OpenMP????

1

• OpenMP????

2
OpenMP????

• OpenMP??
• OpenMP????
• ??????????
• OpenMP????

3
OpenMP??

• OpenMP??????API??????????????????
• ??????(Compiler Directive)??????(Runtime
  Library)?????(Environment Variables)
• ???????(Incremental Parallelization)

4
???OpenMP

• ???OpenMP
• ??????API(Application Programming Interface )
• ?????API??(??????????????)??
• ?C/C ?Fortan????????
• ????????????????????
• OpenMP??????
• ??????????????
• ??????????????
• ????????????????????

5
OpenMP???

• 1994?,???ANSI X3H5????,???
• 1997?,OpenMP????????????ANSI X3H5,?????
• 1997?10?????Fortran???????????? FORTRAN version
  1.0
• 1998?11?9??????C?C?????C/C version 1.0
• 2000?11???FORTRAN version 2.0
• 2002?3???C/C version 2.0
• 2005?5?OpenMP2.5????Fortran?C/C ???????
• ???????http//www.openmp.org/drupal/node/view/8???
  

6
OpenMP???

• ???
• ????
• ????
• ????

7
OpenMP??????

• ???????????(Programming Model)
• OpenMP??Fork-Join??????

8
OpenMP????

• ??Fortran???OpenMP?????
• PROGRAM HELLOINTEGER VAR1, VAR2, VAR3
  !Serial code !Beginning of parallel
  section. Fork a team of threads. !Specify
  variable scoping !OMP PARALLEL PRIVATE(VAR1,
  VAR2) SHARED(VAR3) !Parallel section
  executed by all threads !All threads
  join master thread and disband!OMP END
  PARALLEL !Resume serial code END

9
OpenMP????

• ??c/c???OpenMP?????
• include ltomp.hgt
• main ()
• int var1, var2, var3
• /Serial code/
• /Beginning of parallel section. Fork a team
  ofthreads/
• /Specify variable scoping /
• pragma omp parallel private(var1, var2)
  shared(var3)
• /Parallel section executed by all threads/
  
• /All threads join master thread and disband/
• /Resume serial code /

10
?????OpenMP????

• ??C/C???OpenMP???????????
• include "omp.h//eg1
• int main(int argc, char argv)
• int nthreads, tid
• int nprocs
• char buf32
• / Fork a team of threads /
• pragma omp parallel private(nthreads, tid)
• / Obtain and print thread id /
• tid omp_get_thread_num()
• printf("Hello World from OMP thread d\n",
  tid)
• / Only master thread does this /
• if (tid0)
• nthreads omp_get_num_threads()
• printf("Number of threads d\n",
  nthreads)

11
????

• OpenMP?pragma??????
• pragma omp directive_name

pragma omp directive-name clause, ... newline
???????? ???OpenMP? ???????? ?? OpenMP??? ???????? ???????? ?????? OpenMP??? ?? ???????? ?????,?? ????,??? ?????? ?? ???????? ???????? ????????
12
????

• ???
• ????
• ???????????????,??????????
• ????
• ??OpenMP????????????????
• ????
• ???????????

13
???
???? ????
???? for?????????????? ???? critical?sections?????????????
pragma omp parallel pragma omp for for() sub1() sub2() void sub1() pragma omp critical void sub2() pragma omp sections
14
?????

• ???????????????
• ????
• pragma omp parallel clause,clausenewline
• clause
• if(scalar-expression)
• private(list)
• firstprivate(list)
• default(shared none)
• shared(list)
• copyin(list)
• reduction(operator list)
• num_threads(integer-expression)

15
??????

• ???????????????????????????
• ??for??
• ??sections
• ????

16
for??????

• for??????????????????????
• ????
• pragma omp for clause,clause newline
• clause
• Schedule(type ,chunk)
• ordered
• private (list)
• firstprivate (list)
• lastprivate (list)
• shared (list)
• reduction (operator list)
• nowait

17
for??????

• schedule??????????????????????
• ??????chunk??,????????????????
• type?static,???????? chunk??,???????
• type?dynamic,???????????chunk??,???????

18
Sections??????

• sections?????????????????????????
• ???section????????
• Section????
• pragma omp sections clause,clause
  newline
• pragma omp section newline
• pragma omp section newline

19
Sections??????

• clause
• private (list)
• firstprivate (list)
• lastprivate (list)
• reduction (operator list)
• nowait
• ?sections?????????????,???nowait????

20
Sections??????

• include ltomp.hgt//eg2
• define N 1000
• int main ()
• int i
• float aN, bN, cN
• / Some initializations /
• for (i0 i lt N i)
• ai bi i 1.0
• pragma omp parallel shared(a,b,c) private(i)
• pragma omp sections nowait
• pragma omp section
• for (i0 i lt N/2 i)
• ci ai bi
• pragma omp section
• for (iN/2 i lt N i)
• ci ai bi
• / end of sections /

21
single??????

• single??????????????????????????
• ????????single????????????????,??nowait????
• ????
• pragma omp single clause,clause newline
• clause
• private(list)
• firstprivate(list)
• nowait

22
???????????

• parallel for??????
• parallel sections??????

23
parallel for??????

• Parallel for????????????????????for??
• ????
• pragma omp parallel for clause newline
• clause
• if (scalar_logical_expression)
• default (shared none)
• schedule (type ,chunk)
• shared (list)
• private (list)
• firstprivate (list)
• lastprivate (list)
• reduction (operator list)
• copyin (list)

24
parallel for??????

• include ltomp.hgt //ex2
• define N 1000
• define CHUNKSIZE 100
• int main ()
• int i, chunk
• float aN, bN, cN
• / Some initializations /
• for (i0 i lt N i)
• ai bi i 1.0
• chunk CHUNKSIZE
• pragma omp parallel for \
• shared(a,b,c,chunk) private(i) \
• schedule(static,chunk)
• for (i0 i lt n i)
• ci ai bi

25
parallel sections??????

• parallel sections????????????????????sections??
• ????
• pragma omp parallel sections clause newline
• clause
• default (shared none)
• shared (list)
• private (list)
• firstprivate (list)
• lastprivate (list)
• reduction (operator list)
• copyin (list)
• ordered

26
????

• master ????
• critical????
• barrier????
• atomic????
• flush????
• ordered????

27
master ????

• master????????????????
• ????
• pragma omp master newline

28
critical????

• critical?????????????????????
• ???????????
• ????
• pragma omp critical name newline

29
critical????

• include ltomp.hgt
• main()
• int x
• x 0
• pragma omp parallel shared(x)
• pragma omp critical
• x x 1
• / end of parallel section /

30
barrier????

• barrier???????????????????
• ??????????,??????
• barrier????????????????
• ????
• pragma omp barrier newline

31
barrier????

• barrier?????????

?? ??
if (x 0) pragma omp barrier if (x 0) pragma omp barrier
32
atomic????

• atomic???????????????????
• ????
• pragma omp atomic newline
• atomic?????

33
flush????

• flush?????????????,???????????????????
• ????
• pragma omp flush (list) newline
• flush?????????????,nowait????

barrier critical??????? ordered??????? parallel
???? for???? sections???? single????
34
ordered????

• ordered???????????????
• ??????????????ordered?????
• ?????for??parallel for????????
• ????
• pragma omp ordered newline

35
threadprivate??????

• threadprivate????????????????????????????
• ????????????????
• ????
• pragma omp threadprivate (list) newline

36
threadprivate??????

• int alpha10, beta10, i//eg3
• pragma omp threadprivate(alpha)
• int main ()
• / First parallel region /
• pragma omp parallel private(i,beta)
• for (i0 i lt 10 i)
• alphai betai i
• / Second parallel region /
• pragma omp parallel
• printf("alpha3 d and beta3d\n",alpha3,
  beta3)

37
???????

• ???????
• ???????
• private??
• shared??
• default??
• firstprivate??
• lastprivate??
• copyin??
• reduction??

38
private??

• private???????????????????? ?
• ????
• private(list)
• private?threadprivate??

PRIVATE THREADPRIVATE
???? ?? ??
?? ???????????? ????????????????
??? ? ?
??? ?????- ????????????? ???
??? ?? FIRSTPRIVATE ?? COPYIN
39
shared??

• shared???????????????????????
• ??????????????
• ????
• shared (list)

40
default??

• default??????????????????????????????????
• ????
• default (shared none)

41
firstprivate??

• firstprivate???private?????
• ?????????
• ????
• firstprivate (list)

42
lastprivate??

• lastprivate???private?????
• ?????????????????????
• ????
• lastprivate (list)

43
copyin??

• copyin??????????????threadprivate???????
• ?????????????
• ????
• copyin(list)

44
reduction??

• reduction???????????????????????
• ???,?????????????
• ?????????????????????????,??????????
• ????
• reduction (operator list)

45
reduction??

• include ltomp.hgt//eg4
• int main ()
• int i, n, chunk
• float a100, b100, result
• / Some initializations /
• n 100
• chunk 10
• result 0.0
• for (i0 i lt n i)
• ai i 1.0
• bi i 2.0
• pragma omp parallel for default(shared)
  private(i)\
• schedule(static,chunk) reduction(result)
• for (i0 i lt n i)
• result result (ai bi)
• printf("Final result f\n",result)

46
reduction??

• Reduction?????

xx op expr x expr op x (except subtraction) x
binop expr x x x-- --x
x????? expr??????x????????,????? binop?,,-,/,,
,??,????? op?,,-,/,,,,,or??,?????
47
??/????????
?? ???? ???? ???? ???? ???? ????
?? PARALLEL DO/for SECTIONS SINGLE PARALLEL DO/for PARALLEL SECTIONS
IF v v v
PRIVATE v v v v v v
SHARED v v v v
DEFAULT v v v
FIRSTPRIVATE v v v v v v
LASTPRIVATE v v v v
REDUCTION v v v v v
COPYIN v v v
SCHEDULE v v
ORDERED v v
NOWAIT v v v
48
?????????

• ????
• ??DO/for?SECTIONS?SINGLE?MASTER?BARRIER????????PAR
  ALLEL?,?????????,????????
• ??ORDERED???????DO/for???
• ??ATOMIC??ATOMIC?????????????,??????????
• ??CRITICAL???????CRITICAL???????,??????????
• ?PARALLEL???,????????????????

49
?????????

• ????
• PARALALL ??????????????,?????????????,????????????
  ????,?????????
• DO/for?SECTION?SINGLE????????PARALLEL
  ?,????????????
• DO/for?SECTION?SINGLE???????????CRITICAL?ORDERED?M
  ASTER??
• CRITICAL?????????
• BARRIER???????????DO/for?ORDERED?SECTIONS?SINGLE?M
  ASTER?CRITICAL??
• MASTER???????????DO/for?SECTIONS?SINGLE???
• ORDERED???????????CRITICAL??
• ????????PARALLEL ?????,???????????????????????????
  ,????????????

50
??????????

• ?????
• OpenMP???????????????????????
• ??C/C,???????????omp.h
• ????
• OMP_SCHEDULE????for,parallel for????????????????
• OMP_NUM_THREADS???????????
• OMP_DYNAMIC???????TRUE?FALSE,??????????????????
• OMP_NESTED??????????

51
OpenMP????

• ?????????????Pi??

52
OpenMP????

• C????????
• / Seriel Code /
• static long num_steps 100000
• double step
• void main ()
• int i
• double x, pi, sum 0.0
• step 1.0/(double) num_steps
• for (i0ilt num_steps i)
• x (i0.5)step
• sum sum 4.0/(1.0xx)
• pi step sum

53
OpenMP????

• ???????????//1
• include ltomp.hgt
• static long num_steps 100000
• double step
• define NUM_THREADS 2
• void main ()
• int i
• double x, pi, sumNUM_THREADS
• step 1.0/(double) num_steps
• omp_set_num_threads(NUM_THREADS) //
• pragma omp parallel
• double x
• int id
• id omp_get_thread_num()
• for (iid, sumid0.0ilt num_steps
  iiNUM_THREADS)//
• x (i0.5)step
• sumid 4.0/(1.0xx)

54
OpenMP????

• ?????????????? //2
• include ltomp.hgt
• static long num_steps 100000
• double step
• define NUM_THREADS 2
• void main ()
• int i
• double x, pi, sumNUM_THREADS
• step 1.0/(double) num_steps
• omp_set_num_threads(NUM_THREADS) //
• pragma omp parallel //
• double x
• int id
• id omp_get_thread_num()
• sumid 0 //
• pragma omp for//
• for (i0ilt num_steps i)

55
OpenMP????

• ??private???critical????????
• include ltomp.hgt
• static long num_steps 100000
• double step
• define NUM_THREADS 2
• void main ()
• int i
• double x, sum, pi0.0
• step 1.0/(double) num_steps
• omp_set_num_threads(NUM_THREADS)
• pragma omp parallel private (x, sum)
• id omp_get_thread_num()
• for (iid,sum0.0ilt num_stepsiiNUM_THREADS
  )
• x (i0.5)step
• sum 4.0/(1.0xx)
• pragma omp critical

56
OpenMP????

• ?????????????
• include ltomp.hgt
• static long num_steps 100000
• double step
• define NUM_THREADS 2
• void main ()
• int i
• double x, pi, sum 0.0
• step 1.0/(double) num_steps
• omp_set_num_threads(NUM_THREADS)
• pragma omp parallel for reduction(sum)
  private(x)
• for (i0iltnum_steps i)
• x (i0.5)step
• sum sum 4.0/(1.0xx)
• pi step sum