Accelerated algorithm for S2

INSTALL.rst

@@ -99,7 +99,7 @@ to Parameters (IRP) method.
   the downloaded archive in the :file:`${QP_ROOT}/external` directory
 
 * Extract the archive and go into the :file:`irpf90-*` directory to run
-  :command:`make``
+  :command:`make`
 
 * Create scripts to facilitate the access to :command:`irpf90` and
   :command:`irpman` as follows

config/bull.cfg

@@ -0,0 +1,61 @@
+# Common flags
+##############
+#
+# -mkl=[parallel|sequential] : Use the MKL library
+# --ninja                 : Allow the utilisation of ninja. It is mandatory !
+# --align=32                 : Align all provided arrays on a 32-byte boundary
+#
+[COMMON]
+FC           : mpiifort -shared-libgcc -shared-intel -fpic
+LAPACK_LIB   : -mkl=parallel 
+IRPF90       : irpf90
+IRPF90_FLAGS : --ninja --align=32 -DMPI  
+
+# Global options
+################
+#
+# 1 : Activate
+# 0 : Deactivate
+# 
+[OPTION]
+MODE    : OPT        ; [ OPT | PROFILE | DEBUG ] : Chooses the section below
+CACHE   : 1          ; Enable cache_compile.py
+OPENMP  : 1          ; Append OpenMP flags
+
+# Optimization flags
+####################
+#
+# -xHost                     : Compile a binary optimized for the current architecture
+# -O2                        : O3 not better than O2.
+# -ip                        : Inter-procedural optimizations
+# -ftz                       : Flushes denormal results to zero
+#
+[OPT]
+FCFLAGS  : -xCORE-AVX2 -O2 -ip -ftz -g -traceback -qopt-prefetch=5 -qopt-prefetch-issue-excl-hint -unroll-aggressive 
+# Profiling flags
+#################
+#
+[PROFILE]
+FC       : -p -g
+FCFLAGS  : -xSSE4.2 -O2 -ip -ftz -qopt-prefetch
+
+# Debugging flags
+#################
+#
+# -traceback   : Activate backtrace on runtime
+# -fpe0        : All floating point exaceptions
+# -C           : Checks uninitialized variables,  array subscripts, etc...
+# -g           : Extra debugging information
+# -xSSE2       : Valgrind needs a very simple x86 executable
+#
+[DEBUG]
+FC      : -g -traceback
+FCFLAGS : -xCORE-AVX2 -C -fpe0 -traceback
+
+# OpenMP flags
+#################
+#
+[OPENMP]
+FC           : -qopenmp
+IRPF90_FLAGS : --openmp
+

src/Determinants/occ_pattern.irp.f

@@ -61,13 +61,13 @@ subroutine occ_pattern_to_dets(o,d,sze,n_alpha,Nint)
   integer(bit_kind),intent(in)   :: o(Nint,2)      ! Occ patters
   integer(bit_kind),intent(out)  :: d(Nint,2,sze)  ! Output determinants
   
-  integer                        :: i, k, n, ispin
+  integer                        :: i, k, n, ispin, ispin2
 
   ! Extract list of singly occupied MOs as (int,pos) pairs
   ! ------------------------------------------------------
 
   integer           :: iint(2*n_alpha), ipos(2*n_alpha)
-  integer(bit_kind) :: v, t, tt
+  integer(bit_kind) :: v, t, tt, diff, v_prev
   integer           :: n_alpha_in_single
 
   n=0
@@ -84,26 +84,103 @@ subroutine occ_pattern_to_dets(o,d,sze,n_alpha,Nint)
   enddo
 
   v = shiftl(1,n_alpha_in_single) - 1
-  sze = int(binom_int(n,n_alpha_in_single),4)
-  do i=1,sze
-    ! Initialize with doubly occupied MOs
-    d(:,1,i) = o(:,2)
-    d(:,2,i) = o(:,2)
 
-    do k=1,n
-      if (btest(v,k-1)) then
-        ispin = 1
-      else
-        ispin = 2
-      endif
-      d(iint(k), ispin, i) = ibset( d(iint(k), ispin, i), ipos(k) )
+  sze = int(binom_int(n,n_alpha_in_single),4)
+  if (shiftl(n_alpha_in_single,1) == n) then
+
+    ! Initialize first determinant
+    d(:,1,1) = o(:,2)
+    d(:,2,1) = o(:,2)
+
+    do k=1,n_alpha_in_single
+      d(iint(k),1,1) = ibset( d(iint(k),1,1), ipos(k) )
     enddo
 
-    ! Generate next permutation with Anderson's algorithm
-    t = ior(v,v-1)
-    tt = t+1
-    v = ior(tt, shiftr( and(not(t),tt) - 1, trailz(v)+1) )
-  enddo
+    do k=n_alpha_in_single+1,n
+      d(iint(k),2,1) = ibset( d(iint(k),2,1), ipos(k) )
+    enddo
+
+    ! Time reversal symmetry
+    d(:,1,2) = d(:,2,1)
+    d(:,2,2) = d(:,1,1)
+
+    do i=3,sze,2
+      ! Generate next permutation with Anderson's algorithm
+      v_prev = v
+      t = ior(v,v-1)
+      tt = t+1
+      v = ior(tt, shiftr( and(not(t),tt) - 1, trailz(v)+1) )
+
+      ! Find what has changed between v_prev and v
+      diff = ieor(v,v_prev)
+
+      ! Initialize with previous determinant
+      d(:,1,i) = d(:,1,i-2)
+      d(:,2,i) = d(:,2,i-2)
+
+      ! Swap bits only where they have changed from v_prev to v
+      do while (diff /= 0_bit_kind)
+        k = trailz(diff)+1
+        if (btest(v,k-1)) then
+          d(iint(k),1,i) = ibset( d(iint(k),1,i), ipos(k) )
+          d(iint(k),2,i) = ibclr( d(iint(k),2,i), ipos(k) )
+        else
+          d(iint(k),1,i) = ibclr( d(iint(k),1,i), ipos(k) )
+          d(iint(k),2,i) = ibset( d(iint(k),2,i), ipos(k) )
+        endif
+        diff = iand(diff,diff-1_bit_kind)
+      enddo
+
+      ! Time reversal symmetry
+      d(:,1,i+1) = d(:,2,i)
+      d(:,2,i+1) = d(:,1,i)
+
+    enddo
+
+  else
+
+    ! Initialize first determinant
+    d(:,1,1) = o(:,2)
+    d(:,2,1) = o(:,2)
+
+    do k=1,n_alpha_in_single
+      d(iint(k),1,1) = ibset( d(iint(k),1,1), ipos(k) )
+    enddo
+
+    do k=n_alpha_in_single+1,n
+      d(iint(k),2,1) = ibset( d(iint(k),2,1), ipos(k) )
+    enddo
+
+    do i=2,sze
+      ! Generate next permutation with Anderson's algorithm
+      v_prev = v
+      t = ior(v,v-1)
+      tt = t+1
+      v = ior(tt, shiftr( and(not(t),tt) - 1, trailz(v)+1) )
+
+      ! Find what has changed between v_prev and v
+      diff = ieor(v,v_prev)
+
+      ! Initialize with previous determinant
+      d(:,1,i) = d(:,1,i-1)
+      d(:,2,i) = d(:,2,i-1)
+
+      ! Swap bits only where they have changed from v_prev to v
+      do while (diff /= 0_bit_kind)
+        k = trailz(diff)+1
+        if (btest(v,k-1)) then
+          d(iint(k),1,i) = ibset( d(iint(k),1,i), ipos(k) )
+          d(iint(k),2,i) = ibclr( d(iint(k),2,i), ipos(k) )
+        else
+          d(iint(k),1,i) = ibclr( d(iint(k),1,i), ipos(k) )
+          d(iint(k),2,i) = ibset( d(iint(k),2,i), ipos(k) )
+        endif
+        diff = iand(diff,diff-1_bit_kind)
+      enddo
+
+    enddo
+
+  endif
 
 end
 

src/Hartree_Fock/Roothaan_Hall_SCF.irp.f

@@ -115,7 +115,7 @@ END_DOC
       endif
       dim_DIIS=0
     enddo
-    level_shift = level_shift * 0.5d0
+    level_shift = level_shift * 0.75d0
     SOFT_TOUCH level_shift
     energy_SCF_previous = energy_SCF