From cff6f3b3d6750c47e9f8616bb8b2ec671cfe33d3 Mon Sep 17 00:00:00 2001
From: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Date: Thu, 9 Dec 2021 17:36:25 +0000
Subject: Fix 300% Regression CPU - Change default mws value in Kernel files

Resolves: COMPMID-5001
Change-Id: I13fbe859d2557be0459ba76da0136d0efb15f311
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6809
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>
---
 arm_compute/core/CPP/CPPTypes.h                                     | 1 +
 src/core/NEON/kernels/NEPadLayerKernel.cpp                          | 6 +++++-
 src/cpu/kernels/CpuActivationKernel.cpp                             | 6 +++++-
 src/cpu/kernels/CpuAddKernel.cpp                                    | 6 +++++-
 src/cpu/kernels/CpuIm2ColKernel.cpp                                 | 6 +++++-
 src/cpu/kernels/CpuReshapeKernel.cpp                                | 6 +++++-
 src/cpu/kernels/assembly/CpuGemmAssemblyWrapperKernel.h             | 6 +++++-
 .../kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp    | 6 +++++-
 8 files changed, 36 insertions(+), 7 deletions(-)

diff --git a/arm_compute/core/CPP/CPPTypes.h b/arm_compute/core/CPP/CPPTypes.h
index 76378d27ef..82a6a6c324 100644
--- a/arm_compute/core/CPP/CPPTypes.h
+++ b/arm_compute/core/CPP/CPPTypes.h
@@ -39,6 +39,7 @@ namespace arm_compute
     X(A55r1)                       \
     X(A35)                         \
     X(A73)                         \
+    X(A76)                         \
     X(A510)                        \
     X(X1)                          \
     X(V1)                          \
diff --git a/src/core/NEON/kernels/NEPadLayerKernel.cpp b/src/core/NEON/kernels/NEPadLayerKernel.cpp
index 2e5e9f76be..4e689e74e4 100644
--- a/src/core/NEON/kernels/NEPadLayerKernel.cpp
+++ b/src/core/NEON/kernels/NEPadLayerKernel.cpp
@@ -267,10 +267,14 @@ size_t NEPadLayerKernel::get_mws(const CPUInfo &platform, size_t thread_count) c
     {
         return 10240;
     }
-    else 
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 
 } // namespace arm_compute
diff --git a/src/cpu/kernels/CpuActivationKernel.cpp b/src/cpu/kernels/CpuActivationKernel.cpp
index 4a6468d022..aed73d1fec 100644
--- a/src/cpu/kernels/CpuActivationKernel.cpp
+++ b/src/cpu/kernels/CpuActivationKernel.cpp
@@ -238,10 +238,14 @@ size_t CpuActivationKernel::get_mws(const CPUInfo &platform, size_t thread_count
     {
         return 10240;
     }
-    else 
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 
 void CpuActivationKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info)
diff --git a/src/cpu/kernels/CpuAddKernel.cpp b/src/cpu/kernels/CpuAddKernel.cpp
index 0c3540f0d4..f3ee032ec5 100644
--- a/src/cpu/kernels/CpuAddKernel.cpp
+++ b/src/cpu/kernels/CpuAddKernel.cpp
@@ -298,10 +298,14 @@ size_t CpuAddKernel::get_mws(const CPUInfo &platform, size_t thread_count) const
     {
         return 10240;
     }
-    else
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 
 } // namespace kernels
diff --git a/src/cpu/kernels/CpuIm2ColKernel.cpp b/src/cpu/kernels/CpuIm2ColKernel.cpp
index ecd1748a44..58890fe5a2 100644
--- a/src/cpu/kernels/CpuIm2ColKernel.cpp
+++ b/src/cpu/kernels/CpuIm2ColKernel.cpp
@@ -452,10 +452,14 @@ size_t CpuIm2ColKernel::get_mws(const CPUInfo &platform, size_t thread_count) co
     {
         return 10240;
     }
-    else 
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 } // namespace kernels
 } // namespace cpu
diff --git a/src/cpu/kernels/CpuReshapeKernel.cpp b/src/cpu/kernels/CpuReshapeKernel.cpp
index e19707dd0c..49cb91b330 100644
--- a/src/cpu/kernels/CpuReshapeKernel.cpp
+++ b/src/cpu/kernels/CpuReshapeKernel.cpp
@@ -143,10 +143,14 @@ size_t CpuReshapeKernel::get_mws(const CPUInfo &platform, size_t thread_count) c
     {
         return 10240;
     }
-    else 
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 
 } // namespace kernels
diff --git a/src/cpu/kernels/assembly/CpuGemmAssemblyWrapperKernel.h b/src/cpu/kernels/assembly/CpuGemmAssemblyWrapperKernel.h
index 47548b2538..212fd79306 100644
--- a/src/cpu/kernels/assembly/CpuGemmAssemblyWrapperKernel.h
+++ b/src/cpu/kernels/assembly/CpuGemmAssemblyWrapperKernel.h
@@ -130,10 +130,14 @@ public:
         {
             return 3072;
         }
-        else 
+        else if (platform.get_cpu_model() == CPUModel::A76)
         {
             return 4096;
         }
+        else
+        {
+            return ICPPKernel::default_mws;
+        }
     }
 
 private:
diff --git a/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp b/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
index 934e38b054..6c235df59f 100644
--- a/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
+++ b/src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp
@@ -363,10 +363,14 @@ size_t CpuDepthwiseConv2dAssemblyWrapperKernel::get_mws(const CPUInfo &platform,
     {
         return 10240;
     }
-    else 
+    else if (platform.get_cpu_model() == CPUModel::A76)
     {
         return 9216;
     }
+    else
+    {
+        return ICPPKernel::default_mws;
+    }
 }
 } // namespace kernels
 } // namespace cpu
-- 
cgit v1.2.1