Skip to content

Vertical loop blocking #423

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
Sebastiaan van de Bund (sebvdb) wants to merge 7 commits into
base: main
Choose a base branch
from
Draft

Vertical loop blocking #423

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
e0a8f5a
Add block size parameter constant
sebvdb Jan 14, 2026
d18dca6
Add blocking to vertical loop
sebvdb Jan 14, 2026
2fa2eaf
Fix license file reference in matrix_vector_kernel_mod.F90
sebvdb Jan 15, 2026
b54591c
Fix matrix vector vertical blocking incl. cleanup loop
sebvdb Feb 20, 2026
11c6950
Merge branch 'vertical_loop_blocking' of github.com:sebvdb/lfric_core…
sebvdb May 5, 2026
8ba90cd
Merge branch 'main' into vertical_loop_blocking
sebvdb May 5, 2026
812b8a8
Merge branch 'main' into vertical_loop_blocking
sebvdb May 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 26 additions & 15 deletions components/science/source/kernel/algebra/matrix_vector_kernel_mod.F90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ module matrix_vector_kernel_mod
GH_REAL, GH_READ, GH_INC, &
ANY_SPACE_1, ANY_SPACE_2, &
CELL_COLUMN
use constants_mod, only : i_def, r_single, r_double
use constants_mod, only : i_def, r_single, r_double, BLOCK_SIZE
use kernel_mod, only : kernel_type

implicit none
Expand Down Expand Up @@ -87,15 +87,20 @@ subroutine matrix_vector_code_r_single(cell, &
real(kind=r_single), dimension(ncell_3d,ndf1,ndf2), intent(in) :: matrix

! Internal variables
integer(kind=i_def) :: df, ik, df2, i1, i2, nl
integer(kind=i_def) :: df, ik, df2, i1, i2, nl, k, kk, kend

nl = nlayers-1
ik = (cell-1)*nlayers +1
do df2 = 1, ndf2
i2 = map2(df2)
do df = 1, ndf1
i1 = map1(df)
lhs(i1:i1+nl) = lhs(i1:i1+nl) + matrix(ik:ik+nl,df,df2)*x(i2:i2+nl)
do k = 0, nl, BLOCK_SIZE
ik = (cell-1)*nlayers + 1
kend = min(BLOCK_SIZE-1, nl-k)
do df2 = 1, ndf2
i2 = map2(df2)
do df = 1, ndf1
i1 = map1(df)
do kk = 0, kend
lhs(i1+kk+k) = lhs(i1+kk+k) + matrix(ik+kk+k,df,df2)*x(i2+kk+k)
end do
end do
end do
end do

Expand Down Expand Up @@ -125,15 +130,21 @@ subroutine matrix_vector_code_r_double(cell, &
real(kind=r_double), dimension(undf1), intent(inout) :: lhs
real(kind=r_double), dimension(ncell_3d,ndf1,ndf2), intent(in) :: matrix

integer(kind=i_def) :: df, ik, df2, i1, i2, nl
! Internal variables
integer(kind=i_def) :: df, ik, df2, i1, i2, nl, k, kk, kend

nl = nlayers-1
ik = (cell-1)*nlayers +1
do df2 = 1, ndf2
i2 = map2(df2)
do df = 1, ndf1
i1 = map1(df)
lhs(i1:i1+nl) = lhs(i1:i1+nl) + matrix(ik:ik+nl,df,df2)*x(i2:i2+nl)
do k = 0, nl, BLOCK_SIZE
ik = (cell-1)*nlayers + 1
kend = min(BLOCK_SIZE-1, nl-k)
do df2 = 1, ndf2
i2 = map2(df2)
do df = 1, ndf1
i1 = map1(df)
do kk = 0, kend
lhs(i1+kk+k) = lhs(i1+kk+k) + matrix(ik+kk+k,df,df2)*x(i2+kk+k)
end do
end do
end do
end do

Expand Down
7 changes: 6 additions & 1 deletion infrastructure/source/utilities/constants_mod.F90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ module constants_mod
LARGE_REAL_NEGATIVE, LARGE_DP_NEGATIVE, LARGE_REAL_POSITIVE, &
PI, degrees_to_radians, radians_to_degrees, &
cache_block, PRECISION_REAL, PRECISION_R_SOLVER, &
PRECISION_R_TRAN, EPS_R_TRAN, default_halo_depth
PRECISION_R_TRAN, EPS_R_TRAN, default_halo_depth, BLOCK_SIZE

! Define default application-defined kinds for all intrinsic data types

Expand Down Expand Up @@ -197,4 +197,9 @@ module constants_mod
integer(i_def), parameter :: default_halo_depth = 1 !< Default halo depth for fields and operators
!> @}

!> @name Vertical loop block size
!> @{
integer(i_def), parameter :: BLOCK_SIZE = 1024 !< Vertical OpenMP loop block size
!> @}

end module constants_mod
Loading