Skip to content

Commit

Permalink
Remove reduce from do concurrent loops
Browse files Browse the repository at this point in the history
  • Loading branch information
@gha3mi
gha3mi committed Jan 31, 2024
1 parent e46256d commit 2b0b679
Show file tree
Hide file tree
Showing 2 changed files with 14 additions and 14 deletions.
4 changes: 2 additions & 2 deletions fpm.rsp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ options --flag "-O3 -mtune=native -xHost -qmkl -qopenmp -ipo"
options test
options --compiler ifort
options --profile release
options --flag "-O3 -mtune=native -xHost -qmkl -qopenmp -ipo -coarray -coarray-num-images=4 -DUSE_COARRAY"
options --flag "-O3 -mtune=native -xHost -qmkl -qopenmp -DUSE_DO_CONCURRENT -ipo -coarray -coarray-num-images=4 -DUSE_COARRAY"

@ifx-test
options test
Expand All @@ -32,4 +32,4 @@ options --flag "-O3 -fast -march=native -mtune=native -stdpar=gpu,multicore -lbl
options test
options --compiler gfortran
options --profile release
options --flag "-O3 -march=native -lblas -fopenmp -flto"
options --flag "-O3 -march=native -lblas -fopenmp -DUSE_DO_CONCURRENT -flto"
24 changes: 12 additions & 12 deletions src/formatmul.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -324,7 +324,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,1), size(B,2)), source=0.0_rk)
call compute_block_ranges(size(B,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(:, se:ee) = &
Expand All @@ -343,7 +343,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,2), size(B,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(se:ee, :) = &
Expand All @@ -362,7 +362,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,2), size(B,2)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(se:ee, :) = &
Expand All @@ -381,7 +381,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,1), size(B,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(:, :) = C(:, :) + &
Expand All @@ -403,7 +403,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,2), size(B,2)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(se:ee, :) = &
Expand All @@ -422,7 +422,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,1), size(B,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(:, :) = C(:, :) + &
Expand All @@ -443,7 +443,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,1), size(B,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(:, :) = C(:, :) + &
Expand All @@ -462,7 +462,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,2), size(B,2)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(se:ee, :) = &
Expand All @@ -483,7 +483,7 @@ pure function mat_mat_block_rel(a, b, transA, transB, option, nblock) result(c)
allocate(C(size(A,1), size(B,2)), source=0.0_rk)
call compute_block_ranges(size(B,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:C)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
C(:, se:ee) = &
Expand Down Expand Up @@ -521,7 +521,7 @@ pure function mat_vec_block_rel(A, v, transA, option, nblock) result(w)
allocate(w(size(A,2)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:w)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
w(se:ee) = &
Expand All @@ -540,7 +540,7 @@ pure function mat_vec_block_rel(A, v, transA, option, nblock) result(w)
allocate(w(size(A,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:w)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
w(:) = &
Expand All @@ -560,7 +560,7 @@ pure function mat_vec_block_rel(A, v, transA, option, nblock) result(w)
allocate(w(size(A,1)), source=0.0_rk)
call compute_block_ranges(size(A,2), nblock, block_size, start_elem, end_elem)
#if defined(USE_DO_CONCURRENT)
do concurrent (ib = 1: nblock) reduce(+:w)
do concurrent (ib = 1: nblock)
se = start_elem(ib)
ee = end_elem(ib)
w(:) = &
Expand Down

0 comments on commit 2b0b679

Please sign in to comment.