src/core/NEON/kernels/arm64/NEGEMMAArch64NativeKernel.cpp
deleted file mode 100644
index 0b3212bf55..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMMAArch64NativeKernel.cpp
+++ /dev/null
@@ -1,121 +0,0 @@
-/*
- * Copyright (c) 2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMMAArch64NativeKernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/AccessWindowTranspose.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/NEON/NEFixedPoint.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#include "arm_compute/core/NEON/kernels/convolution/winograd/gemm.hpp"
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-#include <tuple>
-
-namespace arm_compute
-{
-void NEGEMMAArch64NativeKernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0,
-                                                   bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1, output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input0, input1, output);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info(), Steps(16U, 4U));
-
-    const int input0_access_end_x = ceil_to_multiple(input0->info()->tensor_shape().x(), 4);
-    const int input0_access_end_y = ceil_to_multiple(input0->info()->tensor_shape().y(), 4);
-    const int input1_access_end_x = ceil_to_multiple(input1->info()->tensor_shape().x(), 16);
-
-    AccessWindowStatic    input0_access(input0->info(), 0, 0, input0_access_end_x, input0_access_end_y);
-    AccessWindowStatic    input1_access(input1->info(), 0, 0, input1_access_end_x, input1->info()->tensor_shape().y());
-    AccessWindowRectangle output_access(output->info(), 0, 0, 16, 4);
-    update_window_and_padding(win, input0_access, input1_access, output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEGEMMAArch64NativeKernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-    ARM_COMPUTE_UNUSED(info);
-
-    const auto in1_ptr = reinterpret_cast<const float *>(_input1->buffer());
-
-    // Calculate row strides for each matrix
-    const int lda = _input0->info()->strides_in_bytes().y() / sizeof(float);
-    const int ldb = _input1->info()->strides_in_bytes().y() / sizeof(float);
-    const int ldc = _output->info()->strides_in_bytes().y() / sizeof(float);
-
-    // Calculate matrix sizes
-    const int M = std::min(_input0->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int K = _input0->info()->tensor_shape().x();
-    const int N = _input1->info()->tensor_shape().x();
-
-    // Create window (Only iterate over batches)
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    // Create Iterators
-    Iterator in0(_input0, window);
-    Iterator out(_output, window);
-
-    // Execute GEMM
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        BlockedGemm<4, 16, float, float>(reinterpret_cast<const float *>(in0.ptr()),
-                                         reinterpret_cast<const float *>(in1_ptr),
-                                         reinterpret_cast<float *>(out.ptr()),
-                                         M, K, N,
-                                         lda, ldb, ldc);
-    },
-    in0, out);
-}
-} // namespace arm_compute
diff --git a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64A53Kernel.cpp b/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64A53Kernel.cpp
deleted file mode 100644
index 80606dcc07..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64A53Kernel.cpp
+++ /dev/null
@@ -1,201 +0,0 @@
-/*
- * Copyright (c) 2017 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMMLowpAArch64A53Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#include "arm_compute/core/NEON/kernels/assembly/gemm_interleaved.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_s16_12x8.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u16_12x8.hpp"
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-
-// Enable only if compiled for AArch64-V8A targets
-#ifdef ARM_COMPUTE_AARCH64_V8A
-
-namespace arm_compute
-{
-NEGEMMLowpAArch64A53Kernel::NEGEMMLowpAArch64A53Kernel()
-    : _func(nullptr)
-{
-}
-
-void gemm_interleaved_s16_12x8(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1,
-                               const Window     &window,
-                               const ThreadInfo &info)
-{
-    const int lda = input0->info()->strides_in_bytes().y();
-    const int ldb = input1->info()->strides_in_bytes().y();
-    const int ldc = output->info()->strides_in_bytes().y() / sizeof(int32_t);
-
-    const auto in1_ptr = reinterpret_cast<const int8_t *>(input1->buffer());
-
-    const int M = std::min(output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = output->info()->tensor_shape().x();
-    const int K = input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(input0, window);
-    Iterator out(output, window);
-
-    GemmInterleaved<gemm_s16_12x8, int8_t, int32_t> gemm(&info.cpu_info, M, N, K, is_transposed_0, is_transposed_1);
-
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *_workspace     = workspace->buffer() + offset;
-    size_t           workspace_size = workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), _workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const int8_t *>(in0.ptr()), lda,
-                     reinterpret_cast<const int8_t *>(in1_ptr), ldb,
-                     reinterpret_cast<int32_t *>(out.ptr()), ldc,
-                     alpha, beta, _workspace);
-    },
-    in0, out);
-}
-
-void gemm_interleaved_u16_12x8(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1,
-                               const Window     &window,
-                               const ThreadInfo &info)
-{
-    const int lda = input0->info()->strides_in_bytes().y();
-    const int ldb = input1->info()->strides_in_bytes().y();
-    const int ldc = output->info()->strides_in_bytes().y() / sizeof(int32_t);
-
-    const auto in1_ptr = reinterpret_cast<const int8_t *>(input1->buffer());
-
-    const int M = std::min(output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = output->info()->tensor_shape().x();
-    const int K = input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(input0, window);
-    Iterator out(output, window);
-
-    GemmInterleaved<gemm_u16_12x8, uint8_t, uint32_t> gemm(&info.cpu_info, M, N, K, is_transposed_0, is_transposed_1);
-
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *_workspace     = workspace->buffer() + offset;
-    size_t           workspace_size = workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), _workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const uint8_t *>(in0.ptr()), lda,
-                     reinterpret_cast<const uint8_t *>(in1_ptr), ldb,
-                     reinterpret_cast<uint32_t *>(out.ptr()), ldc,
-                     alpha, beta, _workspace);
-    },
-    in0, out);
-}
-
-void NEGEMMLowpAArch64A53Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0,
-                                                    bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::S8, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::S32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    switch(input0->info()->data_type())
-    {
-        case DataType::S8:
-            _func = &gemm_interleaved_s16_12x8;
-            break;
-        case DataType::U8:
-            _func = &gemm_interleaved_u16_12x8;
-            break;
-        default:
-            ARM_COMPUTE_ERROR("Element size not supported");
-            break;
-    }
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info());
-
-    AccessWindowRectangle output_access(output->info(), 0, 0, 12, 8);
-
-    const int input0_access_end = ceil_to_multiple(input0->info()->tensor_shape().x(), 12);
-    const int input1_access_end = ceil_to_multiple(input1->info()->tensor_shape().x(), 12);
-
-    update_window_and_padding(win,
-                              AccessWindowStatic(input0->info(), 0, 0, input0_access_end, input0->info()->tensor_shape().y()),
-                              AccessWindowStatic(input1->info(), 0, 0, input1_access_end, input1->info()->tensor_shape().y()),
-                              output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEGEMMLowpAArch64A53Kernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-    ARM_COMPUTE_ERROR_ON(_func == nullptr);
-
-    (*_func)(_input0, _input1, _output, _workspace, _alpha, _beta, _is_transposed_0, _is_transposed_1, window, info);
-}
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_AARCH64_V8A */
diff --git a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64Kernel.cpp b/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64Kernel.cpp
deleted file mode 100644
index 38f82f0407..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64Kernel.cpp
+++ /dev/null
@@ -1,199 +0,0 @@
-/*
- * Copyright (c) 2017 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMMLowpAArch64Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#include "arm_compute/core/NEON/kernels/assembly/gemm_interleaved.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_s8_4x4.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_4x4.hpp"
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-
-// Enable only if compiled for AArch64-V8A targets
-#ifdef ARM_COMPUTE_AARCH64_V8A
-
-namespace arm_compute
-{
-NEGEMMLowpAArch64Kernel::NEGEMMLowpAArch64Kernel()
-    : _func(nullptr)
-{
-}
-
-void gemm_interleaved_s8(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1, const Window &window,
-                         const ThreadInfo &info)
-{
-    const int lda = input0->info()->strides_in_bytes().y();
-    const int ldb = input1->info()->strides_in_bytes().y();
-    const int ldc = output->info()->strides_in_bytes().y() / sizeof(int32_t);
-
-    const auto in1_ptr = reinterpret_cast<const int8_t *>(input1->buffer());
-
-    const int M = std::min(output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = output->info()->tensor_shape().x();
-    const int K = input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(input0, window);
-    Iterator out(output, window);
-
-    GemmInterleaved<gemm_s8_4x4, int8_t, int32_t> gemm(&info.cpu_info, M, N, K, is_transposed_0, is_transposed_1);
-
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *_workspace     = workspace->buffer() + offset;
-    size_t           workspace_size = workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), _workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const int8_t *>(in0.ptr()), lda,
-                     reinterpret_cast<const int8_t *>(in1_ptr), ldb,
-                     reinterpret_cast<int32_t *>(out.ptr()), ldc,
-                     alpha, beta, _workspace);
-    },
-    in0, out);
-}
-
-void gemm_interleaved_u8(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1, const Window &window,
-                         const ThreadInfo &info)
-{
-    const int lda = input0->info()->strides_in_bytes().y();
-    const int ldb = input1->info()->strides_in_bytes().y();
-    const int ldc = output->info()->strides_in_bytes().y() / sizeof(uint32_t);
-
-    const auto in1_ptr = reinterpret_cast<const uint8_t *>(input1->buffer());
-
-    const int M = std::min(output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = output->info()->tensor_shape().x();
-    const int K = input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(input0, window);
-    Iterator out(output, window);
-
-    GemmInterleaved<gemm_u8_4x4, uint8_t, uint32_t> gemm(&info.cpu_info, M, N, K, is_transposed_0, is_transposed_1);
-
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *_workspace     = workspace->buffer() + offset;
-    size_t           workspace_size = workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), _workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const uint8_t *>(in0.ptr()), lda,
-                     reinterpret_cast<const uint8_t *>(in1_ptr), ldb,
-                     reinterpret_cast<uint32_t *>(out.ptr()), ldc,
-                     alpha, beta, _workspace);
-    },
-    in0, out);
-}
-
-void NEGEMMLowpAArch64Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0,
-                                                 bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::S8, DataType::U8);
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::S32, DataType::U32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    switch(input0->info()->data_type())
-    {
-        case DataType::S8:
-            _func = &gemm_interleaved_s8;
-            break;
-        case DataType::U8:
-            _func = &gemm_interleaved_u8;
-            break;
-        default:
-            ARM_COMPUTE_ERROR("Element size not supported");
-            break;
-    }
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info());
-
-    AccessWindowRectangle output_access(output->info(), 0, 0, 4, 4);
-
-    const int input0_access_end = ceil_to_multiple(input0->info()->tensor_shape().x(), 4);
-    const int input1_access_end = ceil_to_multiple(input1->info()->tensor_shape().x(), 4);
-
-    update_window_and_padding(win,
-                              AccessWindowStatic(input0->info(), 0, 0, input0_access_end, input0->info()->tensor_shape().y()),
-                              AccessWindowStatic(input1->info(), 0, 0, input1_access_end, input1->info()->tensor_shape().y()),
-                              output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEGEMMLowpAArch64Kernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-    ARM_COMPUTE_ERROR_ON(_func == nullptr);
-
-    (*_func)(_input0, _input1, _output, _workspace, _alpha, _beta, _is_transposed_0, _is_transposed_1, window, info);
-}
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_AARCH64_V8A */
diff --git a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64V8P4Kernel.cpp b/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64V8P4Kernel.cpp
deleted file mode 100644
index d4fcf5e3cb..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMMLowpAArch64V8P4Kernel.cpp
+++ /dev/null
@@ -1,198 +0,0 @@
-/*
- * Copyright (c) 2017-2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMMLowpAArch64V8P4Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#include "arm_compute/core/NEON/kernels/assembly/gemm_interleaved.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_s8_12x8.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_gemm_u8_12x8.hpp"
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-
-// Enable only if compiled for AArch64-V8.2-A targets
-#ifdef ARM_COMPUTE_AARCH64_V8_2
-
-namespace
-{
-using namespace arm_compute;
-
-Status validate_arguments(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::QASYMM8, DataType::U8, DataType::S8);
-    ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::S32);
-    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1);
-    ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1);
-
-    return Status{};
-}
-
-std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input0, ITensorInfo *input1, ITensorInfo *output)
-{
-    // Configure kernel window
-    Window win = calculate_max_window(*output);
-
-    AccessWindowRectangle output_access(output, 0, 0, 12, 8);
-
-    const int input0_access_end = ceil_to_multiple(input0->tensor_shape().x(), 8);
-    const int input1_access_end = ceil_to_multiple(input1->tensor_shape().x(), 12);
-
-    bool window_changed = update_window_and_padding(win,
-                                                    AccessWindowStatic(input0, 0, 0, input0_access_end, input0->tensor_shape().y()),
-                                                    AccessWindowStatic(input1, 0, 0, input1_access_end, input1->tensor_shape().y()),
-                                                    output_access);
-
-    Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
-    return std::make_pair(err, win);
-}
-
-template <typename strategy, typename To, typename Tr>
-void *align_workspace(GemmInterleaved<strategy, To, Tr> &gemm, const ThreadInfo &info, ITensor *ws)
-{
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *workspace      = ws->buffer() + offset;
-    size_t           workspace_size = ws->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-    return workspace;
-}
-
-template <typename strategy>
-void execute_gemm(const Window &win, Iterator &in0, Iterator &in1, Iterator &out,
-                  const ThreadInfo &info, ITensor *ws, int M, int N, int K, bool is_transposed_0, bool is_transposed_1,
-                  int lda, int ldb, int ldc, float alpha, float beta)
-{
-    ARM_COMPUTE_UNUSED(M);
-    ARM_COMPUTE_UNUSED(N);
-    ARM_COMPUTE_UNUSED(K);
-    ARM_COMPUTE_UNUSED(is_transposed_0);
-    ARM_COMPUTE_UNUSED(is_transposed_1);
-    GemmInterleaved<strategy, typename strategy::operand_type, typename strategy::result_type> gemm(&info.cpu_info, M, N, K, is_transposed_0, is_transposed_1);
-    void *workspace = align_workspace(gemm, info, ws);
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const typename strategy::operand_type *>(in0.ptr()), lda,
-                     reinterpret_cast<const typename strategy::operand_type *>(in1.ptr()), ldb,
-                     reinterpret_cast<typename strategy::result_type *>(out.ptr()), ldc,
-                     alpha, beta, workspace);
-    },
-    in0, out);
-}
-} // namespace
-
-namespace arm_compute
-{
-void NEGEMMLowpAArch64V8P4Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0,
-                                                     bool is_transposed_1)
-{
-    // Perform validate step
-    ARM_COMPUTE_ERROR_ON_NULLPTR(input0, input1, output);
-    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input0->info(), input1->info(), output->info()));
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    // Configure kernel window
-    auto win_config = validate_and_configure_window(input0->info(), input1->info(), output->info());
-    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
-    INEKernel::configure(win_config.second);
-}
-
-Status NEGEMMLowpAArch64V8P4Kernel::validate(const ITensorInfo *input0, const ITensorInfo *input1, const ITensorInfo *output)
-{
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input0, input1, output));
-    ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input0->clone().get(), input1->clone().get(), output->clone().get()).first);
-
-    return Status{};
-}
-
-void NEGEMMLowpAArch64V8P4Kernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-
-    const int lda = _input0->info()->strides_in_bytes().y();
-    const int ldb = _input1->info()->strides_in_bytes().y();
-    const int ldc = _output->info()->strides_in_bytes().y() / sizeof(uint32_t);
-
-    const int M = std::min(_output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = _output->info()->tensor_shape().x();
-    const int K = _input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(_input0, window);
-    Iterator in1(_input1, window);
-    Iterator out(_output, window);
-
-    switch(_input0->info()->data_type())
-    {
-        case DataType::QASYMM8:
-        case DataType::U8:
-        {
-            execute_gemm<gemm_u8_12x8>(win, in0, in1, out, info, _workspace, M, N, K, _is_transposed_0, _is_transposed_1, lda, ldb, ldc, _alpha, _beta);
-            break;
-        }
-        case DataType::S8:
-        {
-            execute_gemm<gemm_s8_12x8>(win, in0, in1, out, info, _workspace, M, N, K, _is_transposed_0, _is_transposed_1, lda, ldb, ldc, _alpha, _beta);
-            break;
-        }
-        default:
-        {
-            ARM_COMPUTE_ERROR("Not supported.");
-            break;
-        }
-    }
-}
-} // namespace arm_compute
-#endif /* ARM_COMPUTE_AARCH64_V8_2 */
diff --git a/src/core/NEON/kernels/arm64/NEGEMVAArch64Kernel.cpp b/src/core/NEON/kernels/arm64/NEGEMVAArch64Kernel.cpp
deleted file mode 100644
index 163014b04f..0000000000
--- a/src/core/NEON/kernels/arm64/NEGEMVAArch64Kernel.cpp
+++ /dev/null
@@ -1,130 +0,0 @@
-/*
- * Copyright (c) 2017-2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEGEMVAArch64Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/AccessWindowTranspose.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/NEON/NEFixedPoint.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wswitch-default"
-#pragma GCC diagnostic ignored "-Weffc++"
-#include "arm_compute/core/NEON/kernels/assembly/gemv_transposed.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_sgemv_trans.hpp"
-#pragma GCC diagnostic pop
-} // namespace arm_compute
-
-#include <arm_neon.h>
-#include <cstddef>
-#include <cstdint>
-#include <tuple>
-
-namespace arm_compute
-{
-void NEGEMVAArch64Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0, bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F32);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1, output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input0, input1, output);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info());
-
-    AccessWindowRectangle output_access(output->info(), 0, 0, 12, 1);
-
-    const int input0_access_end = ceil_to_multiple(input0->info()->tensor_shape().x(), 12);
-    const int input1_access_end = ceil_to_multiple(input1->info()->tensor_shape().x(), 12);
-
-    update_window_and_padding(win,
-                              AccessWindowStatic(input0->info(), 0, 0, input0_access_end, input0->info()->tensor_shape().y()),
-                              AccessWindowStatic(input1->info(), 0, 0, input1_access_end, input1->info()->tensor_shape().y()),
-                              output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEGEMVAArch64Kernel::run(const Window &window, const ThreadInfo &info)
-{
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-
-    const int lda = _input0->info()->strides_in_bytes().y() / sizeof(sgemv_trans::operand_type);
-    const int ldb = _input1->info()->strides_in_bytes().y() / sizeof(sgemv_trans::operand_type);
-    const int ldc = _output->info()->strides_in_bytes().y() / sizeof(sgemv_trans::result_type);
-
-    const auto in1_ptr = reinterpret_cast<const sgemv_trans::operand_type *>(_input1->buffer());
-
-    const int N = _output->info()->tensor_shape().x();
-    const int K = _input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(_input0, window);
-    Iterator out(_output, window);
-
-    GemvTransposed<sgemv_trans, sgemv_trans::operand_type, sgemv_trans::result_type> gemm(&info.cpu_info, N, K);
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *workspace      = _workspace->buffer() + offset;
-    size_t           workspace_size = _workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const sgemv_trans::operand_type *>(in0.ptr()), lda,
-                     reinterpret_cast<const sgemv_trans::operand_type *>(in1_ptr), ldb,
-                     reinterpret_cast<sgemv_trans::result_type *>(out.ptr()), ldc,
-                     _alpha, _beta, workspace);
-    },
-    in0, out);
-}
-} // namespace arm_compute
diff --git a/src/core/NEON/kernels/arm64/NEHGEMMAArch64FP16Kernel.cpp b/src/core/NEON/kernels/arm64/NEHGEMMAArch64FP16Kernel.cpp
deleted file mode 100644
index e84409cfd2..0000000000
--- a/src/core/NEON/kernels/arm64/NEHGEMMAArch64FP16Kernel.cpp
+++ /dev/null
@@ -1,132 +0,0 @@
-/*
- * Copyright (c) 2017-2018 ARM Limited.
- *
- * SPDX-License-Identifier: MIT
- *
- * Permission is hereby granted, free of charge, to any person obtaining a copy
- * of this software and associated documentation files (the "Software"), to
- * deal in the Software without restriction, including without limitation the
- * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
- * sell copies of the Software, and to permit persons to whom the Software is
- * furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in all
- * copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
- * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
- * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * SOFTWARE.
- */
-#include "arm_compute/core/NEON/kernels/arm64/NEHGEMMAArch64FP16Kernel.h"
-
-#include "arm_compute/core/AccessWindowStatic.h"
-#include "arm_compute/core/AccessWindowTranspose.h"
-#include "arm_compute/core/Error.h"
-#include "arm_compute/core/Helpers.h"
-#include "arm_compute/core/IAccessWindow.h"
-#include "arm_compute/core/ITensor.h"
-#include "arm_compute/core/NEON/NEFixedPoint.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/Types.h"
-#include "arm_compute/core/Utils.h"
-#include "arm_compute/core/Validate.h"
-#include "arm_compute/core/Window.h"
-#include "support/ToolchainSupport.h"
-
-namespace arm_compute
-{
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wswitch-default"
-#include "arm_compute/core/NEON/kernels/assembly/gemm_interleaved.hpp"
-#include "arm_compute/core/NEON/kernels/assembly/kernels/a64_hgemm_24x8.hpp"
-#pragma GCC diagnostic pop
-} // namespace arm_compute
-
-namespace arm_compute
-{
-void NEHGEMMAArch64FP16Kernel::internal_configure(const ITensor *input0, const ITensor *input1, ITensor *output, ITensor *workspace, float alpha, float beta, bool is_transposed_0,
-                                                  bool is_transposed_1)
-{
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input0, 1, DataType::F16);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input0, input1, output);
-    ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input0, input1, output);
-
-    _input0          = input0;
-    _input1          = input1;
-    _output          = output;
-    _workspace       = workspace;
-    _alpha           = alpha;
-    _beta            = beta;
-    _is_transposed_0 = is_transposed_0;
-    _is_transposed_1 = is_transposed_1;
-
-    // Configure kernel window
-    Window win = calculate_max_window(*output->info());
-
-    AccessWindowRectangle output_access(output->info(), 0, 0, 24, 8);
-
-    const int input0_access_end = ceil_to_multiple(input0->info()->tensor_shape().x(), 8);
-    const int input1_access_end = ceil_to_multiple(input1->info()->tensor_shape().x(), 24);
-
-    update_window_and_padding(win,
-                              AccessWindowStatic(input0->info(), 0, 0, input0_access_end, input0->info()->tensor_shape().y()),
-                              AccessWindowStatic(input1->info(), 0, 0, input1_access_end, input1->info()->tensor_shape().y()),
-                              output_access);
-
-    INEKernel::configure(win);
-}
-
-void NEHGEMMAArch64FP16Kernel::run(const Window &window, const ThreadInfo &info)
-{
-#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
-    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
-    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
-
-    const int lda = _input0->info()->strides_in_bytes().y() / sizeof(hgemm_24x8::operand_type);
-    const int ldb = _input1->info()->strides_in_bytes().y() / sizeof(hgemm_24x8::operand_type);
-    const int ldc = _output->info()->strides_in_bytes().y() / sizeof(hgemm_24x8::result_type);
-
-    const auto in1_ptr = reinterpret_cast<const hgemm_24x8::operand_type *>(_input1->buffer());
-
-    const int M = std::min(_output->info()->tensor_shape().y(), static_cast<size_t>(window.y().end())) - window.y().start();
-    const int N = _output->info()->tensor_shape().x();
-    const int K = _input0->info()->tensor_shape().x();
-
-    // Only iterate over batches
-    Window win(window);
-    win.set(0, Window::Dimension(0, 1, 1));
-    win.set(1, Window::Dimension(0, 1, 1));
-
-    Iterator in0(_input0, window);
-    Iterator out(_output, window);
-
-    GemmInterleaved<hgemm_24x8, hgemm_24x8::operand_type, hgemm_24x8::result_type> gemm(&info.cpu_info, M, N, K, _is_transposed_0, _is_transposed_1);
-    constexpr size_t alignment      = 4096;
-    const size_t     offset         = (gemm.get_working_size() + alignment - 1) * info.thread_id;
-    void            *workspace      = _workspace->buffer() + offset;
-    size_t           workspace_size = _workspace->info()->total_size();
-
-    if(support::cpp11::align(alignment, gemm.get_working_size(), workspace, workspace_size) == nullptr)
-    {
-        ARM_COMPUTE_ERROR("Not enough space to align buffer!");
-    }
-
-    execute_window_loop(win, [&](const Coordinates & id)
-    {
-        gemm.execute(reinterpret_cast<const hgemm_24x8::operand_type *>(in0.ptr()), lda,
-                     reinterpret_cast<const hgemm_24x8::operand_type *>(in1_ptr), ldb,
-                     reinterpret_cast<hgemm_24x8::result_type *>(out.ptr()), ldc,
-                     _alpha, 1.f, workspace);
-    },
-    in0, out);
-#else  /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
-    ARM_COMPUTE_UNUSED(window);
-    ARM_COMPUTE_UNUSED(info);
-    ARM_COMPUTE_ERROR("Recompile the library with arch=arm64-v8.2-a to enable support for FP16.");
-#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
-}
-} // namespace arm_compute