root / man / man3 / HPL_reduce.3 @ 1
Historique | Voir | Annoter | Télécharger (1,61 ko)
1 | 1 | equemene | .TH HPL_reduce 3 "September 10, 2008" "HPL 2.0" "HPL Library Functions" |
---|---|---|---|
2 | 1 | equemene | .SH NAME |
3 | 1 | equemene | HPL_reduce \- Reduce operation. |
4 | 1 | equemene | .SH SYNOPSIS |
5 | 1 | equemene | \fB\&#include "hpl.h"\fR |
6 | 1 | equemene | |
7 | 1 | equemene | \fB\&int\fR |
8 | 1 | equemene | \fB\&HPL_reduce(\fR |
9 | 1 | equemene | \fB\&void *\fR |
10 | 1 | equemene | \fI\&BUFFER\fR, |
11 | 1 | equemene | \fB\&const int\fR |
12 | 1 | equemene | \fI\&COUNT\fR, |
13 | 1 | equemene | \fB\&const HPL_T_TYPE\fR |
14 | 1 | equemene | \fI\&DTYPE\fR, |
15 | 1 | equemene | \fB\&const HPL_T_OP \fR |
16 | 1 | equemene | \fI\&OP\fR, |
17 | 1 | equemene | \fB\&const int\fR |
18 | 1 | equemene | \fI\&ROOT\fR, |
19 | 1 | equemene | \fB\&MPI_Comm\fR |
20 | 1 | equemene | \fI\&COMM\fR |
21 | 1 | equemene | \fB\&);\fR |
22 | 1 | equemene | .SH DESCRIPTION |
23 | 1 | equemene | \fB\&HPL_reduce\fR |
24 | 1 | equemene | performs a global reduce operation across all processes of |
25 | 1 | equemene | a group. Note that the input buffer is used as workarray and in all |
26 | 1 | equemene | processes but the accumulating process corrupting the original data. |
27 | 1 | equemene | .SH ARGUMENTS |
28 | 1 | equemene | .TP 8 |
29 | 1 | equemene | BUFFER (local input/output) void * |
30 | 1 | equemene | On entry, BUFFER points to the buffer to be reduced. On |
31 | 1 | equemene | exit, and in process of rank ROOT this array contains the |
32 | 1 | equemene | reduced data. This buffer is also used as workspace during |
33 | 1 | equemene | the operation in the other processes of the group. |
34 | 1 | equemene | .TP 8 |
35 | 1 | equemene | COUNT (global input) const int |
36 | 1 | equemene | On entry, COUNT indicates the number of entries in BUFFER. |
37 | 1 | equemene | COUNT must be at least zero. |
38 | 1 | equemene | .TP 8 |
39 | 1 | equemene | DTYPE (global input) const HPL_T_TYPE |
40 | 1 | equemene | On entry, DTYPE specifies the type of the buffers operands. |
41 | 1 | equemene | .TP 8 |
42 | 1 | equemene | OP (global input) const HPL_T_OP |
43 | 1 | equemene | On entry, OP is a pointer to the local combine function. |
44 | 1 | equemene | .TP 8 |
45 | 1 | equemene | ROOT (global input) const int |
46 | 1 | equemene | On entry, ROOT is the coordinate of the accumulating process. |
47 | 1 | equemene | .TP 8 |
48 | 1 | equemene | COMM (global/local input) MPI_Comm |
49 | 1 | equemene | The MPI communicator identifying the process collection. |
50 | 1 | equemene | .SH SEE ALSO |
51 | 1 | equemene | .BR HPL_broadcast \ (3), |
52 | 1 | equemene | .BR HPL_all_reduce \ (3), |
53 | 1 | equemene | .BR HPL_barrier \ (3), |
54 | 1 | equemene | .BR HPL_min \ (3), |
55 | 1 | equemene | .BR HPL_max \ (3), |
56 | 1 | equemene | .BR HPL_sum \ (3). |