From cbf39c63a6eb89a2c80b2338afc374081803d79d Mon Sep 17 00:00:00 2001
From: Georgios Pinitas <georgios.pinitas@arm.com>
Date: Mon, 10 Sep 2018 15:07:45 +0100
Subject: COMPMID-1566: Add broadcast to CLArithmeticSubtraction

Change-Id: I05d21f9a92013ecfd1128d12cf1561cfd6e5c5e9
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/147983
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
---
 .../CL/functions/CLArithmeticSubtraction.cpp        | 15 +++++++++++++--
 src/runtime/NEON/functions/NEArithmeticAddition.cpp |  5 +++--
 .../NEON/functions/NEArithmeticSubtraction.cpp      | 21 +++++++++++++++++----
 3 files changed, 33 insertions(+), 8 deletions(-)

(limited to 'src/runtime')

diff --git a/src/runtime/CL/functions/CLArithmeticSubtraction.cpp b/src/runtime/CL/functions/CLArithmeticSubtraction.cpp
index 5fca30c4f9..e661f6adc1 100644
--- a/src/runtime/CL/functions/CLArithmeticSubtraction.cpp
+++ b/src/runtime/CL/functions/CLArithmeticSubtraction.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016, 2017 ARM Limited.
+ * Copyright (c) 2016-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -23,6 +23,7 @@
  */
 #include "arm_compute/runtime/CL/functions/CLArithmeticSubtraction.h"
 
+#include "arm_compute/core/CL/ICLTensor.h"
 #include "arm_compute/core/CL/kernels/CLArithmeticSubtractionKernel.h"
 #include "support/ToolchainSupport.h"
 
@@ -30,11 +31,21 @@
 
 using namespace arm_compute;
 
-void CLArithmeticSubtraction::configure(const ICLTensor *input1, const ICLTensor *input2, ICLTensor *output, ConvertPolicy policy)
+void CLArithmeticSubtraction::configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy)
 {
     auto k = arm_compute::support::cpp14::make_unique<CLArithmeticSubtractionKernel>();
     k->configure(input1, input2, output, policy);
     _kernel = std::move(k);
+
+    if(output->info()->dimension(0) > 1)
+    {
+        ICLTensor *broadcasted_info = (input1->info()->dimension(0) == 1) ? input1 : input2;
+
+        if(broadcasted_info->info()->dimension(0) == 1)
+        {
+            _border_handler.configure(broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE);
+        }
+    }
 }
 
 Status CLArithmeticSubtraction::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy)
diff --git a/src/runtime/NEON/functions/NEArithmeticAddition.cpp b/src/runtime/NEON/functions/NEArithmeticAddition.cpp
index 7d8e3cff1c..677e9f676f 100644
--- a/src/runtime/NEON/functions/NEArithmeticAddition.cpp
+++ b/src/runtime/NEON/functions/NEArithmeticAddition.cpp
@@ -29,8 +29,8 @@
 
 #include <utility>
 
-using namespace arm_compute;
-
+namespace arm_compute
+{
 void NEArithmeticAddition::configure(ITensor *input1, ITensor *input2, ITensor *output, ConvertPolicy policy)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEArithmeticAdditionKernel>();
@@ -51,3 +51,4 @@ Status NEArithmeticAddition::validate(const ITensorInfo *input1, const ITensorIn
 {
     return NEArithmeticAdditionKernel::validate(input1, input2, output, policy);
 }
+} // namespace arm_compute
diff --git a/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp b/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
index 5c0491ec6f..ceb4b496bc 100644
--- a/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
+++ b/src/runtime/NEON/functions/NEArithmeticSubtraction.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -23,20 +23,33 @@
  */
 #include "arm_compute/runtime/NEON/functions/NEArithmeticSubtraction.h"
 
+#include "arm_compute/core/ITensor.h"
 #include "arm_compute/core/NEON/kernels/NEArithmeticSubtractionKernel.h"
 #include "support/ToolchainSupport.h"
 
 #include <utility>
 
-using namespace arm_compute;
-
-void NEArithmeticSubtraction::configure(const ITensor *input1, const ITensor *input2, ITensor *output, ConvertPolicy policy)
+namespace arm_compute
+{
+void NEArithmeticSubtraction::configure(ITensor *input1, ITensor *input2, ITensor *output, ConvertPolicy policy)
 {
     auto k = arm_compute::support::cpp14::make_unique<NEArithmeticSubtractionKernel>();
     k->configure(input1, input2, output, policy);
     _kernel = std::move(k);
+
+    if(output->info()->dimension(0) > 1)
+    {
+        ITensor *broadcasted_info = (input1->info()->dimension(0) == 1) ? input1 : input2;
+
+        if(broadcasted_info->info()->dimension(0) == 1)
+        {
+            _border_handler.configure(broadcasted_info, _kernel->border_size(), BorderMode::REPLICATE);
+        }
+    }
 }
+
 Status NEArithmeticSubtraction::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, ConvertPolicy policy)
 {
     return NEArithmeticSubtractionKernel::validate(input1, input2, output, policy);
 }
+} // namespace arm_compute
-- 
cgit v1.2.1