1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
|
// SPDX-FileCopyrightText: Copyright (c) Ken Martin, Will Schroeder, Bill Lorensen
// SPDX-License-Identifier: BSD-3-Clause
/**
* @class vtkDIYDataExchanger
* @brief exchange data-object among ranks.
*
* vtkDIYDataExchanger is a utility to exchange data-objects across multiple
* ranks. The design is based on `MPI_Alltoall` enabling algorithms to exchange
* data-objects between each other. The implementation uses DIY.
*
* Note, current implementation only supports exchanging vtkDataSet and
* subclasses. That may change in the future.
*/
#ifndef vtkDIYDataExchanger_h
#define vtkDIYDataExchanger_h
#include "vtkObject.h"
#include "vtkParallelDIYModule.h" // for export macros
#include "vtkSmartPointer.h" // for vtkSmartPointer
#include <vector> // for std::vector
VTK_ABI_NAMESPACE_BEGIN
class vtkDataSet;
class vtkMultiProcessController;
class VTKPARALLELDIY_EXPORT vtkDIYDataExchanger : public vtkObject
{
public:
static vtkDIYDataExchanger* New();
vtkTypeMacro(vtkDIYDataExchanger, vtkObject);
void PrintSelf(ostream& os, vtkIndent indent) override;
///@{
/**
* Get/Set the controller to use. By default
* vtkMultiProcessController::GetGlobalController is used.
*/
void SetController(vtkMultiProcessController*);
vtkGetObjectMacro(Controller, vtkMultiProcessController);
///@}
/**
* Exchange data between all ranks in the process group defined by the
* `Controller`.
*
* Every rank builds a vector of datasets to send to other ranks. This is
* `sendBuffer`. `sendCounts` lets the current rank specify which dataset is
* targeted for which rank. For example, if an MPI group has 3 ranks and the
* current process wants to send 2 datasets to rank 0, none to rank 1, and 3
* datasets to rank 2, then the `sendBuffer` should be filled with 5 (2 + 3)
* datasets and `sendCounts` should be of size 3 filled with `[2, 0, 3]`. The
* first 2 datasets in sendBuffer will be sent to rank 0 while the following 3
* datasets will be send to rank 2.
*
* This is a collective operation which must be called on all ranks in the MPI
* group irrespective of whether any rank is sending or receiving data.
*
* The `recvBuffer` is filled with datasets received from other ranks.
* `recvCounts` is similar to `sendCounts` in that it helps determine how many
* datasets were received from each rank and their offset in the `recvBuffer`.
*
* @param[in] sendBuffer collection of datasets to exchange
* @param[in] sendCounts array of counts for number of datasets to send to
* each rank.
* @param[out] recvBuffer datasets received from other ranks
* @param[out] recvCounts array of counts for number of datasets received from
* each rank.
*
* @returns true on success, else false
*/
bool AllToAll(const std::vector<vtkSmartPointer<vtkDataSet>>& sendBuffer,
const std::vector<int>& sendCounts, std::vector<vtkSmartPointer<vtkDataSet>>& recvBuffer,
std::vector<int>& recvCounts);
protected:
vtkDIYDataExchanger();
~vtkDIYDataExchanger() override;
private:
vtkDIYDataExchanger(const vtkDIYDataExchanger&) = delete;
void operator=(const vtkDIYDataExchanger&) = delete;
vtkMultiProcessController* Controller;
};
VTK_ABI_NAMESPACE_END
#endif
|
