Hybrid programming with MPI & OpenMP - Prace Training Portal

More documents

Recommendations

Info

Gram-Smith kernel:dealing <strong>with</strong> thread private allocatable array!$omp parallel default(shared), private( temp, k, ig )ALLOCATE( temp( ngw ) )!$omp doDO k = 1, kmaxcsc(k) = 0.0d0IF ( ispin(i) .EQ. ispin(k) ) THENDO ig = 1, ngwtemp(ig) = cp(1,ig,k) * cp(1,ig,i) + cp(2,ig,k) * cp(2,ig,i)END DOcsc(k) = 2.0d0 * SUM(temp)IF (gstart == 2) csc(k) = csc(k) - temp(1)ENDIFEND DO!$omp end doDEALLOCATE( temp )!$omp end parallel32
pDO nt = 1, ntypIF ( upf(nt)%tvanp ) THEN !DO ih = 1, nh(nt) !DO jh = ih, nh(nt)!CALL qvan2( ngm, ih, jh, nt, qmod, qgm, ylmk0 ) !!$omp parallel default(shared), private(na,qgm _ na,is,dtmp,ig,mytid,ntids)#ifdef __OPENMPmytid = omp_get_thread_num()take the thread IDntids = omp_get_num_threads()#endifExample of non trivialloop parallelizationtake the number of threadsALLOCATE( qgm_na( ngm ) ) !DO na = 1, nat !#ifdef __OPEN<strong>MPI</strong>F( MOD( na, ntids ) /= mytid ) CYCLEdistribute atoms round-robin to threads#endifIF ( ityp(na) == nt ) THENqgm_na(1:ngm) = qgm(1:ngm)*eigts1(ig1(1:ngm),na)*eigts2(ig2(1:ngm),na)*eigts3(ig3(1:ngm),na)DO is = 1, nspin_magdtmp = 0.0d0DO ig = 1, ngmdtmp = dtmp + aux( ig, is ) * CONJG( qgm_na( ig ) )END DO!$omp end parallelEND DOEND DOEND IFEND DOdeeq(ih,jh,na,is) , , = fact * omega * DBLE( dtmp )deeq(jh,ih,na,is) = deeq(ih,jh,na,is)END DOEND IFEND DODEALLOCATE( qgm_na )33
Page 1 and 2: workshop on application porting and
Page 3 and 4: Motivation:Paradigm Change in HPC30
Page 5 and 6: Standard de-factoMPI (inter node)Op
Page 7 and 8: Mixed Paradigm• PROS:- Better use
Page 10 and 11: Ab-initio simulations:The “standa
Page 12: Parallelization (before OpenMP)1x1x
Page 15 and 16: HECToR - PSIWAT (DEISA benchmark)Pe
Page 17 and 18: Parallelization Strategy• 3D FFT
Page 19 and 20: Changing architecture: HPCxHPCx - W
Page 21 and 22: Multithread libraries (mkl, acml, e
Page 23 and 24: Understanding QE 3DFFT, Paralleliza
Page 25 and 26: Understanding QE 3DFFT: master-only
Page 27 and 28: QE 3DFFT: multiple • Pros:- Overl
Page 29 and 30: Space Integrals: Electrostatic pote
Page 31: Point Function evaluation:Exchange
Page 35 and 36: PDSYEVX vs ad-hoc subroutineTurboRV
Page 37 and 38: Conclusion• Number of cores doubl

Hybrid programming with MPI & OpenMP - Prace Training Portal

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?