Programming :: Timings Of A Program That Uses 1-8 Threads On A Server That Has 4 D?
Jul 25, 2010
I am runig a program on a server at my university that has 4 Dual-Core AMD Opteron(tm) Processor 2210 HE and the O.S. is Linux version 2.6.27.25-78.2.56.fc9.x86_64. My program implements Conways Game of Life and it runs using pthreads and openmp. I timed the parrallel part of the program using the getimeofday() function using 1-8 threads. But the timings don't seem right. I get the biggest time using 1 thread(as expected), then the time gets smaller. But the smallest time i get is when i use 4 threads.
Here is an example when i use an array 1000x1000.
Using 1 thread~9,62 sec, Using 2 Threads~4,73 sec, Using 3 ~ 3.64 sec, Using 4~2.99 sec, Using 5 ~4,19 sec, Using 6~3.84, Using 7~3.34, Using 8~3.12.The above timings are when i use pthreads. When i use openmp the timing are smaller but follow the same pattern.I expected that the time would decrease from 1-8 because of the 4 Dual core cpus? I thought that because there are 4 cpus with 2 cores each, 8 threads could run at the same time. Does it have to do with the operating system that the server runs?
Also i tested the same programs on another server that has 7 Dual-Core AMD Opteron(tm) Processor 8214 and runs Linux version 2.6.18-194.3.1.el5. There the timings i get are what i expected. The timings get smaller starting from 1(the biggest) to 8(smallest excecution time).The program implements the Game of Life correct, both using pthreads and openmp, i just cant figure out why the timings are like the example i posted. So in conclusion, my questions are:
1) The number of threads that can run at the same time on a system depends by the cores of the cpus?it depends only by the cpus although each cpu has more than one cores? It depends by all the previous and the Operating System?
2) Does it have to do with the way i divide the 1000x1000 array to the number of threads? But if i did then the openmp code wouldn't give the same pattern of timings?
In all the examples I have found the server accepts the client's conection, proccess the data received and close the socket. In an very schematic way it would be something like:
Code:
client_thread{ select to see if there is data to read from socket fd if there is something to read{
[code]....
Should I use mutexs or semaphores to block the socket fd before read and write or it is not necesary?
I have a multithreaded program running. The threads are created by using pthread_create(). The ps shows a pid value for each thread in the process. The name of the thread is all the same as the parent process. I want to match the PID values in the ps output to the threads in my program. But when I used getpid() in each thread, the value I get is different from what ps shows. How do I match the pid value in ps output to the threads in my program.
i want a process that can operate as both a TCP echo server and a UDP echo server. The process can provide service to many clients at the same time, but involves a single process that does not start up any other threads.
I wrote a program that multiplies 2 matrices using multi-threads and another one using multiple processes and shared memory. Both in C.I need to find the total memory usage of these programs. I know of the top command, but when my matrices are relatively small they don't even show up on top because they complete so fast, how can I find the memory usage for these instances?Also, how can I find the total turnaround time of my programs
I wrote a C program using Pthreads to compute the product of 2 matrices. Each element in the product matrix is computed in a separate thread. Eg: Thread (i,j) computes the element C[i][j] of the matrix C, where C=A*B. A is m*n, B is n*p, C is m*p. m,n,p are given as command-line arguments. A and B are initialized to random values from 1 to 10, while all elements of C are initialized to -1.But some threads do not get their arguments (i,j) correctly. So some elements C[i][j] still remain as -1, even after the program is over. My OS is Ubuntu 10.10 (Maverick Meerkat) 32-bit.I ran the program on another computer and it worked correctly. Is it due to a problem in the Pthreads library in my OS? Please help me. I have attached the source code.
I have doubt regarding cpu sharing between process and threads.In my program iam creating 4threads=> 1 process+4 threads. How is cpu alloted to these all tasks. Is here process is getting cpu time like thread or having more cpu time than threads.
I have 2 threads and both of them are deleting memory at the end nedded by both.
My problem is that maybe it can happen that a thread start and finish before the other one starts and so it deletes the memory nedded by the other thread. How can I synchronize them so that this can't happend.
As a design my threads look like this:
Code:
The other thread looks the same, but this isn't unoff to stop thread1 to finish before thread2 starts.
Fedora15 32bit. I write a test program, it creates new thread continually, the thread does nothing but sleep. I find virtual memory increases up almost 10Mb when a new thread is created. and when there's more than 200 threads, the virtual memory used by the program is 3Gb, and now cann't create new thread. but on windows, it costs little memory. What can I do to config the operation system to take less memory on threads?
I've implemented a program URL... which reads digital IF data from a radio receiver through a named pipe, measures power levels, and sends the result to stdout. The program is interactive; there is a thread that reads from stdin to watch for commands, a thread that constantly either reads data from the named pipe or throws data away, and an array of processing threads. The program uses GTK+extra to plot the signals. The IF data stream bandwidth exists at the limits of today's technology (is very very fast).
Problem Statement:The program works fine with a few bugs. I've learned since I've made it that using global state variables to coordinate threads isn't a good way of doing it. I also only had knowledge of mutexes and polled the state variable instead of using other methods.My reimplementation will use the following:
- One "Stdin Command Monitoring" thread - One "Get data from named pipe" thread - One post-processor thread - N Processing threads
All threads are alive during the life of main()There are N buffers. Data will come in from the named pipe, and the "Get data" thread will write the data to an "available" buffer. When the buffer is full it will be marked as "full". There will be N processing threads, one for each buffer. When a processing threads' buffer is full, it will process the buffer and save the result to a final buffer. At the end of a number of averages, the post-processor thread will perform a final process on the final buffer and send the results to stdout.
I'm a bit worried about "too many mutexes" in my little curses-based app and would like to get confirmation/opinions that I'm doing this right. I've got an array: int nums[60] I've got 61 threads. 1-60 are doing math on the value in their array index (ie: thread1 increments nums[1], threadN increments nums[N]), then sleep(1) The 61st thread is my curses thread which does a for-loop over the array and prints out all the values to the screen, then sleep(1)
Right now, I've got 1 mutex which gets locked/unlocked each time one of the 60 threads needs to update its array-index with a new value, and the 61st thread locks the same mutex just before the for-loop beings reading the values and unlocks after ending the loop.
My questions: A) Does the above seem OK? (I know it's ok, cause everything works right now but would like opinions on it) B) Do I even need the mutexes since 1-60 only ever update their own index and 61 just reads? C) If I do need the mutex protection, is there a better, more efficient way?
I am working with a C++ program consisting of two threads. The first threads receives packets through an UDP unicast connection and stores them in a buffer. The second thread reads the packets from the buffer and sends them through an UDP multicast connection. Both use blocking sockets and share a common buffer and a linked list L1, which are protected by mutexes. The program seemed to work just fine, receiving a packet and sending it almost immediately, but started giving some trouble recently. The synchronization between both thread started failing, and I decided to use a non-blocking socket in the sending thread. As a consequence, sendto() doesn't work in some cases, causing an errno 11 (Resource unavailable).
I need to find how many threads are alive with respect to the current process for my further processing. Is there any means to trace this number ?[URL]I referred the above link. But sys/pstat.h is not in my system. Don't know which library gives this header.
I want to communicate between two threads, each belonging to a different process. Iam using message queues for this. I use mq_open()call. I created the queues with the same queue name starting with a '/'. But when I open the queue, the queue ID is different in both the process. What should I do so that both the process have the same queue ID?
I have a linked list that two threads work on simultainiousley.The first thread is adding elements to end the linked list while the second is removing them from the front. Can this be done without a lock on the linked list head when elements are beingadded/removed?
I think this lock is causing a performance hit to my application. If there isnt any safe way without it then thats fine but just thought I would check.The first thread uses this fuction to add elements to the list. Full source here. [URL]
Code:
/* Lets add the new packet to the queue. */ pthread_mutex_lock(&workers[queuenum].queue.lock); // Grab lock on queue. if (workers[queuenum].queue.qlen == 0){ // Check if any packets are in the queue.
I am using GPROF for my code profiling ,but it seems GPROF doesn't supports multithread .Does any one know any other technique for profiling the application code I have checked oprofile but it profile kernel ,I just required for C code profiling technique that supports multiple threads
I have to compute prime numbers with threads in C. So I choose the algorithm "Sieve of Eratosthenes" and I tried to create 2 threads, but I don't know how to optimize my computing function with threads I beg for help. This is my function:
[Code]...
but it seems like the both of the threads are calculating the same function twice? Is it possible to be optimized or not?
If multiple threads operate on a single shared resource, we can lock it using pthread_mutex_lock. Uptill that it is fine, but why again pthread_cond_wait?
int shared=0; // global . . // Thread 1 pthread_mutex_lock(&mutex); shared=1; pthread_mutex_unlock(&mutex);
When thread1 invokes the mutex lock, no other threads can access it in parallel. So why again and what for we use, pthread_cond_wait( ).
I'm trying to add local sockets in my multi-threaded application in order to exchange data between threads. The only problem I got is that most of the information available on the net is related to internet oriented socket programming whileI want to perform local connections. got a thread that does the sniffing via libpcap. And I would like that thread to send each captured packet to a second thread that will analyse the packetof the thread implementations is written in separate .h file.Or maybe there is a more effective method of exchanging data between threads
I'm using 11.04 and when I boot I see post messages and then when it gets to grub I just see black screen with a monitor timing not supported issue. I have a geforce 6100 nforce 405 video card with a Dell U2311H monitor. It does boot to the desktop eventually, and everything from there on works as expected.
This is about an oracle application server running on RHEL platform(Linux prod.xyz.com 2.6.18-194.el5PAE #1 SMP Mon Mar 29 20:19:03 EDT 2010 i686 i686 i386 GNU/Linux). All of a sudden all the application services shut down abnormally on weekend. On the first day of the week, business users were unable to login as a result. For root cause investigation, we were unable to findout anything in the application logs. But were trying to get reboot/shutdown information from the system.....but we are bit confused with the info in /var/log/wtmp by using 'last' command(output given bellow):
[root@prod ~]# last reboot |head -2 reboot system boot 2.6.18-194.el5PA Mon Aug 8 11:26 (1+03:10) reboot system boot 2.6.18-194.el5PA Sun Aug 7 17:27 (17:55)
We doubt the server might have gone down on Sunday(7th Aug) but there's also an entry of 8th Aug, on which day the server was running fine, no reboot/shutdown happened. If there any other way of getting info on reboot timings.
