并行计算

Size: px

Start display at page:

Download "并行计算"

材烘怀
4 years ago
Views:

1 OpenMP OpenMP OpenMP OpenMP OpenMP MPI OpenMP

2 OpenMP

3 OpenMP ( ) OpenMP RedHat Linux Intel C OpenMP

4 OpenMP OpenMP OpenMP OpenMP

5 RedHat Linux Intel C root intel8.1 chmod +x *.sh install.sh /opt/intel_cc_80/lib/lib* /lib PATH /opt/intel_cc_80/bin/ /opt/intel_cc_80/bin/icc icc

6 OpenMP F O R K J O I N F O R K J O I N FORK parallel JOIN parallel

7 ( ) ( )

8 #include <omp.h> int main(int argc, char *argv[]){ int nthreads, ithread; #pragma omp parallel private(nthreads, ithread) { ithread = omp_get_thread_num(); nthreads = omp_get_num_threads(); #pragma omp critical printf( The %d-th thread of %d threads.\n, ithread, nthreads); } return(0); }

9 OpenMP OpenMP parallel parallel parallel OpenMP threadprivate

10 OpenMP C #pragma omp directive-name [clause ][clause ]... new-line Fortran sentinel directive_name [clause, ][clause, ]... sentinel C$OMP *$OMP!$OMP

11 OpenMP C #ifdef_openmp iam=omp_get_thread_num()+index; #endif Fortran C$ iam=omp_get_thread_num()+index

12 parallel #pragma omp parallel [clause ][clause ] parallel clause if num_threads (2.0 ) private firstprivate shared default copyin reduction

13 parallel if( ) num_threads( ) private( ) shared( ) firstprivate( ) private copyin( ) firstprivate threadprivate

14 parallel ( ) default(shared none) private firstprivate default default(shared) shared ( threadprivate ) default(shared) default(none) ( threadprivate for parallel for ) parallel default

15 parallel ( ) reduction(op ) op op + * & ^ && op x = x op expr x <binop> = expr x = expr op x ( - ) x++ x-- ++x --x binop + & ^ reduction

16 private firstprivate reduction shared

17 parallel for sections single

18 for #pragma omp for [clause ][clause ] /* for */ clause private firstprivate lastprivate reduction schedule nowait ordered

19 for clause lastprivate lastprivate firstprivate ordered for ordered nowait for parallel for

20 for clause ( ) nowait #pragma omp parallel shared(m,n,a,b) { } #pragma omp for schedule(static) nowait for (k = 0; k<n; k++) a[k] = a[k] + 1; #pragma omp for schedule(static) for (k = 0; k<m; k++) b[k] = b[k] * 2;

21 for clause ( ) schedule(static dynamic guided runtime[, chunk_size]) Schedule(static[, chunk_size]) Schedule(dynamic[, chunk_size]) static Schedule(guided[, chunk_size]) chunk_size Schedule(runtime) OMP_SCHEDULE schedule schedule(runtime) OMP_SCHEDULE OpenMP

22 for for for break for for for schedule ordered nowait chunk_size

23 sections #pragma omp sections [clause ][clause ] { [#pragma omp section /* */ ] [#pragma omp section /* */ ] }

24 sections ( ) Sections clause private firstprivate lastprivate reduction nowait nowait sections nowait section sections

25 sections ( ) #pragma omp parallel { #pragma omp sections { #pragma omp section { T0(); } #pragma omp section { T1(); } } }

26 single single #pragma omp single [clause ][clause ] /* */ clause private firstprivate nowait nowait single

27 Parallel parallel for #pragma omp parallel for [clause ][clause ] /* for */ clause parallel for nowait

28 Parallel ( ) parallel sections #pragma omp parallel sections [clause ][clause ] { } [#pragma omp section /* */ ] [#pragma omp section /* */ ]

29 OpenMP barrier master critical atomic flush ordered

30 barrier #pragma omp barrier if(x==0) #pragma omp barrier if(x==0){ #pragma omp barrier }

31 master #pragma omp master /* */ single

32 critical #pragma omp critical [(name)] /* */ (name)

33 critical ( ) #pragma omp parallel sections { #pragma omp parallel section {/* producer thread */ } task = produce_task(); #pragma omp critical (queue) { } insert_into_queue(task); #pragma omp parallel section {/* consumer thread */ #pragma omp critical (queue) { task = extract_queue(task); } consume_task(task); } }

34 atomic #pragma omp atomic /* */ x binop = expr ++x x++ --x x-- binop atomic critical atomic x expr

35 flush #pragma omp flush [( )] barrier flush

36 flush ( ) flush barrier critial ordered parallel for sections single

37 ordered ordered #pragma omp ordered /* */ for parallel for ordered

38 ordered ( ) ordered for parallel for ordered

39 threadprivate #pragma omp threadprivate ( ) parallel parallel threadprivate if copyin schedule default threadprivate parallel copyin

40 parallel OMP_NESTED FALSE parallel for sections single critical critical ordered master for sections single

41 ( ) for sections single ordered master critical barrier for sections single master critical ordered

42 OpenMP omp.h

43 void omp_set_num_threads(int num_threads) int omp_get_num_threads(void) int omp_get_max_threads(void) int omp_get_thread_num(void) int omp_get_num_procs(void)

44 ( ) int omp_in_parallel(void) void omp_set_dynamic(int dynamic_threads) int omp_get_dynamic(void) void omp_set_nested(int nested) int omp_get_nested(void)

45 void omp_init_lock(omp_lock_t *lock) void omp_init_nest_lock(omp_nest_lock_t *lock) void omp_destroy_lock(omp_lock_t *lock) void omp_destroy_nest_lock(omp_nest_lock_t *lock) void omp_set_lock(omp_lock_t *lock) void omp_set_nest_lock(omp_nest_lock_t *lock)

46 ( ) void omp_unset_lock(omp_lock_t *lock) void omp_unset_nest_lock(omp_nest_lock_t *lock int omp_test_lock(omp_lock_t *lock) int omp_test_nest_lock(omp_nest_lock_t *lock)

47 OpenMP OMP_SCHEDULE OMP_NUM_THREADS OMP_NESTED OMP_DYNAMIC

48 OMP_SCHEDULE schedule(runtime) for parallel for setenv OMP_SCHEDULE \ static dynamic guide[, chunk_size] 1 setenv OMP_SCHEDULE static 2 setenv OMP_SCHEDULE guided,

49 OMP_NUM_THREADS setenv OMP_NUM_THREADS num_threads OMP_NUM_THREADS omp_set_num_threads parallel parallel for num_threads OMP_DYNAMIC TRUE omp_set_dynamic

50 OMP_NESTED setenv OMP_NESTED TRUE FALSE parallel parallel for parallel sections

51 OMP_DYNAMIC setenv OMP_DYNAMIC TRUE FALSE

52 MPI OpenMP OpenMP MPI

53 OpenMP OpenMP OpenMP

54 ( ) OpenMP OpenMP

55 MPI MPI

56 OpenMP MPI SMP SMP MPI OpenMP

第7章-并行计算.ppt

第7章-并行计算.ppt EFEP90 10CDMP3 CD t 0 t 0 To pull a bigger wagon, it is easier to add more oxen than to grow a gigantic ox 10t 0 t 0 n p Ts Tp if E(n, p) < 1 p, then T (n) < T (n, p) s p S(n,p) = p : f(x)=sin(cos(x))