From 31441595009182c985dacbedc70c41ee6664d070 Mon Sep 17 00:00:00 2001
From: Ryan OShea <ryan.oshea3@arm.com>
Date: Mon, 7 Nov 2022 16:20:48 +0000
Subject: IVGCVSW-7214 Disable BF16-Turbo-Mode and remove conversion layers

 - Remove Bf16ToFp32 Conversion Layer
 - Remove Fp32ToBf16 Conversion Layer
 - Remove B16 Conversion tests
 * Throw exception if m_ReduceFp32ToBf16 optimzer option is set to true
 * Provide comments to enable fast math in order to use bf16
 * Update docs to inform users to enable fast math for bf16

 Execute Network Changes
 * Require bf16_turbo_mode to also have fast_math_enabled set to true
 - Remove setting m_ReduceFp32ToBf16 optimizer option

Signed-off-by: Ryan OShea <ryan.oshea3@arm.com>
Change-Id: Ibaa6da9d29c96a1ce32ff5196b0847fde9f04a1c
---
 src/backends/backendsCommon/WorkloadFactory.cpp | 38 -------------------------
 1 file changed, 38 deletions(-)

(limited to 'src/backends/backendsCommon/WorkloadFactory.cpp')

diff --git a/src/backends/backendsCommon/WorkloadFactory.cpp b/src/backends/backendsCommon/WorkloadFactory.cpp
index 665ab3f86c..1283f67660 100644
--- a/src/backends/backendsCommon/WorkloadFactory.cpp
+++ b/src/backends/backendsCommon/WorkloadFactory.cpp
@@ -227,13 +227,6 @@ bool IWorkloadFactory::IsLayerConfigurationSupported(const BackendId& backendId,
             result = layerSupportObject.IsConstantSupported(OverrideDataType(output, dataType), reason);
             break;
         }
-        case LayerType::ConvertBf16ToFp32:
-        {
-            const TensorInfo& input = layer.GetInputSlot(0).GetConnection()->GetTensorInfo();
-            const TensorInfo& output = layer.GetOutputSlot(0).GetTensorInfo();
-            result = layerSupportObject.IsConvertBf16ToFp32Supported(input, output, reason);
-            break;
-        }
         case LayerType::ConvertFp16ToFp32:
         {
             const TensorInfo& input = layer.GetInputSlot(0).GetConnection()->GetTensorInfo();
@@ -241,13 +234,6 @@ bool IWorkloadFactory::IsLayerConfigurationSupported(const BackendId& backendId,
             result = layerSupportObject.IsConvertFp16ToFp32Supported(input, output, reason);
             break;
         }
-        case LayerType::ConvertFp32ToBf16:
-        {
-            const TensorInfo& input = layer.GetInputSlot(0).GetConnection()->GetTensorInfo();
-            const TensorInfo& output = layer.GetOutputSlot(0).GetTensorInfo();
-            result = layerSupportObject.IsConvertFp32ToBf16Supported(input, output, reason);
-            break;
-        }
         case LayerType::ConvertFp32ToFp16:
         {
             const TensorInfo& input = layer.GetInputSlot(0).GetConnection()->GetTensorInfo();
@@ -1630,24 +1616,12 @@ std::unique_ptr<IWorkload> IWorkloadFactory::CreateWorkload(LayerType type,
             auto constantQueueDescriptor = PolymorphicDowncast<const ConstantQueueDescriptor*>(&descriptor);
             return CreateConstant(*constantQueueDescriptor, info);
         }
-        case LayerType::ConvertBf16ToFp32 :
-        {
-            auto convertBf16ToFp32QueueDescriptor
-                    = PolymorphicDowncast<const ConvertBf16ToFp32QueueDescriptor*>(&descriptor);
-            return CreateConvertBf16ToFp32(*convertBf16ToFp32QueueDescriptor, info);
-        }
         case LayerType::ConvertFp16ToFp32:
         {
             auto convertFp16ToFp32QueueDescriptor
                     = PolymorphicDowncast<const ConvertFp16ToFp32QueueDescriptor*>(&descriptor);
             return CreateConvertFp16ToFp32(*convertFp16ToFp32QueueDescriptor, info);
         }
-        case LayerType::ConvertFp32ToBf16:
-        {
-            auto convertFp32ToBf16QueueDescriptor
-                    = PolymorphicDowncast<const ConvertFp32ToBf16QueueDescriptor*>(&descriptor);
-            return CreateConvertFp32ToBf16(*convertFp32ToBf16QueueDescriptor, info);
-        }
         case LayerType::ConvertFp32ToFp16:
         {
             auto convertFp32ToFp16QueueDescriptor
@@ -1992,24 +1966,12 @@ std::unique_ptr<IWorkload> IWorkloadFactory::CreateConstant(const ConstantQueueD
     return std::unique_ptr<IWorkload>();
 }
 
-std::unique_ptr<IWorkload> IWorkloadFactory::CreateConvertBf16ToFp32(const ConvertBf16ToFp32QueueDescriptor& /*desc*/,
-                                                                     const WorkloadInfo& /*info*/) const
-{
-    return std::unique_ptr<IWorkload>();
-}
-
 std::unique_ptr<IWorkload> IWorkloadFactory::CreateConvertFp16ToFp32(const ConvertFp16ToFp32QueueDescriptor& /*desc*/,
                                                                      const WorkloadInfo& /*info*/) const
 {
     return std::unique_ptr<IWorkload>();
 }
 
-std::unique_ptr<IWorkload> IWorkloadFactory::CreateConvertFp32ToBf16(const ConvertFp32ToBf16QueueDescriptor& /*desc*/,
-                                                                     const WorkloadInfo& /*info*/) const
-{
-    return std::unique_ptr<IWorkload>();
-}
-
 std::unique_ptr<IWorkload> IWorkloadFactory::CreateConvertFp32ToFp16(const ConvertFp32ToFp16QueueDescriptor& /*desc*/,
                                                                      const WorkloadInfo& /*info*/) const
 {
-- 
cgit v1.2.1