Improve filelist for GPU

Resolves: COMPMID-4653

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I4f69d42369bf8ab91cd027acf1c97e92ec1ef554
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5948
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>

5 files changed, 59 insertions, 1123 deletions

diff --git a/Android.bp b/Android.bp
index 83e1e0a338..b32a9da895 100644
--- a/Android.bp
+++ b/Android.bp
@@ -378,8 +378,6 @@ cc_library_static {
         "src/core/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedValhall.cpp",
         "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.cpp",
         "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp",
-        "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyBifrost.cpp",
-        "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyValhall.cpp",
         "src/core/helpers/SoftmaxHelpers.cpp",
         "src/core/helpers/WindowHelpers.cpp",
         "src/core/utils/AssemblyUtils.cpp",
diff --git a/SConscript b/SConscript
index 03b87e737f..886ad083ad 100644
--- a/SConscript
+++ b/SConscript
@@ -272,7 +272,7 @@ runtime_files_hp += Glob('src/runtime/CPP/ICPPSimpleFunction.cpp')
 runtime_files = Glob('src/runtime/CPP/functions/*.cpp')
 
 # C API files
-runtime_files_hp += filelist['c_api']['cpu']
+runtime_files_hp += filelist['c_api']['common']
 
 if env['opencl']:
     runtime_files_hp += filelist['c_api']['gpu']
@@ -297,19 +297,8 @@ if env['openmp']:
      runtime_files_hp += Glob('src/runtime/OMP/OMPScheduler.cpp')
 
 if env['opencl']:
-    core_files += Glob('src/core/CL/*.cpp')
-    core_files += Glob('src/core/gpu/cl/*.cpp')
-
-    runtime_files += Glob('src/runtime/CL/*.cpp')
+    runtime_files_hp += filelist['gpu']['common']
     runtime_files += Glob('src/runtime/CL/functions/*.cpp')
-    runtime_files += Glob('src/runtime/CL/gemm/*.cpp')
-    runtime_files += Glob('src/runtime/CL/tuners/*.cpp')
-    runtime_files += Glob('src/runtime/gpu/cl/*.cpp')
-    runtime_files += Glob('src/runtime/CL/mlgo/*.cpp')
-    runtime_files += Glob('src/runtime/CL/gemm_auto_heuristics/*.cpp')
-
-    runtime_files += Glob('src/gpu/cl/*.cpp')
-    graph_files += Glob('src/graph/backends/CL/*.cpp')
 
     operators = filelist['gpu']['operators']
     for operator in operators:
@@ -317,6 +306,8 @@ if env['opencl']:
         if "kernel" in operators[operator]["files"]:
             core_files += operators[operator]["files"]["kernel"]
 
+    graph_files += Glob('src/graph/backends/CL/*.cpp')
+
 sve_o = []
 core_files_sve = []
 if env['neon']:
@@ -330,8 +321,6 @@ if env['neon']:
                                       "arm_compute/core/NEON/kernels/assembly/",
                                       "src/core/cpu/kernels/assembly/",])
 
-    graph_files += Glob('src/graph/backends/NEON/*.cpp')
-
     # Load files based on user's options
     operators = filelist['cpu']['operators']
     for operator in operators:
@@ -346,9 +335,11 @@ if env['neon']:
             core_files += file_list
             core_files_sve += file_list_sve
 
+    runtime_files_hp += filelist['cpu']['common']
     runtime_files_hp += Glob('src/runtime/NEON/*.cpp')
     runtime_files += Glob('src/runtime/NEON/functions/*.cpp')
-    runtime_files_hp += filelist['cpu']['all']
+
+    graph_files += Glob('src/graph/backends/NEON/*.cpp')
 
 bootcode_o = []
 if env['os'] == 'bare_metal':
diff --git a/filelist.json b/filelist.json
index 6babf70eab..d520f8e1dd 100644
--- a/filelist.json
+++ b/filelist.json
@@ -9,7 +9,7 @@
     "src/common/TensorPack.cpp"
   ],
   "c_api": {
-    "cpu": [
+    "common": [
       "src/c/AclContext.cpp",
       "src/c/AclOperator.cpp",
       "src/c/AclQueue.cpp",
@@ -17,11 +17,61 @@
       "src/c/AclTensorPack.cpp",
       "src/c/AclVersion.cpp"
     ],
+    "cpu": [
+    ],
     "gpu": [
       "src/c/cl/AclOpenClExt.cpp"
     ]
   },
   "gpu": {
+    "common": [
+      "src/core/CL/CLCompileContext.cpp",
+      "src/core/CL/CLHelpers.cpp",
+      "src/core/CL/CLKernelLibrary.cpp",
+      "src/core/CL/CLUtils.cpp",
+      "src/core/CL/ICLKernel.cpp",
+      "src/core/CL/ICLSimple2DKernel.cpp",
+      "src/core/CL/ICLSimple3DKernel.cpp",
+      "src/core/CL/ICLSimpleKernel.cpp",
+      "src/core/CL/ICLTensor.cpp",
+      "src/core/CL/OpenCL.cpp",
+      "src/core/gpu/cl/ClKernelLibrary.cpp",
+      "src/core/gpu/cl/kernels/gemm/ClGemmHelpers.cpp",
+      "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeBifrost.cpp",
+      "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeMidgard.cpp",
+      "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeValhall.cpp",
+      "src/core/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedBifrost.cpp",
+      "src/core/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedValhall.cpp",
+      "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.cpp",
+      "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp",
+      "src/core/CL/kernels/CLFillBorderKernel.cpp",
+      "src/gpu/cl/ClContext.cpp",
+      "src/gpu/cl/ClQueue.cpp",
+      "src/gpu/cl/ClTensor.cpp",
+      "src/runtime/CL/CLBufferAllocator.cpp",
+      "src/runtime/CL/CLGEMMHeuristicsHandle.cpp",
+      "src/runtime/CL/CLHelpers.cpp",
+      "src/runtime/CL/CLMemory.cpp",
+      "src/runtime/CL/CLMemoryRegion.cpp",
+      "src/runtime/CL/CLOperator.cpp",
+      "src/runtime/CL/CLRuntimeContext.cpp",
+      "src/runtime/CL/CLScheduler.cpp",
+      "src/runtime/CL/CLSubTensor.cpp",
+      "src/runtime/CL/CLTensor.cpp",
+      "src/runtime/CL/CLTensorAllocator.cpp",
+      "src/runtime/CL/CLTuner.cpp",
+      "src/runtime/CL/ICLSimpleFunction.cpp",
+      "src/runtime/CL/Utils.cpp",
+      "src/runtime/CL/gemm/CLGEMMDefaultTypeBifrost.cpp",
+      "src/runtime/CL/gemm/CLGEMMDefaultTypeMidgard.cpp",
+      "src/runtime/CL/gemm/CLGEMMDefaultTypeValhall.cpp",
+      "src/runtime/CL/gemm_auto_heuristics/CLGEMMAutoHeuristics.cpp",
+      "src/runtime/CL/mlgo/HeuristicTree.cpp",
+      "src/runtime/CL/mlgo/MLGOHeuristics.cpp",
+      "src/runtime/CL/mlgo/MLGOParser.cpp",
+      "src/runtime/CL/mlgo/Utils.cpp",
+      "src/runtime/CL/tuners/CLTuningParametersList.cpp"
+    ],
     "high_priority": [
       "Activation",
       "DepthwiseConv2d",
@@ -214,16 +264,8 @@
           ],
           "kernel": [
             "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyNativeKernel.cpp",
-            "src/core/gpu/cl/kernels/gemm/ClGemmHelpers.cpp",
-            "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeBifrost.cpp",
-            "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeMidgard.cpp",
-            "src/core/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeValhall.cpp",
             "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp",
-            "src/core/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedBifrost.cpp",
-            "src/core/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedValhall.cpp",
             "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp",
-            "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.cpp",
-            "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp",
             "src/core/gpu/cl/kernels/ClGemmMatrixMultiplyKernel.cpp",
             "src/core/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.cpp",
             "src/core/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.cpp"
@@ -418,13 +460,6 @@
           ]
         }
       },
-      "FillBorder": {
-        "files": {
-          "kernel": [
-            "src/core/CL/kernels/CLFillBorderKernel.cpp"
-          ]
-        }
-      },
       "FuseBatchNormalization": {
         "files": {
           "kernel": [
@@ -629,7 +664,7 @@
     }
   },
   "cpu": {
-    "all": [
+    "common": [
       "src/cpu/CpuContext.cpp",
       "src/cpu/CpuQueue.cpp",
       "src/cpu/CpuTensor.cpp"
diff --git a/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyBifrost.cpp b/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyBifrost.cpp
deleted file mode 100644
index 7ed6b39f3e..0000000000
--- a/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyBifrost.cpp
+++ /dev/null
@@ -1,518 +0,0 @@
-/*
- * Copyright (c) 2019-2021 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 "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/GPUTarget.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/TensorShape.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/gpu/cl/kernels/gemm/ClGemmHelpers.h"
-
-#include <utility>
-
-namespace arm_compute
-{
-namespace opencl
-{
-namespace kernels
-{
-namespace gemm
-{
-using namespace arm_compute::misc::shape_calculator;
-
-ClGemmDefaultConfigReshapedRhsOnlyBifrost::ClGemmDefaultConfigReshapedRhsOnlyBifrost(GPUTarget gpu)
-    : IClGemmKernelConfig(gpu)
-{
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure(unsigned int m, unsigned int n, unsigned int k, unsigned int b, DataType data_type)
-{
-    using ConfigurationFunctionExecutorPtr = std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> (ClGemmDefaultConfigReshapedRhsOnlyBifrost::*)(unsigned int m, unsigned int n, unsigned int k,
-                                             unsigned int b);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G51(&ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_u8);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G52(&ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G52_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G52_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_u8);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G76(&ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_u8);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G7x(&ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_u8);
-
-    ConfigurationFunctionExecutorPtr func = nullptr;
-
-    switch(_target)
-    {
-        case GPUTarget::G76:
-            func = configs_G76.get_function(data_type);
-            break;
-        case GPUTarget::G51:
-            func = configs_G51.get_function(data_type);
-            break;
-        case GPUTarget::G52:
-            func = configs_G52.get_function(data_type);
-            break;
-        default:
-            func = configs_G7x.get_function(data_type);
-            break;
-    }
-
-    ARM_COMPUTE_ERROR_ON_MSG(func == nullptr, "Data type not support for GEMM");
-    return (this->*func)(m, n, k, b);
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        if(n <= 2548)
-        {
-            return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 4, false, true, false, true, false);
-        }
-        else
-        {
-            return configure_lhs_rhs_info(m, n, 1, 4, 16, 1, 8, false, true, false, true, false);
-        }
-    }
-    else
-    {
-        return configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 4, false, true, false, true);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    GEMMLHSMatrixInfo lhs_info_buf;
-    GEMMRHSMatrixInfo rhs_info_buf;
-    GEMMLHSMatrixInfo lhs_info_img;
-    GEMMRHSMatrixInfo rhs_info_img;
-
-    const bool is_workload_big = ((m * n * b) / 16) >= 2048;
-
-    if(m == 1)
-    {
-        if(n >= 8192)
-        {
-            const unsigned int h0 = std::max(n / 4, 1U);
-            return configure_lhs_rhs_info(m, n, 1, 4, 8, 1, h0, false, true, false, true, false);
-        }
-        else
-        {
-            const unsigned int h0 = std::max(n / 2, 1U);
-            if(n <= 204)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, h0, false, true, false, true, false);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, h0, false, true, false, true, false);
-            }
-        }
-    }
-    else
-    {
-        const int h0 = std::max(std::min(static_cast<int>(n / 4), static_cast<int>(16)), static_cast<int>(1));
-        if(is_workload_big)
-        {
-            std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, h0, false, true, false, true);
-        }
-        else
-        {
-            std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 2, 4, 8, 1, h0, false, true, false, true);
-        }
-    }
-
-    // Get lhs_info/rhs_info in case of OpenCL image
-    const int h0 = std::max(std::min(static_cast<int>(n / 4), static_cast<int>(16)), static_cast<int>(1));
-    if(is_workload_big)
-    {
-        std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, h0, false, true, false, false, true);
-    }
-    else
-    {
-        std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 2, 4, 8, 1, h0, false, true, false, true, true);
-    }
-
-    const TensorInfo  tensor_rhs_info(TensorShape(n, k, b), 1, DataType::F32);
-    const TensorShape shape = compute_rhs_reshaped_shape(tensor_rhs_info, rhs_info_img);
-    const TensorInfo  tensor_reshaped_info(shape, 1, DataType::F32);
-
-    // In case of vector by matrix or small workloads, we use the OpenCL buffer rather than the OpenCL image2d
-    const bool use_cl_image2d = ((m == 1) || ((((m * n * b) / 16) < 2048) && n < 128)) ? false : true;
-
-    if(bool(validate_image2d_support_on_rhs(tensor_reshaped_info, rhs_info_img)) && use_cl_image2d)
-    {
-        return std::make_pair(lhs_info_img, rhs_info_img);
-    }
-    else
-    {
-        return std::make_pair(lhs_info_buf, rhs_info_buf);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G52_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-    const float r_nk     = static_cast<float>(n) / static_cast<float>(k);
-
-    GEMMLHSMatrixInfo lhs_info_buf;
-    GEMMRHSMatrixInfo rhs_info_buf;
-    GEMMLHSMatrixInfo lhs_info_img;
-    GEMMRHSMatrixInfo rhs_info_img;
-
-    if(m == 1)
-    {
-        if(r_nk <= 0.4664f)
-        {
-            return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 16, false, true, false, true, false);
-        }
-        else
-        {
-            std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 1, 4, 8, 1, 16, false, true, false, true, true);
-            std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 1, 4, 8, 1, 16, false, true, false, true, false);
-
-            return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                       std::make_pair(lhs_info_buf, rhs_info_buf),
-                                       n, k, b, DataType::F32);
-        }
-    }
-    else
-    {
-        if(workload <= 274.4000f)
-        {
-            return configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 16, false, false, false, true, false);
-        }
-        else
-        {
-            std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 2, false, false, false, true, true);
-            std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 2, false, false, false, true, false);
-
-            return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                       std::make_pair(lhs_info_buf, rhs_info_buf),
-                                       n, k, b, DataType::F32);
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int n0 = n < 1280 ? 2 : 4;
-        const unsigned int h0 = std::max(n / n0, 1U);
-        return configure_lhs_rhs_info(m, n, 1, n0, 4, 1, h0, false, true, false, true);
-    }
-    else
-    {
-        return configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 2, false, true, false, true);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        if(n > 2048)
-        {
-            const unsigned int h0 = std::max(n / 4, 1U);
-            return configure_lhs_rhs_info(m, n, 1, 4, 4, 1, h0, false, true, false, true);
-        }
-        else
-        {
-            const unsigned int h0 = std::max(n / 2, 1U);
-            return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, h0, false, true, false, true);
-        }
-    }
-    else
-    {
-        return configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 4, false, true, false, true);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G52_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    const float r_mn     = static_cast<float>(m) / static_cast<float>(n);
-    const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-    const float r_mk     = static_cast<float>(m) / static_cast<float>(k);
-    const float r_nk     = static_cast<float>(n) / static_cast<float>(k);
-
-    GEMMLHSMatrixInfo lhs_info_buf;
-    GEMMRHSMatrixInfo rhs_info_buf;
-    GEMMLHSMatrixInfo lhs_info_img;
-    GEMMRHSMatrixInfo rhs_info_img;
-
-    if(m == 1)
-    {
-        std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 1, 4, 16, 1, 16, false, true, false, false, false);
-
-        if(r_mk <= 0.0026f)
-        {
-            if(r_nk <= 0.4664f)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 32, false, true, false, true, false);
-            }
-            else
-            {
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 1, 4, 16, 1, 16, false, true, false, false, true);
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F16);
-            }
-        }
-        else
-        {
-            if(r_mk <= 0.0148f)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 32, false, true, false, true, false);
-            }
-            else
-            {
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 1, 4, 16, 1, 16, false, true, false, false, true);
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F16);
-            }
-        }
-    }
-    else
-    {
-        std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 5, 8, 4, 1, 2, false, false, false, false, false);
-
-        if(workload <= 362.6000f)
-        {
-            return configure_lhs_rhs_info(m, n, 2, 2, 8, 1, 16, false, false, false, true, false);
-        }
-        else
-        {
-            if(r_mn <= 22.6067f)
-            {
-                if(workload <= 708.8000f)
-                {
-                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 5, 4, 4, 1, 2, false, false, false, false, true);
-                    return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                               std::make_pair(lhs_info_buf, rhs_info_buf),
-                                               n, k, b, DataType::F16);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 5, 8, 2, 1, 16, false, false, false, false, false);
-                }
-            }
-            else
-            {
-                if(r_nk <= 0.0917f)
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 2, 8, 1, 16, false, false, false, true, false);
-                }
-                else
-                {
-                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 5, 4, 4, 1, 2, false, false, false, false, true);
-                    return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                               std::make_pair(lhs_info_buf, rhs_info_buf),
-                                               n, k, b, DataType::F16);
-                }
-            }
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-
-    if(m == 1)
-    {
-        return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 32, false, true, false, true, false);
-    }
-    else
-    {
-        const float r_mn     = static_cast<float>(m) / static_cast<float>(n);
-        const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-
-        if(workload <= 7449.60f)
-        {
-            if(workload <= 691.60f)
-            {
-                return configure_lhs_rhs_info(m, n, 2, 2, 8, 1, 8, false, false, false, false, false);
-            }
-            else
-            {
-                if(workload <= 4155.20f)
-                {
-                    return configure_lhs_rhs_info(m, n, 5, 2, 8, 1, 16, false, false, false, false, false);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 5, 8, 2, 1, 32, false, false, false, false, false);
-                }
-            }
-        }
-        else
-        {
-            if(workload <= 16300.80f)
-            {
-                if(r_mn <= 44.56f)
-                {
-                    GEMMLHSMatrixInfo lhs_info_buf;
-                    GEMMRHSMatrixInfo rhs_info_buf;
-                    GEMMLHSMatrixInfo lhs_info_img;
-                    GEMMRHSMatrixInfo rhs_info_img;
-
-                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 8, 4, 4, 1, 1, false, true, false, false, true);
-                    std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 5, 2, 8, 1, 16, false, false, false, false, false);
-
-                    return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                               std::make_pair(lhs_info_buf, rhs_info_buf),
-                                               n, k, b, DataType::F16);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 5, 2, 8, 1, 16, false, false, false, false, false);
-                }
-            }
-            else
-            {
-                GEMMLHSMatrixInfo lhs_info_buf;
-                GEMMRHSMatrixInfo rhs_info_buf;
-                GEMMLHSMatrixInfo lhs_info_img;
-                GEMMRHSMatrixInfo rhs_info_img;
-
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 5, 4, 4, 1, 2, false, true, false, false, true);
-                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 5, 2, 8, 1, 16, false, false, false, false, false);
-
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F16);
-            }
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int n0 = n < 1280 ? 2 : 4;
-        const unsigned int h0 = std::max(n / n0, 1U);
-        return configure_lhs_rhs_info(m, n, 1, n0, 8, 1, h0, false, true, false, true);
-    }
-    else
-    {
-        return configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 2, false, true, false, true);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G7x_u8(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(dot8_supported(CLKernelLibrary::get().get_device()))
-    {
-        if(m == 1)
-        {
-            const unsigned int h0 = std::max(n / 2, 1U);
-            return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, h0, false, true, false, true);
-        }
-        else
-        {
-            const unsigned int h0 = std::max(n / 4, 1U);
-            return configure_lhs_rhs_info(m, n, 4, 4, 16, 1, h0, false, true, false, true);
-        }
-    }
-    else
-    {
-        const int h0 = std::max(std::min(static_cast<int>(n / 2), static_cast<int>(128)), static_cast<int>(1));
-        if(m == 1)
-        {
-            return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, h0, false, true, false, true);
-        }
-        else
-        {
-            return configure_lhs_rhs_info(m, n, 4, 2, 16, 1, h0, false, true, false, true);
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G76_u8(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int h0 = std::max(n / 2, 1U);
-        return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, h0, false, true, false, true);
-    }
-    else
-    {
-        return configure_lhs_rhs_info(m, n, 4, 4, 16, 1, 2, false, true, false, true);
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyBifrost::configure_G51_u8(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int h0 = std::max(n / 2, 1U);
-        return configure_lhs_rhs_info(m, n, 1, 4, 16, 1, h0, false, true, false, true);
-    }
-    else
-    {
-        const unsigned int h0 = std::max(n / 2, 1U);
-        return configure_lhs_rhs_info(m, n, 4, 2, 16, 1, h0, false, true, false, true);
-    }
-}
-
-} // namespace gemm
-} // namespace kernels
-} // namespace opencl
-} // namespace arm_compute
diff --git a/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyValhall.cpp b/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyValhall.cpp
deleted file mode 100644
index 4c6e633896..0000000000
--- a/src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultReshapedRhsOnlyValhall.cpp
+++ /dev/null
@@ -1,570 +0,0 @@
-/*
- * Copyright (c) 2020-2021 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 "src/core/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.h"
-
-#include "arm_compute/core/CL/CLHelpers.h"
-#include "arm_compute/core/CL/CLKernelLibrary.h"
-#include "arm_compute/core/GPUTarget.h"
-#include "arm_compute/core/TensorInfo.h"
-#include "arm_compute/core/TensorShape.h"
-#include "arm_compute/core/utils/misc/ShapeCalculator.h"
-#include "src/core/gpu/cl/kernels/gemm/ClGemmHelpers.h"
-
-#include <utility>
-
-namespace arm_compute
-{
-namespace opencl
-{
-namespace kernels
-{
-namespace gemm
-{
-using namespace arm_compute::misc::shape_calculator;
-
-ClGemmDefaultConfigReshapedRhsOnlyValhall::ClGemmDefaultConfigReshapedRhsOnlyValhall(GPUTarget gpu)
-    : IClGemmKernelConfig(gpu)
-{
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure(unsigned int m, unsigned int n, unsigned int k, unsigned int b, DataType data_type)
-{
-    using ConfigurationFunctionExecutorPtr = std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> (ClGemmDefaultConfigReshapedRhsOnlyValhall::*)(unsigned int m, unsigned int n, unsigned int k,
-                                             unsigned int b);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G77(&ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_u8);
-
-    CLGEMMConfigArray<ConfigurationFunctionExecutorPtr> configs_G78(&ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G78_f32,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G78_f16,
-                                                                    &ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_u8);
-
-    ConfigurationFunctionExecutorPtr func = nullptr;
-
-    switch(_target)
-    {
-        case GPUTarget::G78:
-            func = configs_G78.get_function(data_type);
-            break;
-        case GPUTarget::G77:
-        default:
-            func = configs_G77.get_function(data_type);
-            break;
-    }
-
-    ARM_COMPUTE_ERROR_ON_MSG(func == nullptr, "Data type not support for GEMM");
-    return (this->*func)(m, n, k, b);
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    if(m == 1)
-    {
-        const float r_mn = static_cast<float>(m) / static_cast<float>(n);
-        const float r_mk = static_cast<float>(m) / static_cast<float>(k);
-
-        if(r_mk <= 0.0064484127797186375)
-        {
-            if(r_mn <= 0.0028273810748942196)
-            {
-                GEMMLHSMatrixInfo lhs_info_buf;
-                GEMMRHSMatrixInfo rhs_info_buf;
-                GEMMLHSMatrixInfo lhs_info_img;
-                GEMMRHSMatrixInfo rhs_info_img;
-
-                const unsigned int h0 = std::max(n / 4, 1U);
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 1, 4, 8, 1, 16, 0, 1, 0, 0, 1);
-                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 1, 4, 4, 1, h0, 0, 1, 0, 1, 0);
-
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F32);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 8, 0, 1, 0, 0, 0);
-            }
-        }
-        else
-        {
-            if(r_mk <= 0.020312500186264515)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 4, 0, 1, 0, 0, 0);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 1, 4, 16, 1, 16, 0, 1, 0, 1, 0);
-            }
-        }
-    }
-    else
-    {
-        const float r_mn     = static_cast<float>(m) / static_cast<float>(n);
-        const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-        const float r_mk     = static_cast<float>(m) / static_cast<float>(k);
-
-        if(workload <= 1999.2000122070312)
-        {
-            if(workload <= 747.1999816894531)
-            {
-                return configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 8, 0, 1, 0, 1, 0);
-            }
-            else
-            {
-                GEMMLHSMatrixInfo lhs_info_buf;
-                GEMMRHSMatrixInfo rhs_info_buf;
-                GEMMLHSMatrixInfo lhs_info_img;
-                GEMMRHSMatrixInfo rhs_info_img;
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 2, 0, 0, 0, 1, 1);
-                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 8, 0, 1, 0, 1, 0);
-
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F32);
-            }
-        }
-        else
-        {
-            if(r_mn <= 0.03348214365541935)
-            {
-                if(r_mk <= 0.028125000186264515)
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 8, 0, 1, 0, 1, 0);
-                }
-                else
-                {
-                    GEMMLHSMatrixInfo lhs_info_buf;
-                    GEMMRHSMatrixInfo rhs_info_buf;
-                    GEMMLHSMatrixInfo lhs_info_img;
-                    GEMMRHSMatrixInfo rhs_info_img;
-                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 2, 0, 0, 0, 1, 1);
-                    std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 8, 0, 1, 0, 1, 0);
-
-                    return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                               std::make_pair(lhs_info_buf, rhs_info_buf),
-                                               n, k, b, DataType::F32);
-                }
-            }
-            else
-            {
-                GEMMLHSMatrixInfo lhs_info_buf;
-                GEMMRHSMatrixInfo rhs_info_buf;
-                GEMMLHSMatrixInfo lhs_info_img;
-                GEMMRHSMatrixInfo rhs_info_img;
-                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 2, 0, 1, 0, 0, 1);
-                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 16, 0, 1, 0, 1, 0);
-
-                return select_lhs_rhs_info(std::make_pair(lhs_info_img, rhs_info_img),
-                                           std::make_pair(lhs_info_buf, rhs_info_buf),
-                                           n, k, b, DataType::F32);
-            }
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int h0 = std::max(n / 2, 1U);
-        if(n <= 836.0)
-        {
-            return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, h0, 0, 1, 0, 1, 0);
-        }
-        else
-        {
-            return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, h0, 0, 1, 0, 1, 0);
-        }
-    }
-    else if(m < 128)
-    {
-        const int h0 = std::max(std::min(static_cast<int>(n / 4), static_cast<int>(256)), static_cast<int>(1));
-        if(k >= 512)
-        {
-            return configure_lhs_rhs_info(m, n, 2, 4, 16, 1, h0, 0, 1, 0, 0);
-        }
-        else
-        {
-            return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, h0, 0, 1, 0, 0);
-        }
-    }
-    else
-    {
-        const int h0 = std::max(std::min(static_cast<int>(n / 4), static_cast<int>(256)), static_cast<int>(1));
-        if(n >= 64)
-        {
-            return configure_lhs_rhs_info(m, n, 4, 8, 4, 1, h0, 0, 1, 0, 0);
-        }
-        else
-        {
-            if(k >= 512)
-            {
-                return configure_lhs_rhs_info(m, n, 2, 4, 16, 1, h0, 0, 1, 0, 0);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, h0, 0, 1, 0, 0);
-            }
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G77_u8(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    ARM_COMPUTE_UNUSED(k);
-    ARM_COMPUTE_UNUSED(b);
-
-    if(m == 1)
-    {
-        const unsigned int h0 = std::max(n / 2, 1U);
-        return configure_lhs_rhs_info(m, n, 1, 4, 16, 1, h0, 0, 1, 0, 1);
-    }
-    else
-    {
-        const int h0 = std::max(std::min(static_cast<int>(n / 4), static_cast<int>(256)), static_cast<int>(1));
-        if(m >= 28)
-        {
-            return configure_lhs_rhs_info(m, n, 4, 4, 16, 1, h0, 0, 1, 0, 1);
-        }
-        else
-        {
-            return configure_lhs_rhs_info(m, n, 2, 4, 16, 1, h0, 0, 1, 0, 1);
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G78_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    const float r_mn     = static_cast<float>(m) / static_cast<float>(n);
-    const float r_mk     = static_cast<float>(m) / static_cast<float>(k);
-    const float r_nk     = static_cast<float>(n) / static_cast<float>(k);
-    const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-
-    if(m == 1)
-    {
-        if(workload <= 278.7000f)
-        {
-            if(workload <= 7.5000f)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, 2, 0, 1, 1, 0, 0);
-            }
-            else
-            {
-                if(r_mn <= 0.0031f)
-                {
-                    if(workload <= 256.6000f)
-                    {
-                        if(workload <= 16.7500f)
-                        {
-                            if(r_nk <= 1.6671f)
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                            }
-                            else
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, 2, 0, 1, 1, 0, 0);
-                            }
-                        }
-                        else
-                        {
-                            return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                        }
-                    }
-                    else
-                    {
-                        return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                    }
-                }
-                else
-                {
-                    if(r_mk <= 0.0027f)
-                    {
-                        if(r_mk <= 0.0014f)
-                        {
-                            return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                        }
-                        else
-                        {
-                            if(workload <= 8.9500f)
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, 2, 0, 1, 1, 0, 0);
-                            }
-                            else
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                            }
-                        }
-                    }
-                    else
-                    {
-                        if(workload <= 14.1500f)
-                        {
-                            return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, 2, 0, 1, 1, 0, 0);
-                        }
-                        else
-                        {
-                            if(r_mk <= 0.0041f)
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 2, 1, 32, 0, 0, 0, 1, 0);
-                            }
-                            else
-                            {
-                                return configure_lhs_rhs_info(m, n, 1, 2, 8, 1, 2, 0, 1, 1, 0, 0);
-                            }
-                        }
-                    }
-                }
-            }
-        }
-        else
-        {
-            if(workload <= 363.7000f)
-            {
-                if(r_mk <= 0.0031f)
-                {
-                    return configure_lhs_rhs_info(m, n, 1, 4, 2, 1, 32, 0, 1, 0, 1, 0);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 1, 4, 4, 1, 32, 0, 1, 0, 1, 0);
-                }
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 1, 4, 2, 1, 32, 0, 1, 0, 1, 0);
-            }
-        }
-    }
-    else
-    {
-        if(workload <= 1384.8000f)
-        {
-            if(workload <= 704.0000f)
-            {
-                return configure_lhs_rhs_info(m, n, 2, 2, 4, 1, 32, 0, 1, 0, 1, 0);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 4, 0, 0, 0, 1, 1);
-            }
-        }
-        else
-        {
-            if(workload <= 16761.6006f)
-            {
-                if(r_mn <= 187.1250f)
-                {
-                    return configure_lhs_rhs_info(m, n, 4, 4, 4, 1, 16, 0, 0, 0, 1, 1);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 4, 0, 0, 0, 1, 1);
-                }
-            }
-            else
-            {
-                if(r_mk <= 432.4630f)
-                {
-                    return configure_lhs_rhs_info(m, n, 5, 4, 4, 1, 16, 0, 0, 0, 1, 1);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 4, 4, 1, 16, 0, 1, 0, 1, 1);
-                }
-            }
-        }
-    }
-}
-
-std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> ClGemmDefaultConfigReshapedRhsOnlyValhall::configure_G78_f16(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
-{
-    const float r_mn     = static_cast<float>(m) / static_cast<float>(n);
-    const float r_mk     = static_cast<float>(m) / static_cast<float>(k);
-    const float r_nk     = static_cast<float>(n) / static_cast<float>(k);
-    const float workload = (static_cast<float>(m) * static_cast<float>(n) * static_cast<float>(b)) / 20.0f;
-
-    if(m == 1)
-    {
-        if(r_mn <= 0.0038f)
-        {
-            if(workload <= 353.9000f)
-            {
-                if(workload <= 278.7000f)
-                {
-                    return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, 32, 0, 0, 1, 0, 0);
-                }
-                else
-                {
-                    if(r_mk <= 0.0004f)
-                    {
-                        return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, 32, 0, 0, 1, 0, 0);
-                    }
-                    else
-                    {
-                        if(r_mk <= 0.0030f)
-                        {
-                            return configure_lhs_rhs_info(m, n, 1, 8, 4, 1, 8, 0, 1, 1, 0, 1);
-                        }
-                        else
-                        {
-                            return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, 32, 0, 0, 1, 0, 0);
-                        }
-                    }
-                }
-            }
-            else
-            {
-                if(r_nk <= 1.9384f)
-                {
-                    return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, 32, 0, 0, 1, 0, 0);
-                }
-                else
-                {
-                    return configure_lhs_rhs_info(m, n, 1, 8, 4, 1, 8, 0, 1, 1, 0, 1);
-                }
-            }
-        }
-        else
-        {
-            if(r_nk <= 1.0368f)
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 16, 1, 32, 0, 0, 1, 0, 0);
-            }
-            else
-            {
-                return configure_lhs_rhs_info(m, n, 1, 2, 4, 1, 32, 0, 0, 1, 0, 0);
-            }
-        }
-    }
-    else
-    {
-        if(workload <= 1422.4000f)
-        {
-            if(workload <= 704.0000f)
-            {
-                return configure_lhs_rhs_info(m, n, 2, 2, 8, 1, 32, 0, 0, 1, 0, 0);
-            }
-            else
-            {
-                if(workload <= 1197.6000f)
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 8, 0, 1, 1, 0, 1);
-                }
-                else
-                {
-                    if(workload <= 1241.6000f)
-                    {
-                        return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                    }
-                    else
-                    {
-                        return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 8, 0, 1, 1, 0, 1);
-                    }
-                }
-            }
-        }
-        else
-        {
-            if(workload <= 2769.6000f)
-            {
-                if(workload <= 1846.4000f)
-                {
-                    if(r_mn <= 2.4927f)
-                    {
-                        return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                    }
-                    else
-                    {
-                        return configure_lhs_rhs_info(m, n, 4, 4, 8, 1, 32, 0, 1, 1, 0, 0);
-                    }
-                }
-                else
-                {
-                    if(r_mn <= 0.6261f)
-                    {
-                        return configure_lhs_rhs_info(m, n, 4, 4, 8, 1, 32, 0, 1, 1, 0, 0);
-                    }
-                    else
-                    {
-                        if(r_mk <= 3.4453f)
-                        {
-                            if(r_mn <= 1.4135f)
-                            {
-                                return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                            }
-                            else
-                            {
-                                return configure_lhs_rhs_info(m, n, 4, 4, 8, 1, 32, 0, 1, 1, 0, 0);
-                            }
-                        }
-                        else
-                        {
-                            return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                        }
-                    }
-                }
-            }
-            else
-            {
-                if(r_nk <= 0.0302f)
-                {
-                    return configure_lhs_rhs_info(m, n, 2, 4, 8, 1, 8, 0, 1, 1, 0, 1);
-                }
-                else
-                {
-                    if(r_mk <= 181.3750f)
-                    {
-                        return configure_lhs_rhs_info(m, n, 4, 4, 8, 1, 32, 0, 1, 1, 0, 0);
-                    }
-                    else
-                    {
-                        if(workload <= 28035.2002f)
-                        {
-                            return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                        }
-                        else
-                        {
-                            if(r_mk <= 808.6667f)
-                            {
-                                return configure_lhs_rhs_info(m, n, 4, 4, 8, 1, 32, 0, 1, 1, 0, 0);
-                            }
-                            else
-                            {
-                                return configure_lhs_rhs_info(m, n, 2, 8, 8, 1, 16, 0, 1, 1, 0, 0);
-                            }
-                        }
-                    }
-                }
-            }
-        }
-    }
-}
-} // namespace gemm
-} // namespace kernels
-} // namespace opencl
-} // namespace arm_compute