update

quanshengwu · May 7, 2024 · 1b2cd91 · 1b2cd91
1 parent 5ade846
commit 1b2cd91
Show file tree

Hide file tree

Showing 6 changed files with 86 additions and 28 deletions.
diff --git a/src/Makefile.intel-mpi b/src/Makefile.intel-mpi
@@ -1,5 +1,5 @@
-OBJ =  module.o sparse.o wt_aux.o math_lib.o mgmres.o symmetry.o \
-		 c_fortran_zgssv.o readHmnR.o inverse.o proteus.o \
+OBJ =  module.o sparse.o wt_aux.o math_lib.o symmetry.o \
+		 readHmnR.o inverse.o proteus.o \
        eigen.o ham_qlayer2qlayer.o psi.o unfolding.o rand.o \
   		 ham_slab.o ham_bulk.o ek_slab.o ek_bulk_polar.o ek_bulk.o \
        readinput.o fermisurface.o surfgreen.o surfstat.o \
@@ -13,11 +13,9 @@ OBJ =  module.o sparse.o wt_aux.o math_lib.o mgmres.o symmetry.o \
 	main.o
 
 # compiler
-F90  = mpif90 -fpp -DMPI -fpe3 -O3 # -DINTELMKL -DSUPERLU #  -check all -traceback -g
-#F90  = ifort -fpp  -DINTELMKL -fpe3 -check all -traceback -g
-CC = mpicc -fpp -DMPI -fpe3 -O3 # -DINTELMKL -DSUPERLU
+F90  = mpif90 -fpp -DMPI -fpe3 -O3 -DARPACK -DINTELMKL  #  -check all -traceback -g
 
-INCLUDE = #-I${MKLROOT}/include -I/Users/user/quan/work/workplace/github-repositories/superlu/SRC
+INCLUDE = -I${MKLROOT}/include
 WFLAG = -nogen-interface
 OFLAG = -O3 -static-intel
 FFLAG = $(OFLAG) $(WFLAG)
@@ -26,9 +24,6 @@ LFLAG = $(OFLAG)
 # ARPACK LIBRARY
 ARPACK=/Users/user/quan/work/workplace/ARPACK/libarpack_MAC.a
 
-# need to specify if -DSUPERLU privided in the F90 
-SUPERLULIB=#/usr/local/lib/libsuperlu.a
-
 # blas and lapack libraries
 # static linking
 #LIBS = -Wl,--start-group ${MKLROOT}/lib/intel64/libmkl_intel_lp64.a \
@@ -37,7 +32,7 @@ SUPERLULIB=#/usr/local/lib/libsuperlu.a
 		  ${ARPACK}
 
 # dynamic linking
-LIBS = ${ARPACK} -L/${MKLROOT}/lib/intel64 -lmkl_core -lmkl_sequential -lmkl_intel_lp64 -lpthread ${SUPERLULIB}
+LIBS = ${ARPACK} -L/${MKLROOT}/lib/intel64 -lmkl_core -lmkl_sequential -lmkl_intel_lp64 -lpthread 
 
 main : $(OBJ)
 	$(F90) $(LFLAG) $(OBJ) -o wt.x $(LIBS)
@@ -47,8 +42,6 @@ main : $(OBJ)
 
 .f90.o :
 	$(F90) $(FFLAG) $(INCLUDE) -c $*.f90
-.c.o:
-	$(CC) $(CFLAGS) -I$(INCLUDE) -c $< $(VERBOSE)
 
 clean :
 	rm -f *.o *.mod *~ wt.x

diff --git a/src/Makefile.intel-mpi-for-developer b/src/Makefile.intel-mpi-for-developer
@@ -0,0 +1,55 @@
+OBJ =  module.o sparse.o wt_aux.o math_lib.o symmetry.o \
+		 c_fortran_zgssv.o readHmnR.o inverse.o proteus.o \
+       eigen.o ham_qlayer2qlayer.o psi.o unfolding.o rand.o \
+  		 ham_slab.o ham_bulk.o ek_slab.o ek_bulk_polar.o ek_bulk.o \
+       readinput.o fermisurface.o surfgreen.o surfstat.o \
+  		 mat_mul.o ham_ribbon.o ek_ribbon.o \
+       fermiarc.o berrycurvature.o \
+  		 wanniercenter.o dos.o  orbital_momenta.o \
+  		 landau_level_sparse.o landau_level.o lanczos_sparse.o \
+		 berry.o wanniercenter_adaptive.o \
+  		 effective_mass.o findnodes.o \
+		 sigma_OHE.o sigma.o Boltz_transport_anomalous.o \
+	main.o
+
+# compiler
+F90  = mpif90 -fpp -DMPI -fpe3 -O3 -DARPACK #-DINTELMKL -DSUPERLU #  -check all -traceback -g
+#F90  = ifort -fpp  -DINTELMKL -fpe3 -check all -traceback -g
+CC = mpicc -fpp -O3 #-DINTELMKL -DSUPERLU
+
+INCLUDE = -I${MKLROOT}/include -I/Users/user/quan/work/workplace/github-repositories/superlu/SRC
+WFLAG = -nogen-interface
+OFLAG = -O3 -static-intel
+FFLAG = $(OFLAG) $(WFLAG)
+LFLAG = $(OFLAG)
+
+# ARPACK LIBRARY
+ARPACK=/Users/user/quan/work/workplace/ARPACK/libarpack_MAC.a
+
+# need to specify if -DSUPERLU privided in the F90 
+SUPERLULIB=/usr/local/lib/libsuperlu.a
+
+# blas and lapack libraries
+# static linking
+#LIBS = -Wl,--start-group ${MKLROOT}/lib/intel64/libmkl_intel_lp64.a \
+        ${MKLROOT}/lib/intel64/libmkl_sequential.a \
+        ${MKLROOT}/lib/intel64/libmkl_core.a -Wl,--end-group -lpthread -lm -ldl \
+		  ${ARPACK}
+
+# dynamic linking
+LIBS = ${ARPACK} -L/${MKLROOT}/lib/intel64 -lmkl_core -lmkl_sequential -lmkl_intel_lp64 -lpthread ${SUPERLULIB}
+
+main : $(OBJ)
+	$(F90) $(LFLAG) $(OBJ) -o wt.x $(LIBS)
+	cp -f wt.x ../bin
+
+.SUFFIXES: .o .f90 .c
+
+.f90.o :
+	$(F90) $(FFLAG) $(INCLUDE) -c $*.f90
+.c.o:
+	$(CC) $(CFLAGS) -I$(INCLUDE) -c $< $(VERBOSE)
+
+clean :
+	rm -f *.o *.mod *~ wt.x
+
diff --git a/src/module.f90 b/src/module.f90
@@ -587,13 +587,19 @@ module para
      !> specify the atom index that located on the top surface that you want to study
      integer :: topsurface_atom_index
      real(dp) :: shift_to_topsurface_cart(3)
+
+     !> a tag to control how do we call ARPACK to diagonalize a sparse matrix
+     !> value: zndrv1 using A*x
+     !>        zndrv2 using inv(A)
+     !> default "zndrv1"
+     character(20) :: arpack_solver
 
      !> namelist parameters
      namelist /PARAMETERS/ Eta_Arc,EF_broadening, OmegaNum, OmegaNum_unfold, OmegaMin, OmegaMax, &
         E_arc, Nk1, Nk2, Nk3, NP, Gap_threshold, Tmin, Tmax, NumT, &
         NBTau, BTauNum, BTauMax, Rcut, Magp, Magq, Magp_min, Magp_max, Nslice_BTau_Max, &
         wcc_neighbour_tol, wcc_calc_tol, Beta,NumLCZVecs, &
-        Relaxation_Time_Tau,  symprec, &
+        Relaxation_Time_Tau,  symprec, arpack_solver, &
         NumRandomConfs, NumSelectedEigenVals, projection_weight_mode, topsurface_atom_index
 
      real(Dp) :: E_fermi  ! Fermi energy, search E-fermi in OUTCAR for VASP, set to zero for Wien2k

diff --git a/src/readinput.f90 b/src/readinput.f90
@@ -568,6 +568,7 @@ subroutine readinput
    Beta= 100
    Relaxation_Time_Tau= 1d0  ! in ps
    topsurface_atom_index= 0
+   arpack_solver= 'zndrv1'
 
 
    !> by default, we only project on atoms for a given wave function

diff --git a/src/sigma_OHE.f90 b/src/sigma_OHE.f90
@@ -1814,7 +1814,7 @@ subroutine evolve_k_ohe
             open(unit=myfileindex(ib), file=evolvefilename)
             write(myfileindex(ib), '(a10, i5, a16, f16.8)')'# evolve k ', bands_fermi_level(ib), 'energy level', Ek(ib)
             write(myfileindex(ib), '("#", a13, 24a16)')'BTau (T.ps)', &
-               'OmegaTau', 'kx', 'ky', 'kz', 'vx', 'vy', 'vz', 'k1', 'k2','k3', 'Energy(ev)', "vx'", "vy'", "vz'"
+                'kx', 'ky', 'kz', 'vx', 'vy', 'vz', 'k1', 'k2','k3', 'Energy(ev)', "vx'", "vy'", "vz'"
          endif
       enddo
 #if defined (MPI)
@@ -1898,7 +1898,7 @@ subroutine evolve_k_ohe
                     call direct_cart_rec(kout_all(:, it, ik), k)
                     call project_k3_to_kplane_defined_by_direction(v_t, Bdirection, v_t2)
                     write(myfileindex(ib), '(100f16.8)')Btau*Magneticfluxdensity_atomic/Relaxation_Time_Tau, &
-                       Btau*Magneticfluxdensity_atomic/Relaxation_Time_Tau*0.175874356d0, k, v_t, kout_all(:, it, ik), E_iband, v_t2
+                        k, v_t, kout_all(:, it, ik), E_iband, v_t2
                  enddo
               endif
            endif

diff --git a/src/sparse.f90 b/src/sparse.f90
@@ -2302,12 +2302,19 @@ subroutine arpack_sparse_coo_eigs(ndims,nnzmax,nnz,acoo,jcoo,icoo,neval,nvecs,de
          zeigv= 0d0
 
 #if defined (INTELMKL)
+         !> zndrv2 needs a sparse solver to solve (A-sigma*I)*x=B with given A, sigma, and B
          !> get eigenvalues of a sparse matrix by calling arpack subroutine
          !> acoo, jcoo, icoo would be converted in to A-sigma*I, then converted into CSR format
-         call zmat_arpack_zndrv2(ndims, nnzmax, nnz, acoo, jcoo, icoo, sigma, neval, nvecs, deval, zeigv, ritzvec)
+         !> usually zndrv1 is about 10 times faster then zndrv2
+         if (arpack_solver=='zndrv2') then
+            call zmat_arpack_zndrv2(ndims, nnzmax, nnz, acoo, jcoo, icoo, sigma, neval, nvecs, deval, zeigv, ritzvec)
+         else
+            call zmat_arpack_zndrv1(ndims, nnzmax, nnz,  acoo, jcoo, icoo, sigma, neval, nvecs, deval, zeigv, ritzvec)
+         endif
 #else
          !> here acoo, icoo, jcoo are stored in COO format
          !> use matrix vector multiplication
+         !> zndrv1 needs a matrix vector multiplication operator A*x
          call zmat_arpack_zndrv1(ndims, nnzmax, nnz,  acoo, jcoo, icoo, sigma, neval, nvecs, deval, zeigv, ritzvec)
 #endif
 
@@ -2619,7 +2626,7 @@ subroutine zmat_arpack_zndrv4(ndims, nnzmax, nnz, acsr, jcsr, icsr, &
          call znaupd ( ido, bmat, n, which, nev, tol, resid, ncv, &
             zeigv, ldv, iparam, ipntr, workd, workl, lworkl, rwork, info )
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
          call now(time_end)
          time_cost_znaupd=  time_cost_znaupd+ time_end- time_start
@@ -2751,7 +2758,7 @@ subroutine zmat_arpack_zndrv4(ndims, nnzmax, nnz, acsr, jcsr, icsr, &
                zeigv, ldv, iparam, ipntr, workd, workl, lworkl, &
                rwork, ierr)
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
 !
 !        %----------------------------------------------%
@@ -3103,7 +3110,7 @@ subroutine zmat_arpack_zndrv3(ndims, nnzmax, nnz, acsr, jcsr, icsr, snnzmax, snn
          call znaupd ( ido, bmat, n, which, nev, tol, resid, ncv, &
             zeigv, ldv, iparam, ipntr, workd, workl, lworkl, rwork, info )
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
 
          iter = iter + 1
@@ -3197,7 +3204,7 @@ subroutine zmat_arpack_zndrv3(ndims, nnzmax, nnz, acsr, jcsr, icsr, snnzmax, snn
                zeigv, ldv, iparam, ipntr, workd, workl, lworkl, &
                rwork, ierr)
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
 !
 !        %----------------------------------------------%
@@ -3528,7 +3535,7 @@ subroutine zmat_arpack_zndrv1(ndims, nnzmax, nnz, acsr, jcsr, icsr, sigma, neval
          call znaupd ( ido, bmat, n, which, nev, tol, resid, ncv, &
             zeigv, ldv, iparam, ipntr, workd, workl, lworkl, rwork, info )
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
 
          iter = iter + 1
@@ -3662,7 +3669,7 @@ subroutine zmat_arpack_zndrv1(ndims, nnzmax, nnz, acsr, jcsr, icsr, sigma, neval
 !
             if ( info .eq. 1) then
                if (cpuid==0) write(stdout, *) ' '
-               if (cpuid==0) write(stdout, *) ' Maximum number of iterations reached.'
+               if (cpuid==0) write(stdout, *) ' Maximum number of iterations reached. try increasing NCV'
                if (cpuid==0) write(stdout, *) ' '
             else if ( info .eq. 3) then
                if (cpuid==0) write(stdout, *) ' '
@@ -3951,7 +3958,7 @@ subroutine zmat_arpack_zndrv2(ndims, nnzmax, nnz, acsr, jcsr, icsr, sigma,neval,
          call znaupd ( ido, bmat, n, which, nev, tol, resid, ncv, &
             zeigv, ldv, iparam, ipntr, workd, workl, lworkl, rwork,info )
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
 
          iter = iter + 1
@@ -3969,10 +3976,6 @@ subroutine zmat_arpack_zndrv2(ndims, nnzmax, nnz, acsr, jcsr, icsr, sigma,neval,
             !           | the result to workd(ipntr(2)).            |
             !           %-------------------------------------------%
             !
-            !           call zcopy( n, workd(ipntr(1)),1, workd(ipntr(2)), 1)
-            !
-            !           call zgttrs('N', n, 1, dl, dd, du, du2, ipiv, workd(ipntr(2)), n, ierr)
-            !call zmat_mkldss_zgesv(ndims, nnz, acsr, jcsr, icsr, workd(ipntr(1)), workd(ipntr(2)))
             call sparse_solver(ndims, nnz, acsr, icsr, jcsr, workd(ipntr(1)), workd(ipntr(2)))
             !
             !           %-----------------------------------------%
@@ -4019,7 +4022,7 @@ subroutine zmat_arpack_zndrv2(ndims, nnzmax, nnz, acsr, jcsr, icsr, sigma,neval,
                zeigv, ldv, iparam, ipntr, workd, workl, lworkl, &
                rwork, ierr)
 #else
-         STOP "ERROR : Please install WannierTools with ARPACK"
+         STOP "ERROR : Please install WannierTools with ARPACK since you are diagonalizing a large sparse matrix"
 #endif
             !
             !        %----------------------------------------------%