OpenMP. António Abreu. Instituto Politécnico de Setúbal. 1 de Março de 2013

Transcription

1 OpenMP António Abreu Instituto Politécnico de Setúbal 1 de Março de 2013 António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

2 openmp what? It s an Application Program Interface (API) that allows parallel programs to be explicitly and simply developed, in C/C++, for multi-platform, shared memory, multiprocessor computers (including Solaris, AIX, HP-UX, GNU/Linux, Mac OS X, and Windows platforms), supported by the major computer hardware and software vendors (including AMD, IBM, Intel, Cray, HP, Fujitsu, Nvidia, NEC, Microsoft, Texas Instruments, Oracle Corporation, and others.). António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

3 cores and memory Multicore computers have a memory system where some memories are shared while others are not. The next figure makes this distinction clear. TLB stands for Translation Lookaside Buffer, which is an address cache. When making parallel programs one must know which memory is shared and which memory is not. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

4 Fork join OpenMP is based on multithreading, i.e., a form of parallelization whereby a master thread forks a specified number of slave threads, with the runtime environment allocating threads to different processors. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

5 How many cores does my machine have? In linux, the file /proc/cpuinfo contains a lot of information about the hardware of the machine. Typing less /proc/cpuinfo allows one to see it all. To see info about memory, see the contents of the file /proc/meminfo. The first number one wants to see is the one corresponding to MemTotal. In order to use openmp, one has to have a propoer compiler. In linux, GCC 4.2 or higher supports openmp. To see the version of your (linux) compiler, type the command gcc -v. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

6 António num_threads Abreu (Instituto Politécnico (integer-expression) de Setúbal) OpenMP 1 de Março de / 37 parallel directive #pragma omp parallel [clause...] structured_block where clause can be newline if (scalar_expression) private (list) shared (list) default (shared none) firstprivate (list) reduction (operator: list) copyin (list)

7 Hello world #include <stdio.h> #include <omp.h> int main(void) #pragma omp parallel int ID = omp_get_thread_num(); printf("hello (%d)\n",id); printf("world (%d)\n",id); printf("! (%d)\n",id); return 0; António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

8 Compile with gcc -fopenmp hello.c -o hello Hello (0) world (0)! (0) Hello (1) world (1)! (1) Hello (2) world (2)! (2) Hello (3) world (3)! (3) António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

9 The code between the curly brackets (after the pragma directive) is set to execute in a predetermined number of threads. After the first curly bracket there is a fork, i.e., the master thread creates a team of parallel threads, and after the second curly bracket there is a join, i.e., the master thread continues execution after all the slave threads end. The second curly bracket constitutes a barrier, of which only the master thread passes. The number of threads is typically set to the number of cores in the microprocessor; it can be set by the command line export OMP_NUM_THREADS=4. omp_get_thread_num() is a function that returns the Id of the respective thread. The master thread has Id 0 and makes part of the thread team. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

10 We observe an ordered output, but sometimes this may not happen; in fact there is a race condition because the four threads share the standard output. Note that openmp is not necessarily implemented identically by all vendors. Also, it does not provide check for data dependencies, data conflicts, race conditions, or deadlocks. In particular, it does not guarantee that input or output to the same file is synchronous when executed in parallel. It is up to the programmer to synchronize input and output. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

11 Synchronization Constructs barriers #include <omp.h> #include <stdio.h> #include <stdlib.h> int main (int argc, char *argv[]) int th_id, nthreads; #pragma omp parallel private(th_id) th_id = omp_get_thread_num(); printf("hello World from thread %d\n", th_id); #pragma omp barrier if ( th_id == 0 ) nthreads = omp_get_num_threads(); printf("there are %d threads\n",nthreads); return EXIT_SUCCESS; António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

12 Hello World from thread 1 Hello World from thread 3 Hello World from thread 0 Hello World from thread 2 There are 4 threads António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

13 Barriers are a synchronization primitive. This means that all threads in the team wait for the last one to reach the barrier. At that moment, all threads in the team resume execution in parallel. If there is a thread that does not reach the barrier, all threads in the team wait, and the process hangs without any work being produced. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

14 Quiz If we comment the barrier pragma in the code above the output will be, Hello World from thread 0 There are 4 threads Hello World from thread 3 Hello World from thread 1 Hello World from thread 2 Explain why. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

15 Quiz If we add the code printf("bye from thread %d\n", th_id); after the if, what would be the output? António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

16 Workshare directives for #pragma omp for [clause...] for_loop newline where clause can be, schedule (type [,chunk]) ordered private (list) firstprivate (list) lastprivate (list) shared (list) reduction (operator: list) collapse (n) nowait António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

17 parallel for example #include <omp.h> #define CHUNKSIZE 100 #define N 1000 main () int i, chunk = CHUNKSIZE; float a[n], b[n], c[n]; /* Some initializations */ for (i=0; i < N; i++) a[i] = b[i] = i * 1.0; #pragma omp parallel shared(a,b,c,chunk) private(i) #pragma omp for schedule(dynamic,chunk) nowait for (i=0; i < N; i++) c[i] = a[i] + b[i]; /* end of parallel section */ António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

18 The for pragma asks the compiler to create threads from the N iterations of the for loop. The clause schedule informs the OS (operating system) about how to schedule those threads. In this case, the scheduling policy is dynamic, which means that threads are dynamically assigned on a first-come-first-serve basis. In this case each thread will execute chunk (i.e., 100) iterations of the total of 1000 in the loop. The nowait clause makes the implied barrier at the end of the for directive to be ignored. Put differently, if there was not such a clause, all team threads stop at the end of the for primitive, and only thread 0 would continue past this point. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

19 Quiz In the following program, which for cycle is executed in parallel: the first, or both? Before answering, note that the clauses parallel and for are combined in a single one. This is valid. #include <stdio.h> int main(int argc, char *argv[]) const int N = 100; int i, a[n]; #pragma omp parallel for for (i = 0; i < N; i++) a[i] = 2 * i; for (i = 0; i < N; i++) printf("%d ",a[i]); return 0; António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

20 Workshare directives sections #pragma omp sections [clause...] #pragma omp section newline structured_block #pragma omp section newline structured_block newline where clause can be, private (list) firstprivate (list) lastprivate (list) reduction (operator: list) nowait António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

21 section directive example #include <stdio.h> #include <omp.h> int main(void) #pragma omp parallel sections #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); printf("bye from thread %d\n",omp_get_thread_num()); António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

22 A few executions First execution hello from thread 0 hello from thread 0 hello from thread 0 Bye from thread 0 Second execution hello from thread 0 hello from thread 1 hello from thread 3 Bye from thread 0 Third execution hello from thread 2 hello from thread 1 hello from thread 0 Bye from thread 0 António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

23 Another example #include <stdio.h> #include <omp.h> int main(void) int i=0; #pragma omp parallel sections if (i==1) #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); #pragma omp section printf("hello from thread %d\n",omp_get_thread_num()); printf("bye from thread %d\n",omp_get_thread_num()); António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

24 Unique result hello from thread 0 hello from thread 0 hello from thread 0 Bye from thread 0 Since the condition is false, the team of threads is not created; but the master thread stands. Note that the assigned work (three blocks of code) is executed serially; so the if clause permits to parallelize work or not (i.e., to seriallize it), and the decision is made at runtime. Also, there is an implicit barrier at the end of each section. This explains why Bye from... (in the last two examples) is always the last message to print. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

25 Clause reduction reduction (operator: list) At the creation of a team of threads the variables in list are created as private. At the end of the threads in the team, operator is applied to the variables in list, a process known as reduction, and the final result is written back to the variables in list, now seen as global shared variables. Variables in list must be scalar; not arrays or structures. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

26 #include <stdio.h> #include <omp.h> int main(void) int t=0; omp_set_num_threads(4); #pragma omp parallel reduction(+:t) t = omp_get_thread_num() + 1; printf("local %d\n", t); printf("reduction %d\n", t); António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

27 Result local 1 local 2 local 3 local 4 reduction 10 The function of omp_set_num_threads() is self explanatory. As expected, it cannot be called from a parallelized block of code. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

28 Synchronization Constructs atomic Used to identify a memory location that should not be modified simultaneously by more than one thread in the team. In other words, it provides an atomic access to the memory location. #pragma omp atomic <statement_block> The directive applies only to a single statement. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

29 Synchronization Constructs single Used when there is a block of code that must be executed by a single thread in the team. Note that by no means this implies that the code is made atomic. It may happen that other threads (outside this team) access the same memory location, thus creating a race condition. #pragma omp single [clause[[,] clause]...] statement_block Threads in the team that do not execute this directive, wait at the end of the code block, unless a nowait clause is specified. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

30 Synchronization Constructs master Used to identify a block of code that must executed only by the master thread. #pragma omp master statement_block António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

31 Synchronization Constructs critical Specifies a block of code that must be executed by only one thread at a time. In other words, if the code in a critical region is executing, no other thread with that code will run in parallel. #pragma omp critical [(name)] statement_block Different critical regions with the same name are treated as the same region. All unnamed critical regions are treated as the same region. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

32 Example #include <omp.h> main() int x; x = 0; #pragma omp parallel shared(x) #pragma omp critical x = x + 1; /* end of parallel section */ António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

33 Synchronization Constructs flush This directive identifies a point at which a consistent view of memory must exist, i.e., thread-visible variables are written back to memory is response to this directive. #pragma omp flush [ (list) ] Remember the first figure of these course notes. This directive forces the data in the data cache of each core to be written to the shared unified cache memory (and not necessarily to the main memory; that decision is made by the virtual memory system). António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

34 openmp functions about threads #include <stdio.h> #include <omp.h> int main(void) printf("omp_get_max_threads=%d\n",omp_get_max_threads()); omp_set_num_threads(2); printf("omp_get_num_procs=%d\n",omp_get_num_procs()); #pragma omp parallel printf("omp_get_thread_num=%d\n",omp_get_thread_num()); printf("omp_get_thread_num=%d\n",omp_get_thread_num()); António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

35 omp_get_max_threads=4 omp_get_num_procs=2 omp_get_thread_num=0 omp_get_thread_num=1 omp_get_thread_num=0 omp_get_num_procs() returns the number of processors in the machine. António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

36 Synchronization locks omp_lock_t lck; omp_init_lock(&lck); #pragma omp parallel private (tmp,id) id = omp_get_thread_num(); tmp = do_lots_of_work(id); // critical region wrt tmp omp_set_lock(&lck); printf(%d %d",id,tmp); // atomic access to id and tmp omp_unset_lock(&lck); tmp = do_more_lots_of_work(id); // critical region wrt tmp omp_destroy_lock(&lck); António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37

37 Bibliography wikipedia António Abreu (Instituto Politécnico de Setúbal) OpenMP 1 de Março de / 37