Commit 5f26aa45 authored by Pavel Kus's avatar Pavel Kus Committed by Andreas Marek

further polishing elpa2_trans_ev_tridi_to_band_template

parent c20ab6b3
...@@ -2020,8 +2020,7 @@ ...@@ -2020,8 +2020,7 @@
&MATH_DATATYPE& &MATH_DATATYPE&
&_gpu_& &_gpu_&
&PRECISION& &PRECISION&
&( & &(row_group_dev, aIntern_dev, stripe_count, stripe_width, last_stripe_width, a_dim2, l_nev, &
row_group_dev, aIntern_dev, stripe_count, stripe_width, last_stripe_width, a_dim2, l_nev, &
row_group(:, :), j * nblk + a_off, row_group_size) row_group(:, :), j * nblk + a_off, row_group_size)
do i = 1, row_group_size do i = 1, row_group_size
...@@ -2031,24 +2030,18 @@ ...@@ -2031,24 +2030,18 @@
do i = 1, min(na - num_blk*nblk, nblk) do i = 1, min(na - num_blk*nblk, nblk)
#ifdef WITH_OPENMP #ifdef WITH_OPENMP
#if REALCASE == 1 call pack_row_&
call pack_row_real_cpu_openmp_& &MATH_DATATYPE&
#endif &_cpu_openmp_&
#if COMPLEXCASE == 1
call pack_row_complex_cpu_openmp_&
#endif
&PRECISION& &PRECISION&
&(obj,aIntern, row, j*nblk+i+a_off, stripe_width, stripe_count, max_threads, thread_width, l_nev) &(obj,aIntern, row, j*nblk+i+a_off, stripe_width, stripe_count, max_threads, thread_width, l_nev)
#else /* WITH_OPENMP */ #else /* WITH_OPENMP */
#if REALCASE == 1 call pack_row_&
call pack_row_real_cpu_& &MATH_DATATYPE&
#endif &_cpu_&
#if COMPLEXCASE == 1
call pack_row_complex_cpu_&
#endif
&PRECISION& &PRECISION&
&(obj,aIntern, row, j*nblk+i+a_off, stripe_width, last_stripe_width, stripe_count) &(obj,aIntern, row, j*nblk+i+a_off, stripe_width, last_stripe_width, stripe_count)
#endif /* WITH_OPENMP */ #endif /* WITH_OPENMP */
q((num_blk/np_rows)*nblk+i,1:l_nev) = row(:) q((num_blk/np_rows)*nblk+i,1:l_nev) = row(:)
enddo enddo
...@@ -2061,32 +2054,23 @@ ...@@ -2061,32 +2054,23 @@
&MATH_DATATYPE& &MATH_DATATYPE&
&_gpu_& &_gpu_&
&PRECISION& &PRECISION&
&( & &(row_group_dev, aIntern_dev, stripe_count, stripe_width, &
row_group_dev, aIntern_dev, stripe_count, stripe_width, &
last_stripe_width, a_dim2, l_nev, & last_stripe_width, a_dim2, l_nev, &
result_buffer(:, :, nbuf), j * nblk + a_off, nblk) result_buffer(:, :, nbuf), j * nblk + a_off, nblk)
else ! useGPU else ! useGPU
do i = 1, nblk do i = 1, nblk
#if WITH_OPENMP #if WITH_OPENMP
call pack_row_&
#if REALCASE == 1 &MATH_DATATYPE&
call pack_row_real_cpu_openmp_& &_cpu_openmp_&
#endif &PRECISION&
#if COMPLEXCASE == 1 &(obj,aIntern, result_buffer(:,i,nbuf), j*nblk+i+a_off, stripe_width, stripe_count, &
call pack_row_complex_cpu_openmp_&
#endif
&PRECISION&
&(obj,aIntern, result_buffer(:,i,nbuf), j*nblk+i+a_off, stripe_width, stripe_count, &
max_threads, thread_width, l_nev) max_threads, thread_width, l_nev)
#else /* WITH_OPENMP */ #else /* WITH_OPENMP */
call pack_row_&
#if REALCASE == 1 &MATH_DATATYPE&
call pack_row_real_cpu_& &_cpu_&
#endif
#if COMPLEXCASE == 1
call pack_row_complex_cpu_&
#endif
&PRECISION& &PRECISION&
&(obj, aIntern, result_buffer(:,i,nbuf),j*nblk+i+a_off, stripe_width, last_stripe_width, stripe_count) &(obj, aIntern, result_buffer(:,i,nbuf),j*nblk+i+a_off, stripe_width, last_stripe_width, stripe_count)
#endif /* WITH_OPENMP */ #endif /* WITH_OPENMP */
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment