I'm trying to use MPI_Comm_spawn to start a second process. Just for demonstration pourposes. The program is quite simple:
int main(int argc, char* argv[])
{ int my_id, numprocs;
MPI_Comm All;
MPI_Init(&argc, &argv) ;
MPI_Comm_rank(MPI_COMM_WORLD, &my_id) ;
MPI_Comm_size(MPI_COMM_WORLD, &numprocs) ;
cout << "I'm process "<< my_id << " and we are " << numprocs <<endl;
MPI_Comm_spawn("child.exe",MPI_ARGV_NULL,2,MPI_INFO_NULL,my_id,MPI_COMM_WORLD, &All,MPI_ERRCODES_IGNORE);
MPI_Comm_size(All, &numprocs) ;
cout << "I'm process "<< my_id << " and we are " << numprocs <<endl;
{
int i;
cin >> i;
}
MPI_Finalize();
return 0;
}
The child.exe is in the same directory of the compiled program, and is also terribly simple:
int main(int argc, char* argv[])
{
int my_id, numprocs,length;
MPI_Comm Parent;
MPI_Win pwin,gwin;
MPI_Init(&argc, &argv) ;
MPI_Comm_rank(MPI_COMM_WORLD, &my_id) ;
MPI_Comm_size(MPI_COMM_WORLD, &numprocs) ;
MPI_Comm_get_parent(&Parent);
cout << "I'm child process "<< my_id << " and we are " << numprocs <<endl;
MPI_Comm_size(Parent, &numprocs) ;
cout << "My parent communicator size is: "<< numprocs <<endl;
MPI_Finalize();
return 0;
}
The parent process fails with a criptic error:
C:\Users.....\Documents\Visual Studio 2010\Projects\mpi\x64\Release>mpi.exe
I'm process 0 and we are 1
job aborted:
[ranks] message
[0] fatal error
Fatal error in MPI_Comm_spawn: Other MPI error, error stack:
MPI_Comm_spawn(106)..........: MPI_Comm_spawn(cmd="child.exe", argv=0x0000000000
000000, maxprocs=2, MPI_INFO_NULL, root=0, MPI_COMM_WORLD, intercomm=0x000000000
026FC20, errors=0x0000000000000000) failed
MPID_Comm_spawn_multiple(314): Function not implemented
What I'm doing wrong.... seem so simple.... MSMPI does not implement MPI_Comm_spawn ?
According to this document, it is not supported, at least not in Windows HPC Server 2008.
From the document:
MS MPI includes a complete set of MPI2
functionality as implemented in MPICH2
with the exception of dynamic process spawn
and publishing, which may be
included in a later release.
Related
I've written a code to solve Buying new tablet problem from codechef. Although everything goes well in CLION and Visual Studio, even outputs are correct, codechef compiler says runtime error SIGTSTP. I'm new to this site, can not understand what it says. I've done a little research on Google, but nothing.
#include <iostream>
#include <algorithm>
#include <vector>
using namespace std;
int main(){
int T,N,B,W,H,P;
bool isGr = false;
cin >> T;
while (T--){
cin >> N;
cin >> B;
vector <int> tablets;
while (N--){
cin >> W;
cin >> H;
cin >> P;
if (P <= B) {
tablets.push_back(W*H);
isGr = true;
}
else isGr = false;
}
if (isGr){
sort(tablets.begin(),tablets.end());
cout << tablets.back() << endl;
tablets = {};
} else cout << "no tablet" << endl;
isGr = false;
}
}
The condition
'''else isGr=false;'''
in second while loop, is not required....
I'm not very good at character devices, so I need your help. A have a char device(let's call it /dev/my_light) which is a light sensor. I have to read the data from this file and transform it to the brightness value and then pass it to the brightness manager that changes the brightness of my screen. The problem is that when I read the value for some period of time I get old values from the file.I assume there is a buffer(again not sure how character devices exactly work). Whereas when I use cat /dev/my_light I see new data! Is it possible to get rid off the buffer and read new values that were written to the file just right now. Here is my code in Qt:
void MySensor::updateMySensor()
{
Packet packet;
packet.startByte = 0;
packet.mantissa = 0;
packet.exp = 0;
d->device = ::open(d->path.toStdString().c_str(), O_RDONLY);
if (d->device == -1)
{
qDebug() << Q_FUNC_INFO << "can't open the sensor";
return;
}
ssize_t size = ::read(d->device, &packet, sizeof(packet));
close(d->device);
if (size == -1)
{
qDebug() << errno;
return;
}
packet.exp &= 0x0F;
float illumination = pow(2, packet.exp) * packet.mantissa * 0.045;
if(d->singleShot) emit lightSensorIsRunning(true);
emit illuminationRead(illumination);
}
The mySensor function is called every second. I tried to call it each 200 msec but it didn't help. The value of illumination stays old for about 7 seconds(!) whereas the value that I get from cat is new just immediately.
Thank you in advance!
I can't test with your specific device, however, I'm using the keyboard as a read only device.
The program attempts to connect to keyboard and read all keys pressed inside and outside the window. It's a broad solution you'll have to adapt to meet your demands.
Note that I'm opening the file with O_RDONLY | O_NONBLOCK which means open in read only mode and no wait for the event be triggered(some notifier needed to know when data is ready!) respectively.
You'll need super user privilege to run this example!
#include <QtCore>
#include <fcntl.h>
#include <linux/input.h>
int main(int argc, char *argv[])
{
QCoreApplication a(argc, argv);
const char *device_name = "/dev/input/by-path/platform-i8042-serio-0-event-kbd";
int descriptor = open(device_name, O_RDONLY | O_NONBLOCK);
if (descriptor < 0)
{
qDebug() << "Error" << strerror(errno);
return a.exec();
}
QFile device;
if (!device.open(descriptor, QFile::ReadOnly))
{
qDebug() << "Error" << qPrintable(device.errorString());
return a.exec();
}
QSocketNotifier notifier(device.handle(), QSocketNotifier::Read);
QObject::connect(¬ifier, &QSocketNotifier::activated, ¬ifier, [&](int socket){
Q_UNUSED(socket)
struct input_event ev;
QByteArray data = device.readAll();
qDebug() << "Event caught:"
<< "\n\nDATA SIZE" << data.size()
<< "\nSTRUCT COUNT" << data.size() / int(sizeof(input_event))
<< "\nSTRUCT SIZE" << sizeof(input_event);
qDebug() << ""; //New line
while (data.size() >= int(sizeof(input_event)))
{
memcpy(&ev, data.data(), sizeof(input_event));
data.remove(0, int(sizeof(input_event)));
qDebug() << "TYPE" << ev.type << "CODE" << ev.code << "VALUE" << ev.value << "TIME" << ev.time.tv_sec;
}
qDebug() << ""; //New line
});
return a.exec();
}
I want to create MPI communicators linking the process with rank 0 to every other process. Suppose n is the total number of processes. Then the process with rank 0 is supposed to have n-1 communicators while each of the other processes has one communicator. Is this possible, and, if it is, why can I not use the program below to achieve this?
Compiling the code below using the mpic++ compiler terminates without warnings and errors on my computer. But when I run the resulting program using 3 or more processes (mpiexec -n 3), it never terminates.
Likely I'm misunderstanding the concept of communicators in MPI. Maybe someone can help me understand why the program below gets stuck, and what is a better way to create those communicators? Thanks.
#include <iostream>
#include <vector>
#include <thread>
#include <chrono>
#include "mpi.h"
void FinalizeMPI();
void InitMPI(int argc, char** argv);
int main(int argc, char** argv) {
InitMPI(argc, argv);
int rank,comm_size;
MPI_Comm_rank(MPI_COMM_WORLD,&rank);
MPI_Comm_size(MPI_COMM_WORLD,&comm_size);
if(comm_size<2) {
FinalizeMPI();
return 0;
}
MPI_Group GroupAll;
MPI_Comm_group(MPI_COMM_WORLD, &GroupAll);
if(rank==0) {
std::vector<MPI_Group> myGroups(comm_size-1);
std::vector<MPI_Comm> myComms(comm_size-1);
for(int k=1;k<comm_size;++k) {
int ranks[2]{0, k};
MPI_Group_incl(GroupAll, 2, ranks, &myGroups[k-1]);
int err = MPI_Comm_create(MPI_COMM_WORLD, myGroups[k-1], &myComms[k-1]);
std::cout << "Error: " << err << std::endl;
}
} else {
MPI_Group myGroup;
MPI_Comm myComm;
int ranks[2]{0,rank};
MPI_Group_incl(GroupAll, 2, ranks, &myGroup);
int err = MPI_Comm_create(MPI_COMM_WORLD, myGroup, &myComm);
std::cout << "Error: " << err << std::endl;
}
std::cout << "Communicators created: " << rank << std::endl;
std::this_thread::sleep_for(std::chrono::milliseconds(1000));
FinalizeMPI();
return 0;
}
void FinalizeMPI() {
int flag;
MPI_Finalized(&flag);
if(!flag)
MPI_Finalize();
}
void InitMPI(int argc, char** argv) {
int flag;
MPI_Initialized(&flag);
if(!flag) {
int provided_Support;
MPI_Init_thread(&argc, &argv, MPI_THREAD_MULTIPLE, &provided_Support);
if(provided_Support!=MPI_THREAD_MULTIPLE) {
exit(0);
}
}
}
MPI_Comm_create is a collective operation on the initial communicator (MPI_COMM_WORLD) - you must call it on all processes.
The simplest way to fix the issue is to use MPI_Comm_create_group just the way you do it. It is similar to MPI_Comm_create except that it is collective over the group.
I'm absolutely new to OpenCL programming. I have a working installation of OpenCL library and drivers. But the program I'm trying to run is not producing expected output (Output is all zeros). It is just a simple vector_add program.
Thanks in advance for suggestions.
int main(int argc, char** argv)
{
cout << "Hello OpenCL" << endl;
vector<Platform> all_platforms;
int err = Platform::get(&all_platforms);
cout << "Getting Platform ... Error code " << err << endl;
if (all_platforms.size()==0)
(cout << "No platforms" << endl, exit(0));
cout << "Platform info : " << all_platforms[0].getInfo<CL_PLATFORM_NAME>() << endl;
Platform default_platform = all_platforms[0];
cout << "Defaulting to it ..." << endl;
vector<Device> all_devices;
err = default_platform.getDevices(CL_DEVICE_TYPE_GPU, &all_devices);
cout << "Getting devices ... Error code : " << err << endl;
if (all_devices.size()==0)
(cout << "No devices" << endl, exit(0));
Device default_device = all_devices[0];
cout << all_devices.size() << " devices & " << "Device info : " << all_devices[0].getInfo<CL_DEVICE_NAME>() << endl;
cout << "Defaulting to it ..." << endl;
Context context(default_device);
Program::Sources sources;
std::string kernel_code=
" void kernel simple_add(global const int* A, global const int* B, global int* C){"
" unsigned int i = get_global_id(0); "
" C[i]=A[i]+B[i]; "
" } ";
sources.push_back(make_pair(kernel_code.c_str(), kernel_code.length()+1));
Program program(context, sources);
if (program.build(all_devices)==CL_SUCCESS)
cout << "Built Successfully" << endl;
Buffer buffer_A(context,CL_MEM_READ_WRITE,sizeof(int)*10);
Buffer buffer_B(context,CL_MEM_READ_WRITE,sizeof(int)*10);
Buffer buffer_C(context,CL_MEM_READ_WRITE,sizeof(int)*10);
int A[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
int B[] = {0, 1, 2, 0, 1, 2, 0, 1, 2, 0};
CommandQueue queue(context,default_device);
queue.enqueueWriteBuffer(buffer_A,CL_TRUE,0,sizeof(int)*10,A); // load data from host to device
queue.enqueueWriteBuffer(buffer_B,CL_TRUE,0,sizeof(int)*10,B);
Kernel kernel(program, "vector_add");
kernel.setArg(0, buffer_A);
kernel.setArg(1, buffer_B);
kernel.setArg(2, buffer_C);
queue.enqueueNDRangeKernel(kernel,cl::NullRange,cl::NDRange(10),cl::NullRange);
queue.finish();
int *C = new int[10];
queue.enqueueReadBuffer(buffer_C, CL_TRUE, 0, 10 * sizeof(int), C);
for (int i=0;i<10;i++)
std::cout << A[i] << " + " << B[i] << " = " << C[i] << std::endl;
return 0;
}
As pointed out in the comments, you should always check the error codes when using OpenCL API functions. This can be achieved by enabling exception handling in the C++ wrapper:
#define __CL_ENABLE_EXCEPTIONS // with cl.hpp
//#define CL_HPP_ENABLE_EXCEPTIONS // with cl2.hpp
#include <CL/cl.hpp>
int main(int argc, char *argv[])
{
try
{
// OpenCL code here
}
catch (cl::Error& err)
{
cout << err.what() << " failed with error code " << err.err() << endl;
}
}
If you do this, you will receive useful information about a couple of issues with your code.
The clCreateKernel function returns CL_INVALID_NAME. In your kernel, you define the kernel function with the name simple_add, but then you try and create a kernel object using the name vector_add.
If you have an OpenCL platform with multiple devices, you may also receive an error when building your kernel program. This is because you are creating an OpenCL context with a single device, but then trying to build the program for a list of devices:
Context context(default_device);
// ...
if (program.build(all_devices)==CL_SUCCESS)
cout << "Built Successfully" << endl;
The simplest fix is just to remove the argument from the build function, since by default it will build the program for all devices in the context (which is almost always what you actually want):
if (program.build()==CL_SUCCESS)
cout << "Built Successfully" << endl;
//#include <QtCore/QCoreApplication>
#include <stdio.h>
#include <stdlib.h>
#include <iostream>
#include <string>
#include "shift.h"
using namespace::std;
int main(int argc, char *argv[])
{
unsigned char tmpBuf;
FILE* fp;
FILE* fp2;
char fname[50];
static unsigned int lSize, count, num;
cout << "Input the filename:" << endl;
cin >> fname;
fp = fopen(fname,"r");
if(fp == NULL) {
cout << "The file does not exist!" << endl;
exit(1);
}
// obtain file size:
fseek(fp , 0 , SEEK_END);
lSize = ftell(fp);
rewind(fp);
cout << "The intput file's size is: " << lSize << endl;
fp2 = fopen("myfile", "w");
while(1){
num = fread(&tmpBuf, 1, 1, fp);
count += num;
// putchar(tmpBuf);
// tmpBuf = cror(tmpBuf, 4);
// tmpBuf = crol(tmpBuf, 4);
fwrite(&tmpBuf, 1, num, fp2);
cout << tmpBuf << " " << num << " " << count << endl;
if (count == lSize){
printf("over\n");
break;
}
}
fclose(fp);
fclose(fp2);
while(1){}
return 0;
//return a.exec();
}
I made a Qt console program and disable QtCore, like above code. When read some file, e.g. 1.txt (which contains only 1234567890) it succeeds. But when reading some other file, e.g. 1.rar, it failed like below: Why?
Check the value of num coming back from
num = fread(&tmpBuf, 1, 1, fp);
If it comes back as 0, that would explain why
if (count == lSize){
never comes back as true to break you out of the loop.
As to why that could happen, you're opening fname in "r" mode but a rar file would be binary. For that, I'd suggest opening in "rb" mode. If fread expects txt format but hits the EOF indicator, it'll stop advancing so num = 0 and count won't increase.
I don't have windows available to test this, but other questions have come up on stackoverflow for this reason:
fread/ftell apparently broken under Windows, works fine under Linux