<![CDATA[
<p>
This page demonstrate a parallel computing tool implemented in R.
You have to have a MPI environment and a Rmpi package to run this example.
It also show the comparison with other methods.
In general, not all computing problem can be easily parallelized such as MCMC,
and not all parallel computing is more efficient than series computing.
Parallel computing is also highly dependent on programing and algorithm.
<hr>
- <font color="#800000"><b>MPI -- Message Passing Interface</b></font>
With <a href="http://www.lam-mpi.org/" target="_blank">LAM/MPI</a> and
R package <a href="http://www.stats.uwo.ca/faculty/yu/Rmpi/" target="_blank">Rmpi</a>
by <a href="http://www.stats.uwo.ca/faculty/yu/" target="_blank">Dr. Hao Yu</a>.
- <font color="#800000"><b>Download</b></font>
For Mandrake Linux system:<br>
- For <b>LAM/MPI</b>, it requires
- lam-devel-6.5.9-2mdk (my mirror <a href="./source/Rmpi/lam-devel-6.5.9-2mdk.rpm">here</a>),
- lam-runtime-6.5.9-2mdk (my mirror <a href="./source/Rmpi/lam-runtime-6.5.9-2mdk.rpm">here</a>),
- lam-doc-6.5.9-2mdk (my mirror <a href="./source/Rmpi/lam-doc-6.5.9-2mdk.rpm">here</a>),
- liblam0-devel-6.5.9-2mdk (my mirror <a href="./source/Rmpi/liblam0-devel-6.5.9-2mdk.rpm">here</a>). <br>
- For <b>Rmpi</b>, it requires Rmpi_0.4-6.tar.gz (my mirror <a href="./source/Rmpi/Rmpi_0.4-6.tar.gz">here</a>).
- <font color="#800000"><b>Master & Salve</b></font>
The basic idea is <b>Master</b> for MPI rank 0
and <b>Slave</b> for MPI rank from 1 to n, where n is the universal size in your MPI enviroments.<br>
<table>
<tr>
<th valign="top"> Step:
<th align="left">
0. Initial. <br>
1. Master send to Slave. (bcast, send) <br>
2. Slave receive from Master. (bcast, recv) <br>
3. compute. <br>
4. Slave send to Master. (send) <br>
5. Master receive from Slave. (recv) <br>
6. complete and quit.
</table>
Here, create a file "<a href="./example/rmpi/rmpi_ms.r" target="_blank">rmpi_ms.r</a>" as follows,
```
# File name: rmpi_ms.r
call.mpi.master <- function(){
library(Rmpi)
mpi.spawn.Rslaves(needlog = FALSE)
mpi.bcast.Robj2slave(call.mpi.slave)
mpi.bcast.cmd(call.mpi.slave())
x <- 100
mpi.bcast(as.integer(x), type = 1)
mysize <- mpi.universe.size()
y <- 200
for(i in 1 : mysize){
mpi.send(as.integer(y), type = 1, dest = i, tag = 1)
}
ret <- NULL
for(i in 1 : mysize){
ret.slave <- mpi.recv.Robj(source = i, tag = 2)
ret <- rbind(ret, ret.slave)
}
ret
}
call.mpi.slave <- function(){
x <- mpi.bcast(integer(1), type = 1)
y <- mpi.recv(integer(1), type = 1, source = 0, tag = 1)
myrank <- mpi.comm.rank()
ret.slave <- c(myrank, x, y, myrank, x * myrank + y)
mpi.send.Robj(ret.slave, dest = 0, tag = 2)
}
call.mpi.master()
```
The output will like this,
```
[,1] [,2] [,3] [,4] [,5]
ret.slave 1 100 200 1 300
ret.slave 2 100 200 2 400
ret.slave 3 100 200 3 500
ret.slave 4 100 200 4 600
```
- <font color="#800000"><b>Sum by Rmpi</b></font>
As the <a href="./loop.html">Loop</a> page, use MPI to split
the for loop and send to another slave to compute, reduce the computing time. Here are
examples "<a href="./example/rmpi/rmpi_for_1.r" target="_blank">rmpi_for_1.r</a>",
"<a href="./example/rmpi/rmpi_for_2.r" target="_blank">rmpi_for_2.r</a>",
"<a href="./example/rmpi/rmpi_apply.r" target="_blank">rmpi_apply.r</a>" and
"<a href="./example/rmpi/rmpi_rowSums.r" target="_blank">rmpi_rowSums.r</a>".
<br><br>
- <font color="#800000"><b>Computing time</b></font>
For PIII-1.4G PC cluster with 4 nodes, the test computing time as follows,
<table cellpadding="5">
<tr>
<th bgcolor="#d3dce3"> Sum by <br> <font color="red">Rmpi</font>
<th bgcolor="#d3dce3"> for 1
<th bgcolor="#d3dce3"> for 2
<th bgcolor="#d3dce3"> apply
<th bgcolor="#d3dce3"> rowSums
<th bgcolor="#d3dce3"> <br>
<tr>
<th bgcolor="#d3dce3"> Time <br> (secs)
<td bgcolor="#dddddd"> 86
<td bgcolor="#dddddd"> 79
<td bgcolor="#dddddd"> 41
<td bgcolor="#dddddd"> 5
<td bgcolor="#dddddd"> <br>
<tr>
<th bgcolor="#d3dce3"> Sum by <br> <font color="red">Loop</font>
<th bgcolor="#d3dce3"> for 1
<th bgcolor="#d3dce3"> for 2
<th bgcolor="#d3dce3"> apply
<th bgcolor="#d3dce3"> rowSums
<th bgcolor="#d3dce3"> dyn
<tr>
<th bgcolor="#d3dce3"> Time <br> (secs)
<td bgcolor="#dddddd"> 331
<td bgcolor="#dddddd"> 307
<td bgcolor="#dddddd"> 117
<td bgcolor="#dddddd"> 2
<td bgcolor="#dddddd"> 19
</table>
- <font color="#800000"><b>Conclusion</b></font>
The conclusion is the same in <a href="./loop.html">Loop</a> page. <br>
Use MPI to split independent job. <br>
Reduce by the number of CPUs? <br>
Communication times? <br>
---
<div w3-include-html="../preamble_tail_date.html"></div>
]]>