Merge remote-tracking branch 'origin/develop' into batched_gemm_c_permute

include/ck/tensor_operation/gpu/thread/reduction_functions_threadwise.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/reduction_functions_accumulate.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_contraction_dl.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_gemm_dlops_v3.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_THREADWISE_GEMM_DLOPS_V3_HPP
 #define CK_THREADWISE_GEMM_DLOPS_V3_HPP
 

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_set.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"
@@ -233,9 +236,14 @@ template <typename SrcData,
           index_t SrcScalarPerVector,
           index_t SrcScalarStrideInVector,
           bool SrcResetCoordinateAfterRun,
+          bool InvalidElementAsNaN                                        = false,
           typename enable_if<DstDesc::IsKnownAtCompileTime(), bool>::type = false>
 struct ThreadwiseTensorSliceTransfer_v2
 {
+    static_assert((InvalidElementAsNaN && !std::is_integral<DstData>::value) ||
+                      (!InvalidElementAsNaN),
+                  "Filling invalid element as NaN is only for floating point types");
+
     static constexpr index_t nDim = SliceLengths::Size();
 
     using Index = MultiIndex<nDim>;
@@ -315,8 +323,18 @@ struct ThreadwiseTensorSliceTransfer_v2
                     dst_desc.CalculateOffset(to_multi_index(dst_slice_origin_idx) + src_data_idx +
                                              i * src_scalar_step_in_vector);
 
-                dst_buf(Number<dst_offset>{}) =
-                    type_convert<DstData>(src_vector.template AsType<SrcData>()[i]);
+                if constexpr(InvalidElementAsNaN)
+                {
+                    dst_buf(Number<dst_offset>{}) =
+                        is_src_valid
+                            ? type_convert<DstData>(src_vector.template AsType<SrcData>()[i])
+                            : NumericLimits<DstData>::QuietNaN();
+                }
+                else
+                {
+                    dst_buf(Number<dst_offset>{}) =
+                        type_convert<DstData>(src_vector.template AsType<SrcData>()[i]);
+                }
             });
 
             if constexpr(idx_1d.value != num_access - 1)

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v3r1.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v3r3.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_THREADWISE_TENSOR_SLICE_TRANSFER_V3R3_HPP
 #define CK_THREADWISE_TENSOR_SLICE_TRANSFER_V3R3_HPP
 

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v4r1.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v5r1.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v6r1.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v6r2.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v6r3.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v7.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/tensor_operation/gpu/warp/xdlops_gemm.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/utility/common_header.hpp"

include/ck/utility/amd_address_space.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 
 #include "ck/ck.hpp"

include/ck/utility/amd_buffer_addressing.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 #include "data_type.hpp"
 

include/ck/utility/amd_inline_asm.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_AMD_INLINE_ASM_HPP
 #define CK_AMD_INLINE_ASM_HPP
 

include/ck/utility/amd_llvm_intrinsic.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_AMD_LLVM_INTRINSIC_HPP
 #define CK_AMD_LLVM_INTRINSIC_HPP
 

include/ck/utility/amd_xdlops.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_AMD_XDLOPS_HPP
 #define CK_AMD_XDLOPS_HPP
 

include/ck/utility/array.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef CK_ARRAY_HPP
 #define CK_ARRAY_HPP