From c7b183ab741650653289f8ce3bdeb4926521fdbd Mon Sep 17 00:00:00 2001
From: Georgios Pinitas <georgios.pinitas@arm.com>
Date: Fri, 6 Mar 2020 18:12:09 +0000
Subject: COMPMID-3160: Add Bfloat16 support in NEGEMMConvolutionLayer

Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I0e449306c138a562ffc1455e76ec44b2fd059d85
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/2860
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
---
 .../validation/fixtures/ConvolutionLayerFixture.h  | 39 ++++++++++++++++++----
 1 file changed, 32 insertions(+), 7 deletions(-)

(limited to 'tests/validation/fixtures/ConvolutionLayerFixture.h')

diff --git a/tests/validation/fixtures/ConvolutionLayerFixture.h b/tests/validation/fixtures/ConvolutionLayerFixture.h
index 3c4b625ac6..b4abebe18d 100644
--- a/tests/validation/fixtures/ConvolutionLayerFixture.h
+++ b/tests/validation/fixtures/ConvolutionLayerFixture.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 ARM Limited.
+ * Copyright (c) 2017-2020 ARM Limited.
  *
  * SPDX-License-Identifier: MIT
  *
@@ -64,7 +64,9 @@ public:
         _data_type                = data_type;
         _weights_data_type        = weights_data_type;
         _is_quantized             = is_data_type_quantized_asymmetric(data_type);
-        _bias_data_type           = _is_quantized ? DataType::S32 : data_type;
+        _is_bfloat16              = data_type == DataType::BFLOAT16;
+        _bias_data_type           = _is_quantized ? DataType::S32 : (_is_bfloat16 ? DataType::F32 : data_type);
+        _output_data_type         = _is_bfloat16 ? DataType::F32 : data_type;
         _quantization_info        = quantization_info;
         _weight_quantization_info = weight_quantization_info;
         _data_layout              = data_layout;
@@ -74,6 +76,15 @@ public:
     }
 
 protected:
+    void regularize_values(void *values, size_t size)
+    {
+        float *fvalues = static_cast<float *>(values);
+        for(size_t i = 0; i < size; ++i)
+        {
+            fvalues[i] = float(bfloat16(fvalues[i]));
+        }
+    }
+
     template <typename U>
     void fill(U &&tensor, int i)
     {
@@ -119,6 +130,7 @@ protected:
                 library->fill(tensor, distribution, i);
                 break;
             }
+            case DataType::BFLOAT16:
             case DataType::F16:
             case DataType::F32:
             {
@@ -155,7 +167,7 @@ protected:
         TensorType src     = create_tensor<TensorType>(input_shape, _data_type, 1, _quantization_info, _data_layout);
         TensorType weights = create_tensor<TensorType>(reshaped_weights_shape, _weights_data_type, 1, _weight_quantization_info, _data_layout);
         TensorType bias    = create_tensor<TensorType>(bias_shape, _bias_data_type, 1, _quantization_info, _data_layout);
-        TensorType dst     = create_tensor<TensorType>(output_shape, _data_type, 1, _quantization_info, _data_layout);
+        TensorType dst     = create_tensor<TensorType>(output_shape, _output_data_type, 1, _quantization_info, _data_layout);
 
         // Create and configure function
         FunctionType conv;
@@ -195,16 +207,27 @@ protected:
 
         const unsigned int num_groups = input_shape[2] / weights_shape[2];
 
+        // Setup reference data types
+        const DataType src_dt     = _is_bfloat16 ? DataType::F32 : _data_type;
+        const DataType weights_dt = _is_bfloat16 ? DataType::F32 : _weights_data_type;
+        const DataType bias_dt    = _is_bfloat16 ? DataType::F32 : _bias_data_type;
+
         // Create reference
-        SimpleTensor<T>     src{ input_shape, _data_type, 1, _quantization_info };
-        SimpleTensor<TW>    weights{ weights_shape, _weights_data_type, 1, _weight_quantization_info };
-        SimpleTensor<TBias> bias{ bias_shape, _bias_data_type, 1, _quantization_info };
+        SimpleTensor<T>     src{ input_shape, src_dt, 1, _quantization_info };
+        SimpleTensor<TW>    weights{ weights_shape, weights_dt, 1, _weight_quantization_info };
+        SimpleTensor<TBias> bias{ bias_shape, bias_dt, 1, _quantization_info };
 
-        // Fill reference
         fill(src, 0);
         fill(weights, 1);
         fill(bias, 2);
 
+        // Fill with bfloat16 to perform the conversion and reduce the mismatches in the output
+        if(_is_bfloat16)
+        {
+            regularize_values(static_cast<void *>(src.data()), src.num_elements());
+            regularize_values(static_cast<void *>(weights.data()), weights.num_elements());
+        }
+
         return (act_info.enabled()) ? reference::activation_layer<T>(reference::convolution_layer<T>(src, weights, bias, output_shape, info, dilation, num_groups),
                                                                      act_info) :
                reference::convolution_layer<T>(src, weights, bias, output_shape, info, dilation, num_groups);
@@ -215,10 +238,12 @@ protected:
     DataType         _data_type{};
     DataType         _weights_data_type{};
     DataType         _bias_data_type{};
+    DataType         _output_data_type{};
     DataLayout       _data_layout{};
     QuantizationInfo _quantization_info{};
     QuantizationInfo _weight_quantization_info{};
     bool             _is_quantized = false;
+    bool             _is_bfloat16  = false;
 };
 
 template <typename TensorType, typename AccessorType, typename FunctionType, typename T>
-- 
cgit v1.2.1