COMPMID-3959: Update Mali-G52 heuristic for CLGEMM - F32

- Add heuristic in CLGEMMKernelSelection
- Add heuristic in CLGEMMReshapedRHSOnly
- Add heuristic in CLGEMMReshaped

Change-Id: Ibaa13398f7a5976418a0ab1b6696ace09cc480fa
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4366
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 110 insertions, 0 deletions

diff --git a/src/core/CL/gemm/reshaped/CLGEMMReshapedKernelConfigurationBifrost.cpp b/src/core/CL/gemm/reshaped/CLGEMMReshapedKernelConfigurationBifrost.cpp
index c1ca187a70..70992974a3 100644
--- a/src/core/CL/gemm/reshaped/CLGEMMReshapedKernelConfigurationBifrost.cpp
+++ b/src/core/CL/gemm/reshaped/CLGEMMReshapedKernelConfigurationBifrost.cpp
@@ -60,6 +60,17 @@ std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> CLGEMMReshapedKernelConfiguratio
         { DataType::QSYMM8_PER_CHANNEL, &CLGEMMReshapedKernelConfigurationBifrost::configure_G76_u8 }
     };
 
+    // Configurations for Mali-G52
+    static std::map<DataType, ConfigurationFunctionExecutorPtr> gemm_configs_G52 =
+    {
+        { DataType::F32, &CLGEMMReshapedKernelConfigurationBifrost::configure_G52_f32 },
+        { DataType::F16, &CLGEMMReshapedKernelConfigurationBifrost::configure_G7x_f16 },
+        { DataType::QASYMM8, &CLGEMMReshapedKernelConfigurationBifrost::configure_G7x_u8 },
+        { DataType::QSYMM8, &CLGEMMReshapedKernelConfigurationBifrost::configure_G7x_u8 },
+        { DataType::QASYMM8_SIGNED, &CLGEMMReshapedKernelConfigurationBifrost::configure_G7x_u8 },
+        { DataType::QSYMM8_PER_CHANNEL, &CLGEMMReshapedKernelConfigurationBifrost::configure_G7x_u8 }
+    };
+
     // Configurations for Mali-G7x
     static std::map<DataType, ConfigurationFunctionExecutorPtr> gemm_configs_G7x =
     {
@@ -153,6 +164,105 @@ std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> CLGEMMReshapedKernelConfiguratio
     }
 }
 
+std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> CLGEMMReshapedKernelConfigurationBifrost::configure_G52_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 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(workload <= 274.4000f)
+    {
+        if(r_nk <= 0.7461f)
+        {
+            if(r_mn <= 21.1667f)
+            {
+                return configure_lhs_rhs_info(m, n, 4, 2, 4, 4, 4, false, true, true, false, false);
+            }
+            else
+            {
+                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, true, false, true, true);
+                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, 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
+        {
+            std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, true, false, true, true);
+            std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, 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(r_mk <= 17.3926f)
+        {
+            if(workload <= 542.4000f)
+            {
+                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, true, false, true, true);
+                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, 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
+            {
+                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 2, 1, true, true, false, true, true);
+                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 2, 1, true, 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(r_nk <= 0.5463f)
+            {
+                if(workload <= 11767.6001f)
+                {
+                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, true, false, true, true);
+                    std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, 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
+                {
+                    std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 2, 1, true, true, false, true, true);
+                    std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 2, 1, true, 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
+            {
+                std::tie(lhs_info_img, rhs_info_img) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, true, false, true, true);
+                std::tie(lhs_info_buf, rhs_info_buf) = configure_lhs_rhs_info(m, n, 4, 4, 4, 4, 2, true, 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);
+            }
+        }
+    }
+}
+
 std::pair<GEMMLHSMatrixInfo, GEMMRHSMatrixInfo> CLGEMMReshapedKernelConfigurationBifrost::configure_G76_f32(unsigned int m, unsigned int n, unsigned int k, unsigned int b)
 {
     ARM_COMPUTE_UNUSED(k);