diff --git a/src/SAMPLING/srmc_step.irp.f b/src/SAMPLING/srmc_step.irp.f new file mode 100644 index 0000000..4cd823f --- /dev/null +++ b/src/SAMPLING/srmc_step.irp.f @@ -0,0 +1,306 @@ +! Providers of *_srmc_block_walk +!============================== +BEGIN_SHELL [ /usr/bin/python ] +from properties import * + +t = """ + BEGIN_PROVIDER [ $T, $X_srmc_block_walk $D1 ] +&BEGIN_PROVIDER [ $T, $X_srmc_block_walk_kahan $D2 ] +&BEGIN_PROVIDER [ $T, $X_2_srmc_block_walk $D1 ] +&BEGIN_PROVIDER [ $T, $X_2_srmc_block_walk_kahan $D2 ] + implicit none + BEGIN_DOC +! SRMC averages of $X. Computed in E_loc_srmc_block_walk + END_DOC + $X_srmc_block_walk = 0.d0 + $X_srmc_block_walk_kahan = 0.d0 + $X_2_srmc_block_walk = 0.d0 + $X_2_srmc_block_walk_kahan = 0.d0 +END_PROVIDER +""" +for p in properties: + if p[1] != 'e_loc': + if p[2] == "": + D1 = "" + D2 = ", (3)" + else: + D1 = ", ("+p[2][1:-1]+")" + D2 = ", ("+p[2][1:-1]+",3)" + print t.replace("$X",p[1]).replace("$T",p[0]).replace("$D1",D1).replace("$D2",D2) + +END_SHELL + + + + BEGIN_PROVIDER [ double precision, E_loc_srmc_block_walk ] +&BEGIN_PROVIDER [ double precision, E_loc_2_srmc_block_walk ] +&BEGIN_PROVIDER [ double precision, E_loc_srmc_block_walk_kahan, (3) ] +&BEGIN_PROVIDER [ double precision, E_loc_2_srmc_block_walk_kahan, (3) ] + implicit none + include '../types.F' + BEGIN_DOC +! Properties averaged over the block using the SRMC method + END_DOC + + real, allocatable :: elec_coord_tmp(:,:,:) + integer :: mod_align + double precision :: E_loc_save(walk_num_dmc_max) + double precision :: E_loc_save_tmp(walk_num_dmc_max) + double precision :: psi_value_save(walk_num) + double precision :: psi_value_save_tmp(walk_num) + double precision :: srmc_weight(walk_num) + !DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: E_loc_save + !DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: E_loc_save_tmp + !DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: psi_value_save + !DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: psi_value_save_tmp + !DIR$ ATTRIBUTES ALIGN : $IRP_ALIGN :: srmc_weight + allocate ( elec_coord_tmp(mod_align(elec_num+1),3,walk_num) ) + psi_value_save = 0.d0 + psi_value_save_tmp = 0.d0 + srmc_weight = 1.d0 + +! Initialization + if (vmc_algo /= t_Brownian) then + call abrt(irp_here,'SRMC should run with Brownian algorithm') + endif + + integer :: k, i_walk, i_step + +BEGIN_SHELL [ /usr/bin/python ] +from properties import * +t = """ + if (calc_$X) then + !DIR$ VECTOR ALIGNED + $X_srmc_block_walk = 0.d0 + !DIR$ VECTOR ALIGNED + $X_srmc_block_walk_kahan = 0.d0 + !DIR$ VECTOR ALIGNED + $X_2_srmc_block_walk = 0.d0 + !DIR$ VECTOR ALIGNED + $X_2_srmc_block_walk_kahan = 0.d0 + endif +""" +for p in properties: + print t.replace("$X",p[1]) +END_SHELL + + logical :: loop + integer*8 :: cpu0, cpu1, cpu2, count_rate, count_max + + loop = .True. + call system_clock(cpu0, count_rate, count_max) + cpu2 = cpu0 + + block_weight = 0.d0 + + real, external :: accep_rate + double precision :: delta, thr + + thr = 2.d0/time_step_sq + + + do while (loop) + + ! Every walker makes a step + do i_walk=1,walk_num + integer :: i,j,l + do l=1,3 + do i=1,elec_num+1 + elec_coord(i,l) = elec_coord_full(i,l,i_walk) + enddo + enddo + TOUCH elec_coord + + + double precision :: p,q + real :: delta_x + logical :: accepted + call brownian_step(p,q,accepted,delta_x) + + if ( psi_value * psi_value_save(i_walk) >= 0.d0 ) then + delta = ((E_loc+E_loc_save(i_walk))*0.5d0 - E_ref) * p + if ( delta > thr ) then + srmc_weight(i_walk) = dexp(-dtime_step*thr) + else if ( delta < -thr ) then + srmc_weight(i_walk) = dexp(dtime_step*thr) + else + srmc_weight(i_walk) = dexp(-dtime_step*delta) + endif + else + srmc_weight(i_walk) = 0.d0 + endif + + elec_coord(elec_num+1,1) += p*time_step + elec_coord(elec_num+1,2) = E_loc + elec_coord(elec_num+1,3) = srmc_weight(i_walk) + do l=1,3 + do i=1,elec_num+1 + elec_coord_full(i,l,i_walk) = elec_coord(i,l) + enddo + enddo + + psi_value_save(i_walk) = psi_value + E_loc_save(i_walk) = E_loc + +BEGIN_SHELL [ /usr/bin/python ] +from properties import * +t = """ + if (calc_$X) then + ! Kahan's summation algorithm to compute these sums reducing the rounding error: + ! $X_srmc_block_walk += $X * pop_weight_mult * srmc_weight(i_walk) + ! $X_2_srmc_block_walk += $X_2 * pop_weight_mult * srmc_weight(i_walk) + ! see http://en.wikipedia.org/wiki/Kahan_summation_algorithm + + $X_srmc_block_walk_kahan($D2 3) = $X * pop_weight_mult * srmc_weight(i_walk) - $X_srmc_block_walk_kahan($D2 1) + $X_srmc_block_walk_kahan($D2 2) = $X_srmc_block_walk $D1 + $X_srmc_block_walk_kahan($D2 3) + $X_srmc_block_walk_kahan($D2 1) = ($X_srmc_block_walk_kahan($D2 2) - $X_srmc_block_walk $D1 ) & + - $X_srmc_block_walk_kahan($D2 3) + $X_srmc_block_walk $D1 = $X_srmc_block_walk_kahan($D2 2) + + + $X_2_srmc_block_walk_kahan($D2 3) = $X_2 * pop_weight_mult * srmc_weight(i_walk) - $X_2_srmc_block_walk_kahan($D2 1) + $X_2_srmc_block_walk_kahan($D2 2) = $X_2_srmc_block_walk $D1 + $X_2_srmc_block_walk_kahan($D2 3) + $X_2_srmc_block_walk_kahan($D2 1) = ($X_2_srmc_block_walk_kahan($D2 2) - $X_2_srmc_block_walk $D1 ) & + - $X_2_srmc_block_walk_kahan($D2 3) + $X_2_srmc_block_walk $D1 = $X_2_srmc_block_walk_kahan($D2 2) + endif +""" +for p in properties: + if p[2] == "": + D1 = "" + D2 = "" + else: + D1 = "("+":"*(p[2].count(',')+1)+")" + D2 = ":"*(p[2].count(',')+1)+"," + print t.replace("$X",p[1]).replace("$D1",D1).replace("$D2",D2) + +END_SHELL + + block_weight += pop_weight_mult * srmc_weight(i_walk) + + enddo + + ! Move to the next projection step + if (srmc_projection > 0) then + srmc_projection_step = mod(srmc_projection_step,srmc_projection)+1 + else + srmc_projection_step = 1 + endif + + ! Eventually, recompute the weight of the population + if (srmc_projection_step == 1) then + pop_weight_mult = 1.d0 + do k=1,srmc_projection + pop_weight_mult *= pop_weight(k) + enddo + endif + + ! Remove contribution of the old value of the weight at the new + ! projection step + pop_weight_mult *= 1.d0/pop_weight(srmc_projection_step) + + ! Compute the new weight of the population + double precision :: sum_weight + sum_weight = 0.d0 + do k=1,walk_num + sum_weight += srmc_weight(k) + enddo + pop_weight(srmc_projection_step) = sum_weight/dble(walk_num) + + ! Update the running population weight + pop_weight_mult *= pop_weight(srmc_projection_step) + +! Reconfiguration + integer :: ipos(walk_num) + + call reconfigure(ipos,srmc_weight) + + do k=1,walk_num + do l=1,3 + do i=1,elec_num+1 + elec_coord_tmp(i,l,k) = elec_coord_full(i,l,k) + enddo + enddo + psi_value_save_tmp(k) = psi_value_save(k) + E_loc_save_tmp(k) = E_loc_save(k) + enddo + + integer :: ipm + do k=1,walk_num + ipm = ipos(k) + do l=1,3 + do i=1,elec_num+1 + elec_coord_full(i,l,k) = elec_coord_tmp(i,l,ipm) + enddo + enddo + psi_value_save(k) = psi_value_save_tmp(ipm) + E_loc_save(k) = E_loc_save_tmp(ipm) + enddo + + call system_clock(cpu1, count_rate, count_max) + if (cpu1 < cpu0) then + cpu1 = cpu1+cpu0 + endif + loop = dble(cpu1-cpu0) < dble(block_time)*dble(count_rate) + if (cpu1-cpu2 > count_rate) then + integer :: do_run + call get_running(do_run) + loop = do_run == t_Running + cpu2 = cpu1 + endif + + SOFT_TOUCH elec_coord_full psi_value psi_grad_psi_inv_x psi_grad_psi_inv_y psi_grad_psi_inv_z elec_coord pop_weight_mult + + enddo + + double precision :: factor + factor = 1.d0/block_weight + SOFT_TOUCH block_weight + +BEGIN_SHELL [ /usr/bin/python ] +from properties import * +t = """ + if (calc_$X) then + $X_srmc_block_walk *= factor + $X_2_srmc_block_walk *= factor + endif +""" +for p in properties: + print t.replace("$X",p[1]) +END_SHELL + + deallocate ( elec_coord_tmp ) + +END_PROVIDER + + +BEGIN_PROVIDER [ double precision, pop_weight_mult ] + implicit none + BEGIN_DOC +! Population weight of SRMC + END_DOC + pop_weight_mult = pop_weight(srmc_projection) +END_PROVIDER + + BEGIN_PROVIDER [ integer, srmc_projection ] +&BEGIN_PROVIDER [ integer, srmc_projection_step ] + implicit none + BEGIN_DOC +! Number of projection steps for SRMC + END_DOC + real :: srmc_projection_time + srmc_projection_time = 1. + call get_simulation_srmc_projection_time(srmc_projection_time) + srmc_projection = int( srmc_projection_time/time_step) + srmc_projection_step = 0 +END_PROVIDER + +BEGIN_PROVIDER [ double precision, pop_weight, (0:srmc_projection+1) ] + implicit none + BEGIN_DOC +! Population weight of SRMC + END_DOC + pop_weight = 1.d0 +END_PROVIDER + +