From 1813422f5095983fa802e96cb8e9e0d980de82c7 Mon Sep 17 00:00:00 2001
From: Georgios Pinitas <georgios.pinitas@arm.com>
Date: Thu, 3 Sep 2020 21:00:23 +0100
Subject: COMPMID-3583: Add S32 support to NEElementwiseDivision

Division follows the flooring division approach where for example 5/2=2 while
-5/2=-3

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I65756e0b31fe8d97f743a4c13dc5f96304722f75
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/3929
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
---
 tests/validation/NEON/ElementwiseDivision.cpp      | 13 +++++++++++
 .../fixtures/ElementwiseOperationsFixture.h        | 25 ++++++++++++++--------
 .../validation/reference/ElementwiseOperations.cpp | 11 +++++++++-
 3 files changed, 39 insertions(+), 10 deletions(-)

(limited to 'tests')

diff --git a/tests/validation/NEON/ElementwiseDivision.cpp b/tests/validation/NEON/ElementwiseDivision.cpp
index f6e0a65e84..4b613cbd0a 100644
--- a/tests/validation/NEON/ElementwiseDivision.cpp
+++ b/tests/validation/NEON/ElementwiseDivision.cpp
@@ -44,6 +44,9 @@ namespace
 {
 RelativeTolerance<float> tolerance_fp32(0.000001f);
 /** Input data sets **/
+const auto ElementwiseDivisionS32Dataset = combine(combine(framework::dataset::make("DataType", DataType::S32),
+                                                           framework::dataset::make("DataType", DataType::S32)),
+                                                   framework::dataset::make("DataType", DataType::S32));
 #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 RelativeTolerance<half> tolerance_fp16(static_cast<half>(0.01f));
 const auto              ElementwiseDivisionFP16Dataset = combine(combine(framework::dataset::make("DataType", DataType::F16), framework::dataset::make("DataType", DataType::F16)),
@@ -135,6 +138,16 @@ FIXTURE_DATA_TEST_CASE(RunSmallBroadcast, NEElementwiseDivisionBroadcastFixture<
 TEST_SUITE_END() // F32
 TEST_SUITE_END() // Float
 
+TEST_SUITE(Integer)
+TEST_SUITE(S32)
+FIXTURE_DATA_TEST_CASE(RunSmall, NEElementwiseDivisionFixture<int32_t>, framework::DatasetMode::ALL, combine(datasets::SmallShapes(), ElementwiseDivisionS32Dataset))
+{
+    // Validate output
+    validate(Accessor(_target), _reference);
+}
+TEST_SUITE_END() // S32
+TEST_SUITE_END() // Integer
+
 TEST_SUITE_END() // ElementwiseDivision
 TEST_SUITE_END() // NEON
 } // namespace validation
diff --git a/tests/validation/fixtures/ElementwiseOperationsFixture.h b/tests/validation/fixtures/ElementwiseOperationsFixture.h
index ebc52d5083..dcb408c801 100644
--- a/tests/validation/fixtures/ElementwiseOperationsFixture.h
+++ b/tests/validation/fixtures/ElementwiseOperationsFixture.h
@@ -59,16 +59,23 @@ protected:
     template <typename U>
     void fill(U &&tensor, int i)
     {
-        switch(_op)
+        if(is_data_type_float(tensor.data_type()))
         {
-            case ArithmeticOperation::DIV:
-                library->fill_tensor_uniform_ranged(tensor, i, { std::pair<float, float>(-0.001f, 0.001f) });
-                break;
-            case ArithmeticOperation::POWER:
-                library->fill_tensor_uniform(tensor, i, 0.0f, 5.0f);
-                break;
-            default:
-                library->fill_tensor_uniform(tensor, i);
+            switch(_op)
+            {
+                case ArithmeticOperation::DIV:
+                    library->fill_tensor_uniform_ranged(tensor, i, { std::pair<float, float>(-0.001f, 0.001f) });
+                    break;
+                case ArithmeticOperation::POWER:
+                    library->fill_tensor_uniform(tensor, i, 0.0f, 5.0f);
+                    break;
+                default:
+                    library->fill_tensor_uniform(tensor, i);
+            }
+        }
+        else
+        {
+            library->fill_tensor_uniform(tensor, i);
         }
     }
 
diff --git a/tests/validation/reference/ElementwiseOperations.cpp b/tests/validation/reference/ElementwiseOperations.cpp
index aab9d9d00c..f22c84e153 100644
--- a/tests/validation/reference/ElementwiseOperations.cpp
+++ b/tests/validation/reference/ElementwiseOperations.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-2019 Arm Limited.
+ * Copyright (c) 2018-2020 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -74,6 +74,15 @@ T arithm_op(ArithmeticOperation op, T src1, T src2, ConvertPolicy convert_policy
         case ArithmeticOperation::DIV:
         {
             val = (static_cast<intermediate_type>(src1) / static_cast<intermediate_type>(src2));
+            if(std::is_integral<T>::value)
+            {
+                // Implement flooring division
+                val = (src2 == 0) ? 0 : val;
+                if(static_cast<int32_t>(src1) % static_cast<int32_t>(src2) != 0 && ((src1 < 0) != (src2 < 0)))
+                {
+                    --val;
+                }
+            }
             break;
         }
         case ArithmeticOperation::POWER:
-- 
cgit v1.2.1