Clean-up the headers (#713)

* fix headers for gpu instances

* remove unused headers

---------
Co-authored-by: zjing14 <zhangjing14@gmail.com>

include/ck/utility/print.hpp

-// SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
-
-#ifndef CK_PRINT_HPP
-#define CK_PRINT_HPP
-
-#include "array.hpp"
-#include "statically_indexed_array.hpp"
-#include "container_helper.hpp"
-#include "sequence.hpp"
-
-namespace ck {
-
-template <typename T>
-__host__ __device__ void print_array(const char* s, T a)
-{
-    constexpr index_t nsize = a.Size();
-
-    printf("%s size %d, {", s, nsize);
-    static_for<0, nsize, 1>{}([&a](auto i) constexpr { printf("%d, ", int32_t{a[i]}); });
-    printf("}\n");
-}
-
-} // namespace ck
-#endif

library/include/ck/library/reference_tensor_operation/cpu/reference_gemm_bias_2d.hpp

-// SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
-
-#pragma once
-
-#include <iostream>
-#include <sstream>
-
-#include "ck/tensor_operation/gpu/device/device_base.hpp"
-#include "ck/library/utility/host_tensor.hpp"
-
-namespace ck {
-namespace tensor_operation {
-namespace host {
-
-template <typename ADataType,
-          typename BDataType,
-          typename C0DataType,
-          typename CDataType,
-          typename AccDataType,
-          typename AElementwiseOperation,
-          typename BElementwiseOperation,
-          typename CElementwiseOperation>
-struct ReferenceGemmBias2D : public device::BaseOperator
-{
-    // Argument
-    struct Argument : public device::BaseArgument
-    {
-        Argument(const Tensor<ADataType>& a_m_k,
-                 const Tensor<BDataType>& b_k_n,
-                 const Tensor<C0DataType>& c0_m_n,
-                 Tensor<CDataType>& c_m_n,
-                 AElementwiseOperation a_element_op,
-                 BElementwiseOperation b_element_op,
-                 CElementwiseOperation c_element_op)
-            : a_m_k_{a_m_k},
-              b_k_n_{b_k_n},
-              c0_m_n_{c0_m_n},
-              c_m_n_{c_m_n},
-              a_element_op_{a_element_op},
-              b_element_op_{b_element_op},
-              c_element_op_{c_element_op}
-        {
-        }
-
-        const Tensor<ADataType>& a_m_k_;
-        const Tensor<BDataType>& b_k_n_;
-        const Tensor<CDataType>& c0_m_n_;
-        Tensor<CDataType>& c_m_n_;
-
-        AElementwiseOperation a_element_op_;
-        BElementwiseOperation b_element_op_;
-        CElementwiseOperation c_element_op_;
-    };
-
-    // Invoker
-    struct Invoker : public device::BaseInvoker
-    {
-        using Argument = ReferenceGemmBias2D::Argument;
-
-        float Run(const Argument& arg)
-        {
-            auto f_mk_kn_mn = [&](auto m, auto n) {
-                const int K = arg.a_m_k_.mDesc.GetLengths()[1];
-
-                AccDataType a   = 0;
-                AccDataType b   = 0;
-                AccDataType acc = 0;
-
-                for(int k = 0; k < K; ++k)
-                {
-                    arg.a_element_op_(a, ck::type_convert<AccDataType>(arg.a_m_k_(m, k)));
-                    arg.b_element_op_(b, ck::type_convert<AccDataType>(arg.b_k_n_(k, n)));
-                    acc += a * b;
-                }
-
-                CDataType cast_acc = static_cast<CDataType>(acc);
-                arg.c_element_op_(arg.c_m_n_(m, n), cast_acc, arg.c0_m_n_(m, n));
-            };
-
-            make_ParallelTensorFunctor(
-                f_mk_kn_mn, arg.c_m_n_.mDesc.GetLengths()[0], arg.c_m_n_.mDesc.GetLengths()[1])(
-                std::thread::hardware_concurrency());
-
-            return 0;
-        }
-
-        float Run(const device::BaseArgument* p_arg,
-                  const StreamConfig& /* stream_config */ = StreamConfig{}) override
-        {
-            return Run(*dynamic_cast<const Argument*>(p_arg));
-        }
-    };
-
-    static constexpr bool IsValidCompilationParameter()
-    {
-        // TODO: properly implement this check
-        return true;
-    }
-
-    bool IsSupportedArgument(const device::BaseArgument*) override { return true; }
-
-    static auto MakeArgument(const Tensor<ADataType>& a_m_k,
-                             const Tensor<BDataType>& b_k_n,
-                             const Tensor<C0DataType>& c0_m_n,
-                             Tensor<CDataType>& c_m_n,
-                             AElementwiseOperation a_element_op,
-                             BElementwiseOperation b_element_op,
-                             CElementwiseOperation c_element_op)
-    {
-        return Argument{a_m_k, b_k_n, c0_m_n, c_m_n, a_element_op, b_element_op, c_element_op};
-    }
-
-    static auto MakeInvoker() { return Invoker{}; }
-
-    virtual std::unique_ptr<device::BaseInvoker> MakeInvokerPointer()
-    {
-        return std::make_unique<Invoker>(Invoker{});
-    }
-
-    std::string GetTypeString() const override
-    {
-        auto str = std::stringstream();
-
-        // clang-format off
-        str << "ReferenceGemmBias2D"
-            << std::endl;
-        // clang-format on
-
-        return str.str();
-    }
-};
-
-} // namespace host
-} // namespace tensor_operation
-} // namespace ck

library/include/ck/library/reference_tensor_operation/cpu/reference_gemm_bias_activation.hpp

-// SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
-
-#pragma once
-
-#include <iostream>
-#include <sstream>
-
-#include "ck/tensor_operation/gpu/device/device_base.hpp"
-
-#include "ck/library/utility/host_tensor.hpp"
-
-namespace ck {
-namespace tensor_operation {
-namespace host {
-
-template <typename ADataType,
-          typename BDataType,
-          typename CDataType,
-          typename AElementwiseOperation,
-          typename BElementwiseOperation,
-          typename CElementwiseOperation>
-struct ReferenceGemmBiasActivation : public device::BaseOperator
-{
-    // Argument
-    struct Argument : public device::BaseArgument
-    {
-        Argument(const Tensor<ADataType>& a_m_k,
-                 const Tensor<BDataType>& b_k_n,
-                 Tensor<CDataType>& c_m_n,
-                 const Tensor<CDataType>& c0_n,
-                 AElementwiseOperation a_element_op,
-                 BElementwiseOperation b_element_op,
-                 CElementwiseOperation c_element_op)
-            : a_m_k_{a_m_k},
-              b_k_n_{b_k_n},
-              c_m_n_{c_m_n},
-              c0_n_{c0_n},
-              a_element_op_{a_element_op},
-              b_element_op_{b_element_op},
-              c_element_op_{c_element_op}
-        {
-        }
-
-        const Tensor<ADataType>& a_m_k_;
-        const Tensor<BDataType>& b_k_n_;
-        Tensor<CDataType>& c_m_n_;
-        const Tensor<CDataType>& c0_n_;
-
-        AElementwiseOperation a_element_op_;
-        BElementwiseOperation b_element_op_;
-        CElementwiseOperation c_element_op_;
-    };
-
-    // Invoker
-    struct Invoker : public device::BaseInvoker
-    {
-        using Argument = ReferenceGemmBiasActivation::Argument;
-
-        float Run(const Argument& arg)
-        {
-            auto f_mk_kn_mn = [&](auto m, auto n) {
-                const int K = arg.a_m_k_.mDesc.GetLengths()[1];
-
-                float v_acc = 0;
-
-                for(int k = 0; k < K; ++k)
-                {
-                    float v_a;
-                    float v_b;
-
-                    arg.a_element_op_(v_a, static_cast<const float>(arg.a_m_k_(m, k)));
-                    arg.b_element_op_(v_b, static_cast<const float>(arg.b_k_n_(k, n)));
-
-                    v_acc += v_a * v_b;
-                }
-
-                float v_c;
-
-                arg.c_element_op_(v_c, v_acc, static_cast<float>(arg.c0_n_(n)));
-
-                arg.c_m_n_(m, n) = v_c;
-            };
-
-            make_ParallelTensorFunctor(
-                f_mk_kn_mn, arg.c_m_n_.mDesc.GetLengths()[0], arg.c_m_n_.mDesc.GetLengths()[1])(
-                std::thread::hardware_concurrency());
-
-            return 0;
-        }
-
-        float Run(const device::BaseArgument* p_arg,
-                  const StreamConfig& /* stream_config */ = StreamConfig{}) override
-        {
-            return Run(*dynamic_cast<const Argument*>(p_arg));
-        }
-    };
-
-    static constexpr bool IsValidCompilationParameter()
-    {
-        // TODO: properly implement this check
-        return true;
-    }
-
-    bool IsSupportedArgument(const device::BaseArgument*) override { return true; }
-
-    static auto MakeArgument(const Tensor<ADataType>& a_m_k,
-                             const Tensor<BDataType>& b_k_n,
-                             Tensor<CDataType>& c_m_n,
-                             const Tensor<CDataType>& c0_n,
-                             AElementwiseOperation a_element_op,
-                             BElementwiseOperation b_element_op,
-                             CElementwiseOperation c_element_op)
-    {
-        return Argument{a_m_k, b_k_n, c_m_n, c0_n, a_element_op, b_element_op, c_element_op};
-    }
-
-    static auto MakeInvoker() { return Invoker{}; }
-
-    virtual std::unique_ptr<device::BaseInvoker> MakeInvokerPointer()
-    {
-        return std::make_unique<Invoker>(Invoker{});
-    }
-
-    std::string GetTypeString() const override
-    {
-        auto str = std::stringstream();
-
-        // clang-format off
-        str << "ReferenceGemmBiasActivation"
-            << std::endl;
-        // clang-format on
-
-        return str.str();
-    }
-};
-
-} // namespace host
-} // namespace tensor_operation
-} // namespace ck

library/include/ck/library/reference_tensor_operation/cpu/reference_gemm_bias_activation_add.hpp

-// SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
-
-#pragma once
-
-#include <iostream>
-#include <sstream>
-
-#include "ck/tensor_operation/gpu/device/device_base.hpp"
-
-#include "ck/library/utility/host_tensor.hpp"
-
-namespace ck {
-namespace tensor_operation {
-namespace host {
-
-template <typename ADataType,
-          typename BDataType,
-          typename CDataType,
-          typename AElementwiseOperation,
-          typename BElementwiseOperation,
-          typename CElementwiseOperation>
-struct ReferenceGemmBiasActivationAdd : public device::BaseOperator
-{
-    // Argument
-    struct Argument : public device::BaseArgument
-    {
-        Argument(const Tensor<ADataType>& a_m_k,
-                 const Tensor<BDataType>& b_k_n,
-                 Tensor<CDataType>& c_m_n,
-                 const Tensor<CDataType>& c0_n,
-                 const Tensor<CDataType>& c1_m_n,
-                 AElementwiseOperation a_element_op,
-                 BElementwiseOperation b_element_op,
-                 CElementwiseOperation c_element_op)
-            : a_m_k_{a_m_k},
-              b_k_n_{b_k_n},
-              c_m_n_{c_m_n},
-              c0_n_{c0_n},
-              c1_m_n_{c1_m_n},
-              a_element_op_{a_element_op},
-              b_element_op_{b_element_op},
-              c_element_op_{c_element_op}
-        {
-        }
-
-        const Tensor<ADataType>& a_m_k_;
-        const Tensor<BDataType>& b_k_n_;
-        Tensor<CDataType>& c_m_n_;
-        const Tensor<CDataType>& c0_n_;
-        const Tensor<CDataType>& c1_m_n_;
-
-        AElementwiseOperation a_element_op_;
-        BElementwiseOperation b_element_op_;
-        CElementwiseOperation c_element_op_;
-    };
-
-    // Invoker
-    struct Invoker : public device::BaseInvoker
-    {
-        using Argument = ReferenceGemmBiasActivationAdd::Argument;
-
-        float Run(const Argument& arg)
-        {
-            auto f_mk_kn_mn = [&](auto m, auto n) {
-                const int K = arg.a_m_k_.mDesc.GetLengths()[1];
-
-                float v_acc = 0;
-
-                for(int k = 0; k < K; ++k)
-                {
-                    float v_a;
-                    float v_b;
-
-                    arg.a_element_op_(v_a, static_cast<const float>(arg.a_m_k_(m, k)));
-                    arg.b_element_op_(v_b, static_cast<const float>(arg.b_k_n_(k, n)));
-
-                    v_acc += v_a * v_b;
-                }
-
-                float v_c;
-
-                arg.c_element_op_(v_c,
-                                  v_acc,
-                                  static_cast<float>(arg.c0_n_(n)),
-                                  static_cast<float>(arg.c1_m_n_(m, n)));
-
-                arg.c_m_n_(m, n) = v_c;
-            };
-
-            make_ParallelTensorFunctor(
-                f_mk_kn_mn, arg.c_m_n_.mDesc.GetLengths()[0], arg.c_m_n_.mDesc.GetLengths()[1])(
-                std::thread::hardware_concurrency());
-
-            return 0;
-        }
-
-        float Run(const device::BaseArgument* p_arg,
-                  const StreamConfig& /* stream_config */ = StreamConfig{}) override
-        {
-            return Run(*dynamic_cast<const Argument*>(p_arg));
-        }
-    };
-
-    static constexpr bool IsValidCompilationParameter()
-    {
-        // TODO: properly implement this check
-        return true;
-    }
-
-    bool IsSupportedArgument(const device::BaseArgument*) override { return true; }
-
-    static auto MakeArgument(const Tensor<ADataType>& a_m_k,
-                             const Tensor<BDataType>& b_k_n,
-                             Tensor<CDataType>& c_m_n,
-                             const Tensor<CDataType>& c0_n,
-                             const Tensor<CDataType>& c1_m_n,
-                             AElementwiseOperation a_element_op,
-                             BElementwiseOperation b_element_op,
-                             CElementwiseOperation c_element_op)
-    {
-        return Argument{
-            a_m_k, b_k_n, c_m_n, c0_n, c1_m_n, a_element_op, b_element_op, c_element_op};
-    }
-
-    static auto MakeInvoker() { return Invoker{}; }
-
-    virtual std::unique_ptr<device::BaseInvoker> MakeInvokerPointer()
-    {
-        return std::make_unique<Invoker>(Invoker{});
-    }
-
-    std::string GetTypeString() const override
-    {
-        auto str = std::stringstream();
-
-        // clang-format off
-        str << "ReferenceGemmBiasActivationAdd"
-            << std::endl;
-        // clang-format on
-
-        return str.str();
-    }
-};
-
-} // namespace host
-} // namespace tensor_operation
-} // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_batched_gemm.hpp"

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_add_relu_gemm_add.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_batched_gemm_multiple_d_gemm_multiple_d.hpp"

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_bias_softmax_gemm_permute.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_batched_gemm_softmax_gemm_permute.hpp"

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_gemm.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_batched_gemm_gemm.hpp"

library/include/ck/library/tensor_operation_instance/gpu/batched_gemm_softmax_gemm_permute.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_batched_gemm_softmax_gemm_permute.hpp"

library/include/ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp

@@ -3,10 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
 #include <vector>
 #include <memory>
-
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"

library/include/ck/library/tensor_operation_instance/gpu/contraction_scale.hpp

@@ -3,10 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
 #include <vector>
 #include <memory>
-
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_contraction_multiple_d.hpp"

library/include/ck/library/tensor_operation_instance/gpu/convolution_backward_data.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_conv_bwd_data.hpp"

library/include/ck/library/tensor_operation_instance/gpu/convolution_forward.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_conv_fwd.hpp"

library/include/ck/library/tensor_operation_instance/gpu/device_elementwise_instance.hpp

@@ -3,8 +3,7 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_elementwise_impl.hpp"

library/include/ck/library/tensor_operation_instance/gpu/device_gemm_mean_squaremean_instance.hpp

@@ -4,7 +4,7 @@
 #pragma once
 
 #include <cstdlib>
-
+#include <vector>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_reduce.hpp"

library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp

@@ -3,10 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
 #include <memory>
 #include <vector>
-
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm.hpp"

library/include/ck/library/tensor_operation_instance/gpu/gemm_add_add_fastgelu.hpp

@@ -3,10 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
 #include <vector>
 #include <memory>
-
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d.hpp"

library/include/ck/library/tensor_operation_instance/gpu/gemm_bilinear.hpp

@@ -3,10 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
 #include <vector>
 #include <memory>
-
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_multiple_d.hpp"

library/include/ck/library/tensor_operation_instance/gpu/gemm_splitk.hpp

@@ -3,8 +3,8 @@
 
 #pragma once
 
-#include <cstdlib>
-
+#include <vector>
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_gemm_splitk.hpp"

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp

@@ -4,7 +4,7 @@
 #pragma once
 
 #include <vector>
-
+#include <memory>
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"