include/ck/tensor_operation/gpu/device/device_batched_gemm.hpp Source File

include/ck/tensor_operation/gpu/device/device_batched_gemm.hpp Source File#

Composable Kernel: include/ck/tensor_operation/gpu/device/device_batched_gemm.hpp Source File
device_batched_gemm.hpp
Go to the documentation of this file.
1 // SPDX-License-Identifier: MIT
2 // Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
3 
4 #pragma once
5 
6 #include <iostream>
7 #include <vector>
8 
9 #include "device_base.hpp"
10 
11 namespace ck {
12 namespace tensor_operation {
13 namespace device {
14 
15 template <typename ALayout,
16  typename BLayout,
17  typename CLayout,
18  typename ADataType,
19  typename BDataType,
20  typename CDataType,
21  typename AElementwiseOperation,
22  typename BElementwiseOperation,
23  typename CElementwiseOperation>
24 struct DeviceBatchedGemm : public BaseOperator
25 {
26  virtual std::unique_ptr<BaseArgument>
27  MakeArgumentPointer(const void* p_a,
28  const void* p_b,
29  void* p_c,
30  ck::index_t M,
31  ck::index_t N,
32  ck::index_t K,
33  ck::index_t StrideA,
34  ck::index_t StrideB,
35  ck::index_t StrideC,
36  ck::index_t BatchStrideA,
37  ck::index_t BatchStrideB,
38  ck::index_t BatchStrideC,
39  ck::index_t Batch,
40  AElementwiseOperation a_element_op,
41  BElementwiseOperation b_element_op,
42  CElementwiseOperation c_element_op) = 0;
43 
44  virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
45 };
46 
47 template <typename ALayout,
48  typename BLayout,
49  typename CLayout,
50  typename ADataType,
51  typename BDataType,
52  typename CDataType,
53  typename AElementwiseOperation,
54  typename BElementwiseOperation,
55  typename CElementwiseOperation>
56 using DeviceBatchedGemmPtr = std::unique_ptr<DeviceBatchedGemm<ALayout,
57  BLayout,
58  CLayout,
59  ADataType,
60  BDataType,
61  CDataType,
62  AElementwiseOperation,
63  BElementwiseOperation,
64  CElementwiseOperation>>;
65 
66 } // namespace device
67 } // namespace tensor_operation
68 } // namespace ck
device_base.hpp
ck::tensor_operation::device::DeviceBatchedGemmPtr
std::unique_ptr< DeviceBatchedGemm< ALayout, BLayout, CLayout, ADataType, BDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation > > DeviceBatchedGemmPtr
Definition: device_batched_gemm.hpp:64
ck
Definition: ck.hpp:264
ck::index_t
int32_t index_t
Definition: ck.hpp:289
ck::tensor_operation::device::BaseOperator
Definition: device_base.hpp:76
ck::tensor_operation::device::DeviceBatchedGemm
Definition: device_batched_gemm.hpp:25
ck::tensor_operation::device::DeviceBatchedGemm::MakeArgumentPointer
virtual std::unique_ptr< BaseArgument > MakeArgumentPointer(const void *p_a, const void *p_b, void *p_c, ck::index_t M, ck::index_t N, ck::index_t K, ck::index_t StrideA, ck::index_t StrideB, ck::index_t StrideC, ck::index_t BatchStrideA, ck::index_t BatchStrideB, ck::index_t BatchStrideC, ck::index_t Batch, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op)=0
ck::tensor_operation::device::DeviceBatchedGemm::MakeInvokerPointer
virtual std::unique_ptr< BaseInvoker > MakeInvokerPointer()=0