Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "167b6c69229873a6928ff90b2eafd2e37e727d56"
Commit 06d2c7b1 authored by Jing Zhang's avatar Jing Zhang Committed by root
Browse files

clean

parents b27909a0 3b18f1e3
Hide whitespace changes
Inline Side-by-side
Showing with 81 additions and 80 deletions
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_add.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_amax.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_avg.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_max.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_min.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_f64_f64_f64_norm2.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_i8_i32_i8_add.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_i8_i32_i8_avg.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_i8_i8_i8_amax.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_i8_i8_i8_max.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/device_reduce_instance_threadwise_i8_i8_i8_min.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/reduce/reduce.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......
library/include/ck/library/tensor_operation_instance/gpu/softmax.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -9,34 +9,33 @@
#include "ck/ck.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
#include "ck/tensor_operation/gpu/device/device_softmax.hpp"
#include "ck/library/tensor_operation_instance/gpu/softmax/device_softmax_instance.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank3_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3>>&);
void add_device_softmax_f16_f16_rank4_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4>>&);
void add_device_softmax_f32_f32_rank3_instances(
std::vector<DeviceSoftmaxPtr<F32, F32, F32, PassThrough, PassThrough, 3>>&);
void add_device_softmax_f32_f32_rank4_instances(
std::vector<DeviceSoftmaxPtr<F32, F32, F32, PassThrough, PassThrough, 4>>&);
void add_device_softmax_i8_i8_rank3_instances(
std::vector<DeviceSoftmaxPtr<I8, F32, I8, PassThrough, PassThrough, 3>>&);
void add_device_softmax_i8_i8_rank4_instances(
std::vector<DeviceSoftmaxPtr<I8, F32, I8, PassThrough, PassThrough, 4>>&);
template <typename InDataType, typename AccDataType, typename OutDataType, index_t Rank>
struct DeviceOperationInstanceFactory<
ck::tensor_operation::device::
DeviceSoftmax<InDataType, AccDataType, OutDataType, PassThrough, PassThrough, Rank>>
template <typename InDataType,
typename AccDataType,
typename OutDataType,
index_t Rank,
index_t NumReduceDim>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceSoftmax<InDataType,
AccDataType,
OutDataType,
PassThrough,
PassThrough,
Rank,
NumReduceDim>>
{
using DeviceOp =
DeviceSoftmax<InDataType, AccDataType, OutDataType, PassThrough, PassThrough, Rank>;
using DeviceOp = DeviceSoftmax<InDataType,
AccDataType,
OutDataType,
PassThrough,
PassThrough,
Rank,
NumReduceDim>;
static auto GetInstances()
{
......@@ -46,25 +45,49 @@ struct DeviceOperationInstanceFactory<
std::is_same_v<OutDataType, F16>)
{
if constexpr(Rank == 3)
add_device_softmax_f16_f16_rank3_instances(op_ptrs);
{
if constexpr(NumReduceDim == 1)
add_device_softmax_f16_f16_rank3_reduce1_instances(op_ptrs);
else if constexpr(NumReduceDim == 2)
add_device_softmax_f16_f16_rank3_reduce2_instances(op_ptrs);
else if constexpr(NumReduceDim == 3)
add_device_softmax_f16_f16_rank3_reduce3_instances(op_ptrs);
}
else if constexpr(Rank == 4)
add_device_softmax_f16_f16_rank4_instances(op_ptrs);
{
if constexpr(NumReduceDim == 1)
add_device_softmax_f16_f16_rank4_reduce1_instances(op_ptrs);
else if constexpr(NumReduceDim == 2)
add_device_softmax_f16_f16_rank4_reduce2_instances(op_ptrs);
else if constexpr(NumReduceDim == 3)
add_device_softmax_f16_f16_rank4_reduce3_instances(op_ptrs);
else if constexpr(NumReduceDim == 4)
add_device_softmax_f16_f16_rank4_reduce4_instances(op_ptrs);
}
}
else if constexpr(std::is_same_v<InDataType, F32> && std::is_same_v<AccDataType, F32> &&
std::is_same_v<OutDataType, F32>)
{
if constexpr(Rank == 3)
add_device_softmax_f32_f32_rank3_instances(op_ptrs);
else if constexpr(Rank == 4)
add_device_softmax_f32_f32_rank4_instances(op_ptrs);
}
else if constexpr(std::is_same_v<InDataType, I8> && std::is_same_v<AccDataType, F32> &&
std::is_same_v<OutDataType, I8>)
{
if constexpr(Rank == 3)
add_device_softmax_i8_i8_rank3_instances(op_ptrs);
{
if constexpr(NumReduceDim == 1)
add_device_softmax_f32_f32_rank3_reduce1_instances(op_ptrs);
else if constexpr(NumReduceDim == 2)
add_device_softmax_f32_f32_rank3_reduce2_instances(op_ptrs);
else if constexpr(NumReduceDim == 3)
add_device_softmax_f32_f32_rank3_reduce3_instances(op_ptrs);
}
else if constexpr(Rank == 4)
add_device_softmax_i8_i8_rank4_instances(op_ptrs);
{
if constexpr(NumReduceDim == 1)
add_device_softmax_f32_f32_rank4_reduce1_instances(op_ptrs);
else if constexpr(NumReduceDim == 2)
add_device_softmax_f32_f32_rank4_reduce2_instances(op_ptrs);
else if constexpr(NumReduceDim == 3)
add_device_softmax_f32_f32_rank4_reduce3_instances(op_ptrs);
else if constexpr(NumReduceDim == 4)
add_device_softmax_f32_f32_rank4_reduce4_instances(op_ptrs);
}
}
return op_ptrs;
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance.hpp deleted 100644 → 0
View file @ b27909a0
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
#include "ck/tensor_operation/gpu/device/device_softmax.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank3_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3>>& instances);
void add_device_softmax_f16_f16_rank4_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4>>& instances);
} // namespace instance
} // namespace device
} // namespace tensor_operation
} // namespace ck
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank3_reduce1.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank3_reduce1_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3, 1>>& instances);
} // namespace instance
} // namespace device
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank3_reduce2.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank3_reduce2_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3, 2>>& instances);
} // namespace instance
} // namespace device
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank3_reduce3.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank3_reduce3_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 3, 3>>& instances);
} // namespace instance
} // namespace device
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank4_reduce1.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank4_reduce1_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4, 1>>& instances);
} // namespace instance
} // namespace device
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank4_reduce2.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank4_reduce2_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4, 2>>& instances);
} // namespace instance
} // namespace device
......
library/include/ck/library/tensor_operation_instance/gpu/softmax/device_softmax_f16_f16_instance_rank4_reduce3.hpp
View file @ 06d2c7b1
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
......@@ -14,7 +14,7 @@ namespace device {
namespace instance {
void add_device_softmax_f16_f16_rank4_reduce3_instances(
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4>>& instances);
std::vector<DeviceSoftmaxPtr<F16, F32, F16, PassThrough, PassThrough, 4, 3>>& instances);
} // namespace instance
} // namespace device
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment