Skip to content

GitLab

Commit f515b7b2 authored by Andreas Marek's avatar Andreas Marek
Browse files

Rename successCUDA -> successGPU

parent f7a2f3d2
Hide whitespace changes
Inline Side-by-side
src/elpa1/elpa1_trans_ev_template.F90
View file @ f515b7b2
......@@ -139,7 +139,7 @@ subroutine trans_ev_&
integer(kind=c_intptr_t) :: num
integer(kind=C_intptr_T) :: q_dev, tmp_dev, hvm_dev, tmat_dev
integer(kind=ik) :: blockStep
logical :: successCUDA
logical :: successGPU
integer(kind=c_intptr_t), parameter :: size_of_datatype = size_of_&
&PRECISION&
&_&
......@@ -242,45 +242,45 @@ subroutine trans_ev_&
!&MATH_DATATYPE&
!&", "hvm1", istat, errorMessage)
num = (max_local_rows*max_stored_rows) * size_of_datatype
successCUDA = gpu_malloc_host(hvm1_host,num)
check_alloc_cuda("trans_ev: hvm1_host", successCUDA)
successGPU = gpu_malloc_host(hvm1_host,num)
check_alloc_cuda("trans_ev: hvm1_host", successGPU)
call c_f_pointer(hvm1_host,hvm1,(/(max_local_rows*max_stored_rows)/))
num = (max_stored_rows*max_stored_rows) * size_of_datatype
successCUDA = gpu_malloc_host(tmat_host,num)
check_alloc_cuda("trans_ev: tmat_host", successCUDA)
successGPU = gpu_malloc_host(tmat_host,num)
check_alloc_cuda("trans_ev: tmat_host", successGPU)
call c_f_pointer(tmat_host,tmat,(/max_stored_rows,max_stored_rows/))
num = (max_local_cols*max_stored_rows) * size_of_datatype
successCUDA = gpu_malloc_host(tmp1_host,num)
check_alloc_cuda("trans_ev: tmp1_host", successCUDA)
successGPU = gpu_malloc_host(tmp1_host,num)
check_alloc_cuda("trans_ev: tmp1_host", successGPU)
call c_f_pointer(tmp1_host,tmp1,(/(max_local_cols*max_stored_rows)/))
num = (max_local_cols*max_stored_rows) * size_of_datatype
successCUDA = gpu_malloc_host(tmp2_host,num)
check_alloc_cuda("trans_ev: tmp2_host", successCUDA)
successGPU = gpu_malloc_host(tmp2_host,num)
check_alloc_cuda("trans_ev: tmp2_host", successGPU)
call c_f_pointer(tmp2_host,tmp2,(/(max_local_cols*max_stored_rows)/))
successCUDA = gpu_malloc(tmat_dev, max_stored_rows * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successCUDA)
successGPU = gpu_malloc(tmat_dev, max_stored_rows * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successGPU)
successCUDA = gpu_malloc(hvm_dev, max_local_rows * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successCUDA)
successGPU = gpu_malloc(hvm_dev, max_local_rows * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successGPU)
successCUDA = gpu_malloc(tmp_dev, max_local_cols * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successCUDA)
successGPU = gpu_malloc(tmp_dev, max_local_cols * max_stored_rows * size_of_datatype)
check_alloc_cuda("trans_ev", successGPU)
num = ldq * matrixCols * size_of_datatype
successCUDA = gpu_malloc(q_dev, num)
check_alloc_cuda("trans_ev", successCUDA)
successGPU = gpu_malloc(q_dev, num)
check_alloc_cuda("trans_ev", successGPU)
successCUDA = gpu_host_register(int(loc(q_mat),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(q_mat),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("trans_ev: q_mat", successCUDA)
check_host_register_cuda("trans_ev: q_mat", successGPU)
successCUDA = gpu_memcpy(q_dev, int(loc(q_mat(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(q_dev, int(loc(q_mat(1,1)),kind=c_intptr_t), &
num, gpuMemcpyHostToDevice)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
endif ! useGPU
do istep = 1, na, blockStep
......@@ -387,15 +387,15 @@ subroutine trans_ev_&
hvm1(1:hvm_ubnd*nstor) = reshape(hvm(1:hvm_ubnd,1:nstor), (/ hvm_ubnd*nstor /))
!hvm_dev(1:hvm_ubnd*nstor) = hvm1(1:hvm_ubnd*nstor)
successCUDA = gpu_memcpy(hvm_dev, int(loc(hvm1(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(hvm_dev, int(loc(hvm1(1)),kind=c_intptr_t), &
hvm_ubnd * nstor * size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
!tmat_dev = tmat
successCUDA = gpu_memcpy(tmat_dev, int(loc(tmat(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(tmat_dev, int(loc(tmat(1,1)),kind=c_intptr_t), &
max_stored_rows * max_stored_rows * size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
endif
! Q = Q - V * T * V**T * Q
......@@ -421,8 +421,8 @@ subroutine trans_ev_&
else !l_rows>0
if (useGPU) then
successCUDA = gpu_memset(tmp_dev, 0, l_cols * nstor * size_of_datatype)
check_memcpy_cuda("trans_ev", successCUDA)
successGPU = gpu_memset(tmp_dev, 0, l_cols * nstor * size_of_datatype)
check_memcpy_cuda("trans_ev", successGPU)
else
tmp1(1:l_cols*nstor) = 0
endif
......@@ -432,9 +432,9 @@ subroutine trans_ev_&
! In the legacy GPU version, this allreduce was ommited. But probably it has to be done for GPU + MPI
! todo: does it need to be copied whole? Wouldn't be a part sufficient?
if (useGPU) then
successCUDA = gpu_memcpy(int(loc(tmp1(1)),kind=c_intptr_t), tmp_dev, &
successGPU = gpu_memcpy(int(loc(tmp1(1)),kind=c_intptr_t), tmp_dev, &
max_local_cols * max_stored_rows * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
endif
call obj%timer%start("mpi_communication")
call mpi_allreduce(tmp1, tmp2, int(nstor*l_cols,kind=MPI_KIND), MPI_MATH_DATATYPE_PRECISION, MPI_SUM, &
......@@ -442,9 +442,9 @@ subroutine trans_ev_&
call obj%timer%stop("mpi_communication")
! copy back tmp2 - after reduction...
if (useGPU) then
successCUDA = gpu_memcpy(tmp_dev, int(loc(tmp2(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(tmp_dev, int(loc(tmp2(1)),kind=c_intptr_t), &
max_local_cols * max_stored_rows * size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
endif ! useGPU
......@@ -498,27 +498,27 @@ subroutine trans_ev_&
if (useGPU) then
!q_mat = q_dev
successCUDA = gpu_memcpy(int(loc(q_mat(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(int(loc(q_mat(1,1)),kind=c_intptr_t), &
q_dev, ldq * matrixCols * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("trans_ev", successCUDA)
check_memcpy_cuda("trans_ev", successGPU)
successCUDA = gpu_host_unregister(int(loc(q_mat),kind=c_intptr_t))
check_host_unregister_cuda("trans_ev: q_mat", successCUDA)
successGPU = gpu_host_unregister(int(loc(q_mat),kind=c_intptr_t))
check_host_unregister_cuda("trans_ev: q_mat", successGPU)
successCUDA = gpu_free_host(hvm1_host)
check_host_dealloc_cuda("trans_ev: hvm1_host", successCUDA)
successGPU = gpu_free_host(hvm1_host)
check_host_dealloc_cuda("trans_ev: hvm1_host", successGPU)
nullify(hvm1)
successCUDA = gpu_free_host(tmat_host)
check_host_dealloc_cuda("trans_ev: tmat_host", successCUDA)
successGPU = gpu_free_host(tmat_host)
check_host_dealloc_cuda("trans_ev: tmat_host", successGPU)
nullify(tmat)
successCUDA = gpu_free_host(tmp1_host)
check_host_dealloc_cuda("trans_ev: tmp1_host", successCUDA)
successGPU = gpu_free_host(tmp1_host)
check_host_dealloc_cuda("trans_ev: tmp1_host", successGPU)
nullify(tmp1)
successCUDA = gpu_free_host(tmp2_host)
check_host_dealloc_cuda("trans_ev: tmp2_host", successCUDA)
successGPU = gpu_free_host(tmp2_host)
check_host_dealloc_cuda("trans_ev: tmp2_host", successGPU)
nullify(tmp2)
!deallocate(hvm1, stat=istat, errmsg=errorMessage)
......@@ -530,17 +530,17 @@ subroutine trans_ev_&
!endif
!deallocate(q_dev, tmp_dev, hvm_dev, tmat_dev)
successCUDA = gpu_free(q_dev)
check_dealloc_cuda("trans_ev", successCUDA)
successGPU = gpu_free(q_dev)
check_dealloc_cuda("trans_ev", successGPU)
successCUDA = gpu_free(tmp_dev)
check_dealloc_cuda("trans_ev", successCUDA)
successGPU = gpu_free(tmp_dev)
check_dealloc_cuda("trans_ev", successGPU)
successCUDA = gpu_free(hvm_dev)
check_dealloc_cuda("trans_ev", successCUDA)
successGPU = gpu_free(hvm_dev)
check_dealloc_cuda("trans_ev", successGPU)
successCUDA = gpu_free(tmat_dev)
check_dealloc_cuda("trans_ev", successCUDA)
successGPU = gpu_free(tmat_dev)
check_dealloc_cuda("trans_ev", successGPU)
else
deallocate(tmat, tmp1, tmp2, stat=istat, errmsg=errorMessage)
check_deallocate("trans_ev_&
......
src/elpa1/elpa1_tridiag_template.F90
View file @ f515b7b2
......@@ -139,7 +139,7 @@ subroutine tridiag_&
integer(kind=C_intptr_T) :: a_dev, v_row_dev, v_col_dev, u_row_dev, u_col_dev, vu_stored_rows_dev, &
uv_stored_cols_dev
logical :: successCUDA
logical :: successGPU
integer(kind=ik) :: istep, i, j, l_col_beg, l_col_end, l_row_beg, l_row_end
integer(kind=ik) :: tile_size, l_rows_per_tile, l_cols_per_tile
......@@ -290,52 +290,52 @@ subroutine tridiag_&
if (useGPU) then
num = (max_local_rows+1) * size_of_datatype
successCUDA = gpu_malloc_host(v_row_host, num)
check_host_alloc_cuda("tridiag: v_row_host", successCUDA)
successGPU = gpu_malloc_host(v_row_host, num)
check_host_alloc_cuda("tridiag: v_row_host", successGPU)
call c_f_pointer(v_row_host,v_row,(/(max_local_rows+1)/))
num = (max_local_cols) * size_of_datatype
successCUDA = gpu_malloc_host(v_col_host,num)
check_host_alloc_cuda("tridiag: v_col_host", successCUDA)
successGPU = gpu_malloc_host(v_col_host,num)
check_host_alloc_cuda("tridiag: v_col_host", successGPU)
call c_f_pointer(v_col_host,v_col,(/(max_local_cols)/))
num = (max_local_cols) * size_of_datatype
successCUDA = gpu_malloc_host(u_col_host,num)
check_host_alloc_cuda("tridiag: u_col_host", successCUDA)
successGPU = gpu_malloc_host(u_col_host,num)
check_host_alloc_cuda("tridiag: u_col_host", successGPU)
call c_f_pointer(u_col_host,u_col,(/(max_local_cols)/))
num = (max_local_rows) * size_of_datatype
successCUDA = gpu_malloc_host(u_row_host,num)
check_host_alloc_cuda("tridiag: u_row_host", successCUDA)
successGPU = gpu_malloc_host(u_row_host,num)
check_host_alloc_cuda("tridiag: u_row_host", successGPU)
call c_f_pointer(u_row_host,u_row,(/(max_local_rows)/))
num = (max_local_rows * 2*max_stored_uv) * size_of_datatype
successCUDA = gpu_host_register(int(loc(vu_stored_rows),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(vu_stored_rows),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("tridiag: vu_stored_roes", successCUDA)
check_host_register_cuda("tridiag: vu_stored_roes", successGPU)
num = (max_local_cols * 2*max_stored_uv) * size_of_datatype
successCUDA = gpu_host_register(int(loc(uv_stored_cols),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(uv_stored_cols),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("tridiag: uv_stored_cols", successCUDA)
check_host_register_cuda("tridiag: uv_stored_cols", successGPU)
#if defined(DOUBLE_PRECISION_REAL) || defined(DOUBLE_PRECISION_COMPLEX)
num = na * 8
#else
num = na * 4
#endif
successCUDA = gpu_host_register(int(loc(e_vec),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(e_vec),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("tridiag: e_vec", successCUDA)
check_host_register_cuda("tridiag: e_vec", successGPU)
#if defined(DOUBLE_PRECISION_REAL) || defined(DOUBLE_PRECISION_COMPLEX)
num = na * 8
#else
num = na * 4
#endif
successCUDA = gpu_host_register(int(loc(d_vec),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(d_vec),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("tridiag: d_vec", successCUDA)
check_host_register_cuda("tridiag: d_vec", successGPU)
else
allocate(v_row(max_local_rows+1), stat=istat, errmsg=errorMessage)
......@@ -373,24 +373,24 @@ subroutine tridiag_&
u_col = 0
if (useGPU) then
successCUDA = gpu_malloc(v_row_dev, max_local_rows * size_of_datatype)
check_alloc_cuda("tridiag: v_row_dev", successCUDA)
successGPU = gpu_malloc(v_row_dev, max_local_rows * size_of_datatype)
check_alloc_cuda("tridiag: v_row_dev", successGPU)
successCUDA = gpu_malloc(u_row_dev, max_local_rows * size_of_datatype)
successGPU = gpu_malloc(u_row_dev, max_local_rows * size_of_datatype)
check_alloc_cuda("tridiag: u_row_dev", successCUDA)
check_alloc_cuda("tridiag: u_row_dev", successGPU)
successCUDA = gpu_malloc(v_col_dev, max_local_cols * size_of_datatype)
check_alloc_cuda("tridiag: v_col_dev", successCUDA)
successGPU = gpu_malloc(v_col_dev, max_local_cols * size_of_datatype)
check_alloc_cuda("tridiag: v_col_dev", successGPU)
successCUDA = gpu_malloc(u_col_dev, max_local_cols * size_of_datatype)
check_alloc_cuda("tridiag: u_col_dev", successCUDA)
successGPU = gpu_malloc(u_col_dev, max_local_cols * size_of_datatype)
check_alloc_cuda("tridiag: u_col_dev", successGPU)
successCUDA = gpu_malloc(vu_stored_rows_dev, max_local_rows * 2 * max_stored_uv * size_of_datatype)
check_alloc_cuda("tridiag: vu_stored_rows_dev", successCUDA)
successGPU = gpu_malloc(vu_stored_rows_dev, max_local_rows * 2 * max_stored_uv * size_of_datatype)
check_alloc_cuda("tridiag: vu_stored_rows_dev", successGPU)
successCUDA = gpu_malloc(uv_stored_cols_dev, max_local_cols * 2 * max_stored_uv * size_of_datatype)
check_alloc_cuda("tridiag: vu_stored_rows_dev", successCUDA)
successGPU = gpu_malloc(uv_stored_cols_dev, max_local_cols * 2 * max_stored_uv * size_of_datatype)
check_alloc_cuda("tridiag: vu_stored_rows_dev", successGPU)
endif !useGPU
......@@ -416,16 +416,16 @@ subroutine tridiag_&
num = matrixRows * matrixCols * size_of_datatype
successCUDA = gpu_malloc(a_dev, num)
check_alloc_cuda("tridiag: a_dev", successCUDA)
successGPU = gpu_malloc(a_dev, num)
check_alloc_cuda("tridiag: a_dev", successGPU)
successCUDA = gpu_host_register(int(loc(a_mat),kind=c_intptr_t),num,&
successGPU = gpu_host_register(int(loc(a_mat),kind=c_intptr_t),num,&
gpuHostRegisterDefault)
check_host_register_cuda("tridiag: a_mat", successCUDA)
check_host_register_cuda("tridiag: a_mat", successGPU)
successCUDA = gpu_memcpy(a_dev, int(loc(a_mat(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(a_dev, int(loc(a_mat(1,1)),kind=c_intptr_t), &
num, gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: a_dev", successCUDA)
check_memcpy_cuda("tridiag: a_dev", successGPU)
endif
! main cycle of tridiagonalization
......@@ -448,12 +448,12 @@ subroutine tridiag_&
! copy l_cols + 1 column of A to v_row
if (useGPU) then
a_offset = l_cols * matrixRows * size_of_datatype
! we use v_row on the host at the moment! successCUDA = cuda_memcpy(v_row_dev, a_dev + a_offset,
! we use v_row on the host at the moment! successGPU = cuda_memcpy(v_row_dev, a_dev + a_offset,
! (l_rows)*size_of_PRECISION_real, cudaMemcpyDeviceToDevice)
successCUDA = gpu_memcpy(int(loc(v_row),kind=c_intptr_t), &
successGPU = gpu_memcpy(int(loc(v_row),kind=c_intptr_t), &
a_dev + a_offset, (l_rows)* size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag a_dev 1", successCUDA)
check_memcpy_cuda("tridiag a_dev 1", successGPU)
else
v_row(1:l_rows) = a_mat(1:l_rows,l_cols+1)
endif
......@@ -564,20 +564,20 @@ subroutine tridiag_&
u_row(1:l_rows) = 0
if (l_rows > 0 .and. l_cols> 0 ) then
if (useGPU) then
successCUDA = gpu_memset(u_col_dev, 0, l_cols * size_of_datatype)
check_memcpy_cuda("tridiag: u_col_dev", successCUDA)
successGPU = gpu_memset(u_col_dev, 0, l_cols * size_of_datatype)
check_memcpy_cuda("tridiag: u_col_dev", successGPU)
successCUDA = gpu_memset(u_row_dev, 0, l_rows * size_of_datatype)
check_memcpy_cuda("tridiag: u_row_dev", successCUDA)
successGPU = gpu_memset(u_row_dev, 0, l_rows * size_of_datatype)
check_memcpy_cuda("tridiag: u_row_dev", successGPU)
successCUDA = gpu_memcpy(v_col_dev, int(loc(v_col(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(v_col_dev, int(loc(v_col(1)),kind=c_intptr_t), &
l_cols * size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: v_col_dev", successCUDA)
check_memcpy_cuda("tridiag: v_col_dev", successGPU)
successCUDA = gpu_memcpy(v_row_dev, int(loc(v_row(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(v_row_dev, int(loc(v_row(1)),kind=c_intptr_t), &
l_rows * size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: v_row_dev", successCUDA)
check_memcpy_cuda("tridiag: v_row_dev", successGPU)
endif ! useGU
#ifdef WITH_OPENMP_TRADITIONAL
......@@ -589,7 +589,7 @@ subroutine tridiag_&
!$omp num_threads(max_threads) &
!$omp default(none) &
!$omp private(my_thread,n_threads,n_iter,i,l_col_beg,l_col_end,j,l_row_beg,l_row_end) &
!$omp shared(useGPU, isSkewsymmetric, gpuMemcpyDeviceToHost, successCuda, u_row, u_row_dev, &
!$omp shared(useGPU, isSkewsymmetric, gpuMemcpyDeviceToHost, successGPU, u_row, u_row_dev, &
!$omp & v_row, v_row_dev, v_col, v_col_dev, u_col, u_col_dev, a_dev, a_offset, &
!$omp& max_local_cols, max_local_rows, obj, wantDebug, l_rows_per_tile, l_cols_per_tile, &
!$omp& matrixRows, istep, tile_size, l_rows, l_cols, ur_p, uc_p, a_mat)
......@@ -741,13 +741,13 @@ subroutine tridiag_&
enddo
end if !multiplication as one block / per stripes
successCUDA = gpu_memcpy(int(loc(u_col(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(int(loc(u_col(1)),kind=c_intptr_t), &
u_col_dev, l_cols * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: u_col_dev 1", successCUDA)
check_memcpy_cuda("tridiag: u_col_dev 1", successGPU)
successCUDA = gpu_memcpy(int(loc(u_row(1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(int(loc(u_row(1)),kind=c_intptr_t), &
u_row_dev, l_rows * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: u_row_dev 1", successCUDA)
check_memcpy_cuda("tridiag: u_row_dev 1", successGPU)
endif ! useGPU
#ifdef WITH_OPENMP_TRADITIONAL
......@@ -873,15 +873,15 @@ subroutine tridiag_&
if (n_stored_vecs == max_stored_uv .or. istep == 3) then
if (useGPU) then
successCUDA = gpu_memcpy(vu_stored_rows_dev, int(loc(vu_stored_rows(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(vu_stored_rows_dev, int(loc(vu_stored_rows(1,1)),kind=c_intptr_t), &
max_local_rows * 2 * max_stored_uv * &
size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: uv_stored_rows_dev", successCUDA)
check_memcpy_cuda("tridiag: uv_stored_rows_dev", successGPU)
successCUDA = gpu_memcpy(uv_stored_cols_dev, int(loc(uv_stored_cols(1,1)),kind=c_intptr_t), &
successGPU = gpu_memcpy(uv_stored_cols_dev, int(loc(uv_stored_cols(1,1)),kind=c_intptr_t), &
max_local_cols * 2 * max_stored_uv * &
size_of_datatype, gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: uv_stored_cols_dev", successCUDA)
check_memcpy_cuda("tridiag: uv_stored_cols_dev", successGPU)
endif
do i = 0, (istep-2)/tile_size
......@@ -944,9 +944,9 @@ subroutine tridiag_&
!a_mat(l_rows,l_cols) = a_dev(l_rows,l_cols)
a_offset = ((l_rows - 1) + matrixRows * (l_cols - 1)) * size_of_datatype
successCUDA = gpu_memcpy(int(loc(a_mat(l_rows, l_cols)),kind=c_intptr_t), a_dev + a_offset, &
successGPU = gpu_memcpy(int(loc(a_mat(l_rows, l_cols)),kind=c_intptr_t), a_dev + a_offset, &
1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 3", successCUDA)
check_memcpy_cuda("tridiag: a_dev 3", successGPU)
endif
if (n_stored_vecs > 0) then
......@@ -966,12 +966,12 @@ subroutine tridiag_&
if (useGPU) then
!a_dev(l_rows,l_cols) = a_mat(l_rows,l_cols)
!successCUDA = cuda_threadsynchronize()
!check_memcpy_cuda("tridiag: a_dev 4a5a", successCUDA)
!successGPU = cuda_threadsynchronize()
!check_memcpy_cuda("tridiag: a_dev 4a5a", successGPU)
successCUDA = gpu_memcpy(a_dev + a_offset, int(loc(a_mat(l_rows, l_cols)),kind=c_intptr_t), &
successGPU = gpu_memcpy(a_dev + a_offset, int(loc(a_mat(l_rows, l_cols)),kind=c_intptr_t), &
int(1 * size_of_datatype, kind=c_intptr_t), gpuMemcpyHostToDevice)
check_memcpy_cuda("tridiag: a_dev 4", successCUDA)
check_memcpy_cuda("tridiag: a_dev 4", successGPU)
endif
endif
......@@ -984,9 +984,9 @@ subroutine tridiag_&
if (my_prow==prow(1, nblk, np_rows)) then
! We use last l_cols value of loop above
if (useGPU) then
successCUDA = gpu_memcpy(int(loc(aux3(1)),kind=c_intptr_t), a_dev + (matrixRows * (l_cols - 1)) * size_of_datatype, &
successGPU = gpu_memcpy(int(loc(aux3(1)),kind=c_intptr_t), a_dev + (matrixRows * (l_cols - 1)) * size_of_datatype, &
1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 5", successCUDA)
check_memcpy_cuda("tridiag: a_dev 5", successGPU)
vrl = aux3(1)
else !useGPU
vrl = a_mat(1,l_cols)
......@@ -1020,9 +1020,9 @@ subroutine tridiag_&
#endif /* WITH_MPI */
if (my_prow == prow(1, nblk, np_rows) .and. my_pcol == pcol(1, nblk, np_cols)) then
if (useGPU) then
successCUDA = gpu_memcpy(int(loc(aux3(1)),kind=c_intptr_t), a_dev, &
successGPU = gpu_memcpy(int(loc(aux3(1)),kind=c_intptr_t), a_dev, &
1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 6", successCUDA)
check_memcpy_cuda("tridiag: a_dev 6", successGPU)
d_vec(1) = PRECISION_REAL(aux3(1))
else !useGPU
d_vec(1) = PRECISION_REAL(a_mat(1,1))
......@@ -1036,9 +1036,9 @@ subroutine tridiag_&
if (my_prow==prow(1, nblk, np_rows) .and. my_pcol==pcol(2, nblk, np_cols)) then
if (useGPU) then
successCUDA = gpu_memcpy(int(loc(e_vec(1)),kind=c_intptr_t), a_dev + (matrixRows * (l_cols - 1)) * size_of_datatype, &
successGPU = gpu_memcpy(int(loc(e_vec(1)),kind=c_intptr_t), a_dev + (matrixRows * (l_cols - 1)) * size_of_datatype, &
1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 7", successCUDA)
check_memcpy_cuda("tridiag: a_dev 7", successGPU)
else !useGPU
e_vec(1) = a_mat(1,l_cols) ! use last l_cols value of loop above
endif !useGPU
......@@ -1047,8 +1047,8 @@ subroutine tridiag_&
! Store d_vec(1)
if (my_prow==prow(1, nblk, np_rows) .and. my_pcol==pcol(1, nblk, np_cols)) then
if(useGPU) then
successCUDA = gpu_memcpy(int(loc(d_vec(1)),kind=c_intptr_t), a_dev, 1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 8", successCUDA)
successGPU = gpu_memcpy(int(loc(d_vec(1)),kind=c_intptr_t), a_dev, 1 * size_of_datatype, gpuMemcpyDeviceToHost)
check_memcpy_cuda("tridiag: a_dev 8", successGPU)
else !useGPU
if (isSkewsymmetric) then
d_vec(1) = 0.0_rk
......@@ -1064,26 +1064,26 @@ subroutine tridiag_&
if (useGPU) then
! todo: should we leave a_mat on the device for further use?
successCUDA = gpu_free(a_dev)
check_dealloc_cuda("tridiag: a_dev 9", successCUDA)
successGPU = gpu_free(a_dev)
check_dealloc_cuda("tridiag: a_dev 9", successGPU)
successCUDA = gpu_free(v_row_dev)
check_dealloc_cuda("tridiag: v_row_dev", successCUDA)
successGPU = gpu_free(v_row_dev)
check_dealloc_cuda("tridiag: v_row_dev", successGPU)
successCUDA = gpu_free(u_row_dev)
check_dealloc_cuda("tridiag: (u_row_dev", successCUDA)
successGPU = gpu_free(u_row_dev)
check_dealloc_cuda("tridiag: (u_row_dev", successGPU)
successCUDA = gpu_free(v_col_dev)
check_dealloc_cuda("tridiag: v_col_dev", successCUDA)
successGPU = gpu_free(v_col_dev)
check_dealloc_cuda("tridiag: v_col_dev", successGPU)
successCUDA = gpu_free(u_col_dev)
check_dealloc_cuda("tridiag: u_col_dev ", successCUDA)
successGPU = gpu_free(u_col_dev)
check_dealloc_cuda("tridiag: u_col_dev ", successGPU)
successCUDA = gpu_free(vu_stored_rows_dev)
check_dealloc_cuda("tridiag: vu_stored_rows_dev ", successCUDA)
successGPU = gpu_free(vu_stored_rows_dev)
check_dealloc_cuda("tridiag: vu_stored_rows_dev ", successGPU)
successCUDA = gpu_free(uv_stored_cols_dev)
check_dealloc_cuda("tridiag:uv_stored_cols_dev ", successCUDA)
successGPU = gpu_free(uv_stored_cols_dev)
check_dealloc_cuda("tridiag:uv_stored_cols_dev ", successGPU)
endif
! distribute the arrays d_vec and e_vec to all processors
......@@ -1112,36 +1112,36 @@ subroutine tridiag_&
check_deallocate("tridiag: tmp_real", istat, errorMessage)
if (useGPU) then
successCUDA = gpu_host_unregister(int(loc(a_mat),kind=c_intptr_t))
check_host_unregister_cuda("tridiag: a_mat", successCUDA)
successGPU = gpu_host_unregister(int(loc(a_mat),kind=c_intptr_t))
check_host_unregister_cuda("tridiag: a_mat", successGPU)
successCUDA = gpu_free_host(v_row_host)
check_host_dealloc_cuda("tridiag: v_row_host", successCUDA)
successGPU = gpu_free_host(v_row_host)
check_host_dealloc_cuda("tridiag: v_row_host", successGPU)
nullify(v_row)
successCUDA = gpu_free_host(v_col_host)
check_host_dealloc_cuda("tridiag: v_col_host", successCUDA)
successGPU = gpu_free_host(v_col_host)
check_host_dealloc_cuda("tridiag: v_col_host", successGPU)
nullify(v_col)
successCUDA = gpu_free_host(u_col_host)
check_host_dealloc_cuda("tridiag: u_col_host", successCUDA)
successGPU = gpu_free_host(u_col_host)
check_host_dealloc_cuda("tridiag: u_col_host", successGPU)
nullify(u_col)
successCUDA = gpu_free_host(u_row_host)
check_host_dealloc_cuda("tridiag: u_row_host", successCUDA)
successGPU = gpu_free_host(u_row_host)
check_host_dealloc_cuda